Julius: julius/global.h Source File

00001 
00023 /*
00024  * Copyright (c) 1991-2006 Kawahara Lab., Kyoto University
00025  * Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and Technology
00026  * Copyright (c) 2005-2006 Julius project team, Nagoya Institute of Technology, Nagoya Institute of Technology
00027  * All rights reserved
00028  */
00029 
00030 #ifndef __SENT_EXTERNAL_DEFINITION__
00031 #define __SENT_EXTERNAL_DEFINITION__
00032 
00033 #include <sent/stddefs.h>
00034 #include <sent/hmm.h>
00035 #include <sent/vocabulary.h>
00036 #ifdef USE_NGRAM
00037 #include <sent/ngram2.h>
00038 #else  /* USE_DFA */
00039 #include <sent/dfa.h>
00040 #endif
00041 #include "wchmm.h"
00042 #include "search.h"
00043 
00049 #ifdef GLOBAL_VARIABLE_DEFINE
00050 #define GLOBAL /*  */
00051 #define GLOBAL_VAL(v) = (v)
00052 #else
00053 #define GLOBAL extern
00054 #define GLOBAL_VAL(v) /*  */
00055 #endif
00056 
00057 
00058 /* -------------------------------------------------------------- */
00059 /* ------- User-specified filenames ----------------------------- */
00060 /* -------------------------------------------------------------- */
00061 
00062 /********************************************/
00063 /* File names (NULL if not specified) */
00064 GLOBAL char *hmmfilename GLOBAL_VAL(NULL);
00065 GLOBAL char *mapfilename GLOBAL_VAL(NULL); 
00066 GLOBAL char *hmm_gs_filename GLOBAL_VAL(NULL); 
00067 GLOBAL char *dictfilename GLOBAL_VAL(NULL); 
00068 #ifdef USE_NGRAM
00069 GLOBAL char *ngram_filename GLOBAL_VAL(NULL); 
00070 GLOBAL char *ngram_filename_lr_arpa GLOBAL_VAL(NULL); 
00071 GLOBAL char *ngram_filename_rl_arpa GLOBAL_VAL(NULL); 
00072 #endif
00073 #ifdef USE_DFA
00074 GLOBAL char *dfa_filename GLOBAL_VAL(NULL); 
00075 GLOBAL GRAMLIST *gramlist_root GLOBAL_VAL(NULL); 
00076 #endif
00077 
00078 GLOBAL char *inputlist_filename GLOBAL_VAL(NULL); 
00079 GLOBAL char *cmnload_filename GLOBAL_VAL(NULL); 
00080 GLOBAL char *cmnsave_filename GLOBAL_VAL(NULL); 
00081 GLOBAL char *ssload_filename GLOBAL_VAL(NULL); 
00082 GLOBAL char *record_dirname GLOBAL_VAL(NULL); 
00083 GLOBAL char *gmm_filename GLOBAL_VAL(NULL); 
00084 
00085 
00086 /* -------------------------------------------------------------- */
00087 /* ------- User-specified parameters --------- ------------------ */
00088 /* -------------------------------------------------------------- */
00089 
00090 /********************************************/
00091 /* Input source */
00092 GLOBAL int speech_input GLOBAL_VAL(SP_MFCFILE); 
00093 GLOBAL int adinnet_port GLOBAL_VAL(ADINNET_PORT); 
00094 #ifdef USE_NETAUDIO
00095 GLOBAL char *netaudio_devname GLOBAL_VAL("localhost:0"); 
00096 #endif
00097 
00098 /********************************************/
00099 /* Input triggering and silence cutting */
00100 GLOBAL int silence_cut GLOBAL_VAL(2); 
00101 GLOBAL int level_thres GLOBAL_VAL(2000); 
00102 GLOBAL int zero_cross_num GLOBAL_VAL(60); 
00103 GLOBAL int head_margin_msec GLOBAL_VAL(300); 
00104 GLOBAL int tail_margin_msec GLOBAL_VAL(400); 
00105 
00106 GLOBAL boolean strip_zero_sample GLOBAL_VAL(TRUE); 
00107 GLOBAL boolean use_zmean GLOBAL_VAL(FALSE); 
00108 
00109 /********************************************/
