Download as
Browse Code »

Commit 0e854e1e1dab832696362b8210d202475bd32de7

Authored by Erickson Silva
1 parent e1cf2b2f
Exists in master and in 1 other branch devel

Remove arquivos binários (dependências) desnecessarios

Showing 22 changed files with 0 additions and 9606 deletions   Show diff stats
aelius-install.tar.gz
View file @e1cf2b2
No preview for this file type
aelius_install.sh
View file @e1cf2b2
... ... @@ -1,36 +0,0 @@
1   -#!/bin/bash
2   -
3   -#Instalador do Tradutor Aelius
4   -
5   -cd ~/vlibras-core
6   -
7   -echo "\n# Extraindo...\n"
8   -tar -xf aelius-install.tar.gz -C ~/.
9   -
10   -echo "# Instalando dependências...\n"
11   -sudo apt-get install python-dev python-yaml python-numpy python-matplotlib
12   -
13   -cd ~/nltk-2.0.1rc1/
14   -
15   -sudo python setup.py install
16   -
17   -echo "\n# Finalizando...\n"
18   -
19   -# path to HunPos binaries
20   -echo "\nPATH=\"${PATH}:$HOME/Applications/bin\"" >> ~/.bashrc
21   -echo "export PATH\n" >> ~/.bashrc
22   -
23   -# path to Aelius and Translate package
24   -echo "PYTHONPATH=\"${PYTHONPATH}:$HOME/Applications:$HOME/vlibras-core/tradutor/src/py\"" >> ~/.bashrc
25   -echo "export PYTHONPATH\n" >> ~/.bashrc
26   -
27   -cd ..
28   -
29   -sudo mv usr-local-bin.tar.gz /usr/local/bin/
30   -
31   -cd /usr/local/bin/
32   -
33   -sudo tar -xf usr-local-bin.tar.gz
34   -sudo rm usr-local-bin.tar.gz
35   -
36   -echo "### Instalação finalizada! \n## Execute o seguinte comando para concluir:\n\n$ source ~/.bashrc\n"
recognize/src/julius/doc/Juliusbook-4.1.5.pdf
View file @e1cf2b2
No preview for this file type
recognize/src/julius/doc/manuals/accept_check.txt
View file @e1cf2b2
... ... @@ -1,80 +0,0 @@
1   - accept_check
2   -
3   -ACCEPT_CHECK(1) ACCEPT_CHECK(1)
4   -
5   -
6   -
7   -NAME
8   - accept_check
9   - - Check whether a grammar accept / reject given word sequences
10   -
11   -SYNOPSIS
12   - accept_check [-t] [-s spname] [-v] {prefix}
13   -
14   -DESCRIPTION
15   - accept_check is a tool to check whether a sentence can be accepted or
16   - rejected on a grammar (prefix.dfa and prefix.dict). The sentence should
17   - be given from standard input. You can do a batch check by preparing all
18   - test sentence at each line of a text file, and give it as standard
19   - input of accept_check.
20   -
21   - This tool needs .dfa, .dict and .term files. You should convert a
22   - written grammar file to generate them by mkdfa.pl.
23   -
24   - A sentence should be given as space-separated word sequence. It may be
25   - required to add head / tail silence word like sil, depending on your
26   - grammar. And should not contain a short-pause word.
27   -
28   - When a word belongs to various category in a grammar, accept_check will
29   - check all the possible sentence patterns, and accept it if any of those
30   - is acceptable.
31   -
32   -OPTIONS
33   - -t
34   - Use category name as input instead of word.
35   -
36   - -s spname
37   - Short-pause word name to be skipped. (default: "sp")
38   -
39   - -v
40   - Debug output.
41   -
42   -EXAMPLES
43   - An output for "date" grammar:
44   -
45   - % echo '<s> NEXT SUNDAY </s>' | accept_check date
46   - Reading in dictionary...
47   - 143 words...done
48   - Reading in DFA grammar...done
49   - Mapping dict item <-> DFA terminal (category)...done
50   - Reading in term file (optional)...done
51   - 27 categories, 143 words
52   - DFA has 35 nodes and 71 arcs
53   - -----
54   - wseq: <s> NEXT SUNDAY </s>
55   - cate: NS_B (NEXT|NEXT) (DAYOFWEEK|DAYOFWEEK|DAY|DAY) NS_E
56   - accepted
57   -
58   -
59   -SEE ALSO
60   - mkdfa.pl ( 1 ) ,
61   - generate ( 1 ) ,
62   - nextword ( 1 )
63   -
64   -COPYRIGHT
65   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
66   -
67   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
68   -
69   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
70   - Technology
71   -
72   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
73   - Technology
74   -
75   -LICENSE
76   - The same as Julius.
77   -
78   -
79   -
80   - 10/02/2008 ACCEPT_CHECK(1)
recognize/src/julius/doc/manuals/adinrec.txt
View file @e1cf2b2
... ... @@ -1,130 +0,0 @@
1   - adinrec
2   -
3   -ADINREC(1) ADINREC(1)
4   -
5   -
6   -
7   -NAME
8   - adinrec
9   - - record audio device and save one utterance to a file
10   -
11   -SYNOPSIS
12   - adinrec [options...] {filename}
13   -
14   -DESCRIPTION
15   - adinrec opens an audio stream, detects an utterance input and store it
16   - to a specified file. The utterance detection is done by level and
17   - zero-cross thresholds. Default input device is microphone, but other
18   - audio input source, including Julius A/D-in plugin, can be used by
19   - using "-input" option.
20   -
21   - The audio format is 16 bit, 1 channel, in Microsoft WAV format. If the
22   - given filename already exists, it will be overridden.
23   -
24   - If filename is "-" , the captured data will be streamed into standard
25   - out, with no header (raw format).
26   -
27   -OPTIONS
28   - adinrec uses JuliusLib and adopts Julius options. Below is a list of
29   - valid options.
30   -
31   - adinrec specific options
32   - -freq Hz
33   - Set sampling rate in Hz. (default: 16,000)
34   -
35   - -raw
36   - Output in raw file format.
37   -
38   - JuliusLib options
39   - -input {mic|rawfile|adinnet|stdin|netaudio|esd|alsa|oss}
40   - Choose speech input source. Specify 'file' or 'rawfile' for waveform
41   - file. On file input, users will be prompted to enter the file name
42   - from stdin.
43   -
44   - 'mic' is to get audio input from a default live microphone device,
45   - and 'adinnet' means receiving waveform data via tcpip network from
46   - an adinnet client. 'netaudio' is from DatLink/NetAudio input, and
47   - 'stdin' means data input from standard input.
48   -
49   - At Linux, you can choose API at run time by specifying alsa, oss and
50   - esd.
51   -
52   - -lv thres
53   - Level threshold for speech input detection. Values should be in
54   - range from 0 to 32767. (default: 2000)
55   -
56   - -zc thres
57   - Zero crossing threshold per second. Only input that goes over the
58   - level threshold (-lv) will be counted. (default: 60)
59   -
60   - -headmargin msec
61   - Silence margin at the start of speech segment in milliseconds.
62   - (default: 300)
63   -
64   - -tailmargin msec
65   - Silence margin at the end of speech segment in milliseconds.
66   - (default: 400)
67   -
68   - -zmean
69   - This option enables DC offset removal.
70   -
71   - -smpFreq Hz
72   - Set sampling rate in Hz. (default: 16,000)
73   -
74   - -48
75   - Record input with 48kHz sampling, and down-sample it to 16kHz
76   - on-the-fly. This option is valid for 16kHz model only. The
77   - down-sampling routine was ported from sptk. (Rev. 4.0)
78   -
79   - -NA devicename
80   - Host name for DatLink server input (-input netaudio).
81   -
82   - -adport port_number
83   - With -input adinnet, specify adinnet port number to listen.
84   - (default: 5530)
85   -
86   - -nostrip
87   - Julius by default removes successive zero samples in input speech
88   - data. This option stop it.
89   -
90   - -C jconffile
91   - Load a jconf file at here. The content of the jconffile will be
92   - expanded at this point.
93   -
94   - -plugindir dirlist
95   - Specify which directories to load plugin. If several direcotries
96   - exist, specify them by colon-separated list.
97   -
98   -ENVIRONMENT VARIABLES
99   - ALSADEV
100   - Device name string for ALSA. (default: "default")
101   -
102   - AUDIODEV
103   - Device name string for OSS. (default: "/dev/dsp")
104   -
105   - LATENCY_MSEC
106   - Input latency of microphone input in milliseconds. Smaller value
107   - will shorten latency but sometimes make process unstable. Default
108   - value will depend on the running OS.
109   -
110   -SEE ALSO
111   - julius ( 1 ) ,
112   - adintool ( 1 )
113   -
114   -COPYRIGHT
115   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
116   -
117   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
118   -
119   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
120   - Technology
121   -
122   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
123   - Technology
124   -
125   -LICENSE
126   - The same as Julius.
127   -
128   -
129   -
130   - 10/02/2008 ADINREC(1)
recognize/src/julius/doc/manuals/adintool.txt
View file @e1cf2b2
... ... @@ -1,214 +0,0 @@
1   - adintool
2   -
3   -ADINTOOL(1) ADINTOOL(1)
4   -
5   -
6   -
7   -NAME
8   - adintool
9   - - a tool to record / split / send / receive audio streams
10   -
11   -SYNOPSIS
12   - adintool {-in inputdev} {-out outputdev} [options...]
13   -
14   -DESCRIPTION
15   - adintool analyzes speech input, finds speech segments skipping silence,
16   - and records the detected segments in various ways. It performs speech
17   - detection based on zerocross number and power (level), and records the
18   - detected parts to files or other output devices sucessively.
19   -
20   -
21   - adintool is a upper version of adinrec with various functions.
22   - Supported input device are: microphone input, a speech file, standard
23   - tty input, and network socket (called adin-net server mode). Julius
24   - plugin can be also used. Detected speech segments will be saved to
25   - output devices: speech files, standard tty output, and network socket
26   - (called adin-net client mode). For example, you can split the incoming
27   - speech to segments and send them to Julius to be recognized.
28   -
29   - Output format is WAV, 16bit (signed short), monoral. If the file
30   - already exist, it will be overridden.
31   -
32   -OPTIONS
33   - All Julius options can be set. Only audio input related options are
34   - treated and others are silently skipped. Below is a list of options.
35   -
36   - adintool specific options
37   - -freq Hz
38   - Set sampling rate in Hz. (default: 16,000)
39   -
40   - -in inputdev
41   - Audio input device. "mic" to capture via microphone input, "file"
42   - for audio file input, and "stdin" to read raw data from
43   - standard-input. For file input, file name prompt will appear after
44   - startup. Use "adinnet" to make adintool as "adinnet server",
45   - receiving data from client via network socket. Default port number
46   - is 5530, which can be altered by option "-inport".
47   -
48   - Alternatively, input device can be set by "-input" option, in which
49   - case you can use plugin input.
50   -
51   - -out outputdev
52   - Audio output device store the data. Specify "file" to save to file,
53   - in which the output filename should be given by "-filename". Use
54   - "stdout" to standard out. "adinnet" will make adintool to be an
55   - adinnet client, sending speech data to a server via tcp/ip socket.
56   - When using "adinnet" output, the server name to send data should be
57   - specified by "-server". The default port number is 5530, which can
58   - be changed by "-port" option.
59   -
60   - -inport num
61   - When adintool becomes adinnet server to receive data (-in adinnet),
62   - set the port number to listen. (default: 5530)
63   -
64   - -server [host] [,host...]
65   - When output to adinnet server (-out adinnet), set the hostname. You
66   - can send to multiple hosts by specifying their hostnames as
67   - comma-delimited list like "host1,host2,host3".
68   -
69   - -port [num] [,num...]
70   - When adintool send a data to adinnet server (-out adinnet), set the
71   - port number to connect. (default: 5530) For multiple servers,
72   - specify port numbers for all servers like "5530,5530,5531".
73   -
74   - -filename file
75   - When output to file (-out file), set the output filename. The actual
76   - file name will be as "file.0000.wav" , "file.0001.wav" and so on,
77   - where the four digit number increases as speech segment detected.
78   - The initial number will be set to 0 by default, which can be changed
79   - by "-startid" option. When using "-oneshot" option to save only the
80   - first segment, the input will be saved as "file".
81   -
82   - -startid number
83   - At file output, set the initial file number. (default: 0)
84   -
85   - -oneshot
86   - Exit after the end of first speech segment.
87   -
88   - -nosegment
89   - Do not perform speech detection for input, just treat all the input
90   - as a single valid segment.
91   -
92   - -raw
93   - Output as RAW file (no header).
