1 /*---------------------------------------------------------------------------* 2 * simapi.h * 3 * * 4 * Copyright 2007, 2008 Nuance Communciations, Inc. * 5 * * 6 * Licensed under the Apache License, Version 2.0 (the 'License'); * 7 * you may not use this file except in compliance with the License. * 8 * * 9 * You may obtain a copy of the License at * 10 * http://www.apache.org/licenses/LICENSE-2.0 * 11 * * 12 * Unless required by applicable law or agreed to in writing, software * 13 * distributed under the License is distributed on an 'AS IS' BASIS, * 14 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * 15 * See the License for the specific language governing permissions and * 16 * limitations under the License. * 17 * * 18 *---------------------------------------------------------------------------*/ 19 20 21 22 #ifndef _h_simapi_ 23 #define _h_simapi_ 24 25 #ifdef SET_RCSID 26 static const char simapi_h[] = "$Id: simapi.h,v 1.22.4.25 2008/03/07 17:37:14 dahan Exp $"; 27 #endif 28 29 #include "creccons.h" /* CREC Public Constants */ 30 31 #include "c42mul.h" 32 #include "caexcept.h" 33 #include "setting.h" 34 #include "mutualob.h" 35 #include "srec_sizes.h" 36 #include "search_network.h" 37 #include "srec.h" 38 #include "srec_eosd.h" 39 #include "srec_results.h" 40 #include "swimodel.h" 41 #include "ESR_Locale.h" 42 43 /** 44 ************************************************************************ 45 * SwigImageBegin <- DO NOT MOVE THIS LINE ! 46 * 47 * This is a CADOC Keyword section. 48 * 49 * If CADOC is instructed to create a SWIG I-File and this is one of the 50 * files in the input list, everything between the 'SwigImage Begin' and 51 * 'SwigImage End' keywords comment blocks will be copied 'as-is' to the 52 * SWIG I-File specified on the CADOC command line. 53 * 54 * The purpose of the following marked section is to identify CA 55 * structures that SWIG may need to complete the CREC API interface. 56 * 57 * Refer to the document "DOC\CADOC.TXT" for details for the CADOC Tool. 58 * 59 ************************************************************************ 60 */ 61 62 63 #ifdef __cplusplus 64 extern "C" 65 { 66 #endif 67 68 69 #ifndef SWIGBUILD 70 71 typedef featdata CA_Feature; 72 73 /** 74 * @todo document 75 */ 76 #define NCONFPARS 6 77 78 typedef struct 79 { 80 double scale[NCONFPARS]; 81 double offset[NCONFPARS]; 82 double weight[NCONFPARS]; 83 } Confidence_model_parameters; 84 85 typedef struct 86 { 87 Confidence_model_parameters one_nbest; 88 Confidence_model_parameters many_nbest; 89 } 90 CA_ConfidenceScorer; 91 92 /** 93 * @todo document 94 */ 95 typedef struct 96 { 97 int ca_rtti; 98 booldata is_resultBlocked; 99 booldata is_configured; 100 int setup_count; 101 booldata is_running; 102 multi_srec *recm; /* contains recognition structure */ 103 srec_eos_detector_parms* eosd_parms;/* contains ep parameters */ 104 } 105 CA_Recog; 106 107 /** 108 * @todo document 109 */ 110 typedef struct 111 { 112 int ca_rtti; 113 booldata is_loaded; 114 int pattern_setup_count; 115 int use_dim; 116 int partial_distance_calc_dim; 117 prdata imelda_scale; 118 const SWIModel *swimodel; /* owning pointer to compact acoustic models */ 119 } 120 CA_Acoustic; 121 122 /** 123 * @todo document 124 */ 125 typedef struct 126 { 127 int ca_rtti; 128 /* IMPORTANT NOTE 129 * 130 * This structure is used by CA_WriteSyntaxImage 131 * in the file "ca/cod_imag.c". Certain assumptions 132 * about the member types and order are made. 133 * DON'T change this structure without making 134 * appropriate changes to the above function. 135 */ 136 137 booldata has_groups; 138 booldata has_rules; 139 int setup_count; 140 /* syntax_info* synx; */ 141 srec_context* synx; 142 } 143 CA_Syntax; 144 145 /** 146 * @todo document 147 */ 148 typedef struct 149 { 150 int dummy; 151 } 152 CA_Arbdata; 153 154 /** 155 * @todo document 156 */ 157 typedef struct 158 { 159 int ca_rtti; 160 /* IMPORTANT NOTE 161 * 162 * This structure is used by CA_WritePatternImage 163 * in the file "ca/cod_imag.c". Certain assumptions 164 * about the member types and order are made. 165 * DON'T change this structure without making 166 * appropriate changes to the above function. 167 */ 168 169 booldata is_loaded; 170 //model_info *setup_whole; 171 //model_info *setup_sub; 172 booldata has_cache; 173 booldata true_accumulates; 174 booldata recog_terminated; 175 pattern_info data; 176 } 177 CA_Pattern; 178 179 180 /** 181 * @todo document 182 */ 183 typedef struct 184 { 185 int dim; 186 transform_info imelda_acc; 187 transform_info mllr_acc; 188 booldata do_mllr; 189 booldata do_imelda; 190 booldata is_loaded; 191 booldata is_setup; 192 } 193 CA_Transform; 194 195 196 /** 197 * @todo document 198 */ 199 typedef struct 200 { 201 int ca_rtti; 202 booldata is_loaded; /* generates VOCAB_NOT_CONFIGURED */ 203 vocab_info voc; 204 } 205 CA_Vocab; 206 207 /** 208 * @todo document 209 */ 210 typedef struct 211 { 212 int ca_rtti; 213 short type; 214 unsigned long startTime; 215 unsigned long endTime; 216 short wordID; 217 long score; 218 char *label; 219 } 220 CA_Result; 221 222 /* typedef nbest_list CA_NBestList; */ 223 224 /** 225 * @todo document 226 */ 227 typedef struct 228 { 229 asr_int32_t dummy[2]; 230 } 231 CA_NBestList; 232 233 /** 234 * @todo document 235 */ 236 typedef struct 237 { 238 int ca_rtti; 239 booldata is_loaded; 240 int dimen; 241 int stay_penalty; 242 int skip_penalty; 243 int whole_stay_penalty; 244 int whole_skip_penalty; 245 int dur_scale; 246 int frame_period; 247 int min_var; 248 int max_var; 249 booldata load_all; 250 char load_models[MAX_LABEL]; 251 } 252 CA_AcoustInputParams; 253 254 255 /** 256 * @todo document 257 */ 258 typedef struct 259 { 260 int ca_rtti; 261 booldata is_loaded; 262 ESR_BOOL do_early; 263 ESR_BOOL do_partial; 264 int word_penalty; 265 int top_choices; 266 int reject_score; 267 int traceback_freq; 268 int viterbi_prune_thresh; /*score-based pruning threshold - only keep paths within this delta of best cost*/ 269 int max_hmm_tokens; /*controls the maximum number of HMM's alive in any frame. If number 270 exceeded, pruning gets tightened. So, this threshold can be used 271 to tradeoff accuracy for computation an memory*/ 272 int max_fsmnode_tokens; /*controls the maximum number of FSMs alive in any frame. If number, 273 exceeded, pruning gets tightened. So, this threshold can be used 274 to tradeoff accuracy for computation an memory*/ 275 int max_word_tokens; /*controls the maximum number of word tokens kept in the word lattice. 276 if number exceeded, the word lattice is pruned more tightly (less word 277 ends per frame*/ 278 int max_altword_tokens; /*controls the maximum number of alternative paths to propagate, for nbest */ 279 int num_wordends_per_frame; /*controls the size of the word lattice - the number of word ends to 280 keep at each time frame*/ 281 int max_fsm_nodes; /*allocation size of a few arrays in the search - needs to be big enough 282 to handle any grammar that the search needs to run. Initialization fails 283 if num exceeded*/ 284 int max_fsm_arcs; /*allocation size of a few arrays in the search - needs to be big enough 285 to handle any grammar that the search needs to run. Initialization fails 286 if num exceeded*/ 287 int max_searches; /* now: 2 searches for 2 genders */ 288 int terminal_timeout; /* end of utterance timeout at terminal node */ 289 int optional_terminal_timeout; /* end of utterance timeout at optional terminal node */ 290 int non_terminal_timeout; /* end of utterance timeout at non terminal node */ 291 int eou_threshold; /* select hypotheses for end-of-utterance tests */ 292 int stats_enabled; /* enable frame-by-frame recognizer stats */ 293 int max_frames; /* max number of frames in for searching */ 294 int max_model_states; /* indicates largest acoustic model this search can use */ 295 } 296 CA_RecInputParams; 297 298 /** 299 * @todo document 300 */ 301 typedef struct 302 { 303 int ca_rtti; 304 booldata is_loaded; 305 int dimen; 306 int whole_dimen; 307 int feat_start; 308 float mix_score_scale; 309 float imelda_scale; 310 float uni_score_scale; 311 float uni_score_offset; 312 int forget_speech; 313 int forget_background; 314 int rel_low; 315 int rel_high; 316 int gap_period; 317 int click_period; 318 int breath_period; 319 int extend_annotation; 320 int param; 321 #if PARTIAL_DISTANCE_APPROX 322 int partial_distance_dim; 323 int partial_distance_threshold; 324 int partial_distance_offset; 325 int global_model_means[MAX_DIMEN]; 326 #endif 327 int min_initial_quiet_frames; /* num silence frames needed before input */ 328 int min_annotation_frames; /* minimum overall length */ 329 int max_annotation_frames; /* maximum overall length */ 330 int delete_leading_segments; /* num segments to delete. 0=no action */ 331 int leading_segment_accept_if_not_found; /* Do not reject segmentation if not found */ 332 int leading_segment_min_frames; /* remove unless shorter */ 333 int leading_segment_max_frames; /* remove unless exceeded */ 334 int leading_segment_min_silence_gap_frames;/* remove if good silence gap to next segment */ 335 int beep_size; /*X201 beep filter */ 336 int beep_threshold; /*X201 beep filter */ 337 int min_segment_rel_c0; /* Any segment gets deleted whose peak c0 is < max - min_segment_rel_c0 */ 338 339 #if DO_SUBTRACTED_SEGMENTATION 340 int snr_holdoff; /* Ignore first n frames when estimating speech level for SNR measure */ 341 int min_acceptable_snr; /* for successful segmentation */ 342 #endif 343 int stats_enabled; /* enable frame-by-frame pattern stats */ 344 } 345 CA_PatInputParams; 346 347 348 349 #endif 350 351 #define IMELDA_ACC_BETWEEN 1 352 #define IMELDA_ACC_WITHIN 2 353 #define IMELDA_ACC_BMEAN 3 354 #define IMELDA_ACC_WMEAN 4 355 #define MLLR_ACC_BETWEEN 5 356 #define MLLR_ACC_WITHIN 6 357 #define MLLR_ACC_BMEAN 7 358 #define MLLR_ACC_WMEAN 8 359 #define IMELDA_MATRIX 9 360 #define INVERSE_IMELDA_MATRIX 10 361 #define MLLR_MATRIX 11 362 #define TRANSFORM_MATRIX 12 363 #define TRANSFORM_OFFSET 13 364 #define CHAN_OFFSET 14 365 366 /* for fast match debugging 367 ** 5/1/99 SL 368 */ 369 #define FRAME_VOICED 1 370 #define FRAME_QUIET 2 371 #define FRAME_UNCERTAIN 3 372 373 int CA_GetVoicingStatus(CA_Pattern *hPattern); 374 /** 375 * 376 * Params hPattern Handle to valid object 377 * Returns int See below 378 * 379 ************************************************************************ 380 * Returns the voicing status of the current frame contained in the 381 * Pattern object. 