Changes between Initial Version and Version 1 of Ticket #62


Ignore:
Timestamp:
09/13/06 00:48:15 (16 years ago)
Author:
kmaclean
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #62 – Description

    initial v1  
    55}} 
    66 
    7 julian.jconf file 
    87 
    9 {{{ 
    10 # Copyright (c) 1991-2005 Kawahara Lab., Kyoto University 
    11 # Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and Technology 
    12 # Copyright (c) 2005      Julius project team, Nagoya Institute of Technology 
    13 # All rights reserved 
    14 # 
    15 # Example of Runtime Configuration File for Julian rev.3.5 
    16 # 
    17 # 1) NOTICE: relative paths must be relative to THIS FILE. 
    18 #            Current working directory does not affect. 
    19 # 2) Blank lines and comments after '#' are ignored. 
    20 #    Use `\#' to specify `#' in an argument,  
    21 # 3) Each line should be shorter than 512 bytes. 
    22 # 4) Syntax of each arguments are same as corresponding command options. 
    23 # 5) The commented-out values in this file are default values in Julian. 
    24 # 
    25  
    26 ###################################################################### 
    27 #### Files 
    28 ###################################################################### 
    29 ## 
    30 ## Grammar definition file (DFA and dictionary) 
    31 ## 
    32  
    33 #### There are three ways to specify the grammar files. 
    34 #### (1) and (2) can be used multiple times. 
    35  
    36 #### (1) Specify by common prefix of .dfa and .dict files. Comma-separated 
    37 #### prefixes can be specified for multiple grammar recognition 
    38 #-gram /cdrom/testrun/sample_grammars/vfr/vfr 
    39  
    40 #### (2) Or you can give Julian a text file which contains list of grammar 
    41 #### prefixes one per line. 
    42 #-gramlist file 
    43  
    44 #### (3) Classic way to specify a grammar. 
    45 -dfa sample.dfa 
    46 -v sample.dict 
    47  
    48 #### If you want to clear previously specified grammars, use this at the 
    49 #### point. 
    50 #-nogram 
    51  
    52 ## 
    53 ## Acoustic HMM file 
    54 ## 
    55 # support ascii hmmdefs or binary format (converted by "mkbinhmm") 
    56 # format (ascii/binary) will be automatically detected 
    57 -h hmmdefs 
    58  
    59 ## triphone model needs HMMList that maps logical triphone to physical ones. 
    60 -hlist tiedlist 
    61  
    62 ###################################################################### 
    63 #### Multiple grammar recognition 
    64 ###################################################################### 
    65 #-multigramout          # Output results for each grammar 
    66  
    67 ###################################################################### 
    68 #### Language Model  
    69 ###################################################################### 
    70 ## 
    71 ## word insertion penalty 
    72 ## 
    73 #-penalty1 0.0          # first pass 
    74 #-penalty2 0.0          # second pass 
    75  
    76 ###################################################################### 
    77 #### Dictionary 
    78 ###################################################################### 
    79 ## 
    80 ## do not giveup startup on error words 
    81 ## 
    82 #-forcedict 
    83  
    84 ###################################################################### 
    85 #### Acoustic Model 
    86 ###################################################################### 
    87 ## 
    88 ## Context-dependency handling will be enabled according to the model type. 
    89 ## Try below if julius wrongly detect the type of hmmdefs 
    90 ## 
    91 #-no_ccd                # disable context-dependency handling 
    92 #-force_ccd             # enable context-dependency handling 
    93  
    94 ##  
    95 ## If julius go wrong with checking parameter type, try below. 
    96 ## 
    97 #-notypecheck 
    98 # 
    99  
    100 ## 
    101 ## (PTM/triphone) switch computation method of IWCD on 1st pass 
    102 ## 
    103 #-iwcd1 best N  # assign average of N-best likelihood of the same context 
    104 -iwcd1 max      # assign maximum likelihood of the same context 
    105 #-iwcd1 avg     # assign average likelihood of the same context (default) 
    106  
    107 ###################################################################### 
    108 #### Gaussian Pruning 
    109 ###################################################################### 
    110 ## Number of mixtures to select in a mixture pdf. 
