void TextToSpeech::load(
char *dn_mecab, char *fn_ms_dur, char *fn_ts_dur,
char *fn_ms_mgc, char *fn_ts_mgc, char **fn_ws_mgc, int num_ws_mgc,
char *fn_ms_lf0, char *fn_ts_lf0, char **fn_ws_lf0, int num_ws_lf0,
char *fn_ms_lpf, char *fn_ts_lpf, char **fn_ws_lpf, int num_ws_lpf,
char *fn_ms_gvm, char *fn_ts_gvm, char *fn_ms_gvl, char *fn_ts_gvl,
char *fn_ms_gvf, char *fn_ts_gvf, char *fn_gv_switch)
{
Mecab_load(&open_jtalk_.mecab, dn_mecab);
HTS_Engine_load_duration_from_fn(&open_jtalk_.engine, &fn_ms_dur, &fn_ts_dur, 1);
HTS_Engine_load_parameter_from_fn(&open_jtalk_.engine, &fn_ms_mgc, &fn_ts_mgc, fn_ws_mgc, 0, FALSE, num_ws_mgc, 1);
HTS_Engine_load_parameter_from_fn(&open_jtalk_.engine, &fn_ms_lf0, &fn_ts_lf0, fn_ws_lf0, 1, TRUE, num_ws_lf0, 1);
if (HTS_Engine_get_nstream(&open_jtalk_.engine) == 3)
HTS_Engine_load_parameter_from_fn(&open_jtalk_.engine, &fn_ms_lpf, &fn_ts_lpf, fn_ws_lpf, 2, FALSE, num_ws_lpf, 1);
if (fn_ms_gvm != NULL) {
if (fn_ts_gvm != NULL)
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine, &fn_ms_gvm, &fn_ts_gvm, 0, 1);
else
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine, &fn_ms_gvm, NULL, 0, 1);
}
if (fn_ms_gvl != NULL) {
if (fn_ts_gvl != NULL)
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine, &fn_ms_gvl, &fn_ts_gvl, 1, 1);
else
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine, &fn_ms_gvl, NULL, 1, 1);
}
if (HTS_Engine_get_nstream(&open_jtalk_.engine) == 3 && fn_ms_gvf != NULL) {
if (fn_ts_gvf != NULL)
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine, &fn_ms_gvf, &fn_ts_gvf, 2, 1);
else
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine, &fn_ms_gvf, NULL, 2, 1);
}
if (fn_gv_switch != NULL)
HTS_Engine_load_gv_switch_from_fn(&open_jtalk_.engine, fn_gv_switch);
}
void TextToSpeech::load(
const char *dn_mecab, const char *fn_ms_dur, const char *fn_ts_dur,
const char *fn_ms_mgc, const char *fn_ts_mgc, const char **fn_ws_mgc, int num_ws_mgc,
const char *fn_ms_lf0, const char *fn_ts_lf0, const char **fn_ws_lf0, int num_ws_lf0,
const char *fn_ms_lpf, const char *fn_ts_lpf, const char **fn_ws_lpf, int num_ws_lpf,
const char *fn_ms_gvm, const char *fn_ts_gvm, const char *fn_ms_gvl, const char *fn_ts_gvl,
const char *fn_ms_gvf, const char *fn_ts_gvf, const char *fn_gv_switch)
{
Mecab_load(&open_jtalk_.mecab, const_cast<char *>(dn_mecab));
HTS_Engine_load_duration_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_dur),
const_cast<char **>(&fn_ts_dur), 1);
HTS_Engine_load_parameter_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_mgc),
const_cast<char **>(&fn_ts_mgc),
const_cast<char **>(fn_ws_mgc), 0, FALSE, num_ws_mgc, 1);
HTS_Engine_load_parameter_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_lf0),
const_cast<char **>(&fn_ts_lf0),
const_cast<char **>(fn_ws_lf0), 1, TRUE, num_ws_lf0, 1);
if (HTS_Engine_get_nstream(&open_jtalk_.engine) == 3)
HTS_Engine_load_parameter_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_lpf),
const_cast<char **>(&fn_ts_lpf),
const_cast<char **>(fn_ws_lpf), 2, FALSE, num_ws_lpf, 1);
if (fn_ms_gvm != NULL) {
if (fn_ts_gvm != NULL)
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_gvm),
const_cast<char **>(&fn_ts_gvm), 0, 1);
else
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_gvm), NULL, 0, 1);
}
if (fn_ms_gvl != NULL) {
if (fn_ts_gvl != NULL)
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_gvl),
const_cast<char **>(&fn_ts_gvl), 1, 1);
else
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_gvl), NULL, 1, 1);
}
if (HTS_Engine_get_nstream(&open_jtalk_.engine) == 3 && fn_ms_gvf != NULL) {
if (fn_ts_gvf != NULL)
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_gvf),
const_cast<char **>(&fn_ts_gvf), 2, 1);
else
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_gvf), NULL, 2, 1);
}
if (fn_gv_switch != NULL)
HTS_Engine_load_gv_switch_from_fn(&open_jtalk_.engine,
const_cast<char *>(fn_gv_switch));
}
int OpenJTalk_Create(OpenJTalk** openjtalk,Mecab* mecab,int argc,const char **argv)
{
/* file names of models */
char *fn_ms_dur = NULL;
char *fn_ms_mgc = NULL;
char *fn_ms_lf0 = NULL;
char *fn_ms_lpf = NULL;
/* file names of trees */
char *fn_ts_dur = NULL;
char *fn_ts_mgc = NULL;
char *fn_ts_lf0 = NULL;