382 * The function will return one of the following values :- 383 * 384 * FRAME_VOICED 1 385 * FRAME_QUIET 2 386 * FRAME_UNCERTAIN 3 387 * 388 * TODO: Integrate these defines into CR_type.h 389 ************************************************************************ 390 */ 391 392 393 394 /* 395 ** File: simapi.c 396 */ 397 void CA_GetVersion(char *vertab, 398 int len); 399 /** 400 * Params *vertab Version info 401 * len length of vertab 402 * 403 * Returns nothing 404 * 405 ************************************************************************ 406 * There doesn't seem to be anything using this function. Has it been 407 * replaced by CrecGetVersion?? If so, there should be a CA interface 408 * to it. 409 ************************************************************************ 410 */ 411 412 413 void CA_Log(const char *string); 414 /** 415 * 416 * Params string Single string for output log 417 * 418 * Returns nothing 419 * 420 * 421 ************************************************************************ 422 * NOTE swig dosen't allow <varargs> for function arguments hence, you 423 * will have to format the string yourself prior to calling this function 424 ************************************************************************ 425 */ 426 427 428 void CA_Init(void); 429 /** 430 * 431 * Params void 432 * 433 * Returns Nothing, the error callback routine is registered 434 * 435 * See CA_Term 436 * 437 ************************************************************************ 438 * Initializes the CREC (CR) interface. A callback routine is supplied 439 * by the user. This is used as the registered callback function. 440 * Any errors will vector through this method. The prototype for the 441 * callback funtion is defined in CR_TYPE.H 442 * 443 * This method must be the first CR call. It can only be called again 444 * during the lifetime of a process after a call to CA_Term(). 445 * 446 * The user can specify NULL as the handler funtion, this will cause 447 * the default error handler, crDefaultHandler(), to run when errors 448 * are encountered. This allows the easy use of PERL scripts and the CR 449 * interface. Generally a handler should be supplied. 450 ************************************************************************ 451 */ 452 453 void CA_Term(void); 454 /** 455 * 456 * Params void 457 * 458 * Returns Nothing, CR is terminated. No CR method can be called 459 * until another call to CA_Init() 460 * 461 * See CA_Init 462 * 463 ************************************************************************ 464 * Signals that the CREC (CR) interface is no longer needed. This should 465 * be the last method called in a CR application. 466 * 467 * The callback routine registered with CA_Init() is released. 468 * 469 * A test is made to ensure that all the handles associated with 470 * CR objects have been de-allocated. If an object handle still 471 * exists when CA_Term() is called this will generate an error. 472 ************************************************************************ 473 */ 474 475 476 477 478 #ifdef USE_EXCEPTION_HANDLING 479 /* 480 ** CA_Exception methods 481 */ 482 int CA_ExceptionGetCode(CA_Exception* e); 483 int CA_ExceptionGetMessage(CA_Exception* e , char* buff, int len, 484 char *filebuff, int filebufflen, int* line); 485 486 int CA_ExceptionGetHandleType(CA_Exception* e); 487 int CA_ExceptionGetHandle(CA_Exception* e, void **handle); 488 int CA_ExceptionDelete(CA_Exception* e); 489 #endif 490 491 492 /* 493 ** CA_Acoustic methods 494 */ 495 /* 496 ** File: acc_basi.c 497 */ 498 CA_Acoustic *CA_AllocateAcoustic(void); 499 /** 500 * 501 * Params void 502 * 503 * Returns Handle to new acoustic structure, or NULL on error. 504 * 505 * See CA_FreeAcoustic 506 * 507 ************************************************************************ 508 * Allocates a new Acoustic structure (object) 509 ************************************************************************ 510 */ 511 512 513 void CA_FreeAcoustic(CA_Acoustic *hAcoust); 514 /** 515 * 516 * Params hAcoust Handle to previously allocated acoustic structure 517 * 518 * Returns void 519 * 520 * See CA_AllocateAcoustic 521 * 522 ************************************************************************ 523 * Frees a previously allocated acoustic object. 524 ************************************************************************ 525 */ 526 527 528 void CA_UnloadAcoustic(CA_Acoustic *hAcoust); 529 530 531 /* 532 ** File: acc_sub.c 533 */ 534 int CA_LoadAcousticSub(CA_Acoustic *hAcoust, 535 char *subname, 536 CA_AcoustInputParams *hAcoustInp); 537 /** 538 * 539 * Params hAcoust point to a valid acoustic object 540 * subname subword acoustic model file 541 * do_arb 0 for pcx, 1 for arb 542 * do_uni 1 for unimodal models 543 * hAcoustInp Handle to valid acoustic input object 544 * 545 * Returns True/False 546 * 547 * See CA_UnloadAcoustic 548 * CA_LoadAcousticWhole 549 * CA_LoadDictionary 550 * 551 ************************************************************************ 552 * The acoustic sub-word models, whether speaker dependent or independent 553 * must be loaded with this method. 554 * 555 * The model files must be in the 'Tracy' format (.PDF, .MIX etc.). 556 * It is possible to load the ARB files, PCX files and CLV files. 557 * 558 * Note: The phoneme table located with the ARB file is checked with 559 * other phoneme tables in the system (such as the one specified 560 * in the OK file). This is why the Vocabulary class is also 561 * required. 562 ************************************************************************ 563 */ 564 565 566 567 568 /* 569 ** CA_Vocab methods 570 */ 571 /* 572 ** File: voc_basi.c 573 */ 574 CA_Vocab *CA_AllocateVocabulary(void); 575 /** 576 * 577 * Params void 578 * 579 * Returns Handle to allocated vocabulary structure, or NULL on error. 580 * 581 * See CA_FreeVocabulary 582 * 583 ************************************************************************ 584 * This method creates a new vocabulary structure (object). The new 585 * structure is supplied to the caller by the method's return value. 586 ************************************************************************ 587 */ 588 589 590 void CA_FreeVocabulary(CA_Vocab *hVocab); 591 /** 592 * 593 * Params hVocab Handle to a previously allocated vocabulary 594 * 595 * Returns void 596 * 597 * See CA_AllocateVocabulary 598 * 599 ************************************************************************ 600 * Frees a previously allocated vocabular structure (object). 601 ************************************************************************ 602 */ 603 void CA_LoadDictionary(CA_Vocab *hVocab, 604 const LCHAR* vocname, 605 char *phtname, 606 ESR_Locale* locale); 607 /** 608 * 609 * Params hVocab Handle to vocabulary object 610 * vocFile Source vocabulary file 611 * phtname Source of phoneme table file 612 * locale Locale associated with dictionary 613 * 614 * Returns void 615 * 616 * See CA_UnloadDictionary 617 * 618 ************************************************************************ 619 * Loads vocabulary 'words' from an OK file. 620 ************************************************************************ 621 */ 622 623 624 void CA_UnloadDictionary(CA_Vocab *hVocab); 625 /** 626 * 627 * Params hVocab Handle to vocabulary object 628 * 629 * Returns void 630 * 631 * See CA_LoadDictionary 632 * 633 ************************************************************************ 634 * Unloads a previously loaded dictionary 635 ************************************************************************ 636 */ 637 638 639 int CA_CheckEntryInDictionary(CA_Vocab *hVocab, 640 const char *label); 641 /** 642 * 643 * Params hVocab Handle to vocabulary object 644 * label textual trasncription of requested word 645 * 646 * Returns ? 647 * 648 ************************************************************************ 649 * Checks the existance of a pronunciation 650 ************************************************************************ 651 */ 652 653 654 int CA_GetEntryInDictionary(CA_Vocab *hVocab, 655 const char *label, 656 char *pron, 657 int *pronSize, 658 int pronMaxSize); 659 /** 660 * 661 * Params hVocab Handle to vocabulary object 662 * label textual trasncription of requested word 663 * pron buffer to put pronunciation in. The format 664 * of this buffer is a series of prons terminated 665 * by a null all terminated by a double null 666 * pronSize the number of bytes returned in the pron buffer 667 * pronMaxSiz this size of "pron" 668 * 669 * Returns true if the word exists in the dictionary 670 * 671 ************************************************************************ 672 * Get the pronunciation of a word 673 ************************************************************************ 674 */ 675 676 677 /* 678 ** CA_NGram methods 679 */ 680 /* 681 ** File: lmf_basi.c 682 */ 683 /* 684 ** Gethin 25Aug00 685 ** Functions excluded from compilation in ca.mak, 686 ** therefore commented here 687 ** 688 ** CA_NGram *CA_AllocateNGram (void); 689 ** void CA_FreeNGram (CA_NGram *hNGram); 690 ** void CA_LoadNGram (CA_NGram *hNGram, char *lmfname, char *trnname); 691 ** void CA_SaveNGram (CA_NGram *hNGram, char *lmfname); 692 ** void CA_UnloadNGram (CA_NGram *hNGram); 693 ** int CA_ScoreSentence (CA_NGram *hNGram, char *sentence); 694 ** int CA_SetupNGram (CA_NGram *hNGram, CA_Vocab *hVocab, int type); 695 */ 696 697 /* 698 ** CA_Pattern methods 699 */ 700 /* 701 ** File: pat_basi.c 702 */ 703 704 CA_Pattern *CA_AllocatePattern(void); 705 /** 706 * 707 * Params void 708 * 709 * Returns Handle to new pattern object 710 * 711 * See CA_FreePattern 712 * 713 ************************************************************************ 714 * Creates a new pattern structure (object) 715 ************************************************************************ 716 */ 717 718 719 void CA_FreePattern(CA_Pattern *hPattern); 720 /** 721 * 722 * Params hPattern valid pattern handle 723 * 724 * Returns Nothing, the pattern handle is not longer valid. 725 * 726 * See CA_AllocatePattern 727 * 728 ************************************************************************ 729 * Deletes a previously allocated pattern structure 730 ************************************************************************ 731 */ 732 733 734 int CA_LoadPattern(CA_Pattern *hPattern, 735 CA_PatInputParams *hPatInput, 736 int dimen, 737 char *multable, 738 char *imelda); 739 /** 740 * 741 * Params hPattern valid pattern handle 742 * hPatInput valid pattern input paramater handle 743 * dimen 744 * multable 745 * imelda 746 * 747 * Returns true/false 748 * 749 * See CA_UnloadPattern 750 * CA_SetupPatternForAcoustic 751 * CA_LoadPatternParameters 752 * 753 ************************************************************************ 754 * Loads a pattern object. Once loaded it need to be setup 755 * with an acoustic model before any further setups involving 756 * the recognizer. The imelda parameters and optionally 757 * the multables are loaded into the object. Setup operations 758 * based on parameters are also done. 759 * 760 * It is an error to call this function without first loading 761 * pattern input parameters. 