00110 /* Speech analysis parameters */
00111 GLOBAL int smpPeriod GLOBAL_VAL(625); 
00112 GLOBAL int smpFreq GLOBAL_VAL(16000); 
00113 GLOBAL int fshift GLOBAL_VAL(DEF_FRAMESHIFT); 
00114 GLOBAL int fsize GLOBAL_VAL(DEF_FRAMESIZE); 
00115 GLOBAL int delwin GLOBAL_VAL(DEF_DELWIN); 
00116 GLOBAL int accwin GLOBAL_VAL(DEF_ACCWIN); 
00117 /* Lo-pass/hi-pass filter */
00118 GLOBAL int hipass GLOBAL_VAL(-1); 
00119 GLOBAL int lopass GLOBAL_VAL(-1); 
00120 GLOBAL float preemph GLOBAL_VAL(DEF_PREENPH); 
00121 GLOBAL int fbank_num GLOBAL_VAL(DEF_FBANK); 
00122 GLOBAL int ceplifter GLOBAL_VAL(DEF_CEPLIF); 
00123 GLOBAL boolean rawe_required GLOBAL_VAL(FALSE); 
00124 GLOBAL boolean enormal_required GLOBAL_VAL(FALSE); 
00125 GLOBAL float enormal_escale GLOBAL_VAL(1.0); 
00126 GLOBAL float enormal_silfloor GLOBAL_VAL(DEF_SILFLOOR); 
00127 
00128 /********************************************/
00129 /* Spectral subtraction */
00130 GLOBAL boolean sscalc GLOBAL_VAL(FALSE); 
00131 GLOBAL int sscalc_len GLOBAL_VAL(300); 
00132 GLOBAL float ssalpha GLOBAL_VAL(DEF_SSALPHA); 
00133 GLOBAL float ssfloor GLOBAL_VAL(DEF_SSFLOOR); 
00134 
00135 /********************************************/
00136 /* LM parameters (default will be set in m_bootup.c) */
00137 #ifdef USE_NGRAM
00138 GLOBAL LOGPROB lm_weight;       
00139 GLOBAL LOGPROB lm_penalty;      
00140 GLOBAL LOGPROB lm_weight2;      
00141 GLOBAL LOGPROB lm_penalty2;     
00142 GLOBAL LOGPROB lm_penalty_trans GLOBAL_VAL(0.0); 
00143 #endif /* USE_NGRAM */
00144 #ifdef USE_DFA
00145 GLOBAL LOGPROB penalty1 GLOBAL_VAL(0.0); 
00146 GLOBAL LOGPROB penalty2 GLOBAL_VAL(0.0); 
00147 #endif
00148 
00149 #ifdef CONFIDENCE_MEASURE
00150 /********************************************/
00151 /* Confidence measure */
00152 #ifdef CM_MULTIPLE_ALPHA        /* Test multiple alpha coef. for confidence scoring */
00153 GLOBAL LOGPROB cm_alpha_bgn GLOBAL_VAL(0.03); 
00154 GLOBAL LOGPROB cm_alpha_end GLOBAL_VAL(0.15); 
00155 GLOBAL LOGPROB cm_alpha_step GLOBAL_VAL(0.03); 
00156 GLOBAL int cm_alpha_num GLOBAL_VAL(5); 
00157 #else  /* single value (default) */
00158 GLOBAL LOGPROB cm_alpha GLOBAL_VAL(0.05); 
00159 #endif
00160 #ifdef CM_SEARCH_LIMIT
00161 GLOBAL LOGPROB cm_cut_thres GLOBAL_VAL(0.03); 
00162 #endif
00163 #ifdef CM_SEARCH_LIMIT_POP
00164 GLOBAL LOGPROB cm_cut_thres_pop GLOBAL_VAL(0.1); 
00165 #endif
00166 #endif /* CONFIDENCE_MEASURE */
00167 
00168 /********************************************/
00169 /* Pause (silence) model related */
00170 #ifdef USE_NGRAM
00171 GLOBAL char *head_silname GLOBAL_VAL(BEGIN_WORD_DEFAULT); 
00172 GLOBAL char *tail_silname GLOBAL_VAL(END_WORD_DEFAULT); 
00173 /* Short pause word, that will be added to dictionary when "-iwspword" is spcified */
00174 GLOBAL boolean enable_iwspword GLOBAL_VAL(FALSE); 
00175 GLOBAL char *iwspentry GLOBAL_VAL(IWSPENTRY_DEFAULT); 
00176 #endif
00177 GLOBAL char *spmodel_name GLOBAL_VAL(SPMODEL_NAME_DEFAULT); 
00178 
00179 #ifdef MULTIPATH_VERSION
00180 /* 1) in DFA mode, a word with only "spmodel_name" model as a pronunciation will be specially handled as "short-pause word" */
00181 /* 2) if "-iwsp" enabled, the "spmodel_name" model will be attached to every word end within the dictionary */