char *fn_ts_lpf = NULL;
/* file names of windows */
int num_ws_mgc = 0, num_ws_lf0 = 0, num_ws_lpf = 0;
/* file names of global variance */
char *fn_ms_gvm = NULL;
char *fn_ms_gvl = NULL;
char *fn_ms_gvf = NULL;
/* file names of global variance trees */
char *fn_ts_gvm = NULL;
char *fn_ts_gvl = NULL;
char *fn_ts_gvf = NULL;
/* file names of global variance switch */
char *fn_gv_switch = NULL;
/* global parameter */
int sampling_rate = 16000;
int fperiod = 80;
double alpha = 0.42;
int stage = 0; /* gamma = -1.0/stage */
double beta = 0.0;
int audio_buff_size = 1600;
double uv_threshold = 0.5;
double gv_weight_mgc = 1.0;
double gv_weight_lf0 = 1.0;
double gv_weight_lpf = 1.0;
HTS_Boolean use_log_gain = FALSE;
HTS_Boolean use_lpf = FALSE;
/* alloc OpenJTalk ythis pointer! */
*openjtalk = (struct OpenJTalk*)malloc(sizeof(struct OpenJTalk));
(*openjtalk)->mecab = mecab;
(*openjtalk)->errorout[0] = '\0';
/* delta window handler for mel-cepstrum */
(*openjtalk)->fn_ws_mgc = (char **) calloc(argc, sizeof(char *));
/* delta window handler for log f0 */
(*openjtalk)->fn_ws_lf0 = (char **) calloc(argc, sizeof(char *));
/* delta window handler for low-pass filter */
(*openjtalk)->fn_ws_lpf = (char **) calloc(argc, sizeof(char *));
/* parse command line */
if (argc == 1)
{
sprintf((*openjtalk)->errorout,"option error!");
return -1;
}
/* read command */
while (--argc) {
if (**++argv == '-') {
switch (*(*argv + 1)) {
case 't':
switch (*(*argv + 2)) {
case 'd':
fn_ts_dur = (char*)*(++argv);
break;
case 'm':
fn_ts_mgc = (char*)*(++argv);
break;
case 'f':
case 'p':
fn_ts_lf0 = (char*)*(++argv);
break;
case 'l':
fn_ts_lpf = (char*)*(++argv);
break;
default:
sprintf((*openjtalk)->errorout,"Invalid option '-t%c'.\n", *(*argv + 2));
return 0;
}
--argc;
break;
case 'm':
switch (*(*argv + 2)) {
case 'd':
fn_ms_dur = (char*)*(++argv);
break;
case 'm':
fn_ms_mgc = (char*)*(++argv);
break;
case 'f':
case 'p':
fn_ms_lf0 = (char*)*(++argv);
break;
case 'l':
fn_ms_lpf = (char*)*(++argv);
break;
default:
sprintf((*openjtalk)->errorout,"Invalid option '-m%c'.\n", *(*argv + 2));
return 0;
}
--argc;
break;
case 'd':
switch (*(*argv + 2)) {
case 'm':
(*openjtalk)->fn_ws_mgc[num_ws_mgc++] = (char*)*(++argv);
break;
case 'f':
case 'p':
(*openjtalk)->fn_ws_lf0[num_ws_lf0++] = (char*)*(++argv);
break;
case 'l':
(*openjtalk)->fn_ws_lpf[num_ws_lpf++] = (char*)*(++argv);
break;
default:
sprintf((*openjtalk)->errorout,"Invalid option '-d%c'.\n", *(*argv + 2));
return 0;
}
--argc;
break;
case 's':
sampling_rate = atoi(*++argv);
--argc;
break;
case 'p':
fperiod = atoi(*++argv);
--argc;
break;
case 'a':
alpha = atof(*++argv);
--argc;
break;
case 'g':
stage = atoi(*++argv);
--argc;
break;
case 'l':
use_log_gain = TRUE;
break;
case 'b':
beta = atof(*++argv);
--argc;
break;
case 'u':
uv_threshold = atof(*++argv);
--argc;
break;
case 'e':
switch (*(*argv + 2)) {
case 'm':
fn_ts_gvm = (char*)*(++argv);
break;
case 'f':
case 'p':
fn_ts_gvl = (char*)*(++argv);
break;
case 'l':
fn_ts_gvf = (char*)*(++argv);
break;
default:
sprintf((*openjtalk)->errorout,"Invalid option '-e%c'.\n", *(*argv + 2));
return 0;
}
--argc;
break;
case 'c':
switch (*(*argv + 2)) {
case 'm':
fn_ms_gvm = (char*)*(++argv);
break;
case 'f':
case 'p':
fn_ms_gvl = (char*)*(++argv);
break;
case 'l':
fn_ms_gvf = (char*)*(++argv);
break;
default:
sprintf((*openjtalk)->errorout,"Invalid option '-c%c'.\n", *(*argv + 2));
return 0;
}
--argc;
break;
case 'j':
switch (*(*argv + 2)) {
case 'm':
gv_weight_mgc = atof(*(++argv));
break;
case 'f':
case 'p':
gv_weight_lf0 = atof(*(++argv));
break;
case 'l':
gv_weight_lpf = atof(*(++argv));
break;
default:
sprintf((*openjtalk)->errorout,"Invalid option '-j%c'.\n", *(*argv + 2));
return 0;
}
--argc;
break;
case 'k':
fn_gv_switch = (char*)*++argv;
--argc;
break;
case 'z':
audio_buff_size = atoi(*++argv);
--argc;
break;
default:
sprintf((*openjtalk)->errorout,"Invalid option '-%c'.\n", *(*argv + 1));
return 0;
}
}
}
/* number of models,trees check */
if (fn_ms_dur == NULL || fn_ms_mgc == NULL || fn_ms_lf0 == NULL ||
fn_ts_dur == NULL || fn_ts_mgc == NULL || fn_ts_lf0 == NULL ||
num_ws_mgc <= 0 || num_ws_lf0 <= 0) {
sprintf((*openjtalk)->errorout,"Specify models (trees) for each parameter.\n");