762 ************************************************************************ 763 */ 764 765 766 void CA_UnloadPattern(CA_Pattern *hPattern); 767 /** 768 * 769 * Params hPattern valid pattern handle 770 * 771 * Returns Nothing, loaded object is unloaded 772 * 773 * See CA_LoadPattern 774 * 775 ************************************************************************ 776 * Unloads a previously loaded pattern object 777 ************************************************************************ 778 */ 779 780 781 void CA_SetupPatternForAcoustic(CA_Pattern *hPattern, 782 CA_Acoustic *hAcoustic); 783 /** 784 * 785 * Params hPattern valid pattern handle 786 * hAcoustic valid acoustic handle 787 * 788 * Returns Nothing, the pattern object is setup and ready for use 789 * 790 * See CA_ClearPatternForAcoustic 791 * CA_LoadPattern 792 * 793 ************************************************************************ 794 * This sets up a previously allocated pattern object with acoustic 795 * model information from the appropriate acoustic object. 796 * It is necessary to call this function before the acoustic 797 * object can be used any recognition or adaptation. 798 * 799 * If you have not called CA_LoadPattern() an error will be generated. 800 ************************************************************************ 801 */ 802 803 804 void CA_ClearPatternForAcoustic(CA_Pattern *hPattern, 805 CA_Acoustic *hAcoustic); 806 /** 807 * 808 * Params hPattern valid pattern handle 809 * hAcoustic valid acoustic handle 810 * 811 * Returns Nothing 812 * 813 * See CA_SetupPatternForAcoustic 814 * 815 ************************************************************************ 816 * un-sets a previously prepared pattern object 817 ************************************************************************ 818 */ 819 820 821 int CA_NextPatternFrame(CA_Pattern *hPattern, 822 CA_Utterance *hUtt); 823 /** 824 * 825 * Params hPattern valid pattern handle 826 * hUtt valid utterance handle 827 * 828 * Returns status code 829 * 830 * See CA_NextUtteranceFrame 831 * CA_MakePatternFrame 832 * CA_AdvanceRecognitionByFrame 833 * 834 ************************************************************************ 835 * Frames can be accessed, on-by-one, from an utterance object into 836 * a pattern object with this method. If this routine returns a non-zero 837 * value then an utterance frame is available for processing and a call 838 * to the recognition method CA_AdvanceRecognitionByFrame as shown below: 839 * 840 * while ( CA_NextPatternFrame( hPattern, hUtt ) ) 841 * { 842 * CA_AdvanceRecognitionByFrame( hRecog, hPattern, hUtt ); 843 * } 844 * 845 * Frames copied in with this method are transformed (post-IMELDA) 846 * 847 * It is now advised that instead of this function, 848 * one uses CA_GetUtteranceFrame and CA_AdvanceUtteranceFrame 849 * to access frames in sequence. 850 ************************************************************************ 851 */ 852 853 int CA_MakePatternFrame(CA_Pattern *hPattern, 854 CA_Utterance *hUtt); 855 /** 856 * 857 * Params hPattern valid pattern handle 858 * hUtt valid utterance object handle 859 * 860 * Returns ? 861 * 862 * See CA_NextUtteranceFrame 863 * CA_NextPatternFrame 864 * 865 ************************************************************************ 866 * The current frames is pulled in from an utterance object into 867 * a pattern object with this method. The frame position is not 868 * moved on. 869 * 870 * Frames copied in with this method are IMELDA transformed 871 ************************************************************************ 872 */ 873 874 875 /* 876 ** File: pat_tran.c 877 */ 878 void CA_SetCepstrumOffsetInPattern(CA_Pattern *hPattern, 879 int index, 880 int value); 881 /** 882 * 883 * Params hPattern valid pattern handle 884 * hAcoust valid acoustic handle 885 * relevance relevance coefficient 886 * 887 * Returns Nothing 888 * 889 ************************************************************************ 890 * Accumulates the relevance of hAcoustic into hPattern 891 ************************************************************************ 892 */ 893 894 895 void CA_SetupNoiseConditioning(CA_Pattern *hPattern, 896 CA_FrontendInputParams *hFrontpar, 897 int do_wholeword, 898 int do_subword); 899 /** 900 * 901 * Params hPattern valid pattern handle 902 * hFrontpar valid frontend parameter object handle 903 * do_wholeword true or false 904 * do_subword true or false 905 * 906 * Returns ? 907 * 908 * See CA_ClearNoiseConditioning 909 * CA_ConditionAcousticToNoise 910 * 911 ************************************************************************ 912 * The initial setup for Extended Noise Conditioning (ENC). 913 * 914 ************************************************************************ 915 */ 916 917 918 void CA_ClearNoiseConditioning(CA_Pattern *hPattern); 919 /** 920 * 921 * Params hPattern valid pattern handle 922 * 923 * Returns ? 924 * 925 * See CA_SetupNoiseConditioning 926 * CA_ConditinAcousticToNoise 927 * 928 ************************************************************************ 929 * To free data structures constructed for Extended Noise Conditioning. 930 * 931 ************************************************************************ 932 */ 933 934 935 void CA_WriteMLLRCovariates(CA_Pattern *hPattern, 936 CA_Acoustic *hAcoust, 937 char *basename); 938 /** 939 No CR equivalent function 940 */ 941 942 943 void CA_ReadMLLRTransforms(CA_Pattern *hPattern, 944 CA_Acoustic *hAcoust, 945 char *basename, 946 int ntrans); 947 /** 948 No CR equivalent function 949 */ 950 951 952 /* 953 ** File: pat_dump.c 954 */ 955 void CA_SetupTransformDump(CA_Pattern *hPattern, 956 char* basename); 957 /** 958 * 959 * Params hPattern valid pattern handle 960 * basename file base name 961 * 962 * Returns Nothing 963 * 964 * See CA_ClearTransformDump 965 * CA_CalculateTransforms 966 * CA_ClearTransformAccumulates 967 * 968 ************************************************************************ 969 * Sets up the facility to save the transform accumulates to a file. 970 ************************************************************************ 971 */ 972 973 974 void CA_ClearTransformDump(CA_Pattern *hPattern); 975 /** 976 * 977 * Params hPattern valid pattern handle 978 * 979 * Returns Nothing 980 * 981 * See CA_SetupTransformDump 982 * CA_CalculateTransforms 983 * CA_ClearTransformAccumulates 984 * 985 ************************************************************************ 986 * Clears the facility to save the transform accumulates to a file. 987 ************************************************************************ 988 */ 989 990 991 /* 992 * Pattern Array access methods 993 */ 994 /* 995 ** File: pat_swig.c 996 */ 997 998 /* 999 * Transform methods 1000 */ 1001 /* 1002 ** File: catrans.c 1003 */ 1004 /* 1005 * dave 08-Oct-99: These functions are now out of sync with CA. 1006 * 1007 * CA_Transform *CA_AllocateTransform (void); 1008 1009 * void CA_FreeTransform (CA_Transform *hTransform); 1010 * int CA_LoadTransform (CA_Transform *hTransform, int dimen); 1011 * void CA_UnloadTransform (CA_Transform *hTransform); 1012 * void CA_ConfigureTransform (CA_Transform *hTransform,int do_mllr,int do_imelda); 1013 * void CA_ClearTransform (CA_Transform *hTransform); 1014 * void CA_InheritAccumulates (CA_Transform *hTransform, CA_Pattern *hPattern); 1015 * void CA_LoadTransformAccumulates( CA_Pattern *hPattern, 1016 * CA_Transform *hTransform); 1017 */ 1018 1019 /* 1020 ** CA_Utterance methods 1021 */ 1022 /* 1023 ** File: utt_basi.c 1024 */ 1025 CA_Utterance *CA_AllocateUtterance(void); 1026 /** 1027 * 1028 * Params void 1029 * 1030 * Returns Handle to a new utterance object 1031 * 1032 * See CA_FreeUtterance 1033 * 1034 ************************************************************************ 1035 * Creates a new utterance object 1036 ************************************************************************ 1037 */ 1038 1039 1040 void CA_FreeUtterance(CA_Utterance *hUtt); 1041 /** 1042 * 1043 * Params hUtt valid utterance handle 1044 * 1045 * Returns void the object is no longer valid 1046 * 1047 * See CA_AllocateUtterance 1048 * 1049 ************************************************************************ 1050 * Destroys a previously allocated utterance object 1051 ************************************************************************ 1052 */ 1053 1054 1055 int CA_InitUtteranceForFrontend(CA_Utterance *hUtt, 1056 CA_FrontendInputParams *hFrontInp); 1057 /** 1058 * 1059 * Params hUtt valid utterance handle 1060 * hFrontPar valid frontend input parameter handle 1061 * 1062 * Returns non-zero on error 1063 * 1064 * See CA_ClearUtterance 1065 * CA_GetFrontendUtteranceDimension 1066 * CA_LoadUtteranceFrame 1067 * 1068 ************************************************************************ 1069 * Creates an utterance object ready for the frontend to insert frames. 1070 * This method should also be used to create utterances which will be 1071 * used for raw utterance (frame-by-frame) recognition - See 1072 * CA_LoadUtteranceFrame(). 1073 * 1074 * Use the result from CA_GetFrontendUtteranceDimension to specify 1075 * the dimension. 1076 ************************************************************************ 1077 */ 1078 1079 1080 void CA_ClearUtterance(CA_Utterance *hUtt); 1081 /** 1082 * 1083 * Params hUtt valid utterance handle 1084 * 1085 * Returns void 1086 * 1087 * See CA_InitUtteranceForFrontend 1088 * 1089 ************************************************************************ 1090 * Closes a previously opened utterance and frame buffer. 1091 ************************************************************************ 1092 */ 1093 1094 1095 int CA_NextUtteranceFrame(CA_Pattern *hPattern, 1096 CA_Utterance *hUtt); 1097 /** 1098 * 1099 * Params hPattern valid pattern handle 1100 * hUtt valid utterance handle 1101 * 1102 * Returns status code 1103 * 1104 * See CA_NextPatternFrame 1105 * CA_MakePatternFrame 1106 * CA_AdvanceUtteranceFrame 1107 * 1108 ************************************************************************ 1109 * Frames can be accessed, on-by-one, from an utterance object into 1110 * a pattern object with this method. It is advised that instead of 1111 * this function, one use CA_MakePatternFrame and CA_AdvanceUtteranceFrame 1112 * to access frames in sequence. 1113 * 1114 * Frames copied in with this method are un-transformed (pre-IMELDA) 1115 ************************************************************************ 1116 */ 1117 1118 1119 int CA_AdvanceUtteranceFrame(CA_Utterance *hUtt); 1120 /** 1121 * 1122 * Params hUtt valid utterance handle 1123 * 1124 * Returns True if it successfully advances the frame positon by one. 1125 * 1126 * See CA_AdvanceRecognitionByFrame 1127 * 1128 ************************************************************************ 1129 ************************************************************************ 1130 */ 1131 1132 int CA_UtteranceHasVoicing(CA_Utterance *hUtt); 1133 /** 1134 * 1135 * Params hUtt valid utterance handle 1136 * 1137 * Returns non-zero if utterance has voicing data 1138 * 1139 ************************************************************************ 1140 * This method can be polled to interrogate the state of the utterance. 