3rdparty/flite_hts_engine/lib/flite_hts_engine.c

   1 /* ----------------------------------------------------------------- */
   2 /*           The English TTS System "Flite+hts_engine"               */
   3 /*           developed by HTS Working Group                          */
   4 /*           http://hts-engine.sourceforge.net/                      */
   5 /* ----------------------------------------------------------------- */
   6 /*                                                                   */
   7 /*  Copyright (c) 2005-2013  Nagoya Institute of Technology          */
   8 /*                           Department of Computer Science          */
   9 /*                                                                   */
  10 /*                2005-2008  Tokyo Institute of Technology           */
  11 /*                           Interdisciplinary Graduate School of    */
  12 /*                           Science and Engineering                 */
  13 /*                                                                   */
  14 /* All rights reserved.                                              */
  15 /*                                                                   */
  16 /* Redistribution and use in source and binary forms, with or        */
  17 /* without modification, are permitted provided that the following   */
  18 /* conditions are met:                                               */
  19 /*                                                                   */
  20 /* - Redistributions of source code must retain the above copyright  */
  21 /*   notice, this list of conditions and the following disclaimer.   */
  22 /* - Redistributions in binary form must reproduce the above         */
  23 /*   copyright notice, this list of conditions and the following     */
  24 /*   disclaimer in the documentation and/or other materials provided */
  25 /*   with the distribution.                                          */
  26 /* - Neither the name of the HTS working group nor the names of its  */
  27 /*   contributors may be used to endorse or promote products derived */
  28 /*   from this software without specific prior written permission.   */
  29 /*                                                                   */
  30 /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND            */
  31 /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,       */
  32 /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF          */
  33 /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE          */
  34 /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */
  35 /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,          */
  36 /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED   */
  37 /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,     */
  38 /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */
  39 /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,   */
  40 /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY    */
  41 /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE           */
  42 /* POSSIBILITY OF SUCH DAMAGE.                                       */
  43 /* ----------------------------------------------------------------- */
  44
  45 #include "cst_synth.h"
  46 #include "cst_utt_utils.h"
  47 #include "cst_math.h"
  48 #include "cst_file.h"
  49 #include "cst_val.h"
  50 #include "cst_string.h"
  51 #include "cst_alloc.h"
  52 #include "cst_item.h"
  53 #include "cst_relation.h"
  54 #include "cst_utterance.h"
  55 #include "cst_tokenstream.h"