00182 /* short pause special handling */
00183 GLOBAL boolean enable_iwsp GLOBAL_VAL(FALSE); /* enable inter-word short pause handling */
00184 GLOBAL LOGPROB iwsp_penalty GLOBAL_VAL(IWSP_PENALTY_DEFAULT); /* transition penalty of inter-word short pause */
00185 #endif
00186 
00187 /**********************************************/
00188 /* Search parameters for acoustic computation */
00189 GLOBAL int gprune_method GLOBAL_VAL(GPRUNE_SEL_UNDEF); 
00190 GLOBAL int mixnum_thres GLOBAL_VAL(2);  
00191 GLOBAL int gs_statenum GLOBAL_VAL(24);  
00192 
00193 /********************************************/
00194 /* Input rejection by GMM and input length */
00195 GLOBAL int gmm_gprune_num GLOBAL_VAL(10); 
00196 GLOBAL char *gmm_reject_cmn_string GLOBAL_VAL(NULL); 
00197 GLOBAL int rejectshortlen GLOBAL_VAL(0); 
00198 
00199 /********************************************/
00200 /* Search parameters for 1st pass */
00201 GLOBAL int specified_trellis_beam_width GLOBAL_VAL(-1); 
00202 GLOBAL int trellis_beam_width GLOBAL_VAL(-1); 
00203 #ifdef SEPARATE_BY_UNIGRAM
00204 GLOBAL int separate_wnum GLOBAL_VAL(DEFAULT_SEPARATE_WNUM); 
00205 #endif
00206 #ifdef WPAIR
00207 # ifdef WPAIR_KEEP_NLIMIT
00208 GLOBAL int wpair_keep_nlimit GLOBAL_VAL(3); 
00209 # endif
00210 #endif
00211 #ifdef HASH_CACHE_IW
00212 GLOBAL int iw_cache_rate GLOBAL_VAL(10); 
00213 #endif
00214 
00215 /********************************************/
00216 /* Search parameters for 2nd pass */
00217 GLOBAL int enveloped_bestfirst_width GLOBAL_VAL(30); 
00218 #ifdef SCAN_BEAM
00219 GLOBAL LOGPROB scan_beam_thres GLOBAL_VAL(80.0); 
00220 #endif
00221 GLOBAL int hypo_overflow GLOBAL_VAL(2000); 
00222 GLOBAL int stack_size GLOBAL_VAL(500); 
00223 GLOBAL int lookup_range GLOBAL_VAL(5); 
00224 GLOBAL int nbest;               
00225 GLOBAL int output_hypo_maxnum GLOBAL_VAL(1); 
00226 
00227 #ifdef GRAPHOUT
00228 /********************************************/
00229 /* Word graph output */
00230 GLOBAL int graph_merge_neighbor_range GLOBAL_VAL(0); 
00231 GLOBAL int graph_totalwordnum GLOBAL_VAL(0); 
00232 #endif
00233 
00234 #ifdef SP_BREAK_CURRENT_FRAME
00235 /********************************************/
00236 /* Short-pause segmentation */
00237 GLOBAL int sp_frame_duration GLOBAL_VAL(10);
00238 #endif /* SP_BREAK_CURRENT_FRAME */
00239 
00240 /********************************************/
00241 /* Output parameters */
00242 GLOBAL int result_output GLOBAL_VAL(SP_RESULT_TTY);
00243 GLOBAL int progout_interval GLOBAL_VAL(300); 
00244 #ifdef CHARACTER_CONVERSION
00245 GLOBAL char *from_code GLOBAL_VAL(NULL); 
00246 GLOBAL char *to_code GLOBAL_VAL(NULL); 
00247 #endif
00248 
00249 /********************************************/
00250 /* Module mode */
00251 GLOBAL int module_port GLOBAL_VAL(DEFAULT_MODULEPORT); 
00252 
00253 /********************************************/
00254 /* MAP-CMN */
00255 GLOBAL float cmn_map_weight GLOBAL_VAL(100.0); 
00256 
00257 /* -------------------------------------------------------------- */
00258 /* -------- User-specified switches ----------------------------- */
00259 /* -------------------------------------------------------------- */
00260 
00261 /********************************************/