1141 ************************************************************************ 1142 */ 1143 1144 ESR_BOOL CA_IsUtteranceLockedForInput(CA_Utterance *hUtt); 1145 /** 1146 * 1147 * Params hUtt valid utterance handle 1148 * 1149 * Returns TRUE if utterance is locked 1150 * 1151 * See CA_LockUtteranceFromInput 1152 * 1153 ************************************************************************ 1154 * Indicates if utterance is locked. 1155 ************************************************************************ 1156 */ 1157 1158 void CA_UnlockUtteranceForInput(CA_Utterance *hUtt); 1159 /** 1160 * 1161 * Params hUtt valid utterance handle 1162 * 1163 * Returns void 1164 * 1165 * See CA_LockUtteranceFromInput 1166 * CA_ResetVoicing 1167 * 1168 ************************************************************************ 1169 * This primes the utterance object by setting it's status to ACTIVE 1170 ************************************************************************ 1171 */ 1172 1173 1174 void CA_LockUtteranceFromInput(CA_Utterance *hUtt); 1175 /** 1176 * 1177 * Params hUtt valid utterance handle 1178 * 1179 * Returns void 1180 * 1181 * See CA_UnlockUtteranceForInput 1182 * 1183 ************************************************************************ 1184 * This signals that the utterance is complete by setting it's status to IDLE 1185 ************************************************************************ 1186 */ 1187 1188 1189 void CA_ResetVoicing(CA_Utterance *hUtt); 1190 /** 1191 * 1192 * Params hUtt valid utterance handle 1193 * 1194 * Returns void 1195 * 1196 * See CA_UnlockUtteranceForInput 1197 * CA_LockUtteranceFromInput 1198 * 1199 ************************************************************************ 1200 * This prepares the utterance object for re-use. The 'utterance has 1201 * started' and 'utterance has ended' flags are cleared. 1202 * Unlike CA_UnlockUtteranceForInput, this does not fully clear the utterance. 1203 ************************************************************************ 1204 */ 1205 1206 1207 void CA_FlushUtteranceFrames(CA_Utterance *hUtt); 1208 /** 1209 * 1210 * Params hUtt valid utterance handle 1211 * 1212 * Returns void 1213 * 1214 * See CA_InitUtterance 1215 * 1216 ************************************************************************ 1217 * Clears the utterance object's frame buffer 1218 ************************************************************************ 1219 */ 1220 1221 1222 void CA_SetEndOfUtteranceByLevelTimeout(CA_Utterance *hUtt, 1223 long timeout, 1224 long holdOff); 1225 /** 1226 * 1227 * Params hUtt valid utterance handle 1228 * timeout timeout before EOU decision 1229 * holdOff Period which must pass before EOU can be signalled 1230 * 1231 * Returns void 1232 * 1233 * See CA_UtteranceHasEnded 1234 * CA_SeekStartOfUtterance 1235 * 1236 ************************************************************************ 1237 * Sets the duration (of silence) which should pass before making 1238 * and end-of-utterance (EOU) decision. 1239 * 1240 * 'holdOff' is the period which must elapse before an EOU decision can 1241 * be made. An EOU will not occure before 'holdOff'. 1242 * 1243 * 'timeout' anf 'holdOff' are measured in frames. 1244 * 1245 * Caution: if an utterance can be spoken within 'holdOff' frames 1246 * it may be missed by the recognizer. 1247 ************************************************************************ 1248 */ 1249 1250 1251 int CA_UtteranceHasEnded(CA_Utterance *hUtt); 1252 /** 1253 * 1254 * Params hUtt valid utterance handle 1255 * 1256 * Returns non-zero if utterance has finished 1257 * (we've processed the last frame and we're locked from input) 1258 * 1259 * See CA_SeekStartOfUtterance 1260 * CA_SetEndOfUtteranceByLevelTimeout 1261 * 1262 ************************************************************************ 1263 * This method can be polled to interrogate the state of the utterance. 1264 ************************************************************************ 1265 */ 1266 1267 /* 1268 ** File: utt_file.c 1269 */ 1270 int CA_SeekUtteranceFrame(CA_Utterance *hUtt, 1271 int frame_no); 1272 /** 1273 * 1274 * Params hUtt valid utterance handle 1275 * frame_no new frame number (should this be unsigned long?) 1276 * 1277 * Returns -1 on error 1278 * 1279 * See CA_NextPatternFrame 1280 * 1281 ************************************************************************ 1282 * The 'file pointer' is set to an explicit location 1283 ************************************************************************ 1284 */ 1285 1286 1287 int CA_SaveUtterance(CA_Utterance *hUtt); 1288 /** 1289 * 1290 * Params hUtt valid utterance handle 1291 * 1292 * Returns int 1293 * 1294 ************************************************************************ 1295 * Saves an utterance 1296 ************************************************************************ 1297 */ 1298 1299 1300 void CA_SaveUtteranceHeader(CA_Utterance *hUtt, int frames, 1301 int samplerate, int framerate); 1302 /** 1303 * 1304 * Params hUtt valid utterance handle 1305 * frames frames which have been saved in this utt 1306 * samplerate sampling rate (typically 22050Hz) 1307 * framerate frame rate (typically 100 frames per second) 1308 * Returns void 1309 * 1310 ************************************************************************ 1311 * Saves an utterance 1312 ************************************************************************ 1313 */ 1314 1315 1316 /* 1317 ** File: utt_data.c 1318 */ 1319 int CA_SeekStartOfUtterance(CA_Utterance *hUtt); 1320 /** 1321 * 1322 * Params hUtt valid utterance handle 1323 * 1324 * Returns non-zero if utterance has begun 1325 * 1326 * See CA_UtterancehasEnded 1327 * 1328 ************************************************************************ 1329 * This method can be polled to interrogate the state of the utterance. 1330 ************************************************************************ 1331 */ 1332 1333 1334 int CA_GetUnprocessedFramesInUtterance(CA_Utterance *hUtt); 1335 /** 1336 * 1337 * Params hUtt valid utterance handle 1338 * 1339 * Returns number of un-read frames in the utterance's frame buffer 1340 * 1341 * See CA_AdvanceRecognitionByFrame 1342 * 1343 ************************************************************************ 1344 * Returns the number of un-read frames in the utterance's frame buffer, 1345 * i.e. those that have been inserted but have not yet been processed 1346 * by the recognizer object. 1347 ************************************************************************ 1348 */ 1349 1350 1351 /* 1352 ** File: utt_proc.c 1353 */ 1354 int CA_LoadUtteranceFrame(CA_Utterance *hUtt, 1355 unsigned char* pUttFrame, 1356 int uttFrameLen); 1357 /** 1358 * 1359 * Params hUtt valid utterance handle 1360 * pUttFrame Pointer utterance data 1361 * uttFrameLen Number of items (features) in the frame 1362 * 1363 * Returns zero on error 1364 * 1365 * See CA_InitUtteranceForFrontend 1366 * CA_GetFrontendUtteranceDimension 1367 * 1368 ************************************************************************ 1369 * Inserts the frame data pointed to by 'pUttFrame' into the utterance 1370 * object 'hUtt'. 1371 * 1372 * The size of the frame must agree with the value used to initialise 1373 * the utterance object. Utterances suitable for injection are 1374 * ones created for the Frontend. 1375 ************************************************************************ 1376 */ 1377 1378 1379 int CA_CopyUtteranceFrame(CA_Utterance *hUtt1, 1380 CA_Utterance *hUtt2); 1381 /** 1382 * 1383 * Params hUtt1 Destination valid utterance handle 1384 * hUtt2 Source valid utterance handle 1385 * 1386 * Returns zero on error 1387 * 1388 ************************************************************************ 1389 * Copies utterance frames from one utterance object to another 1390 ************************************************************************ 1391 */ 1392 1393 1394 void CA_CopyPatternFrame(CA_Utterance *hUtt, 1395 CA_Pattern *hPattern); 1396 /** 1397 No CR equivalent function 1398 */ 1399 1400 1401 int CA_CalculateUtteranceStatistics(CA_Utterance *hUtt, 1402 int start, 1403 int end); 1404 /** 1405 * 1406 * Params hUtt valid utterance handle 1407 * start starting frame ID (timecode) 1408 * end final frame ID (timecode) 1409 * 1410 * Returns Number of frames used in analysis 1411 * 1412 * 1413 ************************************************************************ 1414 * Signals the collection of statistics over a pre-defined range of 1415 * utterance frames in the utterance's frame buffer. do_filter flag 1416 * selects the frames that are likely to have just the ambient noise 1417 * only. 1418 * returns the number of frames used in the analysis. 1419 ************************************************************************ 1420 */ 1421 1422 1423 /* 1424 ** CA_Syntax methods 1425 */ 1426 1427 CA_Syntax *CA_AllocateSyntax(void); 1428 /** 1429 * 1430 * Params void 1431 * 1432 * Returns Handle to new syntax structure (object) 1433 * 1434 * See CA_FreeSyntax 1435 * 1436 ************************************************************************ 1437 * creates a new syntax object 1438 ************************************************************************ 1439 */ 1440 1441 1442 void CA_FreeSyntax(CA_Syntax *hSyntax); 1443 /** 1444 * 1445 * Params hSyntax valid syntax handle 1446 * 1447 * Returns void 1448 * 1449 * See CA_AllocateSyntax 1450 * 1451 ************************************************************************ 1452 * Frees an allocated syntax object 1453 ************************************************************************ 1454 */ 1455 1456 int CA_IsEnrollmentSyntax(CA_Syntax *hSyntax); 1457 /** 1458 * 1459 * Params hSyntax valid syntax handle 1460 * 1461 * Returns int 1462 * 1463 ************************************************************************ 1464 * tells whether a syntax is to be used for voice-enrollment 1465 ************************************************************************ 1466 */ 1467 1468 int CA_SetupSyntaxForRecognizer(CA_Syntax *hSyntax, 1469 CA_Recog *hRecog); 1470 /** 1471 * 1472 * Params hSyntax valid syntax handle 1473 * hRecog valid recog handle 1474 * 1475 * Returns The number of Syntax groups located 1476 * 1477 * See CA_CeilingSyntaxForRecognizer 1478 * 1479 ************************************************************************ 1480 * This is an essential, pre-recognition, setup method on the syntax object. 1481 * A syntax can be set up using more than one recognizer, provided the 1482 * active models have matching word indices. 1483 * now deprecated 1484 ************************************************************************ 1485 */ 1486 int CA_CeilingSyntaxForRecognizer(CA_Syntax *hSyntax, 1487 CA_Recog *hRecog); 1488 /** 1489 * 1490 * Params hSyntax valid syntax handle 1491 * hRecog valid recog handle 1492 * 1493 * Returns The number of Syntax groups located 1494 * 1495 * See CA_ClearSyntaxForRecognizer 1496 * 1497 ************************************************************************ 1498 * This is an essential, pre-recognition, setup method on the syntax object. 