return 0;
}
if (fn_ms_lpf != NULL && fn_ts_lpf != NULL && num_ws_lpf > 0)
use_lpf = TRUE;
/* initialize */
NJD_initialize(&(*openjtalk)->njd);
JPCommon_initialize(&(*openjtalk)->jpcommon);
if (use_lpf)
HTS_Engine_initialize(&(*openjtalk)->engine, 3);
else
HTS_Engine_initialize(&(*openjtalk)->engine, 2);
HTS_Engine_set_sampling_rate(&(*openjtalk)->engine, sampling_rate);
HTS_Engine_set_fperiod(&(*openjtalk)->engine, fperiod);
HTS_Engine_set_alpha(&(*openjtalk)->engine, alpha);
HTS_Engine_set_gamma(&(*openjtalk)->engine, stage);
HTS_Engine_set_log_gain(&(*openjtalk)->engine, use_log_gain);
HTS_Engine_set_beta(&(*openjtalk)->engine, beta);
HTS_Engine_set_audio_buff_size(&(*openjtalk)->engine, audio_buff_size);
HTS_Engine_set_msd_threshold(&(*openjtalk)->engine, 1, uv_threshold);
HTS_Engine_set_gv_weight(&(*openjtalk)->engine, 0, gv_weight_mgc);
HTS_Engine_set_gv_weight(&(*openjtalk)->engine, 1, gv_weight_lf0);
if (use_lpf)
{
HTS_Engine_set_gv_weight(&(*openjtalk)->engine, 2, gv_weight_lpf);
}
/* Load! */
HTS_Engine_load_duration_from_fn(&(*openjtalk)->engine, &fn_ms_dur, &fn_ts_dur, 1);
HTS_Engine_load_parameter_from_fn(&(*openjtalk)->engine, &fn_ms_mgc, &fn_ts_mgc, (*openjtalk)->fn_ws_mgc, 0,
FALSE, num_ws_mgc, 1);
HTS_Engine_load_parameter_from_fn(&(*openjtalk)->engine, &fn_ms_lf0, &fn_ts_lf0, (*openjtalk)->fn_ws_lf0, 1,
TRUE, num_ws_lf0, 1);
if (HTS_Engine_get_nstream(&(*openjtalk)->engine) == 3)
HTS_Engine_load_parameter_from_fn(&(*openjtalk)->engine, &fn_ms_lpf, &fn_ts_lpf, (*openjtalk)->fn_ws_lpf, 2,
FALSE, num_ws_lpf, 1);
if (fn_ms_gvm != NULL) {
if (fn_ts_gvm != NULL)
HTS_Engine_load_gv_from_fn(&(*openjtalk)->engine, &fn_ms_gvm, &fn_ts_gvm, 0, 1);
else
HTS_Engine_load_gv_from_fn(&(*openjtalk)->engine, &fn_ms_gvm, NULL, 0, 1);
}
if (fn_ms_gvl != NULL) {
if (fn_ts_gvl != NULL)
HTS_Engine_load_gv_from_fn(&(*openjtalk)->engine, &fn_ms_gvl, &fn_ts_gvl, 1, 1);
else
HTS_Engine_load_gv_from_fn(&(*openjtalk)->engine, &fn_ms_gvl, NULL, 1, 1);
}
if (HTS_Engine_get_nstream(&(*openjtalk)->engine) == 3 && fn_ms_gvf != NULL) {
if (fn_ts_gvf != NULL)
HTS_Engine_load_gv_from_fn(&(*openjtalk)->engine, &fn_ms_gvf, &fn_ts_gvf, 2, 1);
else
HTS_Engine_load_gv_from_fn(&(*openjtalk)->engine, &fn_ms_gvf, NULL, 2, 1);
}
if (fn_gv_switch != NULL)
HTS_Engine_load_gv_switch_from_fn(&(*openjtalk)->engine, fn_gv_switch);
return 1;
}
static void load_hts_engine_data(const SMap *data,
SHTSEngineMESynthUttProc105 *HTSsynth,
const char *voice_base_path, s_erc *error)
{
const SMap *tmp;
const SList *trees;
const SList *pdfs;
const SList *windows;
int num;
int i;
char **ctrees;
char **cpdfs;
char **cwindows;
int num_win;
const char *gv;
const char *gv_tree;
const char *gv_switch;
const char *pd_filter;
HTS_Engine *engine = &(HTSsynth->engine);
S_CLR_ERR(error);
/* duration */
tmp = S_MAP(SMapGetObjectDef(data, "duration", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (tmp == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'duration' HTS Engine data");
return;
}
trees = S_LIST(SMapGetObjectDef(tmp, "trees", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (trees == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'duration:trees' HTS Engine data");
return;
}
pdfs = S_LIST(SMapGetObjectDef(tmp, "pdfs", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (pdfs == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'duration:pdfs' HTS Engine data");
return;
}
get_trees_pdfs(trees, pdfs, &ctrees, &cpdfs, &num, voice_base_path, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"get_trees_pdfs\" failed"))
return;
HTS_Engine_load_duration_from_fn(engine, cpdfs, ctrees, 1);
for (i = 0; i < num; i++)
{
S_FREE(ctrees[i]);
S_FREE(cpdfs[i]);
}
S_FREE(ctrees);
S_FREE(cpdfs);
/* log F0 */
tmp = S_MAP(SMapGetObjectDef(data, "log F0", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (tmp == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'log F0' HTS Engine data");
return;
}
trees = S_LIST(SMapGetObjectDef(tmp, "trees", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (trees == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'log F0:trees' HTS Engine data");
return;
}