94   -
95   - -autopause
96   - When output to adinnet server, adintool enter pause state at every
97   - end of speech segment. It will restart when the destination adinnet
98   - server sends it a resume signal.
99   -
100   - -loosesync
101   - When output to multiple adinnet server, not to do strict
102   - synchronization for restart. By default, when adintool has entered
103   - pause state, it will not restart until resume commands are received
104   - from all servers. This option will allow restart at least one
105   - restart command has arrived.
106   -
107   - -rewind msec
108   - When input is a live microphone device, and there has been some
109   - continuing input at the moment adintool resumes, it start recording
110   - backtracking by the specified milliseconds.
111   -
112   - Concerning Julius options
113   - -input {mic|rawfile|adinnet|stdin|netaudio|esd|alsa|oss}
114   - Choose speech input source. Specify 'file' or 'rawfile' for waveform
115   - file. On file input, users will be prompted to enter the file name
116   - from stdin.
117   -
118   - 'mic' is to get audio input from a default live microphone device,
119   - and 'adinnet' means receiving waveform data via tcpip network from
120   - an adinnet client. 'netaudio' is from DatLink/NetAudio input, and
121   - 'stdin' means data input from standard input.
122   -
123   - At Linux, you can choose API at run time by specifying alsa, oss and
124   - esd.
125   -
126   - -lv thres
127   - Level threshold for speech input detection. Values should be in
128   - range from 0 to 32767. (default: 2000)
129   -
130   - -zc thres
131   - Zero crossing threshold per second. Only input that goes over the
132   - level threshold (-lv) will be counted. (default: 60)
133   -
134   - -headmargin msec
135   - Silence margin at the start of speech segment in milliseconds.
136   - (default: 300)
137   -
138   - -tailmargin msec
139   - Silence margin at the end of speech segment in milliseconds.
140   - (default: 400)
141   -
142   - -zmean
143   - This option enables DC offset removal.
144   -
145   - -smpFreq Hz
146   - Set sampling rate in Hz. (default: 16,000)
147   -
148   - -48
149   - Record input with 48kHz sampling, and down-sample it to 16kHz
150   - on-the-fly. This option is valid for 16kHz model only. The
151   - down-sampling routine was ported from sptk. (Rev. 4.0)
152   -
153   - -NA devicename
154   - Host name for DatLink server input (-input netaudio).
155   -
156   - -adport port_number
157   - With -input adinnet, specify adinnet port number to listen.
158   - (default: 5530)
159   -
160   - -nostrip
161   - Julius by default removes successive zero samples in input speech
162   - data. This option stop it.
163   -
164   - -C jconffile
165   - Load a jconf file at here. The content of the jconffile will be
166   - expanded at this point.
167   -
168   - -plugindir dirlist
169   - Specify which directories to load plugin. If several direcotries
170   - exist, specify them by colon-separated list.
171   -
172   -ENVIRONMENT VARIABLES
173   - ALSADEV
174   - (using mic input with alsa device) specify a capture device name. If
175   - not specified, "default" will be used.
176   -
177   - AUDIODEV
178   - (using mic input with oss device) specify a capture device path. If
179   - not specified, "/dev/dsp" will be used.
180   -
181   - LATENCY_MSEC
182   - Try to set input latency of microphone input in milliseconds.
183   - Smaller value will shorten latency but sometimes make process
184   - unstable. Default value will depend on the running OS.
185   -
186   -EXAMPLES
187   - Record microphone input to files: "data.0000.wav", "data.0001.wav" and
188   - so on:
189   - Split a long speech file "foobar.raw" into "foobar.1500.wav",
190   - "foobar.1501.wav" ...:
191   - Copy an entire audio file via network socket.
192   - Detect speech segment, send to Julius via network and recognize it:
193   -
194   -SEE ALSO
195   - julius ( 1 ) ,
196   - adinrec ( 1 )
197   -
198   -COPYRIGHT
199   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
200   -
201   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
202   -
203   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
204   - Technology
205   -
206   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
207   - Technology
208   -
209   -LICENSE
210   - The same as Julius.
211   -
212   -
213   -
214   - 10/02/2008 ADINTOOL(1)
recognize/src/julius/doc/manuals/dfa_determinize.txt
View file @e1cf2b2
... ... @@ -1,51 +0,0 @@
1   - dfa_determinize
2   -
3   -DFA_DETERMINIZE(1) DFA_DETERMINIZE(1)
4   -
5   -
6   -
7   -NAME
8   - dfa_determinize
9   - - Determinize NFA grammar network.
10   -
11   -SYNOPSIS
12   - dfa_determinize [-o outfile] {dfafile}
13   -
14   -DESCRIPTION
15   - dfa_determinize converts a non-deterministic .dfa file into
16   - deterministic DFA. Output to standard output, or file specified by "-o"
17   - option.
18   -
19   - This additional tool is not necessary on a grammar building procedure
20   - in Julius, since the grammar network generated by mkdfa.pl is always
21   - determinized.
22   -
23   -OPTIONS
24   - -o outfile
25   - Outout file. If not specified, output to stdout.
26   -
27   -EXAMPLES
28   - Determinize foo.dfa to bar.dfa:
29   - Another way:
30   -
31   -SEE ALSO
32   - mkdfa.pl ( 1 ) ,
33   - dfa_minimize ( 1 )
34   -
35   -COPYRIGHT
36   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
37   -
38   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
39   -
40   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
41   - Technology
42   -
43   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
44   - Technology
45   -
46   -LICENSE
47   - The same as Julius.
48   -
49   -
50   -
51   - 10/02/2008 DFA_DETERMINIZE(1)
recognize/src/julius/doc/manuals/dfa_minimize.txt
View file @e1cf2b2
... ... @@ -1,49 +0,0 @@
1   - dfa_minimize
2   -
3   -DFA_MINIMIZE(1) DFA_MINIMIZE(1)
4   -
5   -
6   -
7   -NAME
8   - dfa_minimize
9   - - Minimize a DFA grammar network
10   -
11   -SYNOPSIS
12   - dfa_minimize [-o outfile] {dfafile}
13   -
14   -DESCRIPTION
15   - dfa_minimize will convert an .dfa file to an equivalent minimal form.
16   - Output to standard output, or to a file specified by "-o" option.
17   -
18   - On version 3.5.3 and later, mkdfa.pl invokes this tool inside, and the
19   - output .dfa file will be always minimized, so you do not need to use
20   - this manually.
21   -
22   -OPTIONS
23   - -o outfile
24   - Output file. If not specified output to standard output.
25   -
26   -EXAMPLES
27   - Minimize foo.dfa to bar.dfa:
28   - Another way:
29   -
30   -SEE ALSO
31   - mkdfa.pl ( 1 )
32   -
33   -COPYRIGHT
34   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
35   -
36   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
37   -
38   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
39   - Technology
40   -
41   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
42   - Technology
43   -
44   -LICENSE
45   - The same as Julius.
46   -
47   -
48   -
49   - 10/02/2008 DFA_MINIMIZE(1)
recognize/src/julius/doc/manuals/generate-ngram.txt
View file @e1cf2b2
... ... @@ -1,62 +0,0 @@
1   - generate-ngram
2   -
3   -GENERATE-NGRAM(1) GENERATE-NGRAM(1)
4   -
5   -
6   -
7   -NAME
8   - generate-ngram
9   - - random sentence generator from N-gram
10   -
11   -SYNOPSIS
12   - generate-ngram [options...] {binary_ngram}
13   -
14   -DESCRIPTION
15   - generate-ngram is a tool to generate sentences randomly according to
16   - the given N-gram language model. The N-gram model file binary_ngram
17   - should be an binary format.
18   -
19   -OPTIONS
20   - -n num
21   - Number of sentences to generate (default: 10)
22   -
23   - -N
24   - Specify which length of N-gram to use (default: available max in the
25   - given model)
26   -
27   - -bos
28   - Beginning-of-sentence word (default: "<s>")
29   -
30   - -eos
31   - End-of-sentence word (default: "</s>")
32   -
33   - -ignore
34   - Specify a word to be supressed from output (default: "<UNK")
35   -
36   - -v
37   - Verbose output.
38   -
39   - -debug
40   - Debug output.
41   -
42   -SEE ALSO
43   - julius ( 1 ) ,
44   - mkbingram ( 1 )
45   -
46   -COPYRIGHT
47   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
48   -
49   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
50   -
51   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
52   - Technology
53   -
54   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
55   - Technology
56   -
57   -LICENSE
58   - The same as Julius.
59   -
60   -
61   -
62   - 10/02/2008 GENERATE-NGRAM(1)
recognize/src/julius/doc/manuals/generate.txt
View file @e1cf2b2
... ... @@ -1,76 +0,0 @@
1   - generate
2   -
3   -GENERATE(1) GENERATE(1)
4   -
5   -
6   -
7   -NAME
8   - generate
9   - - random sentence generator from a grammar
10   -
11   -SYNOPSIS
12   - generate [-v] [-t] [-n num] [-s spname] {prefix}
13   -
14   -DESCRIPTION
15   - This small program randomly generates sentences that are acceptable by
16   - the given grammar.
17   -
18   -
19   - .dfa, .dict and .term files are needed to execute. They can be
20   - generated from .grammar and .voca file by mkdfa.pl.
21   -
22   -OPTIONS
23   - -t
24   - Output in word's category name.
25   -
26   - -n num
27   - Set number of sentences to be generated (default: 10)
28   -
29   - -s spname
30   - the name string of short-pause word to be supressed (default: "sp")
31   -
32   - -v
33   - Debug output mode.
34   -
35   -EXAMPLES
36   - Exmple output of a sample grammar "fruit":
37   -
38   - % generate fruit
39   - Stat: init_voca: read 36 words
40   - Reading in term file (optional)...done
41   - 15 categories, 36 words
42   - DFA has 26 nodes and 42 arcs
43   - -----
44   - <s> I WANT ONE APPLE </s>
45   - <s> I WANT TEN PEARS </s>
46   - <s> CAN I HAVE A PINEAPPLE </s>
47   - <s> I WANT ONE PEAR </s>
48   - <s> COULD I HAVE A BANANA </s>
49   - <s> I WANT ONE APPLE PLEASE </s>
50   - <s> I WANT NINE APPLES </s>
51   - <s> NINE APPLES </s>
52   - <s> I WANT ONE PINEAPPLE </s>
53   - <s> I WANT A PEAR </s>
54   -
55   -
56   -SEE ALSO
57   - mkdfa.pl ( 1 ) ,
58   - generate-ngram ( 1 )
59   -
60   -COPYRIGHT
61   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
62   -
63   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
64   -
65   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
66   - Technology
67   -
68   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
69   - Technology
70   -
71   -LICENSE
72   - The same as Julius.
73   -
74   -
75   -
76   - 10/02/2008 GENERATE(1)
recognize/src/julius/doc/manuals/gram2sapixml.pl.txt
View file @e1cf2b2
... ... @@ -1,47 +0,0 @@
1   - gram2sapixml.pl
2   -
3   -GRAM2SAPIXML.PL(1) GRAM2SAPIXML.PL(1)
4   -
5   -
6   -
7   -NAME
8   - gram2sapixml.pl
9   - - convert Julius grammar to SAPI XML grammar format
10   -
11   -SYNOPSIS
12   - gram2sapixml.pl [prefix...]
13   -
14   -DESCRIPTION
15   - gram2sapixml.pl converts a recognition grammar file of Julius
16   - (.grammar, .voca) to Microsoft SAPI XML grammar format. prefix should
17   - be a file name of target grammar, excluding suffixes. If multiple
18   - argument is given, each will be process sequencialy in turn.
19   -
20   - The internal character set should be in UTF-8 format. By default
21   - gram2sapixml.pl assume input in EUC-JP encoding and tries to convert it
22   - to UTF-8 using iconv. You may want to disable this feature within the
23   - script.
24   -
25   - It will fail to convert a left recursive rule in the grammar. When
26   - fails, it will leave the source rules in the target .xml file, so you
27   - should modify the output manually to solve it.
28   -
29   -SEE ALSO
30   - mkdfa.pl ( 1 )
31   -
32   -DIAGNOSTICS
33   - The conversion procedure is somewhat dumb one, only converting the
34   - non-terminal symbols and terminal symbols (=word category name) into
35   - corresponding rules one by one. This is only a help tool, and you will
36   - need a manual inspection and editing to use it on a real SAPI
37   - application.
38   -
39   -COPYRIGHT
40   - Copyright (c) 2002 Takashi Sumiyoshi
41   -
42   -LICENSE
43   - The same as Julius.