1499 * A syntax can be set up using more than one recognizer, provided the 1500 * active models have matching word indices. 1501 * now deprecated 1502 ************************************************************************ 1503 */ 1504 1505 1506 1507 void CA_ClearSyntaxForRecognizer(CA_Syntax *hSyntax, 1508 CA_Recog *hRecog); 1509 /** 1510 * 1511 * Params hSyntax valid syntax handle 1512 * hRecog valid recog handle 1513 * 1514 * Returns void 1515 * 1516 * See CA_SetupSyntaxForRecognizer 1517 * CA_UnloadRecognitionModels 1518 * 1519 ************************************************************************ 1520 * Clears a previous syntax setup. 1521 ************************************************************************ 1522 */ 1523 1524 1525 int CA_CompileSyntax(CA_Syntax *hSyntax); 1526 /** 1527 * 1528 * Params hSyntax valid syntax handle 1529 * 1530 * Returns 0 if successful, otherwise 1 to indicate faliure 1531 * 1532 * See CA_AddWordToSyntax 1533 * 1534 ************************************************************************ 1535 * Compiles a syntax after words have been added dynamically 1536 * 1537 ************************************************************************ 1538 */ 1539 1540 /* 1541 ** File: syn_grou.c 1542 */ 1543 1544 1545 void CA_GetGroupName(CA_Syntax *hSyntax, 1546 int group_id, 1547 char *buff, 1548 int buff_len); 1549 /** 1550 No CR equivalent function. Not called by any function 1551 */ 1552 1553 void CA_DeleteSyntaxGroups(CA_Syntax *hSyntax); 1554 /** 1555 CR function has no prototype 1556 */ 1557 1558 /* 1559 ** File: syn_rule.c 1560 */ 1561 1562 int CA_CheckTranscription(CA_Syntax *hSyntax, 1563 const char *trans, 1564 int allow_group_labels); 1565 /** 1566 * 1567 * Params hSyntax valid syntax handle 1568 * trans character string phrase 1569 allow_group_labels If 1, permits either words or group labels 1570 * 1571 * Returns -1 if successful, 1 if failed, 0 if all words parsed, 1572 but not on a valid terminal node. 1573 * 1574 ************************************************************************ 1575 * Use CA_CheckTranscription to check whether a phrase is valid for the 1576 * given syntax. Set allow_group_labels to 0 if you use single-word per group. 1577 * 1578 * NB When allow_group_labels is set to 1, there is a danger that a group 1579 * label might match an invalid word. 1580 ************************************************************************ 1581 */ 1582 1583 /* It's nice to keep the Arbdata separate from the models and from the graphs 1584 because we may have two or more acoustic models that use the same arbdata, 1585 which is also justification for having the model durations in the 1586 acoustic model file. Different grammars should also be able to 1587 share an Arbdata */ 1588 /* new CA_Syntax functions */ 1589 int CA_AttachArbdataToSyntax(CA_Syntax* hSyntax, CA_Arbdata* arbdata); 1590 CA_Arbdata* CA_LoadArbdata(const char* filename); 1591 unsigned int CA_ArbdataGetModelVersionID(CA_Arbdata* ca_arbdata); 1592 int CA_ArbdataGetModelIdsForPron(CA_Arbdata* ca_arbdata, 1593 const char* pronunciation, /* WB assumed at the edges */ 1594 int pronunciation_len, 1595 modelID* pmodelIds); 1596 int CA_ArbdataGetModelIdsForPIC(CA_Arbdata* ca_arbdata, const char lphon, const char cphon, const char rphon); 1597 1598 /** 1599 * Destroys arbdata. 1600 * 1601 * @param arbdata CA_Arbdata handle 1602 */ 1603 void CA_FreeArbdata(CA_Arbdata* arbdata); 1604 int CA_AddWordToSyntax(CA_Syntax* syntax, const char* slot, 1605 const char *phrase, const char* pronunciation, const int weight); 1606 int CA_ResetSyntax(CA_Syntax* syntax); 1607 int CA_LoadSyntaxAsExtensible(CA_Syntax *hSyntax, char *synbase, 1608 int num_words_to_add); 1609 int CA_DumpSyntaxAsImage(CA_Syntax *hSyntax, const char *imagename, int version_number); 1610 int CA_DumpSyntax(CA_Syntax *hSyntax, const char *basename); 1611 int CA_LoadSyntaxFromImage(CA_Syntax *hSyntax, const LCHAR* filename); 1612 1613 1614 /* 1615 ** CA_Recog methods 1616 */ 1617 /* 1618 ** File: rec_basi.c 1619 */ 1620 CA_Recog *CA_AllocateRecognition(void); 1621 /** 1622 * 1623 * Params void 1624 * 1625 * Returns Handle to valid recognizer object, NULL on error. 1626 * 1627 * See CA_FreeRecognition 1628 * 1629 ************************************************************************ 1630 * Creates a reconition object. 1631 * This only creates the object and internal storage structures. 1632 * The pointer can be re-used but must be deleted when finished with. 1633 ************************************************************************ 1634 */ 1635 1636 1637 int CA_ConfigureRecognition(CA_Recog *hRecoc, 1638 CA_RecInputParams *hRecInput); 1639 1640 1641 1642 void CA_UnconfigureRecognition(CA_Recog *hRecog); 1643 /** 1644 * 1645 * Params hRecog valid recog handle 1646 * 1647 * Returns void 1648 * 1649 * See CA_ConfigureRecognition 1650 * 1651 ************************************************************************ 1652 * Unconfigures recognizer. 1653 ************************************************************************ 1654 */ 1655 1656 1657 void CA_FreeRecognition(CA_Recog *hRecog); 1658 /** 1659 * 1660 * Params hRecog valid recog handle 1661 * 1662 * Returns void 1663 * 1664 * See CA_AllocateRecognition 1665 * 1666 ************************************************************************ 1667 * Deletes a previously allocated recognizer 1668 ************************************************************************ 1669 */ 1670 1671 1672 void CA_BeginRecognition(CA_Recog *hRecog, 1673 CA_Syntax *hSyntax, 1674 int first_syntax_node, 1675 CA_RecInputParams *hRecInput); 1676 /** 1677 * 1678 * Params hRecog valid recog handle 1679 * hSyntax valid syntax handle 1680 * first_syntax_node starting syntax node 1681 * hRecInput valid recog input parameter handle 1682 * 1683 * Returns void 1684 * 1685 * See CA_EndRecognition 1686 * CA_LoadRecognitionParameters 1687 * 1688 ************************************************************************ 1689 * Before a recognizer object can be used it must be initialized. 1690 * This method performs this task and is given an appropriate 1691 * syntax (and starting node) along with the required recognition 1692 * input parameters. 1693 * 1694 * 'first_syntax_node' should be set to '1' (the root of the syntax object). 1695 * Only experienced users should set this to any other value. It is 1696 * a start node ID returned by CA_AddNewNode(). 1697 * 1698 * It is an error to call this function without first loading 1699 * recognizer input parameters. 1700 ************************************************************************ 1701 */ 1702 1703 1704 void CA_AdvanceRecognitionByFrame(CA_Recog *hRecog, 1705 CA_Pattern *hPattern, 1706 CA_Utterance *hUtterance); 1707 /** 1708 * 1709 * Params hRecog valid recog handle 1710 * hPattern valid pattern handle 1711 * hUtterance valid utterance handle 1712 * 1713 * Returns void 1714 * 1715 * See CA_BeginRecognition 1716 * CA_EndRecognition 1717 * CA_AllocatePattern 1718 * CA_AllocateUtterance 1719 * CA_NextPatternFrame 1720 * 1721 ************************************************************************ 1722 * The vocabulary and acoustic object are required by the recognition 1723 * object to process input frames should already have been loaded with 1724 * a call to one of the CA_LoadModels methods. 1725 * 1726 * This method advances the recognizer by a single frame from the 1727 * input stream. This should be called when time permits and frames 1728 * are available in the input stream (utterance). Typically one would 1729 * use the CA_Pattern object to detect the presence of unused frames, 1730 * This is shown below: 1731 * 1732 * while ( CA_NextPatternFrame( hPattern, hUtt ) ) 1733 * { 1734 * CA_AdvanceRecognitionByFrame( hRecog, hPattern, hUtt ); 1735 * } 1736 * 1737 * This frame-by-frame approach provides the developer with fine control 1738 * of the recognition task. In a multi-process real-time enviornment 1739 * where a sophisticated OS may not be present, one might want to 1740 * schedule other tasks between individual frames. This was certainly 1741 * the case on early DSP boards chosen to host the CA recognizer. 1742 * The developer may, of course, decide that such fine control is not 1743 * required. In such cases a higher level convenience function can easily 1744 * be constructed and 'wrapped' around this method and 1745 * which prepares the recognizer, processes all frames and then returns to 1746 * the caller. 1747 * 1748 * Recognition results may be issued at any time within this method, 1749 * It is partial traceback controls the time when results can be issued. 1750 * Currently results are placed in a file or to the screen - the 1751 * CA_Result methods provide additional flexibility. 1752 ************************************************************************ 1753 */ 1754 1755 1756 int CA_EndRecognition(CA_Recog *hRecog, 1757 CA_Pattern *hPattern, 1758 CA_Utterance *hUtterance); 1759 /** 1760 * 1761 * Params hRecog valid recog handle 1762 * hPattern valid pattern handle 1763 * hUtterance valid utterance handle 1764 * 1765 * Returns non-ZERO if correct termination 1766 * 1767 * See CA_BeginRecognition 1768 * CA_AdvanceRecognitionByFrame 1769 * CA_DiscardRecognition 1770 * 1771 ************************************************************************ 1772 * When input processing is deemed to be complete, this mthod must 1773 * be called. It should be called after all the required input frames 1774 * have been advanced through. Not more than one recognizer may be 1775 * terminated in this fashion when multiple recognizers are employed on 1776 * given utterance. All others should be discarded. 1777 ************************************************************************ 1778 */ 1779 1780 1781 int CA_DiscardRecognition(CA_Recog *hRecog, 1782 CA_Pattern *hPattern, 1783 CA_Utterance *hUtterance); 1784 /** 1785 * 1786 * Params hRecog valid recog handle 1787 * hPattern valid pattern handle 1788 * hUtterance valid utterance handle 1789 * 1790 * 1791 * Returns non-ZERO if successful 1792 * 1793 * See CA_BeginRecognition 1794 * CA_AdvanceRecognitionByFrame 1795 * CA_EnddRecognition 1796 * 1797 ************************************************************************ 1798 * To discard the current recognition process. Not more than one 1799 * recognizer may be terminated in this fashion when multiple recognizers 1800 * are employed on given utterance. All others should be discarded. 1801 ************************************************************************ 1802 */ 1803 1804 1805 void CA_UpdateCMSAccumulates(CA_Utterance *hUtt, 1806 CA_Recog *hRecog); 1807 /** 1808 * 1809 * Params hUtt valid utterance handle 1810 * hRecog valid recog handle 1811 * 1812 * Returns Nothing 1813 * 1814 * See CA_AttachCMStoUtterane 1815 * CA_CalculateCMSParameters 1816 * 1817 ************************************************************************ 1818 * Copies channel normalization accumulates from recog to utterance. 