  56 #include "cst_string.h"
  57 #include "cst_regex.h"
  58 #include "cst_features.h"
  59 #include "cst_utterance.h"
  60 #include "flite.h"
  61 #include "cst_synth.h"
  62 #include "cst_utt_utils.h"
  63
  64 #include "flite_hts_engine.h"
  65
  66 /* HTS_GStreamSet_get_total_nsamples: get total number of sample */
  67 size_t HTS_GStreamSet_get_total_nsamples(HTS_GStreamSet * gss);
  68
  69 /* HTS_GStreamSet_get_speech: get synthesized speech parameter */
  70 double HTS_GStreamSet_get_speech(HTS_GStreamSet * gss, size_t sample_index);
  71
  72 #define REGISTER_VOX register_cmu_us_kal
  73 #define UNREGISTER_VOX unregister_cmu_us_kal
  74
  75 #define MAXBUFLEN 1024
  76
  77 cst_voice *REGISTER_VOX(const char *voxdir);
  78 cst_voice *UNREGISTER_VOX(cst_voice * vox);
  79
  80 /* Flite_HTS_Engine_create_label: create label per phoneme */
  81 static void Flite_HTS_Engine_create_label(Flite_HTS_Engine * f, cst_item * item, char *label)
  82 {
  83    char seg_pp[8];
  84    char seg_p[8];
  85    char seg_c[8];
  86    char seg_n[8];
  87    char seg_nn[8];
  88    char endtone[8];
  89    int sub_phrases = 0;
  90    int lisp_total_phrases = 0;
  91    int tmp1 = 0;
  92    int tmp2 = 0;
  93    int tmp3 = 0;
  94    int tmp4 = 0;
  95
  96    /* load segments */
  97    strcpy(seg_pp, ffeature_string(item, "p.p.name"));
  98    strcpy(seg_p, ffeature_string(item, "p.name"));
  99    strcpy(seg_c, ffeature_string(item, "name"));
 100    strcpy(seg_n, ffeature_string(item, "n.name"));
 101    strcpy(seg_nn, ffeature_string(item, "n.n.name"));
 102
 103    /* load endtone */
 104    strcpy(endtone, ffeature_string(item, "R:SylStructure.parent.parent.R:Phrase.parent.daughtern.R:SylStructure.daughtern.endtone"));
 105
 106    if (strcmp(seg_c, "pau") == 0) {
 107       /* for pause */
 108       if (item_next(item) != NULL) {
 109          sub_phrases = ffeature_int(item, "n.R:SylStructure.parent.R:Syllable.sub_phrases");
 110          tmp1 = ffeature_int(item, "n.R:SylStructure.parent.parent.R:Phrase.parent.lisp_total_syls");
 111          tmp2 = ffeature_int(item, "n.R:SylStructure.parent.parent.R:Phrase.parent.lisp_total_words");
 112          lisp_total_phrases = ffeature_int(item, "n.R:SylStructure.parent.parent.R:Phrase.parent.lisp_total_phrases");
 113       } else {
 114          sub_phrases = ffeature_int(item, "p.R:SylStructure.parent.R:Syllable.sub_phrases");
 115          tmp1 = ffeature_int(item, "p.R:SylStructure.parent.parent.R:Phrase.parent.lisp_total_syls");
 116          tmp2 = ffeature_int(item, "p.R:SylStructure.parent.parent.R:Phrase.parent.lisp_total_words");
 117          lisp_total_phrases = ffeature_int(item, "p.R:SylStructure.parent.parent.R:Phrase.parent.lisp_total_phrases");
 118       }
 119       sprintf(label, "%s^%s-%s+%s=%s@x_x/A:%d_%d_%d/B:x-x-x@x-x&x-x#x-x$x-x!x-x;x-x|x/C:%d+%d+%d/D:%s_%d/E:x+x@x+x&x+x#x+x/F:%s_%d/G:%d_%d/H:x=x^%d=%d|%s/I:%d=%d/J:%d+%d-%d", strcmp(seg_pp, "0") == 0 ? "x" : seg_pp, strcmp(seg_p, "0") == 0 ? "x" : seg_p, seg_c, strcmp(seg_n, "0") == 0 ? "x" : seg_n, strcmp(seg_nn, "0") == 0 ? "x" : seg_nn, ffeature_int(item, "p.R:SylStructure.parent.R:Syllable.stress"), ffeature_int(item, "p.R:SylStructure.parent.R:Syllable.accented"), ffeature_int(item, "p.R:SylStructure.parent.R:Syllable.syl_numphones"), ffeature_int(item, "n.R:SylStructure.parent.R:Syllable.stress"), ffeature_int(item, "n.R:SylStructure.parent.R:Syllable.accented"), ffeature_int(item, "n.R:SylStructure.parent.R:Syllable.syl_numphones"), ffeature_string(item, "p.R:SylStructure.parent.parent.R:Word.gpos"), ffeature_int(item, "p.R:SylStructure.parent.parent.R:Word.word_numsyls"), ffeature_string(item, "n.R:SylStructure.parent.parent.R:Word.gpos"), ffeature_int(item, "n.R:SylStructure.parent.parent.R:Word.word_numsyls"), ffeature_int(item, "p.R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_syls_in_phrase"), ffeature_int(item, "p.R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_words_in_phrase"), sub_phrases + 1, lisp_total_phrases - sub_phrases, endtone, ffeature_int(item, "n.R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_syls_in_phrase"), ffeature_int(item, "n.R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_words_in_phrase"), tmp1, tmp2, lisp_total_phrases);
 120    } else {
 121       /* for no pause */
 122       tmp1 = ffeature_int(item, "R:SylStructure.pos_in_syl");
 123       tmp2 = ffeature_int(item, "R:SylStructure.parent.R:Syllable.syl_numphones");
 124       tmp3 = ffeature_int(item, "R:SylStructure.parent.R:Syllable.pos_in_word");
 125       tmp4 = ffeature_int(item, "R:SylStructure.parent.parent.R:Word.word_numsyls");
 126       sub_phrases = ffeature_int(item, "R:SylStructure.parent.R:Syllable.sub_phrases");
 127       lisp_total_phrases = ffeature_int(item, "R:SylStructure.parent.parent.R:Phrase.parent.lisp_total_phrases");
 128       sprintf(label, "%s^%s-%s+%s=%s@%d_%d/A:%d_%d_%d/B:%d-%d-%d@%d-%d&%d-%d#%d-%d$%d-%d!%d-%d;%d-%d|%s/C:%d+%d+%d/D:%s_%d/E:%s+%d@%d+%d&%d+%d#%d+%d/F:%s_%d/G:%d_%d/H:%d=%d^%d=%d|%s/I:%d=%d/J:%d+%d-%d", strcmp(seg_pp, "0") == 0 ? "x" : seg_pp, strcmp(seg_p, "0") == 0 ? "x" : seg_p, seg_c, strcmp(seg_n, "0") == 0 ? "x" : seg_n, strcmp(seg_nn, "0") == 0 ? "x" : seg_nn, tmp1 + 1, tmp2 - tmp1, ffeature_int(item, "R:SylStructure.parent.R:Syllable.p.stress"), ffeature_int(item, "R:SylStructure.parent.R:Syllable.p.accented"), ffeature_int(item, "R:SylStructure.parent.R:Syllable.p.syl_numphones"), ffeature_int(item, "R:SylStructure.parent.R:Syllable.stress"), ffeature_int(item, "R:SylStructure.parent.R:Syllable.accented"), tmp2, tmp3 + 1, tmp4 - tmp3, ffeature_int(item, "R:SylStructure.parent.R:Syllable.syl_in") + 1, ffeature_int(item, "R:SylStructure.parent.R:Syllable.syl_out") + 1, ffeature_int(item, "R:SylStructure.parent.R:Syllable.ssyl_in") + 1, ffeature_int(item, "R:SylStructure.parent.R:Syllable.ssyl_out") + 1, ffeature_int(item, "R:SylStructure.parent.R:Syllable.asyl_in") + 1, ffeature_int(item, "R:SylStructure.parent.R:Syllable.asyl_out") + 1, ffeature_int(item, "R:SylStructure.parent.R:Syllable.lisp_distance_to_p_stress"), ffeature_int(item, "R:SylStructure.parent.R:Syllable.lisp_distance_to_n_stress"), ffeature_int(item, "R:SylStructure.parent.R:Syllable.lisp_distance_to_p_accent"), ffeature_int(item, "R:SylStructure.parent.R:Syllable.lisp_distance_to_n_accent"), ffeature_string(item, "R:SylStructure.parent.R:Syllable.syl_vowel"), ffeature_int(item, "R:SylStructure.parent.R:Syllable.n.stress"), ffeature_int(item, "R:SylStructure.parent.R:Syllable.n.accented"), ffeature_int(item, "R:SylStructure.parent.R:Syllable.n.syl_numphones"), ffeature_string(item, "R:SylStructure.parent.parent.R:Word.p.gpos"), ffeature_int(item, "R:SylStructure.parent.parent.R:Word.p.word_numsyls"), ffeature_string(item, "R:SylStructure.parent.parent.R:Word.gpos"), tmp4, ffeature_int(item, "R:SylStructure.parent.parent.R:Word.pos_in_phrase") + 1, ffeature_int(item, "R:SylStructure.parent.parent.R:Word.words_out"), ffeature_int(item, "R:SylStructure.parent.parent.R:Word.content_words_in") + 1, ffeature_int(item, "R:SylStructure.parent.parent.R:Word.content_words_out") + 