44   -
45   -
46   -
47   - 10/02/2008 GRAM2SAPIXML.PL(1)
recognize/src/julius/doc/manuals/jclient.pl.txt
View file @e1cf2b2
... ... @@ -1,46 +0,0 @@
1   - jclient.pl
2   -
3   -JCLIENT.PL(1) JCLIENT.PL(1)
4   -
5   -
6   -
7   -NAME
8   - jclient.pl
9   - - sample client for module mode (perl version)
10   -
11   -SYNOPSIS
12   - jclient.pl
13   -
14   -DESCRIPTION
15   - This is yet another sample client written in perl. It will connect to
16   - Julius running in module mode, receive recognition results from Julius,
17   - and cna send commands to control Julius.
18   -
19   - This is a tiny program with only 57 lines. You can use it for free.
20   -
21   -EXAMPLES
22   - Invoke Julius with module mode by specifying "-module" option:
23   - Then, at other terminal or other host, invoke jclient.pl like below.
24   - The default hostname is "localhost", and port number is 10500. You can
25   - change them by editing the top part of the script.
26   - It will then receive the outputs of Julius and output the raw message
27   - to standard out. Also, by inputting a raw module command to the
28   - standard input of jclient.pl, it will be sent to Julius. See manuals
29   - for the specification of module mode.
30   -
31   -SEE ALSO
32   - julius ( 1 ) ,
33   - jcontrol ( 1 )
34   -
35   -COPYRIGHT
36   - "jclient.pl" has been developed by Dr. Ryuichi Nisimura
37   - (nisimura@sys.wakayama-u.ac.jp). Use at your own risk.
38   -
39   - If you have any feedback, comment or request, please contact the E-mail
40   - address above, or look at the Web page below.
41   -
42   - http://w3voice.jp/
43   -
44   -
45   -
46   - 10/02/2008 JCLIENT.PL(1)
recognize/src/julius/doc/manuals/jcontrol.txt
View file @e1cf2b2
... ... @@ -1,173 +0,0 @@
1   - jcontrol
2   -
3   -JCONTROL(1) JCONTROL(1)
4   -
5   -
6   -
7   -NAME
8   - jcontrol
9   - - a sample module client written in C
10   -
11   -SYNOPSIS
12   - jcontrol {hostname} [portnum]
13   -
14   -DESCRIPTION
15   - jcontrol is a simple console program to control julius running on other
16   - host via network API. It can send command to Julius, and receive
17   - messages from Julius.
18   -
19   - When invoked, jcontrol tries to connect to Julius running in "module
20   - mode" on specified hostname. After connection established, jcontrol
21   - waits for user commands from standard input.
22   -
23   - When user types a command to jcontrol, it will be interpreted and cor-
24   - responding API command will be sent to Julius. When a message is
25   - received from Julius, its content will be output to standard output.
26   -
27   - For the details about the API, see the related documents.
28   -
29   -OPTIONS
30   - hostname
31   - Host name where Julius is runnning in module mode.
32   -
33   - portnum
34   - port number (default: 10500)
35   -
36   -COMMANDS
37   - jcontrol interprets commands from standard input. Below is a list of
38   - all commands.
39   -
40   - Engine control
41   - pause
42   - Stop Julius and enter into paused status. In paused status, Julius
43   - will not run recognition even if speech input occurs. When this
44   - command is issued while recognition is running, Julius will stop
45   - after the recognition has been finished.
46   -
47   - terminate
48   - Same as pause, but discard the current speech input when received
49   - command in the middle of recognition process.
50   -
51   - resume
52   - Restart Julius that has been paused or terminated.
53   -
54   - inputparam arg
55   - Tell Julius how to deal with speech input in case grammar is changed
56   - just when recognition is running. Specify one: "TERMINATE", "PAUSE"
57   - or "WAIT".
58   -
59   - version
60   - Tell Julius to send version description string.
61   -
62   - status
63   - Tell Julius to send the system status (active / sleep)
64   -
65   - Grammar handling
66   - changegram prefix
67   - Send a new grammar "prefix.dfa" and "prefix.dict", and tell julius
68   - to use it as a new grammar. All the current grammars used in the
69   - current process of Julius will be deleted and replaced to the
70   - specifed grammar.
71   -
72   - addgram prefix
73   - Send a new grammar "prefix.dfa" and "prefix.dict" and add it to the
74   - current grammar.
75   -
76   - deletegram gramlist
77   - Tell Julius to delete existing grammar. The grammar can be specified
78   - by either prefix name or number ID. The number ID can be determined
79   - from the message sent from Julius at each time grammar information
80   - has changed. When want to delete more than one grammar, specify all
81   - of them as comma-sparated.
82   -
83   - deactivategram gramlist
84   - Tell Julius to de-activate a specified grammar. The specified
85   - grammar will still be kept but will not be used for recognition.
86   -
87   - The target grammar can be specified by either prefix name or number
88   - ID. The number ID can be determined from the message sent from
89   - Julius at each time grammar information has changed. When want to
90   - delete more than one grammar, specify all of them as comma-sparated.
91   -
92   - activategram gramlist
93   - Tell Julius to activate previously de-activated grammar. The target
94   - grammar can be specified by either prefix name or number ID. The
95   - number ID can be determined from the message sent from Julius at
96   - each time grammar information has changed. When want to delete more
97   - than one grammar, specify all of them as comma-sparated.
98   -
99   - addword grammar_name_or_id dictfile
100   - Add the recognition word entries in the specified dictfile to the
101   - specified grammar on current process.
102   -
103   - syncgram
104   - Force synchronize grammar status, like unix command "sync".
105   -
106   - Process management
107   - Julius-4 supports multi-model recognition nad multi decoding. In this
108   - case it is possible to control each recognition process, as defined by
109   - "-SR" option, from module client.
110   -
111   - In multi decoding mode, the module client holds "current process", and
112   - the process commands and grammar related commands will be issued toward
113   - the current process.
114   -
115   - listprocess
116   - Tell Julius to send the list of existing recognition process.
117   -
118   - currentprocess procname
119   - Switch the current process to the process specified by the name.
120   -
121   - shiftprocess
122   - Rotate the current process. At each call the current process will be
123   - changed to the next one.
124   -
125   - addprocess jconffile
126   - Tell Julisu to load a new recognition process into engine. The
127   - argument jconffile should be a jconf file that contains only one set
128   - of LM options and one SR definition. Note that the file should be
129   - visible on the running Julius, since jcontrol only send the path
130   - name and Julius actually read the jconf file.
131   -
132   - The new LM and SR process will have the name of the jconffile.
133   -
134   - delprocess procname
135   - Delete the specified recognition process from the engine.
136   -
137   - deactivateprocess procname
138   - Tell Julius to temporary stop the specified recognition process. The
139   - stopped process will not be executed for the input until activated
140   - again.
141   -
142   - activateprocess procname
143   - Tell Julius to activate the temporarily stopped process.
144   -
145   -EXAMPLES
146   - The dump messages from Julius are output to tty with prefix ">"
147   - appended to each line. Julius can be started in module mode like this:
148   - jcontrolcan be launched with the host name:
149   - It will then receive the outputs of Julius and output the raw message
150   - to standard out. Also, by inputting the commands above to the standard
151   - input of jcontrol, it will be sent to Julius. See manuals for the
152   - specification of module mode.
153   -
154   -SEE ALSO
155   - julius ( 1 )
156   -
157   -COPYRIGHT
158   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
159   -
160   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
161   -
162   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
163   - Technology
164   -
165   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
166   - Technology
167   -
168   -LICENSE
169   - The same as Julius.
170   -
171   -
172   -
173   - 10/02/2008 JCONTROL(1)
recognize/src/julius/doc/manuals/julius.txt
View file @e1cf2b2
... ... @@ -1,1287 +0,0 @@
1   - julius
2   -
3   -JULIUS(1) JULIUS(1)
4   -
5   -
6   -
7   -NAME
8   - julius
9   - - open source multi-purpose LVCSR engine
10   -
11   -SYNOPSIS
12   - julius [-C jconffile] [options...]
13   -
14   -DESCRIPTION
15   - julius is a high-performance, multi-purpose, open-source speech
16   - recognition engine for researchers and developers. It is capable of
17   - performing almost real-time recognition of continuous speech with over
18   - 60k-word 3-gram language model and triphone HMM model, on most current
19   - PCs. julius can perform recognition on audio files, live microphone
20   - input, network input and feature parameter files.
21   -
22   - The core recognition module is implemented as C library called
23   - "JuliusLib". It can also be extended by plug-in facility.
24   -
25   - Supported Models
26   - julius needs a language model and an acoustic model to run as a speech
27   - recognizer. julius supports the following models.
28   -
29   - Acoustic model
30   - Sub-word HMM (Hidden Markov Model) in HTK ascii format are
31   - supported. Phoneme models (monophone), context dependent phoneme
32   - models (triphone), tied-mixture and phonetic tied-mixture models
33   - of any unit can be used. When using context dependent models,
34   - inter-word context dependency is also handled. Multi-stream
35   - feature and MSD-HMM is also supported. You can further use a
36   - tool mkbinhmm to convert the ascii HMM file to a compact binary
37   - format for faster loading.
38   -
39   - Note that julius itself can only extract MFCC features from
40   - speech data. If you use acoustic HMM trained for other feature,
41   - you should give the input in HTK parameter file of the same
42   - feature type.
43   -
44   - Language model: word N-gram
45   - Word N-gram language model, up to 10-gram, is supported. Julius
46   - uses different N-gram for each pass: left-to-right 2-gram on 1st
47   - pass, and right-to-left N-gram on 2nd pass. It is recommended to
48   - use both LR 2-gram and RL N-gram for Julius. However, you can
49   - use only single LR N-gram or RL N-gram. In such case,
50   - approximated LR 2-gram computed from the given N-gram will be
51   - applied at the first pass.
52   -
53   - The Standard ARPA format is supported. In addition, a binary
54   - format is also supported for efficiency. The tool mkbingram(1)
55   - can convert ARPA format N-gram to binary format.
56   -
57   - Language model: grammar
58   - The grammar format is an original one, and tools to create a
59   - recognirion grammar are included in the distribution. A grammar
60   - consists of two files: one is a 'grammar' file that describes
61   - sentence structures in a BNF style, using word 'category' name
62   - as terminate symbols. Another is a 'voca' file that defines
63   - words with its pronunciations (i.e. phoneme sequences) for each
64   - category. They should be converted by mkdfa.pl(1) to a
65   - deterministic finite automaton file (.dfa) and a dictionary file
66   - (.dict), respectively. You can also use multiple grammars.
67   -
68   - Language model: isolated word
69   - You can perform isolated word recognition using only word
70   - dictionary. With this model type, Julius will perform rapid one
71   - pass recognition with static context handling. Silence models
72   - will be added at both head and tail of each word. You can also
73   - use multiple dictionaries in a process.
74   -
75   - Search Algorithm
76   - Recognition algorithm of julius is based on a two-pass strategy. Word
77   - 2-gram and reverse word 3-gram is used on the respective passes. The
78   - entire input is processed on the first pass, and again the final
79   - searching process is performed again for the input, using the result of
80   - the first pass to narrow the search space. Specifically, the
81   - recognition algorithm is based on a tree-trellis heuristic search
82   - combined with left-to-right frame-synchronous beam search and
83   - right-to-left stack decoding search.
84   -
85   - When using context dependent phones (triphones), interword contexts are
86   - taken into consideration. For tied-mixture and phonetic tied-mixture
87   - models, high-speed acoustic likelihood calculation is possible using
88   - gaussian pruning.
89   -
90   - For more details, see the related documents.
91   -
92   -OPTIONS
93   - These options specify the models, system behaviors and various search
94   - parameters to Julius. These option can be set at the command line, but
95   - it is recommended that you write them in a text file as a "jconf file",
96   - and specify it by "-C" option.
97   -
98   - Applications incorporating JuliusLib also use these options to set the
99   - parameters of core recognition engine. For example, a jconf file can be
100   - loaded to the enine by calling j_config_load_file_new() with the jconf
101   - file name as argument.
102   -
103   - Please note that relative paths in a jconf file should be relative to
104   - the jconf file itself, not the current working directory.
105   -
106   - Below are the details of all options, gathered by group.
107   -
108   - Julius application option
109   - These are application options of Julius, outside of JuliusLib. It
110   - contains parameters and switches for result output, character set
111   - conversion, log level, and module mode options. These option are
112   - specific to Julius, and cannot be used at applications using JuliusLib
113   - other than Julius.
114   -
115   - -outfile
116   - On file input, this option write the recognition result of each file
117   - to a separate file. The output file of an input file will be the
118   - same name but the suffix will be changed to ".out". (rev.4.0)
119   -
120   - -separatescore
121   - Output the language and acoustic scores separately.