pdfs = S_LIST(SMapGetObjectDef(tmp, "pdfs", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (pdfs == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'log F0:pdfs' HTS Engine data");
return;
}
get_trees_pdfs(trees, pdfs, &ctrees, &cpdfs, &num, voice_base_path, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"get_trees_pdfs\" failed"))
return;
windows = S_LIST(SMapGetObjectDef(tmp, "windows", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (windows == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'log F0:windows' HTS Engine data");
return;
}
get_windows(windows, &cwindows, &num_win, voice_base_path, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"get_windows\" failed"))
return;
/* log f0 is stream 1, and msd_flag is TRUE */
HTS_Engine_load_parameter_from_fn(engine, cpdfs, ctrees, cwindows,
1, TRUE, num_win, 1);
for (i = 0; i < num; i++)
{
S_FREE(ctrees[i]);
S_FREE(cpdfs[i]);
}
S_FREE(ctrees);
S_FREE(cpdfs);
for (i = 0; i < num_win; i++)
S_FREE(cwindows[i]);
S_FREE(cwindows);
/* log f0 gv */
gv = SMapGetStringDef(tmp, "gv-lf0", NULL, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetStringDef\" failed"))
return;
/* log f0 gv tree */
gv_tree = SMapGetStringDef(tmp, "tree-gv-lf0", NULL, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetStringDef\" failed"))
return;
if (gv != NULL)
{
char *combined_path_gv;
char *combined_path_gv_tree;
/* get data path, the one in the config file may be relative
* to the voice base path
*/
combined_path_gv = s_path_combine(voice_base_path, gv,
error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"s_path_combine\" failed"))
return;
if (gv_tree != NULL)
{
/* get data path, the one in the config file may be relative
* to the voice base path
*/
combined_path_gv_tree = s_path_combine(voice_base_path, gv_tree,
error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"s_path_combine\" failed"))
{
S_FREE(combined_path_gv);
return;
}
}
else
{
combined_path_gv_tree = NULL;
}
HTS_Engine_load_gv_from_fn(engine, (char**)&combined_path_gv,
(char**)&combined_path_gv_tree, 1, 1);
S_FREE(combined_path_gv);
if (combined_path_gv_tree != NULL)
S_FREE(combined_path_gv_tree);
}
/* spectrum */
tmp = S_MAP(SMapGetObjectDef(data, "spectrum", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (tmp == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'spectrum' HTS Engine data");
return;
}
trees = S_LIST(SMapGetObjectDef(tmp, "trees", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (trees == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'spectrum:trees' HTS Engine data");
return;
}
pdfs = S_LIST(SMapGetObjectDef(tmp, "pdfs", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (pdfs == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'spectrum:pdfs' HTS Engine data");
return;
}
get_trees_pdfs(trees, pdfs, &ctrees, &cpdfs, &num, voice_base_path, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"get_trees_pdfs\" failed"))
return;
windows = S_LIST(SMapGetObjectDef(tmp, "windows", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (windows == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'spectrum:windows' HTS Engine data");
return;
}
get_windows(windows, &cwindows, &num_win, voice_base_path, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"get_windows\" failed"))
return;
/* spectrum is stream 0, and msd_flag is FALSE */
HTS_Engine_load_parameter_from_fn(engine, cpdfs, ctrees, cwindows,
0, FALSE, num_win, 1);
for (i = 0; i < num; i++)
{
S_FREE(ctrees[i]);
S_FREE(cpdfs[i]);
}
S_FREE(ctrees);
S_FREE(cpdfs);
for (i = 0; i < num_win; i++)
S_FREE(cwindows[i]);
S_FREE(cwindows);
/* spectrum gv */
gv = SMapGetStringDef(tmp, "gv-mgc", NULL, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetStringDef\" failed"))
return;
/* spectrum gv tree */
gv_tree = SMapGetStringDef(tmp, "tree-gv-mgc", NULL, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetStringDef\" failed"))
return;
if (gv != NULL)
{
char *combined_path_gv;
char *combined_path_gv_tree;
/* get data path, the one in the config file may be relative
* to the voice base path
*/
combined_path_gv = s_path_combine(voice_base_path, gv,
error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"s_path_combine\" failed"))
return;
