src/modules/hts_engine/HTS_label.c

   1 /* ----------------------------------------------------------------- */
   2 /*           The HMM-Based Speech Synthesis System (HTS)             */
   3 /*           hts_engine API developed by HTS Working Group           */
   4 /*           http://hts-engine.sourceforge.net/                      */
   5 /* ----------------------------------------------------------------- */
   6 /*                                                                   */
   7 /*  Copyright (c) 2001-2010  Nagoya Institute of Technology          */
   8 /*                           Department of Computer Science          */
   9 /*                                                                   */
  10 /*                2001-2008  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 #ifndef HTS_LABEL_C
  46 #define HTS_LABEL_C
  47
  48 #ifdef __cplusplus
  49 #define HTS_LABEL_C_START extern "C" {
  50 #define HTS_LABEL_C_END   }
  51 #else
  52 #define HTS_LABEL_C_START
  53 #define HTS_LABEL_C_END
  54 #endif                          /* __CPLUSPLUS */
  55
  56 HTS_LABEL_C_START;
  57
  58 #include <stdlib.h>             /* for atof() */
  59 #include <ctype.h>              /* for isgraph(),isdigit() */
  60
  61 /* hts_engine libraries */
  62 #include "HTS_hidden.h"
  63
  64 static HTS_Boolean isdigit_string(char *str)
  65 {
  66    int i;
  67
  68    if (sscanf(str, "%d", &i) == 1)
  69       return TRUE;
  70    else
  71       return FALSE;
  72 }
  73
  74 /* HTS_Label_initialize: initialize label */
  75 void HTS_Label_initialize(HTS_Label * label)
  76 {
  77    label->head = NULL;
  78    label->size = 0;
  79    label->frame_flag = FALSE;
  80    label->speech_speed = 1.0;
  81 }
  82
  83 /* HTS_Label_check_time: check label */
  84 static void HTS_Label_check_time(HTS_Label * label)
  85 {
  86    HTS_LabelString *lstring = label->head;
  87    HTS_LabelString *next = NULL;
  88
  89    if (lstring)
  90       lstring->start = 0.0;
  91    while (lstring) {
  92       next = lstring->next;
  93       if (!next)
  94          break;
  95       if (lstring->end < 0.0 && next->start >= 0.0)
  96          lstring->end = next->start;
  97       else if (lstring->end >= 0.0 && next->start < 0.0)
  98          next->start = lstring->end;
  99       if (lstring->start < 0.0)
 100          lstring->start = -1.0;
 101       if (lstring->end < 0.0)
 102          lstring->end = -1.0;
 103       lstring = next;
 104    }
 105 }
 106
 107 /* HTS_Label_load_from_fn: load label from file name */
 108 void HTS_Label_load_from_fn(HTS_Label * label, int sampling_rate, int fperiod,
 109                             char *fn)
 110 {
 111    FILE *fp = HTS_get_fp(fn, "r");
 112    HTS_Label_load_from_fp(label, sampling_rate, fperiod, fp);
 113    fclose(fp);
 114 }
 115
 116 /* HTS_Label_load_from_fp: load label from file pointer */
 117 void HTS_Label_load_from_fp(HTS_Label * label, int sampling_rate, int fperiod,
 118                             FILE * fp)
 119 {
 120    char buff[HTS_MAXBUFLEN];
 121    HTS_LabelString *lstring = NULL;
 122    double start, end;
 123    const double rate = (double) sampling_rate / ((double) fperiod * 1e+7);
 124
 125    if (label->head || label->size != 0)
 126       HTS_error(1, "HTS_Label_load_from_fp: label is not initialized.\n");
 127    /* parse label file */
 128    while (HTS_get_token(fp, buff)) {
 129       if (!isgraph((int) buff[0]))
 130          break;
 131       label->size++;
 132
 133       if (lstring) {
 134          lstring->next =
 135              (HTS_LabelString *) HTS_calloc(1, sizeof(HTS_LabelString));
 136          lstring = lstring->next;
 137       } else {                  /* first time */
 138          lstring = (HTS_LabelString *) HTS_calloc(1, sizeof(HTS_LabelString));
 139          label->head = lstring;
 140       }
 141       if (isdigit_string(buff)) {       /* has frame infomation */
 142          start = atof(buff);
 143          HTS_get_token(fp, buff);
 144          end = atof(buff);
 145          HTS_get_token(fp, buff);
 146          lstring->start = rate * start;
 147          lstring->end = rate * end;
 148       } else {
 149          lstring->start = -1.0;
 150          lstring->end = -1.0;
 151       }
 152       lstring->next = NULL;
 153       lstring->name = HTS_strdup(buff);
 154    }
 155    HTS_Label_check_time(label);
 156 }
 157
 158 /* HTS_Label_load_from_string: load label from string */
 159 void HTS_Label_load_from_string(HTS_Label * label, int sampling_rate,
 160                                 int fperiod, char *data)
 161 {
 162    char buff[HTS_MAXBUFLEN];
 163    HTS_LabelString *lstring = NULL;
 164    int data_index = 0;          /* data index */
 165    double start, end;
 166    const double rate = (double) sampling_rate / ((double) fperiod * 1e+7);
 167
 168    if (label->head || label->size != 0)
 169       HTS_error(1, "HTS_Label_load_from_fp: label list is not initialized.\n");
 170    /* copy label */
 171    while (HTS_get_token_from_string(data, &data_index, buff)) {
 172       if (!isgraph((int) buff[0]))
 173          break;
 174       label->size++;
 175
 176       if (lstring) {
 177          lstring->next =