122   -
123   - -callbackdebug
124   - Print the callback names at each call for debug. (rev.4.0)
125   -
126   - -charconv from to
127   - Print with character set conversion. from is the source character
128   - set used in the language model, and to is the target character set
129   - you want to get.
130   -
131   - On Linux, the arguments should be a code name. You can obtain the
132   - list of available code names by invoking the command "iconv --list".
133   - On Windows, the arguments should be a code name or codepage number.
134   - Code name should be one of "ansi", "mac", "oem", "utf-7", "utf-8",
135   - "sjis", "euc". Or you can specify any codepage number supported at
136   - your environment.
137   -
138   - -nocharconv
139   - Disable character conversion.
140   -
141   - -module [port]
142   - Run Julius on "Server Module Mode". After startup, Julius waits for
143   - tcp/ip connection from client. Once connection is established,
144   - Julius start communication with the client to process incoming
145   - commands from the client, or to output recognition results, input
146   - trigger information and other system status to the client. The
147   - default port number is 10500.
148   -
149   - -record dir
150   - Auto-save all input speech data into the specified directory. Each
151   - segmented inputs are recorded each by one. The file name of the
152   - recorded data is generated from system time when the input ends, in
153   - a style of YYYY.MMDD.HHMMSS.wav. File format is 16bit monoral WAV.
154   - Invalid for mfcfile input.
155   -
156   - With input rejection by -rejectshort, the rejected input will also
157   - be recorded even if they are rejected.
158   -
159   - -logfile file
160   - Save all log output to a file instead of standard output. (Rev.4.0)
161   -
162   - -nolog
163   - Disable all log output. (Rev.4.0)
164   -
165   - -help
166   - Output help message and exit.
167   -
168   - Global options
169   - These are model-/search-dependent options relating audio input, sound
170   - detection, GMM, decoding algorithm, plugin facility, and others. Global
171   - options should be placed before any instance declaration (-AM, -LM, or
172   - -SR), or just after "-GLOBAL" option.
173   -
174   - Audio input
175   - -input
176   - {mic|rawfile|mfcfile|adinnet|stdin|netaudio|alsa|oss|esd}
177   - Choose speech input source. Specify 'file' or 'rawfile' for
178   - waveform file, 'htkparam' or 'mfcfile' for HTK parameter
179   - file. On file input, users will be prompted to enter the file
180   - name from stdin, or you can use -filelist option to specify
181   - list of files to process.
182   -
183   - 'mic' is to get audio input from a default live microphone
184   - device, and 'adinnet' means receiving waveform data via tcpip
185   - network from an adinnet client. 'netaudio' is from
186   - DatLink/NetAudio input, and 'stdin' means data input from
187   - standard input.
188   -
189   - For waveform file input, only WAV (no compression) and RAW
190   - (noheader, 16bit, big endian) are supported by default. Other
191   - format can be read when compiled with libsnd library. To see
192   - what format is actually supported, see the help message using
193   - option -help. For stdin input, only WAV and RAW is supported.
194   - (default: mfcfile)
195   -
196   - At Linux, you can choose API at run time by specifying alsa,
197   - oss and esd.
198   -
199   - -filelist filename
200   - (With -input rawfile|mfcfile) perform recognition on all
201   - files listed in the file. The file should contain input file
202   - per line. Engine will end when all of the files are
203   - processed.
204   -
205   - -notypecheck
206   - By default, Julius checks the input parameter type whether it
207   - matches the AM or not. This option will disable the check and
208   - force engine to use the input vector as is.
209   -
210   - -48
211   - Record input with 48kHz sampling, and down-sample it to 16kHz
212   - on-the-fly. This option is valid for 16kHz model only. The
213   - down-sampling routine was ported from sptk. (Rev. 4.0)
214   -
215   - -NA devicename
216   - Host name for DatLink server input (-input netaudio).
217   -
218   - -adport port_number
219   - With -input adinnet, specify adinnet port number to listen.
220   - (default: 5530)
221   -
222   - -nostrip
223   - Julius by default removes successive zero samples in input
224   - speech data. This option inhibits the removal.
225   -
226   - -zmean , -nozmean
227   - This option enables/disables DC offset removal of input
228   - waveform. Offset will be estimated from the whole input. For
229   - microphone / network input, zero mean of the first 48000
230   - samples (3 seconds in 16kHz sampling) will be used for the
231   - estimation. (default: disabled)
232   -
233   - This option uses static offset for the channel. See also
234   - -zmeansource for frame-wise offset removal.
235   -
236   - Speech detection by level and zero-cross
237   - -cutsilence , -nocutsilence
238   - Turn on / off the speech detection by level and zero-cross.
239   - Default is on for mic / adinnet input, and off for files.
240   -
241   - -lv thres
242   - Level threshold for speech input detection. Values should be
243   - in range from 0 to 32767. (default: 2000)
244   -
245   - -zc thres
246   - Zero crossing threshold per second. Only input that goes over
247   - the level threshold (-lv) will be counted. (default: 60)
248   -
249   - -headmargin msec
250   - Silence margin at the start of speech segment in
251   - milliseconds. (default: 300)
252   -
253   - -tailmargin msec
254   - Silence margin at the end of speech segment in milliseconds.
255   - (default: 400)
256   -
257   - Input rejection
258   - Two simple front-end input rejection methods are implemented,
259   - based on input length and average power of detected segment. The
260   - rejection by average power is experimental, and can be enabled
261   - by --enable-power-reject on compilation. Valid for MFCC feature
262   - with power coefficient and real-time input only.
263   -
264   - For GMM-based input rejection see the GMM section below.
265   -
266   - -rejectshort msec
267   - Reject input shorter than specified milliseconds. Search will
268   - be terminated and no result will be output.
269   -
270   - -powerthres thres
271   - Reject the inputted segment by its average energy. If the
272   - average energy of the last recognized input is below the
273   - threshold, Julius will reject the input. (Rev.4.0)
274   -
275   - This option is valid when --enable-power-reject is specified
276   - at compilation time.
277   -
278   - Gaussian mixture model / GMM-VAD
279   - GMM will be used for input rejection by accumulated score, or
280   - for front-end GMM-based VAD when --enable-gmm-vad is specified.
281   -
282   - NOTE: You should also set the proper MFCC parameters required
283   - for the GMM, specifying the acoustic parameters described in AM
284   - section -AM_GMM.
285   -
286   - When GMM-based VAD is enabled, the voice activity score will be
287   - calculated at each frame as front-end processing. The value will
288   - be computed as \[ \max_{m \in M_v} p(x|m) - \max_{m \in M_n}
289   - p(x|m) \] where $M_v$ is a set of voice GMM, and $M_n$ is a set
290   - of noise GMM whose names should be specified by -gmmreject. The
291   - activity score will be then averaged for the last N frames,
292   - where N is specified by -gmmmargin. Julius updates the averaged
293   - activity score at each frame, and detect speech up-trigger when
294   - the value gets higher than a value specified by -gmmup, and
295   - detecgt down-trigger when it gets lower than a value of
296   - -gmmdown.
297   -
298   - -gmm hmmdefs_file
299   - GMM definition file in HTK format. If specified, GMM-based
300   - input verification will be performed concurrently with the
301   - 1st pass, and you can reject the input according to the
302   - result as specified by -gmmreject. The GMM should be defined
303   - as one-state HMMs.
304   -
305   - -gmmnum number
306   - Number of Gaussian components to be computed per frame on GMM
307   - calculation. Only the N-best Gaussians will be computed for
308   - rapid calculation. The default is 10 and specifying smaller
309   - value will speed up GMM calculation, but too small value (1
310   - or 2) may cause degradation of identification performance.
311   -
312   - -gmmreject string
313   - Comma-separated list of GMM names to be rejected as invalid
314   - input. When recognition, the log likelihoods of GMMs
315   - accumulated for the entire input will be computed
316   - concurrently with the 1st pass. If the GMM name of the
317   - maximum score is within this string, the 2nd pass will not be
318   - executed and the input will be rejected.
319   -
320   - -gmmmargin frames
321   - (GMM_VAD) Head margin in frames. When a speech trigger
322   - detected by GMM, recognition will start from current frame
323   - minus this value. (Rev.4.0)
324   -
325   - This option will be valid only if compiled with
326   - --enable-gmm-vad.
327   -
328   - -gmmup value
329   - (GMM_VAD) Up trigger threshold of voice activity score.
330   - (Rev.4.1)
331   -
332   - This option will be valid only if compiled with
333   - --enable-gmm-vad.
334   -
335   - -gmmdown value
336   - (GMM_VAD) Down trigger threshold of voice activity score.
337   - (Rev.4.1)
338   -
339   - This option will be valid only if compiled with
340   - --enable-gmm-vad.
341   -
342   - Decoding option
343   - Real-time processing means concurrent processing of MFCC
344   - computation 1st pass decoding. By default, real-time processing
345   - on the pass is on for microphone / adinnet / netaudio input, and
346   - for others.
347   -
348   - -realtime , -norealtime
349   - Explicitly switch on / off real-time (pipe-line) processing
350   - on the first pass. The default is off for file input, and on
351   - for microphone, adinnet and NetAudio input. This option
352   - relates to the way CMN and energy normalization is performed:
353   - if off, they will be done using average features of whole
354   - input. If on, MAP-CMN and energy normalization to do
355   - real-time processing.
356   -
357   - Misc. options
358   - -C jconffile
359   - Load a jconf file at here. The content of the jconffile will
360   - be expanded at this point.
361   -
362   - -version
363   - Print version information to standard error, and exit.
364   -
365   - -setting
366   - Print engine setting information to standard error, and exit.
367   -
368   - -quiet
369   - Output less log. For result, only the best word sequence will
370   - be printed.
371   -
372   - -debug
373   - (For debug) output enormous internal message and debug
374   - information to log.
375   -
376   - -check {wchmm|trellis|triphone}
377   - For debug, enter interactive check mode.
378   -
379   - -plugindir dirlist
380   - Specify directory to load plugin. If several direcotries
381   - exist, specify them by colon-separated list.
382   -
383   - Instance declaration for multi decoding
384   - The following arguments will create a new configuration set with
385   - default parameters, and switch current set to it. Jconf parameters
386   - specified after the option will be set into the current set.
387   -
388   - To do multi-model decoding, these argument should be specified at the
389   - first of each model / search instances with different names. Any
390   - options before the first instance definition will be IGNORED.
391   -
392   - When no instance definition is found (as older version of Julius), all
393   - the options are assigned to a default instance named _default.
394   -
395   - Please note that decoding with a single LM and multiple AMs is not
396   - fully supported. For example, you may want to construct the jconf file
397   - as following.
398   - This type of model sharing is not supported yet, since some part of LM
399   - processing depends on the assigned AM. Instead, you can get the same
400   - result by defining the same LMs for each AM, like this:
401   -
402   - -AM name
403   - Create a new AM configuration set, and switch current to the new
404   - one. You should give a unique name. (Rev.4.0)
405   -
406   - -LM name
407   - Create a new LM configuration set, and switch current to the new
408   - one. You should give a unique name. (Rev.4.0)
409   -
410   - -SR name am_name lm_name
411   - Create a new search configuration set, and switch current to the new
412   - one. The specified AM and LM will be assigned to it. The am_name and
413   - lm_name can be either name or ID number. You should give a unique
414   - name. (Rev.4.0)
415   -
416   - -AM_GMM
417   - When using GMM for front-end processing, you can specify
418   - GMM-specific acoustic parameters after this option. If you does not
419   - specify -AM_GMM with GMM, the GMM will share the same parameter
420   - vector as the last AM. The current AM will be switched to the GMM
421   - one, so be careful not to confuse with normal AM configurations.
422   - (Rev.4.0)
423   -
424   - -GLOBAL
425   - Start a global section. The global options should be placed before
426   - any instance declaration, or after this option on multiple model
427   - recognition. This can be used multiple times. (Rev.4.1)
428   -
429   - -nosectioncheck , -sectioncheck
430   - Disable / enable option location check in multi-model decoding. When
431   - enabled, the options between instance declaration is treated as
432   - "sections" and only the belonging option types can be written. For
433   - example, when an option -AM is specified, only the AM related option
434   - can be placed after the option until other declaration is found.
435   - Also, global options should be placed at top, before any instance
436   - declarataion. This is enabled by default. (Rev.4.1)
437   -
438   - Language model (-LM)
439   - This group contains options for model definition of each language model
440   - type. When using multiple LM, one instance can have only one LM.
441   -
442   - Only one type of LM can be specified for a LM configuration. If you
443   - want to use multi model, you should define them one as a new LM.