if (gv_tree != NULL)
{
/* get data path, the one in the config file may be relative
* to the voice base path
*/
combined_path_gv_tree = s_path_combine(voice_base_path, gv_tree,
error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"s_path_combine\" failed"))
{
S_FREE(combined_path_gv);
return;
}
}
else
{
combined_path_gv_tree = NULL;
}
HTS_Engine_load_gv_from_fn(engine, (char**)&combined_path_gv,
(char**)&combined_path_gv_tree, 0, 1);
S_FREE(combined_path_gv);
if (combined_path_gv_tree != NULL)
S_FREE(combined_path_gv_tree);
}
/* band strengths */
if (HTSsynth->me == TRUE)
{
const char *me_filter;
tmp = S_MAP(SMapGetObjectDef(data, "strengths", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (tmp == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'strengths' HTS Engine data");
return;
}
trees = S_LIST(SMapGetObjectDef(tmp, "trees", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (trees == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'strengths:trees' HTS Engine data");
return;
}
pdfs = S_LIST(SMapGetObjectDef(tmp, "pdfs", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (pdfs == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'strengths:pdfs' HTS Engine data");
return;
}
get_trees_pdfs(trees, pdfs, &ctrees, &cpdfs, &num, voice_base_path, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"get_trees_pdfs\" failed"))
return;
windows = S_LIST(SMapGetObjectDef(tmp, "windows", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (windows == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'strengths:windows' HTS Engine data");
return;
}
get_windows(windows, &cwindows, &num_win, voice_base_path, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"get_windows\" failed"))
return;
/* strengths is stream 2, and msd_flag is FALSE */
HTS_Engine_load_parameter_from_fn(engine, cpdfs, ctrees, cwindows,
2, FALSE, num_win, 1);
for (i = 0; i < num; i++)
{
S_FREE(ctrees[i]);
S_FREE(cpdfs[i]);
}
S_FREE(ctrees);
S_FREE(cpdfs);
for (i = 0; i < num_win; i++)
S_FREE(cwindows[i]);
S_FREE(cwindows);
/* strengths gv */
gv = SMapGetStringDef(tmp, "gv-str", NULL, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetStringDef\" failed"))
return;
/* strengths gv tree */
gv_tree = SMapGetStringDef(tmp, "tree-gv-str", NULL, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetStringDef\" failed"))
return;
if (gv != NULL)
{
char *combined_path_gv;
char *combined_path_gv_tree;
/* get data path, the one in the config file may be relative
* to the voice base path
*/
combined_path_gv = s_path_combine(voice_base_path, gv,
error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"s_path_combine\" failed"))
return;
if (gv_tree != NULL)
{
/* get data path, the one in the config file may be relative
* to the voice base path
*/
combined_path_gv_tree = s_path_combine(voice_base_path, gv_tree,
error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"s_path_combine\" failed"))
{
S_FREE(combined_path_gv);
return;
}
}
else
{
combined_path_gv_tree = NULL;
}
HTS_Engine_load_gv_from_fn(engine, (char**)&combined_path_gv,
(char**)&combined_path_gv_tree, 2, 1);
S_FREE(combined_path_gv);
if (combined_path_gv_tree != NULL)
S_FREE(combined_path_gv_tree);
}
/* mixed excitation filter */
me_filter = SMapGetStringDef(tmp, "mixed excitation filter", NULL, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetStringDef\" failed"))
return;
if (me_filter != NULL)
{
char *combined_path;
/* get data path, the one in the config file may be relative
* to the voice base path
*/
combined_path = s_path_combine(voice_base_path, me_filter,
error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"s_path_combine\" failed"))
return;
HTS_Engine_load_me_filter_from_fn(combined_path, &(HTSsynth->me_filter),
&(HTSsynth->me_num_filters), &(HTSsynth->me_filter_order));
S_FREE(combined_path);
/* allocate memory for other filter buffers */
filter_constructor(HTSsynth, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"filter_constructor\" failed"))
return;
}
else
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'strengths:mixed excitation filter' HTS Engine data");
return;
}
}
/* gv switch */