    111 ## This default value is optimized for IPA99's PTM, 
    112 ## with 64 Gaussians per codebook 
    113 #-tmix 2 
    114  
    115 ## Select Gaussian pruning algorithm 
    116 ## defulat: beam (standard setting), safe (others) 
    117 -gprune safe            # safe pruning, accurate but slow 
    118 #-gprune heuristic      # heuristic pruning 
    119 #-gprune beam           # beam pruning, fast but sensitive 
    120 #-gprune none           # no pruning 
    121  
    122 ###################################################################### 
    123 #### Gaussian Mixture Selection  
    124 ###################################################################### 
    125 #-gshmm hmmdefs         # monophone HMM for GMS 
    126                         # (OFF when not specified) 
    127 #-gsnum 24              # number of states to be selected on GMS 
    128  
    129 ###################################################################### 
    130 #### Search Parameters 
    131 ###################################################################### 
    132 #-b 2000                 # beam width on 1st pass (#nodes) for monophone 
    133 #-b 800                 # beam width on 1st pass (#nodes) for triphone,PTM 
    134 #-b 1000                # beam width on 1st pass (#nodes) for triphone,PTM,engine=v2.1 
    135 #-b2 100                 # beam width on 2nd pass (#words) 
    136 #-sb 80.0               # score beam envelope threshold 
    137 #-s 500                 # hypotheses stack size on 2nd pass (#hypo) 
    138 #-m 2000                # hypotheses overflow threshold (#hypo) 
    139 #-lookuprange 5         # lookup range for word expansion (#frame) 
    140 #-n 1                   # num of sentences to find (#sentence) 
    141 #-n 10                  #   (default for 'standard' configuration) 
    142 #-output 1              # num of found sentences to output (#sentence) 
    143 #-looktrellis           # search within only backtrellis words 
    144  
    145 ###################################################################### 
    146 #### Inter-word Short Pause Handling 
    147 ###################################################################### 
    148 ## 
    149 ## Specify short pause model name to be treated as special 
    150 ## 
    151 #-spmodel "sp"          # HMM model name 
    152  
    153 ## 
    154 ## For insertion of context-free short-term inter-word pauses between words 
    155 ##  (multi-path version only) 
    156 ## 
    157 #-iwsp                  # append a skippable sp model at all word ends 
    158 #-iwsppenalty 0.0       # transition penalty for the appenede sp models 
    159  
    160 ###################################################################### 
    161 #### Speech Input Source 
    162 ###################################################################### 
    163 ## select one (default: mfcfile) 
    164 #-input mfcfile         # MFCC file in HTK parameter file format 
    165 #-input rawfile         # raw wavefile (auto-detect format) 
    166                         # WAV(16bit) or 
    167                         # RAW(16bit(signed short),mono,big-endian) 
    168                         # AIFF,AU (with libsndfile extension) 
    169                         # other than 16kHz, sampling rate should be specified 
    170                         # by "-smpFreq" option 
    171 #-input mic             # direct microphone input 
    172                         # device name can be specified via env. val. "AUDIODEV" 
    173 #-input netaudio -NA host:0    # direct input from DatLink(NetAudio) host 
    174 #-input adinnet -adport portnum # via adinnet network client 
    175 #-input stdin           # from standard tty input (pipe) 
    176  
    177 #-filelist filename     # specify file list to be recognized in batch mode 
    178  
    179 #-nostrip               # switch OFF dropping of invalid input segment. 
    180                         # (default: strip off invalid segment (0 sequence etc.) 