1, ffeature_int(item, "R:SylStructure.parent.parent.R:Word.lisp_distance_to_p_content"), ffeature_int(item, "R:SylStructure.parent.parent.R:Word.lisp_distance_to_n_content"), ffeature_string(item, "R:SylStructure.parent.parent.R:Word.n.gpos"), ffeature_int(item, "R:SylStructure.parent.parent.R:Word.n.word_numsyls"), ffeature_int(item, "R:SylStructure.parent.parent.R:Phrase.parent.p.lisp_num_syls_in_phrase"), ffeature_int(item, "R:SylStructure.parent.parent.R:Phrase.parent.p.lisp_num_words_in_phrase"), ffeature_int(item, "R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_syls_in_phrase"), ffeature_int(item, "R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_words_in_phrase"), sub_phrases + 1, lisp_total_phrases - sub_phrases, strcmp(endtone, "0") == 0 ? "NONE" : endtone, ffeature_int(item, "R:SylStructure.parent.parent.R:Phrase.parent.n.lisp_num_syls_in_phrase"), ffeature_int(item, "R:SylStructure.parent.parent.R:Phrase.parent.n.lisp_num_words_in_phrase"), ffeature_int(item, "R:SylStructure.parent.parent.R:Phrase.parent.lisp_total_syls"), ffeature_int(item, "R:SylStructure.parent.parent.R:Phrase.parent.lisp_total_words"), lisp_total_phrases);
 129    }
 130 }
 131
 132 /* Flite_HTS_Engine_initialize: initialize system */
 133 void Flite_HTS_Engine_initialize(Flite_HTS_Engine * f)
 134 {
 135    HTS_Engine_initialize(&f->engine);
 136 }
 137
 138 /* Flite_HTS_Engine_load: load HTS voice */
 139 HTS_Boolean Flite_HTS_Engine_load(Flite_HTS_Engine * f, const char *fn)
 140 {
 141    HTS_Boolean result;
 142    char *voices = strdup(fn);
 143    result = HTS_Engine_load(&f->engine, &voices, 1);
 144    free(voices);
 145    return result;
 146 }
 147
 148 /* Flite_HTS_Engine_set_sampling_frequency: set sampling frequency */
 149 void Flite_HTS_Engine_set_sampling_frequency(Flite_HTS_Engine * f, size_t i)
 150 {
 151    HTS_Engine_set_sampling_frequency(&f->engine, i);
 152 }
 153
 154 /* Flite_HTS_Engine_set_fperiod: set frame period */
 155 void Flite_HTS_Engine_set_fperiod(Flite_HTS_Engine * f, size_t i)
 156 {
 157    HTS_Engine_set_fperiod(&f->engine, i);
 158 }
 159
 160 /* Flite_HTS_Engine_set_audio_buff_size: set audio buffer size */
 161 void Flite_HTS_Engine_set_audio_buff_size(Flite_HTS_Engine * f, size_t i)
 162 {
 163    HTS_Engine_set_audio_buff_size(&f->engine, i);
 164 }
 165
 166 /* Flite_HTS_Engine_set_alpha: set alpha */
 167 void Flite_HTS_Engine_set_alpha(Flite_HTS_Engine * f, double d)
 168 {
 169    HTS_Engine_set_alpha(&f->engine, d);
 170 }
 171
 172 /* Flite_HTS_Engine_set_beta: set beta */
 173 void Flite_HTS_Engine_set_beta(Flite_HTS_Engine * f, double d)
 174 {
 175    HTS_Engine_set_beta(&f->engine, d);
 176 }
 177
 178 /* Flite_HTS_Engine_add_half_tone: add half-tone */
 179 void Flite_HTS_Engine_add_half_tone(Flite_HTS_Engine * f, double d)
 180 {
 181    HTS_Engine_add_half_tone(&f->engine, d);
 182 }
 183
 184 /* Flite_HTS_Engine_set_msd_threshold: set MSD threshold */
 185 void Flite_HTS_Engine_set_msd_threshold(Flite_HTS_Engine * f, size_t stream_index, double d)
 186 {
 187    HTS_Engine_set_msd_threshold(&f->engine, stream_index, d);
 188 }
 189
 190 /* Flite_HTS_Engine_set_gv_weight: set GV weight */
 191 void Flite_HTS_Engine_set_gv_weight(Flite_HTS_Engine * f, size_t stream_index, double d)
 192 {
 193    HTS_Engine_set_gv_weight(&f->engine, stream_index, d);
 194 }
 195
 196 /* Flite_HTS_Engine_set_speed: set speech speed */
 197 void Flite_HTS_Engine_set_speed(Flite_HTS_Engine * f, double d)
 198 {
 199    HTS_Engine_set_speed(&f->engine, d);
 200 }
 201
 202 /* Flite_HTS_Engine_synthesize: synthesize speech */