gv_switch = SMapGetStringDef(data, "gv-switch", NULL, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (gv_switch != NULL)
{
char *combined_path;
/* get data path, the one in the config file may be relative
* to the voice base path
*/
combined_path = s_path_combine(voice_base_path, gv_switch,
error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"s_path_combine\" failed"))
return;
HTS_Engine_load_gv_switch_from_fn(engine, combined_path);
S_FREE(combined_path);
}
/* pulse dispersion filter */
pd_filter = SMapGetStringDef(data, "pulse dispersion filter", NULL, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (pd_filter != NULL)
{
char *combined_path;
/* get data path, the one in the config file may be relative
* to the voice base path
*/
combined_path = s_path_combine(voice_base_path, pd_filter,
error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"s_path_combine\" failed"))
return;
HTS_Engine_load_pd_filter_from_fn(combined_path, &(HTSsynth->pd_filter),
&(HTSsynth->pd_filter_order));
S_FREE(combined_path);
}
}
static void load_hts_engine_data(const SMap *data, HTS_Engine *engine,
const char *voice_base_path, s_erc *error)
{
const SMap *tmp;
const SList *trees;
const SList *pdfs;
const SList *windows;
int num;
int i;
char **ctrees;
char **cpdfs;
char **cwindows;
int num_win;
const char *gv;
const char *gv_switch;
S_CLR_ERR(error);
/* duration */
tmp = S_MAP(SMapGetObjectDef(data, "duration", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (tmp == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'duration' HTS Engine data");
return;
}
trees = S_LIST(SMapGetObjectDef(tmp, "trees", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (trees == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'duration:trees' HTS Engine data");
return;
}
pdfs = S_LIST(SMapGetObjectDef(tmp, "pdfs", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (pdfs == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'duration:pdfs' HTS Engine data");
return;
}
get_trees_pdfs(trees, pdfs, &ctrees, &cpdfs, &num, voice_base_path, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"get_trees_pdfs\" failed"))
return;
HTS_Engine_load_duration_from_fn(engine, cpdfs, ctrees, 1);
for (i = 0; i < num; i++)
{
S_FREE(ctrees[i]);
S_FREE(cpdfs[i]);
}
S_FREE(ctrees);
S_FREE(cpdfs);
/* log F0 */
tmp = S_MAP(SMapGetObjectDef(data, "log F0", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (tmp == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'log F0' HTS Engine data");
return;
}
trees = S_LIST(SMapGetObjectDef(tmp, "trees", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (trees == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'log F0:trees' HTS Engine data");
return;
}
pdfs = S_LIST(SMapGetObjectDef(tmp, "pdfs", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (pdfs == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'log F0:pdfs' HTS Engine data");
return;
}
get_trees_pdfs(trees, pdfs, &ctrees, &cpdfs, &num, voice_base_path, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"get_trees_pdfs\" failed"))
return;
windows = S_LIST(SMapGetObjectDef(tmp, "windows", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (windows == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'log F0:windows' HTS Engine data");
return;
}
get_windows(windows, &cwindows, &num_win, voice_base_path, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"get_windows\" failed"))
return;
/* log f0 is stream 1, and msd_flag is TRUE */
HTS_Engine_load_parameter_from_fn(engine, cpdfs, ctrees, cwindows,
1, TRUE, num_win, 1);
for (i = 0; i < num; i++)
{
S_FREE(ctrees[i]);
S_FREE(cpdfs[i]);
}
S_FREE(ctrees);
S_FREE(cpdfs);
for (i = 0; i < num_win; i++)
S_FREE(cwindows[i]);
S_FREE(cwindows);
/* log f0 gv */
gv = SMapGetStringDef(tmp, "gv-lf0", NULL, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetStringDef\" failed"))
return;
if (gv != NULL)
{
char *combined_path;
/* get data path, the one in the config file may be relative
* to the voice base path
*/
combined_path = s_path_combine(voice_base_path, gv,
error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"s_path_combine\" failed"))
return;
HTS_Engine_load_gv_from_fn(engine, (char**)&combined_path, NULL, 1, 1);
S_FREE(combined_path);
}
/* spectrum */