444   -
445   - N-gram
446   - -d bingram_file
447   - Use binary format N-gram. An ARPA N-gram file can be
448   - converted to Julius binary format by mkbingram.
449   -
450   - -nlr arpa_ngram_file
451   - A forward, left-to-right N-gram language model in standard
452   - ARPA format. When both a forward N-gram and backward N-gram
453   - are specified, Julius uses this forward 2-gram for the 1st
454   - pass, and the backward N-gram for the 2nd pass.
455   -
456   - Since ARPA file often gets huge and requires a lot of time to
457   - load, it may be better to convert the ARPA file to Julius
458   - binary format by mkbingram. Note that if both forward and
459   - backward N-gram is used for recognition, they together will
460   - be converted to a single binary.
461   -
462   - When only a forward N-gram is specified by this option and no
463   - backward N-gram specified by -nrl, Julius performs
464   - recognition with only the forward N-gram. The 1st pass will
465   - use the 2-gram entry in the given N-gram, and The 2nd pass
466   - will use the given N-gram, with converting forward
467   - probabilities to backward probabilities by Bayes rule.
468   - (Rev.4.0)
469   -
470   - -nrl arpa_ngram_file
471   - A backward, right-to-left N-gram language model in standard
472   - ARPA format. When both a forward N-gram and backward N-gram
473   - are specified, Julius uses the forward 2-gram for the 1st
474   - pass, and this backward N-gram for the 2nd pass.
475   -
476   - Since ARPA file often gets huge and requires a lot of time to
477   - load, it may be better to convert the ARPA file to Julius
478   - binary format by mkbingram. Note that if both forward and
479   - backward N-gram is used for recognition, they together will
480   - be converted to a single binary.
481   -
482   - When only a backward N-gram is specified by this option and
483   - no forward N-gram specified by -nlr, Julius performs
484   - recognition with only the backward N-gram. The 1st pass will
485   - use the forward 2-gram probability computed from the backward
486   - 2-gram using Bayes rule. The 2nd pass fully use the given
487   - backward N-gram. (Rev.4.0)
488   -
489   - -v dict_file
490   - Word dictionary file.
491   -
492   - -silhead word_string -siltail word_string
493   - Silence word defined in the dictionary, for silences at the
494   - beginning of sentence and end of sentence. (default: "<s>",
495   - "</s>")
496   -
497   - -mapunk word_string
498   - Specify unknown word. Default is "<unk>" or "<UNK>". This
499   - will be used to assign word probability on unknown words,
500   - i.e. words in dictionary that are not in N-gram vocabulary.
501   -
502   - -iwspword
503   - Add a word entry to the dictionary that should correspond to
504   - inter-word pauses. This may improve recognition accuracy in
505   - some language model that has no explicit inter-word pause
506   - modeling. The word entry to be added can be changed by
507   - -iwspentry.
508   -
509   - -iwspentry word_entry_string
510   - Specify the word entry that will be added by -iwspword.
511   - (default: "<UNK> [sp] sp sp")
512   -
513   - -sepnum number
514   - Number of high frequency words to be isolated from the
515   - lexicon tree, to ease approximation error that may be caused
516   - by the one-best approximation on 1st pass. (default: 150)
517   -
518   - Grammar
519   - Multiple grammars can be specified by repeating -gram and
520   - -gramlist. Note that this is unusual behavior from other options
521   - (in normal Julius option, last one will override previous ones).
522   - You can use -nogram to reset the grammars already specified
523   - before the point.
524   -
525   - -gram gramprefix1[,gramprefix2[,gramprefix3,...]]
526   - Comma-separated list of grammars to be used. the argument
527   - should be a prefix of a grammar, i.e. if you have foo.dfa and
528   - foo.dict, you should specify them with a single argument foo.
529   - Multiple grammars can be specified at a time as a
530   - comma-separated list.
531   -
532   - -gramlist list_file
533   - Specify a grammar list file that contains list of grammars to
534   - be used. The list file should contain the prefixes of
535   - grammars, each per line. A relative path in the list file
536   - will be treated as relative to the file, not the current path
537   - or configuration file.
538   -
539   - -dfa dfa_file -v dict_file
540   - An old way of specifying grammar files separately. This is
541   - bogus, and should not be used any more.
542   -
543   - -nogram
544   - Remove the current list of grammars already specified by
545   - -gram, -gramlist, -dfa and -v.
546   -
547   - Isolated word
548   - Dictionary can be specified by using -w and -wlist. When you
549   - specify multiple times, all of them will be read at startup. You
550   - can use -nogram to reset the already specified dictionaries at
551   - that point.
552   -
553   - -w dict_file
554   - Word dictionary for isolated word recognition. File format is
555   - the same as other LM. (Rev.4.0)
556   -
557   - -wlist list_file
558   - Specify a dictionary list file that contains list of
559   - dictionaries to be used. The list file should contain the
560   - file name of dictionaries, each per line. A relative path in
561   - the list file will be treated as relative to the list file,
562   - not the current path or configuration file. (Rev.4.0)
563   -
564   - -nogram
565   - Remove the current list of dictionaries already specified by
566   - -w and -wlist.
567   -
568   - -wsil head_sil_model_name tail_sil_model_name sil_context_name
569   - On isolated word recognition, silence models will be appended
570   - to the head and tail of each word at recognition. This option
571   - specifies the silence models to be appended.
572   - sil_context_name is the name of the head sil model and tail
573   - sil model as a context of word head phone and tail phone. For
574   - example, if you specify -wsil silB silE sp, a word with phone
575   - sequence b eh t will be translated as silB sp-b+eh b-eh+t
576   - eh-t+sp silE. (Rev.4.0)
577   -
578   - User-defined LM
579   - -userlm
580   - Declare to use user LM functions in the program. This option
581   - should be specified if you use user-defined LM functions.
582   - (Rev.4.0)
583   -
584   - Misc. LM options
585   - -forcedict
586   - Skip error words in dictionary and force running.
587   -
588   - Acoustic model and feature analysis (-AM) (-AM_GMM)
589   - This section is about options for acoustic model, feature extraction,
590   - feature normalizations and spectral subtraction.
591   -
592   - After -AM name, an acoustic model and related specification should be
593   - written. You can use multiple AMs trained with different MFCC types.
594   - For GMM, the required parameter condition should be specified just as
595   - same as AMs after -AM_GMM.
596   -
597   - When using multiple AMs, the values of -smpPeriod, -smpFreq, -fsize and
598   - -fshift should be the same among all AMs.
599   -
600   - Acoustic HMM
601   - -h hmmdef_file
602   - Acoustic HMM definition file. It should be in HTK ascii
603   - format, or Julius binary format. You can convert HTK ascii
604   - format to Julius binary format using mkbinhmm.
605   -
606   - -hlist hmmlist_file
607   - HMMList file for phone mapping. This file provides mapping
608   - between logical triphone names generated in the dictionary
609   - and the defined HMM names in hmmdefs. This option should be
610   - specified for context-dependent model.
611   -
612   - -tmix number
613   - Specify the number of top Gaussians to be calculated in a
614   - mixture codebook. Small number will speed up the acoustic
615   - computation, but AM accuracy may get worse with too small
616   - value. See also -gprune. (default: 2)
617   -
618   - -spmodel name
619   - Specify HMM model name that corresponds to short-pause in an
620   - utterance. The short-pause model name will be used in
621   - recognition: short-pause skipping on grammar recognition,
622   - word-end short-pause model insertion with -iwsp on N-gram, or
623   - short-pause segmentation (-spsegment). (default: "sp")
624   -
625   - -multipath
626   - Enable multi-path mode. To make decoding faster, Julius by
627   - default impose a limit on HMM transitions that each model
628   - should have only one transition from initial state and to end
629   - state. On multi-path mode, Julius does extra handling on
630   - inter-model transition to allows model-skipping transition
631   - and multiple output/input transitions. Note that specifying
632   - this option will make Julius a bit slower, and the larger
633   - beam width may be required.
634   -
635   - This function was a compilation-time option on Julius 3.x,
636   - and now becomes a run-time option. By default (without this
637   - option), Julius checks the transition type of specified HMMs,
638   - and enable the multi-path mode if required. You can force
639   - multi-path mode with this option. (rev.4.0)
640   -
641   - -gprune {safe|heuristic|beam|none|default}
642   - Set Gaussian pruning algorithm to use. For tied-mixture
643   - model, Julius performs Gaussian pruning to reduce acoustic
644   - computation, by calculating only the top N Gaussians in each
645   - codebook at each frame. The default setting will be set
646   - according to the model type and engine setting. default will
647   - force accepting the default setting. Set this to none to
648   - disable pruning and perform full computation. safe
649   - guarantees the top N Gaussians to be computed. heuristic and
650   - beam do more aggressive computational cost reduction, but may
651   - result in small loss of accuracy model (default: safe
652   - (standard), beam (fast) for tied mixture model, none for non
653   - tied-mixture model).
654   -
655   - -iwcd1 {max|avg|best number}
656   - Select method to approximate inter-word triphone on the head
657   - and tail of a word in the first pass.
658   -
659   -
660   - max will apply the maximum likelihood of the same context
661   - triphones. avg will apply the average likelihood of the same
662   - context triphones. best number will apply the average of top
663   - N-best likelihoods of the same context triphone.
664   -
665   - Default is best 3 for use with N-gram, and avg for grammar
666   - and word. When this AM is shared by LMs of both type, latter
667   - one will be chosen.
668   -
669   - -iwsppenalty float
670   - Insertion penalty for word-end short pauses appended by
671   - -iwsp.
672   -
673   - -gshmm hmmdef_file
674   - If this option is specified, Julius performs Gaussian Mixture
675   - Selection for efficient decoding. The hmmdefs should be a
676   - monophone model generated from an ordinary monophone HMM
677   - model, using mkgshmm.
678   -
679   - -gsnum number
680   - On GMS, specify number of monophone states to compute
681   - corresponding triphones in detail. (default: 24)
682   -
683   - Speech analysis
684   - Only MFCC feature extraction is supported in current Julius.
685   - Thus when recognizing a waveform input from file or microphone,
686   - AM must be trained by MFCC. The parameter condition should also
687   - be set as exactly the same as the training condition by the
688   - options below.
689   -
690   - When you give an input in HTK Parameter file, you can use any
691   - parameter type for AM. In this case Julius does not care about
692   - the type of input feature and AM, just read them as vector
693   - sequence and match them to the given AM. Julius only checks
694   - whether the parameter types are the same. If it does not work
695   - well, you can disable this checking by -notypecheck.
696   -
697   - In Julius, the parameter kind and qualifiers (as TARGETKIND in
698   - HTK) and the number of cepstral parameters (NUMCEPS) will be set
699   - automatically from the content of the AM header, so you need not
700   - specify them by options.
701   -
702   - Other parameters should be set exactly the same as training
703   - condition. You can also give a HTK Config file which you used to
704   - train AM to Julius by -htkconf. When this option is applied,
705   - Julius will parse the Config file and set appropriate parameter.
706   -
707   - You can further embed those analysis parameter settings to a
708   - binary HMM file using mkbinhmm.
709   -
710   - If options specified in several ways, they will be evaluated in
711   - the order below. The AM embedded parameter will be loaded first
712   - if any. Then, the HTK config file given by -htkconf will be
713   - parsed. If a value already set by AM embedded value, HTK config
714   - will override them. At last, the direct options will be loaded,
715   - which will override settings loaded before. Note that, when the
716   - same options are specified several times, later will override
717   - previous, except that -htkconf will be evaluated first as
718   - described above.
719   -
720   - -smpPeriod period
721   - Sampling period of input speech, in unit of 100 nanoseconds.
722   - Sampling rate can also be specified by -smpFreq. Please note
723   - that the input frequency should be set equal to the training
724   - conditions of AM. (default: 625, corresponds to 16,000Hz)
725   -
726   - This option corresponds to the HTK Option SOURCERATE. The
727   - same value can be given to this option.
728   -
729   - When using multiple AM, this value should be the same among
730   - all AMs.
731   -
732   - -smpFreq Hz
733   - Set sampling frequency of input speech in Hz. Sampling rate
734   - can also be specified using -smpPeriod. Please note that this
735   - frequency should be set equal to the training conditions of
736   - AM. (default: 16,000)
737   -
738   - When using multiple AM, this value should be the same among
739   - all AMs.
740   -
741   - -fsize sample_num
742   - Window size in number of samples. (default: 400)
743   -
744   - This option corresponds to the HTK Option WINDOWSIZE, but
745   - value should be in samples (HTK value / smpPeriod).
746   -
747   - When using multiple AM, this value should be the same among
748   - all AMs.