1819 * These are then used when a call is made to CA_CalculateCMSParameters. 1820 ************************************************************************ 1821 */ 1822 1823 /* 1824 ** File: rec_load.c 1825 */ 1826 void CA_LoadModelsInAcoustic(CA_Recog *hRecog, 1827 CA_Acoustic *hAcoust, 1828 CA_AcoustInputParams *hAcoustInp); 1829 /** 1830 * 1831 * Params hRecog valid recog handle 1832 * hAcoust valid acoutic (model) handle 1833 * hAcoustInp valid acoustic inpit parameter handle 1834 * 1835 * Returns void 1836 * 1837 * See CA_LoadModelsInDictionary 1838 * CA_UnloadRecognitionModels 1839 * CA_LoadAcousticParameters 1840 * 1841 ************************************************************************ 1842 * This method constructs whole or sub-word models from 1843 * the entire model set in preparation for recognition. 1844 * 1845 * It is an error to call this function without first loading 1846 * acoustic input parameters. 1847 ************************************************************************ 1848 */ 1849 1850 1851 void CA_LoadModelsInDictionary(CA_Recog *hRecog, 1852 CA_Vocab *hVocab, 1853 CA_Acoustic *hAcoust, 1854 CA_AcoustInputParams *hAcoustInp); 1855 1856 1857 1858 void CA_LoadCrosswordTriphonesInSyntax(CA_Recog *hRecog, 1859 CA_Syntax *hSyntax, 1860 CA_Acoustic *hAcoust, 1861 CA_AcoustInputParams *hAcoustInp, 1862 int do_generics); 1863 /** 1864 * 1865 * Params hRecog valid recog handle 1866 * hSyntax valid syntax handle 1867 * hVocab valid vocabulary handle 1868 * hAcoust valid acoustic (model) handle 1869 * hAcoustInp valid acoustic input parameter handle 1870 * do-generics 1871 * 1872 * Returns void 1873 * 1874 * See CA_LoadModelsInAcoustic 1875 * CA_LoadModelsInDictionary 1876 * CA_UnloadRecognitionModels 1877 * CA_LoadAcousticParameters 1878 * 1879 ************************************************************************ 1880 * It is an error to call this function without first loading 1881 * acoustic input parameters. 1882 ************************************************************************ 1883 */ 1884 1885 1886 void CA_UnloadRecognitionModels (CA_Recog *hRecog); 1887 /** 1888 * 1889 * Params hRecog valid recog handle 1890 * 1891 * Returns void 1892 * 1893 * See CA_LoadModelsInAcoustic 1894 * CA_LoadModelsInDictionary 1895 * 1896 ************************************************************************ 1897 * Destorys a previously loaded set of models. 1898 * The models could have been constructed from acoustic information, 1899 * dictionary (vocabulary) information or syntax (grammar) information 1900 ************************************************************************ 1901 */ 1902 1903 1904 void CA_LoadMonophonesInAcoustic(CA_Recog *hRecog, 1905 CA_Acoustic *hAcoust, 1906 CA_AcoustInputParams *hAcoustInp); 1907 /** 1908 * 1909 * Params hRecog valid recog handle 1910 * hAcoust valid acoustic (model) handle 1911 * hAcoustInp valid acoustic input parameter handle 1912 * 1913 * Returns void 1914 * 1915 * See CA_LoadModelsInAcoustic 1916 * CA_LoadCrosswordTriphonesInSyntax 1917 * CA_UnloadRecognitionModels 1918 * CA_LoadAcousticParameters 1919 * 1920 ************************************************************************ 1921 * This method constructs models for phoneme recognition. When using it, 1922 * replace the calls to CA_LoadModelsInAcoustic and 1923 * CA_LoadCrosswordTriphonesInSyntax with a call to 1924 * CA_LoadMonophonesInAcoustic. 1925 ************************************************************************ 1926 */ 1927 1928 1929 int CA_CompareModelIndices(CA_Recog *hRecog1, 1930 CA_Recog *hRecog2); 1931 /** 1932 * 1933 * Params hRecog1 valid recog handle 1934 * hRecog2 valid recog handle 1935 * 1936 * Returns 0 if indices match. 1937 * 1938 * See CA_SetupSyntaxForRecognizer 1939 * 1940 ************************************************************************ 1941 * Checks that the model indices for two recognizers that have been set 1942 * up are compatible. This check should be made if the recognizers are 1943 * to use a shared syntax object. The check should be made prior to the 1944 * calls to CA_BeginRecognition. 1945 ************************************************************************ 1946 */ 1947 1948 1949 int CA_FixModelIndices(CA_Recog *hRecogTgt, 1950 CA_Acoustic *hAcousticSub, 1951 CA_Acoustic *hAcousticWhole, 1952 CA_Recog *hRecogSrc); 1953 /** 1954 * 1955 * Params hRecogTgt valid recog handle 1956 * hAcousticSub valid sub-word acoustic handle 1957 * hAcousticWhole valid whole-word acoustic handle 1958 * hRecogSrc valid recog handle 1959 * 1960 * Returns 0 if successful. 1961 * 1962 * See CA_CompareModelIndices 1963 * CA_LoadCrossWordTriphonesInSyntax 1964 * 1965 ************************************************************************ 1966 * Us this function if the model indices for two recognizers that have 1967 * been set up are not compatible. This function modifies the target 1968 * recognizer's model indices to make them compatible with one another. 1969 * This means that they can safely share a syntax object without the 1970 * need for calling CA_LoadCrosswordTriphonesInSyntax. 1971 ************************************************************************ 1972 */ 1973 1974 1975 1976 1977 /* 1978 ** File: rec_addi.c 1979 */ 1980 int CA_GetRecognitionTopScore(CA_Recog *hRecog); 1981 /** 1982 * 1983 * Params hRecog valid recog handle 1984 * 1985 * Returns top score 1986 * 1987 * See CA_SetRecognitionTopScore 1988 * 1989 ************************************************************************ 1990 * Reads top scoring result (used in multi-model configurations) 1991 ************************************************************************ 1992 */ 1993 1994 1995 void CA_SetRecognitionTopScore(CA_Recog *hRecog, 1996 int score); 1997 /** 1998 * 1999 * Params hRecog valid recog handle 2000 * score score 2001 * 2002 * Returns void 2003 * 2004 * See CA_GetRecognitionTopScore 2005 * 2006 ************************************************************************ 2007 * Sets top scoring result (used in multi-model configurations) 2008 ************************************************************************ 2009 */ 2010 2011 2012 int CA_GetRecognitionTerminationScore(CA_Recog *hRecog); 2013 /** 2014 * 2015 * Params hRecog valid recog handle 2016 * 2017 * Returns top score 2018 * 2019 ************************************************************************ 2020 * Reads top scoring final result (used in multi-model configurations) 2021 ************************************************************************ 2022 */ 2023 2024 2025 int CA_TracebackPerformed(CA_Recog* hRecog); 2026 /** 2027 * 2028 * Params hRecog valid recog handle 2029 * 2030 * Returns non-ZERO if the previous recognition frame processing 2031 * resulted in the execution of a partial-traceback 2032 * 2033 * See CA_SetupRecognizerForSyntax 2034 * 2035 ************************************************************************ 2036 * Used to test for the need to get some results 2037 ************************************************************************ 2038 */ 2039 2040 2041 void CA_BlockRecognitionResults(CA_Recog *hRecog); 2042 /** 2043 * 2044 * Params hRecog valid recog handle 2045 * 2046 * Returns void 2047 * 2048 * See CA_UnBlockRecognitionResults 2049 * 2050 ************************************************************************ 2051 * Disables result generation (used in multi-model configurations) 2052 ************************************************************************ 2053 */ 2054 2055 2056 void CA_UnBlockRecognitionResults(CA_Recog *hRecog); 2057 /** 2058 * 2059 * Params hRecog valid recog handle 2060 * 2061 * Returns void 2062 * 2063 * See CA_BlockRecognitionResults 2064 * 2065 ************************************************************************ 2066 * Re-enables result generation (used in multi-model configurations) 2067 ************************************************************************ 2068 */ 2069 2070 2071 /* 2072 ** File: rec_resu.c 2073 */ 2074 int CA_FullResultScore(CA_Recog *hRecog, 2075 int *score, 2076 int do_incsil); 2077 2078 2079 2080 int CA_FullResultDuration(CA_Recog *hRecog, 2081 int *dur, 2082 int do_incsil); 2083 2084 2085 2086 #ifdef SWIGBUILD 2087 int CA_FullResultLabel(CA_Recog *hRecog, 2088 void *label, 2089 int len); 2090 /** 2091 * 2092 * Params hRecog valid recog handle 2093 * label ASCII storage for returned result label 2094 * len Number of charcaters available for label. 2095 * Up to 'len'-1 characters will be returned. 2096 * 2097 * Returns PARTIAL_RESULTS_EMPTY if no result is available for release. 2098 * Otherwize the result FULL_RESULT if a valid result is available, 2099 * or REJECT_RESULT if the reject score is > reject threshold. 2100 * 2101 * See CA_FullResultScore 2102 * CA_PurgeResults 2103 * 2104 ************************************************************************ 2105 * This method returns all the Main ROOT results that have occured 2106 * (and have been unread) up to now. Results reported by this method 2107 * will only be removed from subsequent calls after a 'purge'. 2108 * 2109 * If no results are available a return code of PARTIAL_RESULTS_EMPTY 2110 * is returned. In this situation the label will be set to "<>". 2111 * 2112 * All results between 'head' and 'tail' are reported. 2113 ************************************************************************ 2114 */ 2115 2116 2117 #else 2118 int CA_FullResultLabel(CA_Recog *hRecog, 2119 char *label, 2120 int len); 2121 /** 2122 * 2123 * Params hRecog valid recog handle 2124 * label ASCII storage for returned result label 2125 * len Number of charcaters available for label. 2126 * Up to 'len'-1 characters will be returned. 2127 * 2128 * Returns PARTIAL_RESULTS_EMPTY if no result is available for release. 2129 * Otherwize the result FULL_RESULT if a valid result is available, 2130 * or REJECT_RESULT if the reject score is > reject threshold. 2131 * 2132 * See CA_FullResultScore 2133 * CA_PurgeResults 2134 * 2135 ************************************************************************ 2136 * This method returns all the Main ROOT results that have occured 2137 * (and have been unread) up to now. Results reported by this method 2138 * will only be removed from subsequent calls after a 'purge'. 2139 * 2140 * If no results are available a return code of PARTIAL_RESULTS_EMPTY 2141 * is returned. In this situation the label will be set to "<>". 2142 * 2143 * All results between 'head' and 'tail' are reported. 2144 ************************************************************************ 2145 */ 2146 #endif 2147 2148 /** 2149 * Strips the slot market characters from the utterance string. 