00262 /* Enable/disable extra check mode for debugging */
00263 GLOBAL boolean wchmm_check_flag GLOBAL_VAL(FALSE); 
00264 GLOBAL boolean trellis_check_flag GLOBAL_VAL(FALSE); 
00265 GLOBAL boolean triphone_check_flag GLOBAL_VAL(FALSE); 
00266 
00267 /********************************************/
00268 /* Dictionary */
00269 GLOBAL boolean forcedict_flag GLOBAL_VAL(FALSE); 
00270 
00271 /********************************************/
00272 /* Search */
00273 GLOBAL boolean compute_only_1pass GLOBAL_VAL(FALSE); 
00274 GLOBAL boolean forced_realtime GLOBAL_VAL(FALSE); 
00275 GLOBAL boolean force_realtime_flag GLOBAL_VAL(FALSE); 
00276 #ifdef CATEGORY_TREE
00277 GLOBAL boolean old_tree_function_flag GLOBAL_VAL(FALSE); 
00278 #ifdef PASS1_IWCD
00279 GLOBAL boolean old_iwcd_flag GLOBAL_VAL(FALSE); 
00280 #endif
00281 #endif
00282 #ifdef USE_NGRAM
00283 GLOBAL short iwcdmethod GLOBAL_VAL(IWCD_NBEST); 
00284 #else
00285 GLOBAL short iwcdmethod GLOBAL_VAL(IWCD_AVG); 
00286 #endif
00287 GLOBAL short iwcdmaxn GLOBAL_VAL(3); 
00288 #ifdef USE_DFA
00289 GLOBAL boolean looktrellis_flag GLOBAL_VAL(FALSE); 
00290 GLOBAL boolean multigramout_flag GLOBAL_VAL(FALSE); 
00291 #endif
00292 GLOBAL boolean result_reorder_flag GLOBAL_VAL(TRUE); 
00293 
00294 /********************************************/
00295 /* Forced alignment */
00296 GLOBAL boolean align_result_word_flag GLOBAL_VAL(FALSE); 
00297 GLOBAL boolean align_result_phoneme_flag GLOBAL_VAL(FALSE); 
00298 GLOBAL boolean align_result_state_flag GLOBAL_VAL(FALSE); 
00299 
00300 /********************************************/
00301 /* Module mode */
00302 GLOBAL boolean module_mode GLOBAL_VAL(FALSE);
00303 
00304 /********************************************/
00305 /* Output switches */
00306 GLOBAL boolean verbose_flag GLOBAL_VAL(TRUE); 
00307 GLOBAL boolean debug2_flag GLOBAL_VAL(FALSE); 
00308 GLOBAL boolean paramtype_check_flag GLOBAL_VAL(TRUE); 
00309 GLOBAL boolean progout_flag GLOBAL_VAL(FALSE); 
00310 #ifdef USE_NGRAM
00311 GLOBAL boolean separate_score_flag GLOBAL_VAL(FALSE); 
00312 #endif
00313 
00314 /********************************************/
00315 /* CMN switches */
00316 GLOBAL boolean cmn_update GLOBAL_VAL(TRUE); 
00317 
00318 /* -------------------------------------------------------------- */
00319 /* -------- Inner status holders -------------------------------- */
00320 /* -------------------------------------------------------------- */
00321 /* for LM */
00322 GLOBAL boolean lmp_specified GLOBAL_VAL(FALSE); 
00323 GLOBAL boolean lmp2_specified GLOBAL_VAL(FALSE); 
00324 /* for search */
00325 GLOBAL boolean realtime_flag GLOBAL_VAL(FALSE); 
00326 GLOBAL boolean ccd_flag GLOBAL_VAL(TRUE); 
00327 GLOBAL boolean ccd_flag_force GLOBAL_VAL(FALSE); 
00328 
00329 /* Misc */
00330 GLOBAL boolean catch_intr_flag GLOBAL_VAL(FALSE); 
00331 
00332 /* -------------------------------------------------------------- */
00333 /* -------- Work area for search -------------------------------- */
00334 /* -------------------------------------------------------------- */
00335 
00336 /* Models */
00337 GLOBAL HTK_HMM_INFO *hmminfo GLOBAL_VAL(NULL);
00338 GLOBAL HTK_HMM_INFO *hmm_gs GLOBAL_VAL(NULL); 