 178              (HTS_LabelString *) HTS_calloc(1, sizeof(HTS_LabelString));
 179          lstring = lstring->next;
 180       } else {                  /* first time */
 181          lstring = (HTS_LabelString *) HTS_calloc(1, sizeof(HTS_LabelString));
 182          label->head = lstring;
 183       }
 184       if (isdigit_string(buff)) {       /* has frame infomation */
 185          start = atof(buff);
 186          HTS_get_token_from_string(data, &data_index, buff);
 187          end = atof(buff);
 188          HTS_get_token_from_string(data, &data_index, buff);
 189          lstring->start = rate * start;
 190          lstring->end = rate * end;
 191       } else {
 192          lstring->start = -1.0;
 193          lstring->end = -1.0;
 194       }
 195       lstring->next = NULL;
 196       lstring->name = HTS_strdup(buff);
 197    }
 198    HTS_Label_check_time(label);
 199 }
 200
 201 /* HTS_Label_load_from_string_list: load label from string list */
 202 void HTS_Label_load_from_string_list(HTS_Label * label, int sampling_rate,
 203                                      int fperiod, char **data, int size)
 204 {
 205    char buff[HTS_MAXBUFLEN];
 206    HTS_LabelString *lstring = NULL;
 207    int i;
 208    int data_index;
 209    double start, end;
 210    const double rate = (double) sampling_rate / ((double) fperiod * 1e+7);
 211
 212    if (label->head || label->size != 0)
 213       HTS_error(1, "HTS_Label_load_from_fp: label list is not initialized.\n");
 214    /* copy label */
 215    for (i = 0; i < size; i++) {
 216       if (!isgraph((int) data[i][0]))
 217          break;
 218       label->size++;
 219
 220       if (lstring) {
 221          lstring->next =
 222              (HTS_LabelString *) HTS_calloc(1, sizeof(HTS_LabelString));
 223          lstring = lstring->next;
 224       } else {                  /* first time */
 225          lstring = (HTS_LabelString *) HTS_calloc(1, sizeof(HTS_LabelString));
 226          label->head = lstring;
 227       }
 228       data_index = 0;
 229       if (isdigit_string(data[i])) {    /* has frame infomation */
 230          HTS_get_token_from_string(data[i], &data_index, buff);
 231          start = atof(buff);
 232          HTS_get_token_from_string(data[i], &data_index, buff);
 233          end = atof(buff);
 234          HTS_get_token_from_string(data[i], &data_index, buff);
 235          lstring->name = HTS_strdup(&buff[data_index]);
 236          lstring->start = rate * start;
 237          lstring->end = rate * end;
 238       } else {
 239          lstring->start = -1.0;
 240          lstring->end = -1.0;
 241          lstring->name = HTS_strdup(data[i]);
 242       }
 243       lstring->next = NULL;
 244    }
 245    HTS_Label_check_time(label);
 246 }
 247
 248 /* HTS_Label_set_frame_specified_flag: set frame specified flag */
 249 void HTS_Label_set_frame_specified_flag(HTS_Label * label, HTS_Boolean i)
 250 {
 251    label->frame_flag = i;
 252 }
 253
 254 /* HTS_Label_set_speech_speed: set speech speed rate */
 255 void HTS_Label_set_speech_speed(HTS_Label * label, double f)
 256 {
 257    if (f > 0.0 && f <= 10.0)
 258       label->speech_speed = f;
 259 }
 260
 261 /* HTS_Label_get_size: get number of label string */
 262 int HTS_Label_get_size(HTS_Label * label)
 263 {
 264    return label->size;
 265 }
 266
 267 /* HTS_Label_get_string: get label string */
 268 char *HTS_Label_get_string(HTS_Label * label, int string_index)
 269 {
 270    HTS_LabelString *lstring = label->head;
 271
 272    while (string_index-- && lstring)
 273       lstring = lstring->next;
 274    if (!lstring)
 275       return NULL;
 276    return lstring->name;
 277 }
 278
 279 /* HTS_Label_get_frame_specified_flag: get frame specified flag */
 280 HTS_Boolean HTS_Label_get_frame_specified_flag(HTS_Label * label)
 281 {
 282    return label->frame_flag;
 283 }
 284
 285 /* HTS_Label_get_start_frame: get start frame */
 286 double HTS_Label_get_start_frame(HTS_Label * label, int string_index)
 287 {
 288    HTS_LabelString *lstring = label->head;
 289
 290    while (string_index-- && lstring)
 291       lstring = lstring->next;
 292    if (!lstring)
 293       return -1.0;
 294    return lstring->start;
 295 }
 296
 297 /* HTS_Label_get_end_frame: get end frame */
 298 double HTS_Label_get_end_frame(HTS_Label * label, int string_index)
 299 {
 300    HTS_LabelString *lstring = label->head;
 301
 302    while (string_index-- && lstring)
 303       lstring = lstring->next;
 304    if (!lstring)
 305       return -1.0;
 306    return lstring->end;
 307 }
 308
 309 /* HTS_Label_get_speech_speed: get speech speed rate */
 310 double HTS_Label_get_speech_speed(HTS_Label * label)
 311 {
 312    return label->speech_speed;
 313 }
 314
 315 /* HTS_Label_clear: free label */
 316 void HTS_Label_clear(HTS_Label * label)
 317 {
 318    HTS_LabelString *lstring, *next_lstring;
 319
 320    for (lstring = label->head; lstring; lstring = next_lstring) {
 321       next_lstring = lstring->next;
 322       HTS_free(lstring->name);
 323       HTS_free(lstring);
 324    }
 325    HTS_Label_initialize(label);
 326 }
 327
 328 HTS_LABEL_C_END;
 329
 330 #endif                          /* !HTS_LABEL_C */