749   -
750   - -fshift sample_num
751   - Frame shift in number of samples. (default: 160)
752   -
753   - This option corresponds to the HTK Option TARGETRATE, but
754   - value should be in samples (HTK value / smpPeriod).
755   -
756   - When using multiple AM, this value should be the same among
757   - all AMs.
758   -
759   - -preemph float
760   - Pre-emphasis coefficient. (default: 0.97)
761   -
762   - This option corresponds to the HTK Option PREEMCOEF. The same
763   - value can be given to this option.
764   -
765   - -fbank num
766   - Number of filterbank channels. (default: 24)
767   -
768   - This option corresponds to the HTK Option NUMCHANS. The same
769   - value can be given to this option. Be aware that the default
770   - value not the same as in HTK (22).
771   -
772   - -ceplif num
773   - Cepstral liftering coefficient. (default: 22)
774   -
775   - This option corresponds to the HTK Option CEPLIFTER. The same
776   - value can be given to this option.
777   -
778   - -rawe , -norawe
779   - Enable/disable using raw energy before pre-emphasis (default:
780   - disabled)
781   -
782   - This option corresponds to the HTK Option RAWENERGY. Be aware
783   - that the default value differs from HTK (enabled at HTK,
784   - disabled at Julius).
785   -
786   - -enormal , -noenormal
787   - Enable/disable normalizing log energy. On live input, this
788   - normalization will be approximated from the average of last
789   - input. (default: disabled)
790   -
791   - This option corresponds to the HTK Option ENORMALISE. Be
792   - aware that the default value differs from HTK (enabled at
793   - HTK, disabled at Julius).
794   -
795   - -escale float_scale
796   - Scaling factor of log energy when normalizing log energy.
797   - (default: 1.0)
798   -
799   - This option corresponds to the HTK Option ESCALE. Be aware
800   - that the default value differs from HTK (0.1).
801   -
802   - -silfloor float
803   - Energy silence floor in dB when normalizing log energy.
804   - (default: 50.0)
805   -
806   - This option corresponds to the HTK Option SILFLOOR.
807   -
808   - -delwin frame
809   - Delta window size in number of frames. (default: 2)
810   -
811   - This option corresponds to the HTK Option DELTAWINDOW. The
812   - same value can be given to this option.
813   -
814   - -accwin frame
815   - Acceleration window size in number of frames. (default: 2)
816   -
817   - This option corresponds to the HTK Option ACCWINDOW. The same
818   - value can be given to this option.
819   -
820   - -hifreq Hz
821   - Enable band-limiting for MFCC filterbank computation: set
822   - upper frequency cut-off. Value of -1 will disable it.
823   - (default: -1)
824   -
825   - This option corresponds to the HTK Option HIFREQ. The same
826   - value can be given to this option.
827   -
828   - -lofreq Hz
829   - Enable band-limiting for MFCC filterbank computation: set
830   - lower frequency cut-off. Value of -1 will disable it.
831   - (default: -1)
832   -
833   - This option corresponds to the HTK Option LOFREQ. The same
834   - value can be given to this option.
835   -
836   - -zmeanframe , -nozmeanframe
837   - With speech input, this option enables/disables frame-wise DC
838   - offset removal. This corresponds to HTK configuration
839   - ZMEANSOURCE. This cannot be used together with -zmean.
840   - (default: disabled)
841   -
842   - -usepower
843   - Use power instead of magnitude on filterbank analysis.
844   - (default: disabled)
845   -
846   - Normalization
847   - Julius can perform cepstral mean normalization (CMN) for inputs.
848   - CMN will be activated when the given AM was trained with CMN
849   - (i.e. has "_Z" qualifier in the header).
850   -
851   - The cepstral mean will be estimated in different way according
852   - to the input type. On file input, the mean will be computed from
853   - the whole input. On live input such as microphone and network
854   - input, the ceptral mean of the input is unknown at the start. So
855   - MAP-CMN will be used. On MAP-CMN, an initial mean vector will be
856   - applied at the beginning, and the mean vector will be smeared to
857   - the mean of the incrementing input vector as input goes. Options
858   - below can control the behavior of MAP-CMN.
859   -
860   - -cvn
861   - Enable cepstral variance normalization. At file input, the
862   - variance of whole input will be calculated and then applied.
863   - At live microphone input, variance of the last input will be
864   - applied. CVN is only supported for an audio input.
865   -
866   - -vtln alpha lowcut hicut
867   - Do frequency warping, typically for a vocal tract length
868   - normalization (VTLN). Arguments are warping factor, high
869   - frequency cut-off and low freq. cut-off. They correspond to
870   - HTK Config values, WARPFREQ, WARPHCUTOFF and WARPLCUTOFF.
871   -
872   - -cmnload file
873   - Load initial cepstral mean vector from file on startup. The
874   - file should be one saved by -cmnsave. Loading an initial
875   - cepstral mean enables Julius to better recognize the first
876   - utterance on a real-time input. When used together with
877   - -cmnnoupdate, this initial value will be used for all input.
878   -
879   - -cmnsave file
880   - Save the calculated cepstral mean vector into file. The
881   - parameters will be saved at each input end. If the output
882   - file already exists, it will be overridden.
883   -
884   - -cmnupdate -cmnnoupdate
885   - Control whether to update the cepstral mean at each input on
886   - real-time input. Disabling this and specifying -cmnload will
887   - make engine to always use the loaded static initial cepstral
888   - mean.
889   -
890   - -cmnmapweight float
891   - Specify the weight of initial cepstral mean for MAP-CMN.
892   - Specify larger value to retain the initial cepstral mean for
893   - a longer period, and smaller value to make the cepstral mean
894   - rely more on the current input. (default: 100.0)
895   -
896   - Front-end processing
897   - Julius can perform spectral subtraction to reduce some
898   - stationary noise from audio input. Though it is not a powerful
899   - method, but it may work on some situation. Julius has two ways
900   - to estimate noise spectrum. One way is to assume that the first
901   - short segment of an speech input is noise segment, and estimate
902   - the noise spectrum as the average of the segment. Another way is
903   - to calculate average spectrum from noise-only input using other
904   - tool mkss, and load it in Julius. The former one is popular for
905   - speech file input, and latter should be used in live input. The
906   - options below will switch / control the behavior.
907   -
908   - -sscalc
909   - Perform spectral subtraction using head part of each file as
910   - silence part. The head part length should be specified by
911   - -sscalclen. Valid only for file input. Conflict with -ssload.
912   -
913   - -sscalclen msec
914   - With -sscalc, specify the length of head silence for noise
915   - spectrum estimation in milliseconds. (default: 300)
916   -
917   - -ssload file
918   - Perform spectral subtraction for speech input using
919   - pre-estimated noise spectrum loaded from file. The noise
920   - spectrum file can be made by mkss. Valid for all speech
921   - input. Conflict with -sscalc.
922   -
923   - -ssalpha float
924   - Alpha coefficient of spectral subtraction for -sscalc and
925   - -ssload. Noise will be subtracted stronger as this value gets
926   - larger, but distortion of the resulting signal also becomes
927   - remarkable. (default: 2.0)
928   -
929   - -ssfloor float
930   - Flooring coefficient of spectral subtraction. The spectral
931   - power that goes below zero after subtraction will be
932   - substituted by the source signal with this coefficient
933   - multiplied. (default: 0.5)
934   -
935   - Misc. AM options
936   - -htkconf file
937   - Parse the given HTK Config file, and set corresponding
938   - parameters to Julius. When using this option, the default
939   - parameter values are switched from Julius defaults to HTK
940   - defaults.
941   -
942   - Recognition process and search (-SR)
943   - This section contains options for search parameters on the 1st / 2nd
944   - pass such as beam width and LM weights, configurations for short-pause
945   - segmentation, switches for word lattice output and confusion network
946   - output, forced alignments, and other options relating recognition
947   - process and result output.
948   -
949   - Default values for beam width and LM weights will change according to
950   - compile-time setup of JuliusLib , AM model type, and LM size. Please
951   - see the startup log for the actual values.
952   -
953   - 1st pass parameters
954   - -lmp weight penalty
955   - (N-gram) Language model weights and word insertion penalties
956   - for the first pass.
957   -
958   - -penalty1 penalty
959   - (Grammar) word insertion penalty for the first pass.
960   - (default: 0.0)
961   -
962   - -b width
963   - Beam width in number of HMM nodes for rank beaming on the
964   - first pass. This value defines search width on the 1st pass,
965   - and has dominant effect on the total processing time. Smaller
966   - width will speed up the decoding, but too small value will
967   - result in a substantial increase of recognition errors due to
968   - search failure. Larger value will make the search stable and
969   - will lead to failure-free search, but processing time will
970   - grow in proportion to the width.
971   -
972   - The default value is dependent on acoustic model type: 400
973   - (monophone), 800 (triphone), or 1000 (triphone, setup=v2.1)
974   -
975   - -nlimit num
976   - Upper limit of token per node. This option is valid when
977   - --enable-wpair and --enable-wpair-nlimit are enabled at
978   - compilation time.
979   -
980   - -progout
981   - Enable progressive output of the partial results on the first
982   - pass.
983   -
984   - -proginterval msec
985   - Set the time interval for -progout in milliseconds. (default:
986   - 300)
987   -
988   - 2nd pass parameters
989   - -lmp2 weight penalty
990   - (N-gram) Language model weights and word insertion penalties
991   - for the second pass.
992   -
993   - -penalty2 penalty
994   - (Grammar) word insertion penalty for the second pass.
995   - (default: 0.0)
996   -
997   - -b2 width
998   - Envelope beam width (number of hypothesis) at the second
999   - pass. If the count of word expansion at a certain hypothesis
1000   - length reaches this limit while search, shorter hypotheses
1001   - are not expanded further. This prevents search to fall in
1002   - breadth-first-like situation stacking on the same position,
1003   - and improve search failure mostly for large vocabulary
1004   - condition. (default: 30)
1005   -
1006   - -sb float
1007   - Score envelope width for enveloped scoring. When calculating
1008   - hypothesis score for each generated hypothesis, its trellis
1009   - expansion and Viterbi operation will be pruned in the middle
1010   - of the speech if score on a frame goes under the width.
1011   - Giving small value makes the second pass faster, but
1012   - computation error may occur. (default: 80.0)
1013   -
1014   - -s num
1015   - Stack size, i.e. the maximum number of hypothesis that can be
1016   - stored on the stack during the search. A larger value may
1017   - give more stable results, but increases the amount of memory
1018   - required. (default: 500)
1019   -
1020   - -m count
1021   - Number of expanded hypotheses required to discontinue the
1022   - search. If the number of expanded hypotheses is greater then
1023   - this threshold then, the search is discontinued at that
1024   - point. The larger this value is, The longer Julius gets to
1025   - give up search. (default: 2000)
1026   -
1027   - -n num
1028   - The number of candidates Julius tries to find. The search
1029   - continues till this number of sentence hypotheses have been
1030   - found. The obtained sentence hypotheses are sorted by score,
1031   - and final result is displayed in the order (see also the
1032   - -output). The possibility that the optimum hypothesis is
1033   - correctly found increases as this value gets increased, but
1034   - the processing time also becomes longer. The default value
1035   - depends on the engine setup on compilation time: 10
1036   - (standard) or 1 (fast or v2.1)
1037   -
1038   - -output num
1039   - The top N sentence hypothesis to be output at the end of
1040   - search. Use with -n (default: 1)
1041   -
1042   - -lookuprange frame
1043   - Set the number of frames before and after to look up next
1044   - word hypotheses in the word trellis on the second pass. This
1045   - prevents the omission of short words, but with a large value,
1046   - the number of expanded hypotheses increases and system
1047   - becomes slow. (default: 5)
1048   -
1049   - -looktrellis
1050   - (Grammar) Expand only the words survived on the first pass
1051   - instead of expanding all the words predicted by grammar. This
1052   - option makes second pass decoding faster especially for large
1053   - vocabulary condition, but may increase deletion error of
1054   - short words. (default: disabled)
1055   -
1056   - Short-pause segmentation / decoder-VAD
1057   - When compiled with --enable-decoder-vad, the short-pause
1058   - segmentation will be extended to support decoder-based VAD.
1059   -
1060   - -spsegment
1061   - Enable short-pause segmentation mode. Input will be segmented
1062   - when a short pause word (word with only silence model in
1063   - pronunciation) gets the highest likelihood at certain
1064   - successive frames on the first pass. When detected segment
1065   - end, Julius stop the 1st pass at the point, perform 2nd pass,
1066   - and continue with next segment. The word context will be
1067   - considered among segments. (Rev.4.0)
1068   -
1069   - When compiled with --enable-decoder-vad, this option enables
1070   - decoder-based VAD, to skip long silence.