2150 * 2151 * @param text String to filter 2152 */ 2153 ESR_ReturnCode CA_ResultStripSlotMarkers(char *text); 2154 ESR_ReturnCode CA_FullResultWordIDs(CA_Recog *hRecog, wordID *wordIDs, size_t* len); 2155 2156 /** 2157 * Gets the recognizer model id that was succesful (id is index in swimdllist file) 2158 * 2159 * @param text String to filter 2160 */ 2161 ESR_ReturnCode CA_GetRecogID(CA_Recog *hRecog, int *id); 2162 2163 2164 int CA_GetRejectMargin(CA_Recog *hRecog); 2165 /** 2166 * 2167 * Params hRecog valid recog handle 2168 * 2169 * Returns reject margin, the differnce between the utterance score 2170 * and the reject model score. 2171 * 2172 * See CA_FullResultLabel 2173 * CA_FullResultScore 2174 * 2175 ************************************************************************ 2176 * This method returns the reject margin, the difference between the 2177 * utterance score and the reject model score. 2178 ************************************************************************ 2179 */ 2180 2181 void CA_PurgeResults(CA_Recog *hRecog); 2182 /** 2183 * 2184 * Params hRecog valid recog handle 2185 * 2186 * Returns Nothing, internal result list will be deleted 2187 * 2188 * See CA_FullResultLabel 2189 * CA_FullResultScore 2190 * 2191 ************************************************************************ 2192 ************************************************************************ 2193 */ 2194 2195 2196 void CA_ClearResults(CA_Recog *hRecog); 2197 /** 2198 * 2199 * Params hRecog valid recog handle 2200 * 2201 * Returns void 2202 * 2203 * See CA_RecognitionHasResults 2204 ************************************************************************ 2205 * Clears results in a recognizer. Should do CA_RecognitionHasResults 2206 * first, to see if there are results to clear. 2207 ************************************************************************ 2208 */ 2209 2210 int CA_RecognitionHasResults(CA_Recog *hRecog); 2211 /** 2212 * 2213 * Params hRecog valid recog handle 2214 * num_frames number of silence frames before end of 2215 * utterance is considered 2216 * 2217 * Returns void 2218 * 2219 ************************************************************************ 2220 ************************************************************************ 2221 */ 2222 2223 2224 int CA_IsEndOfUtteranceByResults(CA_Recog *hRecog); 2225 /** 2226 * 2227 * Params hRecog valid recog handle 2228 * 2229 * Returns 2230 * 2231 * See 2232 * 2233 ************************************************************************ 2234 ************************************************************************ 2235 */ 2236 2237 2238 int CA_FullResultWordCount(CA_Recog *hRecog, 2239 int *count); 2240 /** 2241 * 2242 * Params hRecog valid recog handle 2243 * count pointer to storage for results word count 2244 * 2245 * Returns PARTIAL_RESULTS_EMPTY if no result is available for release. 2246 * Otherwize the result FULL_RESULT if a valid result is available, 2247 * or REJECT_RESULT if the reject score is > reject threshold. 2248 * 2249 * See CA_FullResultLabel 2250 * CA_PurgeResults 2251 * 2252 ************************************************************************ 2253 * This method returns the word count of the Main ROOT results that have 2254 * occured (and have been unread) up to now (silence isn't a word). 2255 * Results reported by this method will only be removed from subsequent 2256 * calls after a 'purge'. 2257 * 2258 * If no results are available a return code of PARTIAL_RESULTS_EMPTY 2259 * is returned. In this situation the the count will be set to 0. 2260 * 2261 * All results between 'head' and 'tail' are reported. 2262 ************************************************************************ 2263 */ 2264 2265 2266 int CA_CheckWordHyphenation(CA_Recog *hRecog, 2267 int *hyphen, 2268 int position); 2269 /** 2270 * 2271 * Params hRecog valid recog handle 2272 * hyphen pointer to storage for hyphenation check result 2273 * position first word position to check (zero based) 2274 * 2275 * Returns PARTIAL_RESULTS_EMPTY if no result is available. 2276 * INVALID_REQUEST if position is illegal (negative or too big). 2277 * Otherwize the result FULL_RESULT. 2278 * 2279 * See CA_FullResultLabel 2280 * CA_FullResultWordCount 2281 * CA_PurgeResults 2282 * 2283 ************************************************************************ 2284 * This method checks whether two words should be hyphenated or not. This 2285 * check is done on the Main ROOT results that have occured (and have been 2286 * unread) up to now, excluding silence labels. The words are pointed out 2287 * by giving the position of the first word to check inside the sequence 2288 * (zero based, not counting silence labels). 2289 * 2290 * If no results are available, a return code of PARTIAL_RESULTS_EMPTY is 2291 * returned. If the given word position is illegal (either negative or too 2292 * big), a return code of INVALID_REQUEST is returned. 2293 * In these situations the hyphenation will be set to 0. 2294 ************************************************************************ 2295 */ 2296 2297 2298 /* 2299 ** File: rec_nbes.c 2300 */ 2301 CA_NBestList *CA_PrepareNBestList(CA_Recog *hRecog, 2302 int n, 2303 asr_int32_t *bestScore); 2304 /** 2305 * 2306 * Params hRecog valid recog object handle 2307 * n number of items in list 2308 * bestScore The score associated with the root result 2309 * is returned in this argument. 2310 * 2311 * Returns Handle to new N-Best object 2312 * 2313 * See CA_DeleteNBestList 2314 * 2315 ************************************************************************ 2316 * Creates an N-Best structure for iterating through as results arrive. 2317 ************************************************************************ 2318 */ 2319 2320 2321 void CA_DeleteNBestList(CA_NBestList *hNbest); 2322 /** 2323 * 2324 * Params hNbest valid nbest object handle 2325 * 2326 * Returns Nothing, the object is no longer valid 2327 * 2328 * See CA_PrepareNBestList 2329 * 2330 ************************************************************************ 2331 * Deletes a previously allocated NBest list object 2332 ************************************************************************ 2333 */ 2334 2335 LCHAR* CA_NBestListGetResultWord(CA_NBestList *hNbest, size_t iChoice); 2336 int CA_NBestListGetResultConfidenceValue(CA_NBestList *hNbest, size_t iChoice, int* value); 2337 int CA_NBestListRemoveResult(CA_NBestList *hNbest, int index); 2338 2339 int CA_NBestListCount(CA_NBestList *hNbest); 2340 /** 2341 * 2342 * Params hNbest valid N-Best object handle 2343 * 2344 * Returns Number of entries found in the list. The maximum number 2345 * of entries in the list is determined when the list is created. 2346 * 2347 * See CA_PrepareNBestList 2348 * 2349 ************************************************************************ 2350 * Once an N-Best list has been created this method allows the caller 2351 * to check the number of entries in the list. 2352 ************************************************************************ 2353 */ 2354 2355 /** 2356 * Returns the n'th nbest-list entry. 2357 * 2358 * @param nbest N-best list structure 2359 * @param index N-best list index 2360 * @param wordIDs Array of wordIDs making up the literal 2361 * @param len [in/out] Length of wordIDs argument. If the return code is ESR_BUFFER_OVERFLOW, 2362 * the required length is returned in this variable. 2363 * @param cost Cost associated with literal 2364 */ 2365 ESR_ReturnCode CA_NBestListGetResultWordIDs(CA_NBestList* nbest, size_t iChoice, wordID* wordIDs, size_t* len, asr_int32_t* cost); 2366 2367 2368 /* 2369 ** CA_Params methods 2370 */ 2371 /* 2372 ** File: par_basi.c 2373 */ 2374 CA_AcoustInputParams *CA_AllocateAcousticParameters(void); 2375 /** 2376 * 2377 * Params void 2378 * 2379 * Returns Handle to valid acoustic input object, NULL on error 2380 * 2381 * See CA_FreeAcousticParameters 2382 * CA_LoadAcousticWhole 2383 * CA_LoadAcousticSub 2384 * CA_LoadModelsInAcoustic 2385 * CA_SaveAcousticSub 2386 * CA_LoadModelsInDictionary 2387 * 2388 ************************************************************************ 2389 * Creates a new acoustic input paramater onject 2390 ************************************************************************ 2391 */ 2392 2393 2394 void CA_FreeAcousticParameters(CA_AcoustInputParams *hAcoustInp); 2395 /** 2396 * 2397 * Params hAcoustInp valid acoustic input handle 2398 * 2399 * Returns Nothing, the acoustic input handle is no longer valid 2400 * 2401 * See CA_AllocateAcousticParameters 2402 * 2403 ************************************************************************ 2404 * Removes a previously allocated parameter object 2405 ************************************************************************ 2406 */ 2407 2408 2409 CA_PatInputParams *CA_AllocatePatternParameters(void); 2410 /** 2411 * 2412 * Params void 2413 * 2414 * Returns Handle to a new pattern object, NULL on error 2415 * 2416 * See CA_FreePatternParameters 2417 * CA_LoadPattern 2418 * 2419 ************************************************************************ 2420 * Creates a new pattern input paramater onject 2421 ************************************************************************ 2422 */ 2423 2424 2425 void CA_FreePatternParameters(CA_PatInputParams *hPatInp); 2426 /** 2427 * 2428 * Params hPatInp valid pattern input handle 2429 * 2430 * Returns void The input handle is no longer valid. 2431 * 2432 * See CA_AllocatePatternParameters 2433 * 2434 ************************************************************************ 2435 * Removes a previously allocated parameter object 2436 ************************************************************************ 2437 */ 2438 2439 2440 CA_RecInputParams *CA_AllocateRecognitionParameters(void); 2441 /** 2442 * 2443 * Params void 2444 * 2445 * Returns Handle to new recog input object 2446 * 2447 * See CA_FreeRecognitionParameters 2448 * CA_BeginRecognition 2449 * 2450 ************************************************************************ 2451 * Creates a new recog input paramater onject 2452 ************************************************************************ 2453 */ 2454 2455 2456 void CA_FreeRecognitionParameters(CA_RecInputParams *hRecInp); 2457 /** 2458 * 2459 * Params hRecInp valid recog input handle 2460 * 2461 * Returns void the recognition input object is no longer valid 2462 * 2463 * See CA_AllocateRecognitionParameters 2464 * 2465 ************************************************************************ 2466 * Removes a previously allocated parameter object 2467 ************************************************************************ 2468 */ 2469 2470 2471 2472 /* 2473 ** File: par_load.c 2474 */ 2475 void CA_LoadAcousticParameters(CA_AcoustInputParams *hAcoustInp, 2476 const char *parfile); 2477 /** 2478 * 2479 * Params hAcoustInp valid acoustic input handle 2480 * parfile parameter (.INI) file to read 2481 * 2482 * Returns void 2483 * 2484 * See CA_SaveAcousticParameters 2485 * CA_LoadAcousticWhole 2486 * CA_LoadAcousticSub 2487 * CA_LoadModelsInAcoustic 2488 * CA_SaveAcousticSub 2489 * CA_LoadModelsInDictionary 2490 * 2491 ************************************************************************ 2492 * Loads known acoustic parameters from the given parameter file. 2493 * The file is an ASCII text file usually of the type .PAR or .INI 2494 ************************************************************************ 2495 */ 2496 2497 2498 void CA_SaveAcousticParameters(CA_AcoustInputParams *hAcoustInp, 2499 const char *parfile); 2500 /** 2501 * 2502 * Params hAcoustInp valid acoustic input handle 2503 * parfile parameter (.INI) file to write 2504 * 2505 * Returns void 2506 * 2507 * See CA_LoadAcousticParameters 2508 * 2509 ************************************************************************ 2510 * Saves a previously loaded (modified) parameter file 2511 * 2512 * It is an error to call this function without first loading 2513 * acoustic input parameters. 