    181 #-zmean                 # enable DC offset removal (invalid for mfcfile input) 
    182  
    183 ###################################################################### 
    184 #### Recording 
    185 ###################################################################### 
    186 #-record directory      # auto-save recognized speech data into the dir 
    187  
    188 ###################################################################### 
    189 #### GMM-based Input Verification and Rejection 
    190 ###################################################################### 
    191 #-gmm gmmdefs           # specify GMM definition file in HTK format 
    192 #-gmmnum 10             # num of Gaussians to be computed per mixture 
    193 #-gmmreject "noise,laugh,cough" # list of GMM names to be rejected 
    194  
    195 ###################################################################### 
    196 #### Too Short Input Rejection 
    197 ###################################################################### 
    198 -rejectshort 2000       # reject input shorter than specified millisecond 
    199  
    200 ###################################################################### 
    201 #### Speech Detection 
    202 ###################################################################### 
    203 #-pausesegment          # turn on speech detection by level and zero-cross 
    204 #-nopausesegment        # turn off speech detection by level and zero-cross 
    205                         # (default: on for mic or adinnet, off for file) 
    206 #-lv 2000               # threshold of input level (0-32767) 
    207 #-headmargin 300        # head margin of input segment (msec) 
    208 #-tailmargin 400        # tail margin of input segment (msec) 
    209 #-zc 60                 # threshold of number of zero-cross in a second 
    210  
    211 ###################################################################### 
    212 #### Acoustic Analysis 
    213 ###################################################################### 
    214 -smpFreq 8000           # sampling rate (Hz) 
    215 #-smpPeriod 625         # sampling period (ns) (= 10000000 / smpFreq) 
    216 #-fsize 400             # window size (samples) 
    217 #-fshift 160            # frame shift (samples) 
    218 #-delwin 2              # delta window (frames) 
    219 #-hifreq -1             # cut-off hi frequency (Hz) (-1: disable) 
    220 #-lofreq -1             # cut-off low frequency (Hz) (-1: disable) 
    221 #-cmnsave filename      # save CMN param to file (update per input) 
    222 #-cmnload filename      # load initial CMN param from file on startup 
    223  
    224 ###################################################################### 
    225 #### Spectral Subtraction (SS) 
    226 ###################################################################### 
    227 #-sscalc                # do SS using head silence (file input only) 
    228 #-sscalclen 300         # length of head silence for SS (msec) 
    229 #-ssload filename       # load constant noise spectrum from file for SS 
    230 #-ssalpha 2.0           # alpha coef. for SS 
    231 #-ssfloor 0.5           # spectral floor for SS 
    232  
    233 ###################################################################### 
    234 #### Forced alignment 
    235 ###################################################################### 
    236 #-walign                # do forced alignment with result per word  
    237 #-palign                # do forced alignment with result per phoneme 
    238 #-salign                # do forced alignment with result per HMM state 
    239  
    240 ###################################################################### 
    241 #### Word Confidence Scoring 
    242 ###################################################################### 
    243 #-cmalpha 0.05          # smoothing coef. alpha 
    244  
    245 ###################################################################### 
    246 #### Output  
    247 ###################################################################### 
    248 #-separatescore         # output language and acoustic score separately 
    249 #-progout               # output partial result per a time interval 
    250 #-proginterval 300      # time interval for "-progout" (msec) 
    251 #-quiet                 # output minimal result 
    252 #-demo                  # = "-progout -quiet", suitable for dictation demo 
    253 #-debug                 # output full message for debug 
    254 #-charconv from to      # output character set conversion (see manual for 
    255                         # available code set name) 
    256  
    257 ###################################################################### 
    258 #### Server module mode 
    259 ###################################################################### 
    260 #-module                # Run Julius on "Server module mode" 
    261 #-module 5530           # (when using another port number for connection) 
    262 #-outcode WLPSC         # select output message toward module (WLPSCwlps) 
    263  
    264 ###################################################################### 
    265 #### Misc. 
    266 ###################################################################### 
    267 #-help                  # output help and exit 
    268 #-setting               # output engine configuration and exit 
    269 #-C jconffile           # expand other jconf file in its place 
    270  
    271 ################################################################# end of file 
    272  
    273 }}}