 203 HTS_Boolean Flite_HTS_Engine_synthesize(Flite_HTS_Engine * f, const char *txt, const char *wav)
 204 {
 205    int i;
 206    FILE *fp;
 207    cst_voice *v = NULL;
 208    cst_utterance *u = NULL;
 209    cst_item *s = NULL;
 210    char **label_data = NULL;
 211    int label_size = 0;
 212
 213    if (txt == NULL)
 214       return FALSE;
 215
 216    /* text analysis part */
 217    v = REGISTER_VOX(NULL);
 218    if (v == NULL)
 219       return FALSE;
 220    u = flite_synth_text(txt, v);
 221    if (u == NULL)
 222       return FALSE;
 223    for (s = relation_head(utt_relation(u, "Segment")); s; s = item_next(s))
 224       label_size++;
 225    if (label_size <= 0)
 226       return FALSE;
 227    label_data = (char **) calloc(label_size, sizeof(char *));
 228    for (i = 0, s = relation_head(utt_relation(u, "Segment")); s; s = item_next(s), i++) {
 229       label_data[i] = (char *) calloc(MAXBUFLEN, sizeof(char));
 230       Flite_HTS_Engine_create_label(f, s, label_data[i]);
 231    }
 232
 233    /* speech synthesis part */
 234    HTS_Engine_synthesize_from_strings(&f->engine, label_data, label_size);
 235    if (wav != NULL) {
 236       fp = fopen(wav, "wb");
 237       HTS_Engine_save_riff(&f->engine, fp);
 238       fclose(fp);
 239    }
 240    HTS_Engine_refresh(&f->engine);
 241
 242    for (i = 0; i < label_size; i++)
 243       free(label_data[i]);
 244    free(label_data);
 245
 246    delete_utterance(u);
 247    UNREGISTER_VOX(v);
 248
 249    return TRUE;
 250 }
 251
 252 /* Flite_HTS_Engine_synthesize: synthesize speech */
 253 HTS_Boolean Flite_HTS_Engine_synthesize_samples_mono16(Flite_HTS_Engine * f, const char *txt,
 254                                                 void** samples, int* sampleCount, int* sampleRate)
 255 {
 256     int i;
 257     cst_voice *v = NULL;
 258     cst_utterance *u = NULL;
 259     cst_item *s = NULL;
 260     char **label_data = NULL;
 261     int label_size = 0;
 262     short* samplePtr = NULL;
 263     HTS_GStreamSet *gss;
 264
 265     if (txt == NULL)
 266         return FALSE;
 267
 268     /* text analysis part */
 269     v = REGISTER_VOX(NULL);
 270     if (v == NULL)
 271         return FALSE;
 272     u = flite_synth_text(txt, v);
 273     if (u == NULL)
 274         return FALSE;
 275     for (s = relation_head(utt_relation(u, "Segment")); s; s = item_next(s))
 276         label_size++;
 277     if (label_size <= 0)
 278         return FALSE;
 279     label_data = (char **) calloc(label_size, sizeof(char *));
 280     for (i = 0, s = relation_head(utt_relation(u, "Segment")); s; s = item_next(s), i++) {
 281         label_data[i] = (char *) calloc(MAXBUFLEN, sizeof(char));
 282         Flite_HTS_Engine_create_label(f, s, label_data[i]);
 283     }
 284
 285     /* speech synthesis part */
 286     HTS_Engine_synthesize_from_strings(&f->engine, label_data, label_size);
 287
 288     gss = &f->engine.gss;
 289     *sampleRate = f->engine.condition.sampling_frequency;
 290     *sampleCount = HTS_GStreamSet_get_total_nsamples(gss);
 291     *samples = malloc(sizeof(short) * *sampleCount);
 292     samplePtr = *samples;
 293
 294     for (i=0; i < *sampleCount; ++i) {
 295         *samplePtr++ = (short) HTS_GStreamSet_get_speech(gss, i);
 296     }
 297
 298     HTS_Engine_refresh(&f->engine);
 299
 300     for (i = 0; i < label_size; i++)
 301         free(label_data[i]);
 302     free(label_data);
 303
 304     delete_utterance(u);
 305     UNREGISTER_VOX(v);
 306
 307     return TRUE;
 308 }
 309
 310
 311 /* Flite_HTS_Engine_clear: free system */