2514 ************************************************************************ 2515 */ 2516 2517 2518 void CA_LoadPatternParameters(CA_PatInputParams *hPatInp, 2519 const char *parfile); 2520 /** 2521 * 2522 * Params hPatInp valid pattern handle 2523 * parfile parameter (.INI) file to read 2524 * 2525 * Returns void 2526 * 2527 * See CA_SavePatternParameters 2528 * CA_LoadPattern 2529 * 2530 ************************************************************************ 2531 * Loads known pattern parameters from the given parameter file. 2532 * The file is an ASCII text file usually of the type .PAR or .INI 2533 ************************************************************************ 2534 */ 2535 2536 2537 void CA_SavePatternParameters(CA_PatInputParams *hPatInp, 2538 const char *parfile); 2539 /** 2540 * 2541 * Params hPatInp valid pattern input handle 2542 * parfile parameter (.INI) file to write 2543 * 2544 * Returns void 2545 * 2546 * See CA_LoadPatternParameters 2547 * CA_LoadPattern 2548 * 2549 ************************************************************************ 2550 * Saves a previously loaded (modified) parameter file 2551 * 2552 * It is an error to call this function without first loading 2553 * pattern input parameters. 2554 ************************************************************************ 2555 */ 2556 2557 2558 void CA_LoadRecognitionParameters(CA_RecInputParams *hRecInp, 2559 const char *parfile); 2560 /** 2561 * 2562 * Params hRecInp valid recog input handle 2563 * parfile parameter (.INI) file to read 2564 * 2565 * Returns void 2566 * 2567 * See CA_AllocateRecognitionParameters 2568 * CA_SaveRecognitionParameters 2569 * CA_BeginRecognition 2570 * 2571 ************************************************************************ 2572 * Loads known recognition parameters from the given parameter file. 2573 * The file is an ASCII text file usually of the type .PAR or .INI 2574 ************************************************************************ 2575 */ 2576 2577 2578 void CA_SaveRecognitionParameters(CA_RecInputParams *hRecInp, 2579 const char *parfile); 2580 /** 2581 * 2582 * Params hRecInp valid recog input handle 2583 * parfile parameter (.INI) file to write 2584 * 2585 * Returns void 2586 * 2587 * See CA_LoadRecognitionParameters 2588 * CA_BeginRecognition 2589 * 2590 ************************************************************************ 2591 * Saves a previously loaded (modified) parameter file 2592 ************************************************************************ 2593 */ 2594 2595 2596 /* 2597 ** File: par_set.c 2598 */ 2599 2600 int CA_GetAcousticParameter(CA_AcoustInputParams *hAcoustInp, 2601 char *key, 2602 void *value); 2603 /** 2604 * 2605 * Params hAcoustInp valid acoustic input handle 2606 * key parameter key (text label) 2607 * value pointer to store parameter value (text) 2608 * valueLen size of value buffer 2609 * 2610 * Returns False on error 2611 * 2612 * See CA_GetAcousticIntParameter 2613 * CA_GetAcousticFloatParameter 2614 * CA_LoadAcousticParameters 2615 * 2616 ************************************************************************ 2617 * Gets a known acoustic parameter. 2618 * 2619 * It is an error to call this function without first loading 2620 * acoustic input parameters. 2621 ************************************************************************ 2622 */ 2623 2624 2625 int CA_GetAcousticStringParameter(CA_AcoustInputParams *hAcoustInp, 2626 char *key, 2627 char *value, 2628 int valueLen, 2629 int *bytes_required); 2630 /** 2631 * 2632 * Params hAcoustInp valid acoustic input handle 2633 * key parameter key (text label) 2634 * value pointer to store parameter value (text) 2635 * valueLen size of value buffer 2636 * 2637 * Returns False on error 2638 * 2639 * See CA_GetAcousticIntParameter 2640 * CA_GetAcousticFloatParameter 2641 * CA_LoadAcousticParameters 2642 * 2643 ************************************************************************ 2644 * Gets a known acoustic parameter. 2645 * 2646 * It is an error to call this function without first loading 2647 * acoustic input parameters. 2648 ************************************************************************ 2649 */ 2650 2651 int CA_SetPatternParameter(CA_PatInputParams *hPatInp, 2652 char *key, 2653 char *value); 2654 /** 2655 * 2656 * Params hPatInp valid Pattern Input handle 2657 * key parameter key (text label) 2658 * value new parameter value (text) 2659 * 2660 * Returns Zero on error 2661 * 2662 * See CA_GetPatternStringParameter 2663 * CA_GetPatternIntParameter 2664 * CA_GetPatternFloatParameter 2665 * CA_LoadPatternParameters 2666 * 2667 ************************************************************************ 2668 * Sets/Modifies a known Pattern Input parameter. 2669 * 2670 * It is an error to call this function without first loading 2671 * pattern input parameters. 2672 ************************************************************************ 2673 */ 2674 2675 2676 int CA_GetPatternParameter(CA_PatInputParams *hPatInp, 2677 char *key, 2678 void *value); 2679 /** 2680 * 2681 * Params hPatInp valid Pattern Input handle 2682 * key parameter key (text label) 2683 * value pointer to store parameter value (int) 2684 * 2685 * Returns False on error 2686 * 2687 * See CA_SetPatternParameter 2688 * CA_LoadPatternParameters 2689 * 2690 ************************************************************************ 2691 * Reads a known Pattern Input Parameter. 2692 * 2693 * It is an error to call this function without first loading 2694 * pattern input parameters. 2695 ************************************************************************ 2696 */ 2697 2698 2699 int CA_GetPatternStringParameter(CA_PatInputParams *hPatInp, 2700 char *key, 2701 char *value, 2702 int valueLen, 2703 int *bytes_required); 2704 /** 2705 * 2706 * Params hPatInp valid Pattern Input handle 2707 * key parameter key (text label) 2708 * value pointer to store parameter value (text) 2709 * valueLen size of value buffer 2710 * 2711 * Returns False on error 2712 * 2713 * See CA_SetPatternParameter 2714 * CA_LoadPatternParameters 2715 * 2716 ************************************************************************ 2717 * Reads a known Pattern Input Parameter. 2718 * 2719 * It is an error to call this function without first loading 2720 * pattern input parameters. 2721 ************************************************************************ 2722 */ 2723 2724 2725 int CA_SetRecognitionParameter(CA_RecInputParams *hRecInp, 2726 char *key, 2727 char *value); 2728 /** 2729 * 2730 * Params hRecInp valid recog input handle 2731 * key parameter key (text label) 2732 * value new parameter value (text) 2733 * 2734 * Returns Zero on error 2735 * 2736 * See CA_GetRecognitionStringParameter 2737 * CA_GetRecognitionIntParameter 2738 * CA_GetRecognitionFloatParameter 2739 * CA_LoadRecognitionParameters 2740 * 2741 ************************************************************************ 2742 * Sets/Modifies a known recognition parameter. 2743 * 2744 * It is an error to call this function without first loading 2745 * recognizer input parameters. 2746 ************************************************************************ 2747 */ 2748 2749 int CA_GetRecognitionParameter(CA_RecInputParams *hRecInp, 2750 char *key, 2751 void *value); 2752 /** 2753 * 2754 * Params hRecInp valid recog input handle 2755 * key parameter key (text label) 2756 * value pointer to store parameter value (int) 2757 * 2758 * Returns False on error 2759 * 2760 * See CA_SetRecognitionParameter 2761 * CA_BeginRecognition 2762 * CA_LoadRecognitionParameters 2763 * 2764 ************************************************************************ 2765 * Reads a known recognition parameter. 2766 * 2767 * It is an error to call this function without first loading 2768 * recognizer input parameters. 2769 ************************************************************************ 2770 */ 2771 2772 2773 int CA_GetRecognitionStringParameter(CA_RecInputParams *hRecInp, 2774 char *key, 2775 char *value, 2776 int valueLen, 2777 int *bytes_required); 2778 #if USE_CONFIDENCE_SCORER 2779 2780 CA_ConfidenceScorer* CA_AllocateConfidenceScorer(void); 2781 2782 /** 2783 * 2784 * Params void 2785 * 2786 * Returns Handle to new ConfidenceScorer structure, or NULL on error. 2787 * 2788 * See CR_FreeConfidenceScorer 2789 * 2790 ************************************************************************ 2791 * Allocates a new SymbolMatch structure (object) 2792 ************************************************************************ 2793 */ 2794 2795 void CA_FreeConfidenceScorer(CA_ConfidenceScorer *hConfidenceScorer); 2796 2797 2798 int CA_LoadConfidenceScorer(CA_ConfidenceScorer* hConfidenceScorer); 2799 /** 2800 * 2801 * Params hConfidenceScorer Handle to previously allocated ConfidenceScorer object 2802 * nnet_file Name of the confidence scoring file 2803 * 2804 * Returns True if successful, False in the case of an error 2805 * 2806 * See 2807 * 2808 ************************************************************************ 2809 ************************************************************************ 2810 */ 2811 2812 void CA_UnloadConfidenceScorer(CA_ConfidenceScorer *hConfidenceScorer); 2813 /** 2814 * 2815 * Params hConfidenceScorer Handle to previously allocated ConfidenceScorer object 2816 * 2817 * Returns 2818 * 2819 * See 2820 * 2821 ************************************************************************ 2822 ************************************************************************ 2823 */ 2824 2825 int CA_ComputeConfidenceValues(CA_ConfidenceScorer* hConfidenceScorer, 2826 CA_Recog* recog, CA_NBestList *nbestlist); 2827 /** 2828 * 2829 * Params hConfidenceScorer Handle to previously allocated ConfidenceScorer object 2830 * recog Handle to recognizer that won 2831 * nbestlist Handle to nbest list where confidence values will be inserted 2832 * 2833 * Returns 0 if successful, 1 if not. 2834 * 2835 * 2836 ************************************************************************ 2837 ************************************************************************ 2838 */ 2839 2840 #endif /* #if USE_CONFIDENCE_SCORER */ 2841 2842 #ifdef __cplusplus 2843 } 2844 #endif 2845 2846 #endif 2847