tmp = S_MAP(SMapGetObjectDef(data, "spectrum", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (tmp == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'spectrum' HTS Engine data");
return;
}
trees = S_LIST(SMapGetObjectDef(tmp, "trees", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (trees == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'spectrum:trees' HTS Engine data");
return;
}
pdfs = S_LIST(SMapGetObjectDef(tmp, "pdfs", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (pdfs == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'spectrum:pdfs' HTS Engine data");
return;
}
get_trees_pdfs(trees, pdfs, &ctrees, &cpdfs, &num, voice_base_path, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"get_trees_pdfs\" failed"))
return;
windows = S_LIST(SMapGetObjectDef(tmp, "windows", NULL, error));
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (windows == NULL)
{
S_CTX_ERR(error, S_FAILURE,
"load_hts_engine_data",
"Failed to find 'spectrum:windows' HTS Engine data");
return;
}
get_windows(windows, &cwindows, &num_win, voice_base_path, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"get_windows\" failed"))
return;
/* spectrum is stream 0, and msd_flag is FALSE */
HTS_Engine_load_parameter_from_fn(engine, cpdfs, ctrees, cwindows,
0, FALSE, num_win, 1);
for (i = 0; i < num; i++)
{
S_FREE(ctrees[i]);
S_FREE(cpdfs[i]);
}
S_FREE(ctrees);
S_FREE(cpdfs);
for (i = 0; i < num_win; i++)
S_FREE(cwindows[i]);
S_FREE(cwindows);
/* spectrum gv */
gv = SMapGetStringDef(tmp, "gv-mgc", NULL, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetStringDef\" failed"))
return;
if (gv != NULL)
{
char *combined_path;
/* get data path, the one in the config file may be relative
* to the voice base path
*/
combined_path = s_path_combine(voice_base_path, gv,
error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"s_path_combine\" failed"))
return;
HTS_Engine_load_gv_from_fn(engine, (char**)&combined_path, NULL, 0, 1);
S_FREE(combined_path);
}
/* gv switch */
gv_switch = SMapGetStringDef(data, "gv-switch", NULL, error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"SMapGetObjectDef\" failed"))
return;
if (gv_switch != NULL)
{
char *combined_path;
/* get data path, the one in the config file may be relative
* to the voice base path
*/
combined_path = s_path_combine(voice_base_path, gv_switch,
error);
if (S_CHK_ERR(error, S_CONTERR,
"load_hts_engine_data",
"Call to \"s_path_combine\" failed"))
return;
HTS_Engine_load_gv_switch_from_fn(engine, combined_path);
S_FREE(combined_path);
}
}
int main(int argc, char **argv)
{
int i;
double f;
char *labfn = NULL;
size_t labfn_size=0;
char *guifn = NULL;
FILE *guifp = NULL;
FILE *rawfp = NULL;
/* number of speakers for interpolation */
int num_interp = 0;
double *rate_interp = NULL;
/* file names of models */
char **fn_ms_dur;
char **fn_ms_mgc;
char **fn_ms_lf0;
char **fn_ms_lpf;
/* number of each models for interpolation */
int num_ms_dur = 0, num_ms_mgc = 0, num_ms_lf0 = 0, num_ms_lpf = 0;
/* file names of trees */
char **fn_ts_dur;
char **fn_ts_mgc;
char **fn_ts_lf0;
char **fn_ts_lpf;
/* number of each trees for interpolation */
int num_ts_dur = 0, num_ts_mgc = 0, num_ts_lf0 = 0, num_ts_lpf = 0;
/* file names of windows */
char **fn_ws_mgc;
char **fn_ws_lf0;
char **fn_ws_lpf;
int num_ws_mgc = 0, num_ws_lf0 = 0, num_ws_lpf = 0;
/* file names of global variance */
char **fn_ms_gvm = NULL;
char **fn_ms_gvl = NULL;
char **fn_ms_gvf = NULL;
int num_ms_gvm = 0, num_ms_gvl = 0, num_ms_gvf = 0;
/* file names of global variance trees */
char **fn_ts_gvm = NULL;
char **fn_ts_gvl = NULL;
char **fn_ts_gvf = NULL;
int num_ts_gvm = 0, num_ts_gvl = 0, num_ts_gvf = 0;
/* file name of global variance switch */
char *fn_gv_switch = NULL;
/* global parameter */
int sampling_rate = 16000;
int fperiod = 80;
double alpha = 0.42;
int stage = 0; /* Gamma=-1/stage: if stage=0 then Gamma=0 */
double beta = 0.0;
int audio_buff_size = 1600;
double uv_threshold = 0.5;
double gv_weight_mgc = 1.0;
double gv_weight_lf0 = 1.0;
double gv_weight_lpf = 1.0;
double half_tone = 0.0;
HTS_Boolean phoneme_alignment = FALSE;
double speech_speed = 1.0;
HTS_Boolean use_log_gain = FALSE;
/* engine */