 312 void Flite_HTS_Engine_clear(Flite_HTS_Engine * f)
 313 {
 314    HTS_Engine_clear(&f->engine);
 315 }
 316
 317 typedef struct _Flite_Utterance {
 318    cst_voice *v;
 319    cst_utterance *u;
 320    int nitem;
 321    cst_item **items;
 322 } Flite_Utterance;
 323
 324 /* Flite_Text_Analyzer_initialize: initialize flite front-end */
 325 void Flite_Text_Analyzer_initialize(Flite_Text_Analyzer * analyzer)
 326 {
 327    if (analyzer == NULL)
 328       return;
 329    analyzer->pointer = NULL;
 330 }
 331
 332 /* Flite_Text_Analyzer_analysis: text analysis */
 333 void Flite_Text_Analyzer_analysis(Flite_Text_Analyzer * analyzer, const char *text)
 334 {
 335    int i;
 336    cst_item *s;
 337    Flite_Utterance *fu;
 338
 339    if (analyzer == NULL || text == NULL)
 340       return;
 341
 342    if (analyzer->pointer != NULL)
 343       Flite_Text_Analyzer_clear(analyzer);
 344
 345    /* allocate */
 346    fu = (Flite_Utterance *) malloc(sizeof(Flite_Utterance));
 347
 348    /* create voice */
 349    fu->v = REGISTER_VOX(NULL);
 350    if (fu->v == NULL) {
 351       free(fu);
 352       return;
 353    }
 354
 355    /* create utterance */
 356    fu->u = flite_synth_text(text, fu->v);
 357    if (fu->u == NULL) {
 358       UNREGISTER_VOX(fu->v);
 359       free(fu);
 360       return;
 361    }
 362
 363    /* count number of phonemes */
 364    for (fu->nitem = 0, s = relation_head(utt_relation(fu->u, "Segment")); s; s = item_next(s), fu->nitem++);
 365    if (fu->nitem == 0) {
 366       delete_utterance(fu->u);
 367       UNREGISTER_VOX(fu->v);
 368       free(fu);
 369       return;
 370    }
 371
 372    /* save informations */
 373    fu->items = (cst_item **) malloc(sizeof(cst_item *) * fu->nitem);
 374    for (i = 0, s = relation_head(utt_relation(fu->u, "Segment")); s; s = item_next(s), i++)
 375       fu->items[i] = s;
 376
 377    analyzer->pointer = (void *) fu;
 378 }
 379
 380 /* Flite_Text_Analyzer_get_nphoneme_in_utterance: get number of phonemes */
 381 int Flite_Text_Analyzer_get_nphoneme_in_utterance(Flite_Text_Analyzer * analyzer)
 382 {
 383    Flite_Utterance *fu;
 384
 385    if (analyzer == NULL || analyzer->pointer == NULL)
 386       return 0;
 387
 388    fu = (Flite_Utterance *) analyzer->pointer;
 389    return fu->nitem;
 390 }
 391
 392 /* Flite_Text_Analyzer_get_phoneme: get phoneme identity */
 393 const char *Flite_Text_Analyzer_get_phoneme(Flite_Text_Analyzer * analyzer, int phoneme_index)
 394 {
 395    Flite_Utterance *fu;
 396
 397    if (analyzer == NULL || analyzer->pointer == NULL)
 398       return NULL;
 399    fu = (Flite_Utterance *) analyzer->pointer;
 400    if (phoneme_index < 0 || phoneme_index >= fu->nitem)
 401       return NULL;
 402    return ffeature_string(fu->items[phoneme_index], "name");
 403 }
 404
 405 /* Flite_Text_Analyzer_get_word: get word */
 406 const char *Flite_Text_Analyzer_get_word(Flite_Text_Analyzer * analyzer, int phoneme_index)
 407 {
 408    Flite_Utterance *fu;
 409
 410    if (analyzer == NULL || analyzer->pointer == NULL)
 411       return NULL;
 412    fu = (Flite_Utterance *) analyzer->pointer;
 413    if (phoneme_index < 0 || phoneme_index >= fu->nitem)
 414       return NULL;
 415    return ffeature_string(fu->items[phoneme_index], "R:SylStructure.parent.parent.name");
 416 }
 417
 418 /* Flite_Text_Analyzer_get_nphoneme_in_syllable: get number of phonemes in syllable */
 419 int Flite_Text_Analyzer_get_nphoneme_in_syllable(Flite_Text_Analyzer * analyzer, int phoneme_index)
 420 {
 421    Flite_Utterance *fu;
 422
 423    if (analyzer == NULL || analyzer->pointer == NULL)