1071   -
1072   - -spdur frame
1073   - Short pause duration length to detect end of input segment,
1074   - in number of frames. (default: 10)
1075   -
1076   - -pausemodels string
1077   - A comma-separated list of pause model names to be used at
1078   - short-pause segmentation. The word whose pronunciation
1079   - consists of only the pause models will be treated as "pause
1080   - word" and used for pause detection. If not specified, name of
1081   - -spmodel, -silhead and -siltail will be used. (Rev.4.0)
1082   -
1083   - -spmargin frame
1084   - Back step margin at trigger up for decoder-based VAD. When
1085   - speech up-trigger found by decoder-VAD, Julius will rewind
1086   - the input parameter by this value, and start recognition at
1087   - the point. (Rev.4.0)
1088   -
1089   - This option will be valid only if compiled with
1090   - --enable-decoder-vad.
1091   -
1092   - -spdelay frame
1093   - Trigger decision delay frame at trigger up for decoder-based
1094   - VAD. (Rev.4.0)
1095   -
1096   - This option will be valid only if compiled with
1097   - --enable-decoder-vad.
1098   -
1099   - Word lattice / confusion network output
1100   - -lattice , -nolattice
1101   - Enable / disable generation of word graph. Search algorithm
1102   - also has changed to optimize for better word graph
1103   - generation, so the sentence result may not be the same as
1104   - normal N-best recognition. (Rev.4.0)
1105   -
1106   - -confnet , -noconfnet
1107   - Enable / disable generation of confusion network. Enabling
1108   - this will also activates -lattice internally. (Rev.4.0)
1109   -
1110   - -graphrange frame
1111   - Merge same words at neighbor position at graph generation. If
1112   - the beginning time and ending time of two word candidates of
1113   - the same word is within the specified range, they will be
1114   - merged. The default is 0 (allow merging same words on exactly
1115   - the same location) and specifying larger value will result in
1116   - smaller graph output. Setting this value to -1 will disable
1117   - merging, in that case same words on the same location of
1118   - different scores will be left as they are. (default: 0)
1119   -
1120   - -graphcut depth
1121   - Cut the resulting graph by its word depth at post-processing
1122   - stage. The depth value is the number of words to be allowed
1123   - at a frame. Setting to -1 disables this feature. (default:
1124   - 80)
1125   -
1126   - -graphboundloop count
1127   - Limit the number of boundary adjustment loop at
1128   - post-processing stage. This parameter prevents Julius from
1129   - blocking by infinite adjustment loop by short word
1130   - oscillation. (default: 20)
1131   -
1132   - -graphsearchdelay , -nographsearchdelay
1133   - When this option is enabled, Julius modifies its graph
1134   - generation algorithm on the 2nd pass not to terminate search
1135   - by graph merging, until the first sentence candidate is
1136   - found. This option may improve graph accuracy, especially
1137   - when you are going to generate a huge word graph by setting
1138   - broad search. Namely, it may result in better graph accuracy
1139   - when you set wide beams on both 1st pass -b and 2nd pass -b2,
1140   - and large number for -n. (default: disabled)
1141   -
1142   - Multi-gram / multi-dic recognition
1143   - -multigramout , -nomultigramout
1144   - On grammar recognition using multiple grammars, Julius will
1145   - output only the best result among all grammars. Enabling this
1146   - option will make Julius to output result for each grammar.
1147   - (default: disabled)
1148   -
1149   - Forced alignment
1150   - -walign
1151   - Do viterbi alignment per word units for the recognition
1152   - result. The word boundary frames and the average acoustic
1153   - scores per frame will be calculated.
1154   -
1155   - -palign
1156   - Do viterbi alignment per phone units for the recognition
1157   - result. The phone boundary frames and the average acoustic
1158   - scores per frame will be calculated.
1159   -
1160   - -salign
1161   - Do viterbi alignment per state for the recognition result.
1162   - The state boundary frames and the average acoustic scores per
1163   - frame will be calculated.
1164   -
1165   - Misc. search options
1166   - -inactive
1167   - Start this recognition process instance with inactive state.
1168   - (Rev.4.0)
1169   -
1170   - -1pass
1171   - Perform only the first pass.
1172   -
1173   - -fallback1pass
1174   - When 2nd pass fails, Julius finish the recognition with no
1175   - result. This option tell Julius to output the 1st pass result
1176   - as a final result when the 2nd pass fails. Note that some
1177   - score output (confidence etc.) may not be useful. This was
1178   - the default behavior of Julius-3.x.
1179   -
1180   - -no_ccd , -force_ccd
1181   - Explicitly switch phone context handling at search. Normally
1182   - Julius determines whether the using AM is a context-dependent
1183   - model or not from the model names, i.e., whether the names
1184   - contain character + and -. This option will override the
1185   - automatic detection.
1186   -
1187   - -cmalpha float
1188   - Smoothing parameter for confidence scoring. (default: 0.05)
1189   -
1190   - -iwsp
1191   - (Multi-path mode only) Enable inter-word context-free short
1192   - pause insertion. This option appends a skippable short pause
1193   - model for every word end. The short-pause model can be
1194   - specified by -spmodel.
1195   -
1196   - -transp float
1197   - Additional insertion penalty for transparent words. (default:
1198   - 0.0)
1199   -
1200   - -demo
1201   - Equivalent to -progout -quiet.
1202   -
1203   -ENVIRONMENT VARIABLES
1204   - ALSADEV
1205   - (using mic input with alsa device) specify a capture device name. If
1206   - not specified, "default" will be used.
1207   -
1208   - AUDIODEV
1209   - (using mic input with oss device) specify a capture device path. If
1210   - not specified, "/dev/dsp" will be used.
1211   -
1212   - LATENCY_MSEC
1213   - Try to set input latency of microphone input in milliseconds.
1214   - Smaller value will shorten latency but sometimes make process
1215   - unstable. Default value will depend on the running OS.
1216   -
1217   -EXAMPLES
1218   - For examples of system usage, refer to the tutorial section in the
1219   - Julius documents.
1220   -
1221   -NOTICE
1222   - Note about jconf files: relative paths in a jconf file are interpreted
1223   - as relative to the jconf file itself, not to the current directory.
1224   -
1225   -SEE ALSO
1226   - julian(1), jcontrol(1), adinrec(1), adintool(1), mkbingram(1),
1227   - mkbinhmm(1), mkgsmm(1), wav2mfcc(1), mkss(1)
1228   -
1229   - http://julius.sourceforge.jp/en/
1230   -
1231   -DIAGNOSTICS
1232   - Julius normally will return the exit status 0. If an error occurs,
1233   - Julius exits abnormally with exit status 1. If an input file cannot be
1234   - found or cannot be loaded for some reason then Julius will skip
1235   - processing for that file.
1236   -
1237   -BUGS
1238   - There are some restrictions to the type and size of the models Julius
1239   - can use. For a detailed explanation refer to the Julius documentation.
1240   - For bug-reports, inquires and comments please contact julius-info at
1241   - lists.sourceforge.jp.
1242   -
1243   -COPYRIGHT
1244   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
1245   -
1246   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
1247   -
1248   - Copyright (c) 2000-2008 Shikano Lab., Nara Institute of Science and
1249   - Technology
1250   -
1251   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
1252   - Technology
1253   -
1254   -AUTHORS
1255   - Rev.1.0 (1998/02/20)
1256   - Designed by Tatsuya KAWAHARA and Akinobu LEE (Kyoto University)
1257   -
1258   - Development by Akinobu LEE (Kyoto University)
1259   -
1260   - Rev.1.1 (1998/04/14), Rev.1.2 (1998/10/31), Rev.2.0 (1999/02/20),
1261   - Rev.2.1 (1999/04/20), Rev.2.2 (1999/10/04), Rev.3.0 (2000/02/14),
1262   - Rev.3.1 (2000/05/11)
1263   - Development of above versions by Akinobu LEE (Kyoto University)
1264   -
1265   - Rev.3.2 (2001/08/15), Rev.3.3 (2002/09/11), Rev.3.4 (2003/10/01),
1266   - Rev.3.4.1 (2004/02/25), Rev.3.4.2 (2004/04/30)
1267   - Development of above versions by Akinobu LEE (Nara Institute of
1268   - Science and Technology)
1269   -
1270   - Rev.3.5 (2005/11/11), Rev.3.5.1 (2006/03/31), Rev.3.5.2 (2006/07/31),
1271   - Rev.3.5.3 (2006/12/29), Rev.4.0 (2007/12/19), Rev.4.1 (2008/10/03)
1272   - Development of above versions by Akinobu LEE (Nagoya Institute of
1273   - Technology)
1274   -
1275   -THANKS TO
1276   - From rev.3.2, Julius is released by the "Information Processing
1277   - Society, Continuous Speech Consortium".
1278   -
1279   - The Windows DLL version was developed and released by Hideki BANNO
1280   - (Nagoya University).
1281   -
1282   - The Windows Microsoft Speech API compatible version was developed by
1283   - Takashi SUMIYOSHI (Kyoto University).
1284   -
1285   -
1286   -
1287   - 02/11/2009 JULIUS(1)
recognize/src/julius/doc/manuals/mkbingram.txt
View file @e1cf2b2
... ... @@ -1,97 +0,0 @@
1   - mkbingram
2   -
3   -MKBINGRAM(1) MKBINGRAM(1)
4   -
5   -
6   -
7   -NAME
8   - mkbingram
9   - - make binary N-gram from ARPA N-gram file
10   -
11   -SYNOPSIS
12   - mkbingram [-nlr forward_ngram.arpa] [-nrl backward_ngram.arpa]
13   - [-d old_bingram_file] {output_bingram_file}
14   -
15   -DESCRIPTION
16   - mkbingram is a tool to convert N-gram definition file(s) in ARPA
17   - standard format to a compact Julius binary format. It will speed up the
18   - initial loading time of N-gram much faster. It can read gzipped file
19   - directly.
20   -
21   - From rev.4.0, Julius can deal with forward N-gram, backward N-gram and
22   - their combinations. So, mkbingram now generates binary N-gram file from
23   - one of them, or combining them two to produce one binary N-gram.
24   -
25   - When only a forward N-gram is specified, mkbingram generates binary
26   - N-gram from only the forward N-gram. When using this binary N-gram at
27   - Julius, it performs the 1st pass with the 2-gram probabilities in the
28   - N-gram, and run the 2nd pass with the given N-gram fully, with
29   - converting forward probabilities to backward probabilities by Bayes
30   - rule.
31   -
32   - When only a backward N-gram is specified, mkbingram generates an binary
33   - N-gram file that contains only the backward N-gram. The 1st pass will
34   - use forward 2-gram probabilities that can be computed from the backward
35   - 2-gram using Bayes rule, and the 2nd pass use the given backward N-gram
36   - fully.
37   -
38   - When both forward and backward N-grams are specified, the 2-gram part
39   - in the forward N-gram and all backward N-gram will be combined into
40   - single bingram file. The forward 2-gram will be applied for the 1st
41   - pass and backward N-gram for the 2nd pass. Note that both N-gram should
42   - be trained in the same corpus with same parameters (i.e. cut-off
43   - thresholds), with same vocabulary.
44   -
45   - The old binary N-gram produced by mkbingram of version 3.x and earlier
46   - can be used in Julius-4, but you can convert the old version to the new
47   - version by specifying it as input of current mkbingram by option "-d".
48   -
49   - Please note that binary N-gram file converted by mkbingram of version
50   - 4.0 and later cannot be read by older Julius 3.x.
51   -
52   -OPTIONS
53   - -nlr forward_ngram.arpa
54   - Read in a forward (left-to-right) word N-gram file in ARPA standard
55   - format.
56   -
57   - -nrl backward_ngram.arpa
58   - Read in a backward (right-to-left) word N-gram file in ARPA standard
59   - format.
60   -
61   - -d old_bingram_file
62   - Read in a binary N-gram file.
63   -
64   - -swap
65   - Swap BOS word <s> and EOS word </s> in N-gram.
66   -
67   - output_bingram_file
68   - binary N-gram file name to output.
69   -
70   -EXAMPLES
71   - Convert a set of forward and backward N-gram in ARPA format into Julius
72   - binary form:
73   - Convert a single forward 4-gram in ARPA format into a binary file:
74   - Convert old binary N-gram file to current format:
75   -
76   -SEE ALSO
77   - julius ( 1 ) ,
78   - mkbinhmm ( 1 ) ,
79   - mkbinhmmlist ( 1 )
80   -
81   -COPYRIGHT
82   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
83   -
84   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
85   -
86   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
87   - Technology
88   -
89   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
90   - Technology
91   -
92   -LICENSE
93   - The same as Julius.