00339 GLOBAL WORD_INFO *winfo GLOBAL_VAL(NULL); 
00340 #ifdef USE_NGRAM
00341 GLOBAL NGRAM_INFO *ngram GLOBAL_VAL(NULL); 
00342 #else
00343 GLOBAL MULTIGRAM *gramlist GLOBAL_VAL(NULL); 
00344 GLOBAL DFA_INFO *dfa GLOBAL_VAL(NULL); 
00345 #endif
00346 GLOBAL HTK_HMM_INFO *gmm GLOBAL_VAL(NULL); 
00347 
00348 /* Tree lexicon HMM */
00349 GLOBAL WCHMM_INFO *wchmm GLOBAL_VAL(NULL);
00350 GLOBAL BACKTRELLIS backtrellis; 
00351 GLOBAL LOGPROB backmax;         
00352 
00353 /* Misc. for search */
00354 GLOBAL int peseqlen;            
00355 
00356 /* Score envelope beaming */
00357 #ifdef SCAN_BEAM
00358 GLOBAL LOGPROB *framemaxscore;
00359 #endif
00360 
00361 /* Input speech related */
00362 GLOBAL SP16 speech[MAXSPEECHLEN]; 
00363 GLOBAL int speechlen GLOBAL_VAL(0);             
00364 GLOBAL boolean cmn_loaded GLOBAL_VAL(FALSE); 
00365 GLOBAL boolean delta_required GLOBAL_VAL(TRUE); 
00366 GLOBAL boolean acc_required GLOBAL_VAL(FALSE); 
00367 GLOBAL boolean energy_required GLOBAL_VAL(TRUE); 
00368 GLOBAL boolean abs_energy_suppress GLOBAL_VAL(TRUE); 
00369 GLOBAL boolean c0_required GLOBAL_VAL(FALSE); 
00370 GLOBAL boolean cmn_required GLOBAL_VAL(TRUE); 
00371 GLOBAL int model_mfcc_dim GLOBAL_VAL(12); 
00372 GLOBAL boolean zmean_frame GLOBAL_VAL(FALSE); 
00373 
00374 /* Spectral subtraction */
00375 GLOBAL float *ssbuf GLOBAL_VAL(NULL);
00376 GLOBAL int sslen;               
00377 
00378 #ifdef SP_BREAK_CURRENT_FRAME
00379 /* Short-pause segmentation */
00380 GLOBAL HTK_Param *rest_param GLOBAL_VAL(NULL);
00381 GLOBAL WORD_ID sp_break_last_word GLOBAL_VAL(WORD_INVALID); 
00382 GLOBAL WORD_ID sp_break_last_nword GLOBAL_VAL(WORD_INVALID); 
00383 GLOBAL boolean sp_break_last_nword_allow_override GLOBAL_VAL(TRUE); 
00384 GLOBAL WORD_ID sp_break_2_begin_word GLOBAL_VAL(WORD_INVALID); 
00385 GLOBAL WORD_ID sp_break_2_end_word GLOBAL_VAL(WORD_INVALID); 
00386 #endif
00387 
00388 /* Module */
00389 GLOBAL int module_sd GLOBAL_VAL(-1);    
00390 
00391 /* Output */
00392 GLOBAL WORD_ID pass1_wseq[MAXSEQNUM]; 
00393 GLOBAL int pass1_wnum; 
00394 GLOBAL LOGPROB pass1_score;     
00395 
00396 /* Pointer to text output functions (will be set by 'result_output') */
00397 GLOBAL void (*status_process_online)();
00398 GLOBAL void (*status_process_offline)();
00399 GLOBAL void (*status_recready)();
00400 GLOBAL void (*status_recstart)();
00401 GLOBAL void (*status_recend)();
00402 GLOBAL void (*status_param)(HTK_Param *p);
00403 GLOBAL void (*result_pass1_begin)();
00404 GLOBAL void (*result_pass1_current)(int t, WORD_ID *seq, int num, LOGPROB score, LOGPROB LMscore, WORD_INFO *winfo);
00405 GLOBAL void (*result_pass1_final)(WORD_ID *seq, int num, LOGPROB score, LOGPROB LMscore, WORD_INFO *winfo);
00406 GLOBAL void (*result_pass1_end)();
00407 GLOBAL void (*result_pass2_begin)();
00408 GLOBAL void (*result_pass2)(NODE *hypo, int rank, WORD_INFO *winfo);
00409 GLOBAL void (*result_pass2_end)();
00410 GLOBAL void (*result_pass2_failed)(WORD_INFO *winfo);
00411 GLOBAL void (*result_rejected)(const char *);
00412 GLOBAL void (*result_gmm)();
00413 #ifdef GRAPHOUT
00414 GLOBAL void (*result_graph)(WordGraph *root, WORD_INFO *winfo);
00415 #endif
00416 
00417 #endif /* __SENT_EXTERNAL_DEFINITION__ */