 424       return 0;
 425    fu = (Flite_Utterance *) analyzer->pointer;
 426    if (phoneme_index < 0 || phoneme_index >= fu->nitem)
 427       return 0;
 428    return ffeature_int(fu->items[phoneme_index], "R:SylStructure.parent.R:Syllable.syl_numphones");
 429 }
 430
 431 /* Flite_Text_Analayzer_get_nsyllable_in_word: get number of syllables in word */
 432 int Flite_Text_Analyzer_get_nsyllable_in_word(Flite_Text_Analyzer * analyzer, int phoneme_index)
 433 {
 434    Flite_Utterance *fu;
 435
 436    if (analyzer == NULL || analyzer->pointer == NULL)
 437       return 0;
 438    fu = (Flite_Utterance *) analyzer->pointer;
 439    if (phoneme_index < 0 || phoneme_index >= fu->nitem)
 440       return 0;
 441    return ffeature_int(fu->items[phoneme_index], "R:SylStructure.parent.parent.R:Word.word_numsyls");
 442 }
 443
 444 /* Flite_Text_Analyzer_get_nword_in_phrase: get number of words in phrase */
 445 int Flite_Text_Analyzer_get_nword_in_phrase(Flite_Text_Analyzer * analyzer, int phoneme_index)
 446 {
 447    Flite_Utterance *fu;
 448
 449    if (analyzer == NULL || analyzer->pointer == NULL)
 450       return 0;
 451    fu = (Flite_Utterance *) analyzer->pointer;
 452    if (phoneme_index < 0 || phoneme_index >= fu->nitem)
 453       return 0;
 454    return ffeature_int(fu->items[phoneme_index], "R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_words_in_phrase");
 455 }
 456
 457 /* Flite_Text_Analyzer_get_nphrase_in_utterance: get number of phrases in utterance */
 458 int Flite_Text_Analyzer_get_nphrase_in_utterance(Flite_Text_Analyzer * analyzer, int phoneme_index)
 459 {
 460    Flite_Utterance *fu;
 461
 462    if (analyzer == NULL || analyzer->pointer == NULL)
 463       return 0;
 464    fu = (Flite_Utterance *) analyzer->pointer;
 465    if (phoneme_index < 0 || phoneme_index >= fu->nitem)
 466       return 0;
 467    return ffeature_int(fu->items[phoneme_index], "R:SylStructure.parent.parent.R:Phrase.parent.lisp_total_phrases");
 468 }
 469
 470 /* Flite_Text_Analyzer_get_accent: get accent */
 471 int Flite_Text_Analyzer_get_accent(Flite_Text_Analyzer * analyzer, int phoneme_index)
 472 {
 473    Flite_Utterance *fu;
 474
 475    if (analyzer == NULL || analyzer->pointer == NULL)
 476       return 0;
 477    fu = (Flite_Utterance *) analyzer->pointer;
 478    if (phoneme_index < 0 || phoneme_index >= fu->nitem)
 479       return 0;
 480    return ffeature_int(fu->items[phoneme_index], "R:SylStructure.parent.R:Syllable.accented");
 481 }
 482
 483 /* Flite_Text_Analyzer_get_stress: get stress */
 484 int Flite_Text_Analyzer_get_stress(Flite_Text_Analyzer * analyzer, int phoneme_index)
 485 {
 486    Flite_Utterance *fu;
 487
 488    if (analyzer == NULL || analyzer->pointer == NULL)
 489       return 0;
 490    fu = (Flite_Utterance *) analyzer->pointer;
 491    if (phoneme_index < 0 || phoneme_index >= fu->nitem)
 492       return 0;
 493    return ffeature_int(fu->items[phoneme_index], "R:SylStructure.parent.R:Syllable.stress");
 494 }
 495
 496 /* Flite_Text_Analyzer_clear: finalize flite front-end */
 497 void Flite_Text_Analyzer_clear(Flite_Text_Analyzer * analyzer)
 498 {
 499    Flite_Utterance *fu;
 500
 501    if (analyzer == NULL || analyzer->pointer == NULL)
 502       return;
 503
 504    fu = (Flite_Utterance *) analyzer->pointer;
 505    if (fu->items != NULL)
 506       free(fu->items);
 507    if (fu->u != NULL)
 508       delete_utterance(fu->u);
 509    if (fu->v != NULL)
 510       UNREGISTER_VOX(fu->v);
 511    free(fu);
 512
 513    analyzer->pointer = NULL;
 514 }