94   -
95   -
96   -
97   - 02/11/2009 MKBINGRAM(1)
recognize/src/julius/doc/manuals/mkbinhmm.txt
View file @e1cf2b2
... ... @@ -1,78 +0,0 @@
1   - mkbinhmm
2   -
3   -MKBINHMM(1) MKBINHMM(1)
4   -
5   -
6   -
7   -NAME
8   - mkbinhmm
9   - - convert HMM definition file in HTK ascii format to Julius binary
10   - format
11   -
12   -SYNOPSIS
13   - mkbinhmm [-htkconf HTKConfigFile] {hmmdefs_file} {binhmm_file}
14   -
15   -DESCRIPTION
16   - mkbinhmm convert an HMM definition file in HTK ascii format into a
17   - binary HMM file for Julius. It will greatly speed up the launch
18   - process.
19   -
20   - You can also embed acoustic analysis condition parameters needed for
21   - recognition into the output file. To embed the parameters, specify the
22   - HTK Config file you have used to extract acoustic features for training
23   - the HMM by the optione "-htkconf".
24   -
25   - The embedded parameters in a binary HMM format will be loaded into
26   - Julius automatically, so you do not need to specify the acoustic
27   - feature options at run time. It will be convenient when you deliver an
28   - acoustic model.
29   -
30   - You can also specify binary file as the input. This can be used to
31   - update the old binary format into new one, or to embed the config
32   - parameters into the already existing binary files. If the input binhmm
33   - already has acoustic analysis parameters embedded, they will be
34   - overridden by the specified values.
35   -
36   -
37   - mkbinhmm can read gzipped file as input.
38   -
39   -OPTIONS
40   - -htkconf HTKConfigFile
41   - HTK Config file you used at training time. If specified, the values
42   - are embedded to the output file.
43   -
44   - hmmdefs_file
45   - The source HMm definitino file in HTK ascii format or Julius binary
46   - format.
47   -
48   - hmmdefs_file
49   - Output file.
50   -
51   -EXAMPLES
52   - Convert HTK ascii format HMM definition file into Julius binary file:
53   - Furthermore, embed acoustic feature parameters as specified by Config
54   - file
55   - Embed the acoustic parameters into an existing binary file
56   -
57   -SEE ALSO
58   - julius ( 1 ) ,
59   - mkbingram ( 1 ) ,
60   - mkbinhmmlist ( 1 )
61   -
62   -COPYRIGHT
63   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
64   -
65   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
66   -
67   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
68   - Technology
69   -
70   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
71   - Technology
72   -
73   -LICENSE
74   - The same as Julius.
75   -
76   -
77   -
78   - 10/02/2008 MKBINHMM(1)
recognize/src/julius/doc/manuals/mkbinhmmlist.txt
View file @e1cf2b2
... ... @@ -1,64 +0,0 @@
1   - mkbinhmmlist
2   -
3   -MKBINHMMLIST(1) MKBINHMMLIST(1)
4   -
5   -
6   -
7   -NAME
8   - mkbinhmmlist
9   - - convert HMMList file into binary format
10   -
11   -SYNOPSIS
12   - mkbinhmmlist {hmmdefs_file} {HMMList_file} {output_binhmmlist_file}
13   -
14   -DESCRIPTION
15   - mkbinhmmlist converts a HMMList file to binary format. Since the index
16   - trees for lookup are also stored in the binary format, it will speed up
17   - the startup of Julius, namely when using big HMMList file.
18   -
19   - For conversion, HMM definition file hmmdefs_file that will be used
20   - together at Julius needs to be specified. The format of the HMM
21   - definition file can be either ascii or Julius binary format.
22   -
23   - The output binary file can be used in Julius as the same by "-hlist".
24   - The format wil be auto-detected by Julius.
25   -
26   -
27   - mkbinhmmlist can read gzipped file.
28   -
29   -OPTIONS
30   - hmmdefs_file
31   - Acoustic HMM definition file, in HMM ascii format or Julius binary
32   - format.
33   -
34   - HMMList_file
35   - Source HMMList file
36   -
37   - output_binhmmlist_file
38   - Output file, will be overwritten if already exist.
39   -
40   -EXAMPLES
41   - Convert a HMMList file logicalTri into binary format and store to
42   - logicalTri.bin:
43   -
44   -SEE ALSO
45   - julius ( 1 ) ,
46   - mkbinhmm ( 1 )
47   -
48   -COPYRIGHT
49   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
50   -
51   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
52   -
53   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
54   - Technology
55   -
56   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
57   - Technology
58   -
59   -LICENSE
60   - The same as Julius.
61   -
62   -
63   -
64   - 10/02/2008 MKBINHMMLIST(1)
recognize/src/julius/doc/manuals/mkdfa.pl.txt
View file @e1cf2b2
... ... @@ -1,67 +0,0 @@
1   - mkdfa.pl
2   -
3   -MKDFA.PL(1) MKDFA.PL(1)
4   -
5   -
6   -
7   -NAME
8   - mkdfa.pl
9   - - grammar compiler
10   -
11   -SYNOPSIS
12   - mkdfa.pl [options...] {prefix}
13   -
14   -DESCRIPTION
15   - mkdfa.pl compiles the Julian format grammar (.grammar and .voca) to
16   - Julian native formats (.dfa and .dict). In addition, ".term" will be
17   - also generated that stores correspondence of category ID used in the
18   - output files to the source category name.
19   -
20   -
21   - prefix should be the common file name prefix of ".grammar" and "voca"
22   - file. From prefix.grammar and prefix.voca file, prefix.dfa, prefix.dict
23   - and prefix.term will be output.
24   -
25   -OPTIONS
26   - -n
27   - Not process dictionary. You can only convert .grammar file to .dfa
28   - file without .voca file.
29   -
30   -ENVIRONMENT VARIABLES
31   - TMP or TEMP
32   - Set directory to store temporal file. If not specified, one of them
33   - on the following list will be used: /tmp, /var/tmp, /WINDOWS/Temp,
34   - /WINNT/Temp.
35   -
36   -EXAMPLES
37   - Convert a grammar foo.grammar and foo.voca to foo.dfa, foo.voca and
38   - foo.term.
39   -
40   -SEE ALSO
41   - julius ( 1 ) ,
42   - generate ( 1 ) ,
43   - nextword ( 1 ) ,
44   - accept_check ( 1 ) ,
45   - dfa_minimize ( 1 )
46   -
47   -DIAGNOSTICS
48   - mkdfa.pl invokes mkfa and dfa_minimize internally. They should be
49   - placed at the same directory as mkdfa.pl.
50   -
51   -COPYRIGHT
52   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
53   -
54   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
55   -
56   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
57   - Technology
58   -
59   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
60   - Technology
61   -
62   -LICENSE
63   - The same as Julius.
64   -
65   -
66   -
67   - 10/02/2008 MKDFA.PL(1)
recognize/src/julius/doc/manuals/mkgshmm.txt
View file @e1cf2b2
... ... @@ -1,50 +0,0 @@
1   - mkgshmm
2   -
3   -MKGSHMM(1) MKGSHMM(1)
4   -
5   -
6   -
7   -NAME
8   - mkgshmm
9   - - convert monophone HMM to GS HMM for Julius
10   -
11   -SYNOPSIS
12   - mkgshmm {monophone_hmmdefs}
13   - >
14   - {outputfile}
15   -
16   -DESCRIPTION
17   - mkgshmm converts monophone HMM definition file in HTK format into a
18   - special format for Gaussian Mixture Selection (GMS) in Julius.
19   -
20   - GMS is an algorithm to reduce the amount of acoustic computation with
21   - triphone HMM, by pre-selection of promising gaussian mixtures using
22   - likelihoods of corresponding monophone mixtures.
23   -
24   -EXAMPLES
25   - (1) Prepare a monophone model which was trained by the same corpus as
26   - target triphone model.
27   -
28   - (2) Convert the monophone model using mkgshmm.
29   - (3) Specify the output file in Julius with option "-gshmm"
30   -
31   -SEE ALSO
32   - julius ( 1 )
33   -
34   -COPYRIGHT
35   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
36   -
37   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
38   -
39   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
40   - Technology
41   -
42   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
43   - Technology
44   -
45   -LICENSE
46   - The same as Julius.
47   -
48   -
49   -
50   - 10/02/2008 MKGSHMM(1)
recognize/src/julius/doc/manuals/mkss.txt
View file @e1cf2b2
... ... @@ -1,55 +0,0 @@
1   - mkss
2   -
3   -MKSS(1) MKSS(1)
4   -
5   -
6   -
7   -NAME
8   - mkss
9   - - calculate average spectrum for spectral subtraction
10   -
11   -SYNOPSIS
12   - mkss [options...] {filename}
13   -
14   -DESCRIPTION
15   - mkss is a tool to estimate noise spectrum for spectral subtraction on
16   - Julius. It reads a few seconds of sound data from microphone input,
17   - calculate the average spectrum and save it to a file. The output file
18   - can be used as a noise spectrum data in Julius (option "-ssload").
19   -
20   - The recording will start immediately after startup. Sampling format is
21   - 16bit, monoral. If outpue file already exist, it will be overridden.
22   -
23   -OPTIONS
24   - -freq Hz
25   - Sampling frequency in Hz (default: 16,000)
26   -
27   - -len msec
28   - capture length in milliseconds (default: 3000)
29   -
30   - -fsize sample_num
31   - frame size in number of samples (default: 400)
32   -
33   - -fshift sample_num
34   - frame shift in number of samples (default: 160)
35   -
36   -SEE ALSO
37   - julius ( 1 )
38   -
39   -COPYRIGHT
40   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
41   -
42   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
43   -
44   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
45   - Technology
46   -
47   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
48   - Technology
49   -
50   -LICENSE
51   - The same as Julius.
52   -
53   -
54   -
55   - 10/02/2008 MKSS(1)
recognize/src/julius/doc/manuals/nextword.txt
View file @e1cf2b2
... ... @@ -1,82 +0,0 @@
1   - nextword
2   -
3   -NEXTWORD(1) NEXTWORD(1)
4   -
5   -
6   -
7   -NAME
8   - nextword
9   - - display next predicted words (in reverse order)
10   -
11   -SYNOPSIS
12   - nextword [-t] [-r] [-s spname] [-v] {prefix}
13   -
14   -DESCRIPTION
15   - Given a partial (part of) sentence from the end, it outputs the next
16   - words allowed in the specified grammar.
17   -
18   -
19   - .dfa, .dict and .term files are needed to execute. They can be
20   - generated from .grammar and .voca file by mkdfa.pl.
21   -
22   - Please note that the latter part of sentence should be given, since the
23   - main 2nd pass does a right-to-left parsing.
24   -
25   -OPTIONS
26   - -t
27   - Input / Output in category name. (default: word)
28   -
29   - -r
30   - Enter in reverse order
31   -
32   - -s spname
33   - the name string of short-pause word to be supressed (default: "sp")
34   -
35   - -v
36   - Debug output.
37   -
38   -EXAMPLES
39   - Exmple output of a sample grammar "fruit":
40   -
41   - % nextword fruit
42   - Stat: init_voca: read 36 words
43   - Reading in term file (optional)...done
44   - 15 categories, 36 words
45   - DFA has 26 nodes and 42 arcs
46   - -----
47   - command completion is disabled
48   - -----
49   - wseq > A BANANA </s>
50   - [wseq: A BANANA </s>]
51   - [cate: (NUM_1|NUM_1|A|A) FRUIT_SINGULAR NS_E]
52   - PREDICTED CATEGORIES/WORDS:
53   - NS_B (<s> )
54   - HAVE (HAVE )
55   - WANT (WANT )
56   - NS_B (<s> )
57   - HAVE (HAVE )
58   - WANT (WANT )
59   -
60   -
61   -SEE ALSO
62   - mkdfa.pl ( 1 ) ,
63   - generate ( 1 ) ,
64   - accept_check ( 1 )
65   -
66   -COPYRIGHT
67   - Copyright (c) 1997-2000 Information-technology Promotion Agency, Japan
68   -
69   - Copyright (c) 1991-2008 Kawahara Lab., Kyoto University
70   -
71   - Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and
72   - Technology
73   -
74   - Copyright (c) 2005-2008 Julius project team, Nagoya Institute of
75   - Technology
76   -
77   -LICENSE
78   - The same as Julius.
79   -
80   -
81   -
82   - 10/02/2008 NEXTWORD(1)
recognize/src/julius/install/julius-4.2.2.tar.gz
View file @e1cf2b2
No preview for this file type