HTS_Engine engine;
/* parse command line */
if (argc == 1)
Usage();
/* delta window handler for mel-cepstrum */
fn_ws_mgc = (char **) calloc(argc, sizeof(char *));
/* delta window handler for log f0 */
fn_ws_lf0 = (char **) calloc(argc, sizeof(char *));
/* delta window handler for low-pass filter */
fn_ws_lpf = (char **) calloc(argc, sizeof(char *));
/* prepare for interpolation */
num_interp = GetNumInterp(argc, argv);
rate_interp = (double *) calloc(num_interp, sizeof(double));
for (i = 0; i < num_interp; i++)
rate_interp[i] = 1.0;
fn_ms_dur = (char **) calloc(num_interp, sizeof(char *));
fn_ms_mgc = (char **) calloc(num_interp, sizeof(char *));
fn_ms_lf0 = (char **) calloc(num_interp, sizeof(char *));
fn_ms_lpf = (char **) calloc(num_interp, sizeof(char *));
fn_ts_dur = (char **) calloc(num_interp, sizeof(char *));
fn_ts_mgc = (char **) calloc(num_interp, sizeof(char *));
fn_ts_lf0 = (char **) calloc(num_interp, sizeof(char *));
fn_ts_lpf = (char **) calloc(num_interp, sizeof(char *));
fn_ms_gvm = (char **) calloc(num_interp, sizeof(char *));
fn_ms_gvl = (char **) calloc(num_interp, sizeof(char *));
fn_ms_gvf = (char **) calloc(num_interp, sizeof(char *));
fn_ts_gvm = (char **) calloc(num_interp, sizeof(char *));
fn_ts_gvl = (char **) calloc(num_interp, sizeof(char *));
fn_ts_gvf = (char **) calloc(num_interp, sizeof(char *));
/* read command */
while (--argc) {
if (**++argv == '-') {
switch (*(*argv + 1)) {
case 'v':
switch (*(*argv + 2)) {
case 'p':
phoneme_alignment = TRUE;
break;
default:
Error(1, "hts_engine: Invalid option '-v%c'.\n", *(*argv + 2));
}
break;
case 't':
switch (*(*argv + 2)) {
case 'd':
fn_ts_dur[num_ts_dur++] = *++argv;
break;
case 'm':
fn_ts_mgc[num_ts_mgc++] = *++argv;
break;
case 'f':
case 'p':
fn_ts_lf0[num_ts_lf0++] = *++argv;
break;
case 'l':
fn_ts_lpf[num_ts_lpf++] = *++argv;
break;
default:
Error(1, "hts_engine: Invalid option '-t%c'.\n", *(*argv + 2));
}
--argc;
break;
case 'm':
switch (*(*argv + 2)) {
case 'd':
fn_ms_dur[num_ms_dur++] = *++argv;
break;
case 'm':
fn_ms_mgc[num_ms_mgc++] = *++argv;
break;
case 'f':
case 'p':
fn_ms_lf0[num_ms_lf0++] = *++argv;
break;
case 'l':
fn_ms_lpf[num_ms_lpf++] = *++argv;
break;
default:
Error(1, "hts_engine: Invalid option '-m%c'.\n", *(*argv + 2));
}
--argc;
break;
case 'd':
switch (*(*argv + 2)) {
case 'm':
fn_ws_mgc[num_ws_mgc++] = *++argv;
break;
case 'f':
case 'p':
fn_ws_lf0[num_ws_lf0++] = *++argv;
break;
case 'l':
fn_ws_lpf[num_ws_lpf++] = *++argv;
break;
default:
Error(1, "hts_engine: Invalid option '-d%c'.\n", *(*argv + 2));
}
--argc;
break;
case 'o':
switch (*(*argv + 2)) {
case 'r':
rawfp = Getfp(*++argv, "ab");
break;
default:
Error(1, "festcat-hts_engine: Invalid option '-o%c'.\n", *(*argv + 2));
}
--argc;
break;
case 'h':
Usage();
break;
case 's':
sampling_rate = atoi(*++argv);
--argc;
break;
case 'p':
fperiod = atoi(*++argv);
--argc;
break;
case 'a':
alpha = atof(*++argv);
--argc;
break;
case 'g':
stage = atoi(*++argv);
--argc;
break;
case 'l':
use_log_gain = TRUE;
break;
case 'b':
beta = atof(*++argv);
--argc;
break;
case 'r':
speech_speed = atof(*++argv);
--argc;
break;
case 'f':
switch (*(*argv + 2)) {
case 'm':
f = atof(*++argv);
if (f < -24.0)
f = -24.0;
if (f > 24.0)
f = 24.0;
half_tone = f;
break;
default:
Error(1, "hts_engine: Invalid option '-f%c'.\n", *(*argv + 2));
}
--argc;
break;
case 'u':
uv_threshold = atof(*++argv);
--argc;
break;
case 'i':
++argv;
argc--;
for (i = 0; i < num_interp; i++) {
rate_interp[i] = atof(*++argv);
argc--;
}
break;
case 'e':
switch (*(*argv + 2)) {
case 'm':
fn_ts_gvm[num_ts_gvm++] = *++argv;
break;
case 'f':
case 'p':
fn_ts_gvl[num_ts_gvl++] = *++argv;
break;
case 'l':
fn_ts_gvf[num_ts_gvf++] = *++argv;
break;
default:
Error(1, "hts_engine: Invalid option '-e%c'.\n", *(*argv + 2));
}
--argc;
break;
case 'c':
switch (*(*argv + 2)) {
case 'm':
fn_ms_gvm[num_ms_gvm++] = *++argv;
break;
case 'f':
case 'p':
fn_ms_gvl[num_ms_gvl++] = *++argv;
break;
case 'l':
fn_ms_gvf[num_ms_gvf++] = *++argv;
break;
default:
Error(1, "hts_engine: Invalid option '-c%c'.\n", *(*argv + 2));
}
--argc;
break;
case 'j':
switch (*(*argv + 2)) {
case 'm':
gv_weight_mgc = atof(*++argv);
break;
case 'f':
case 'p':
gv_weight_lf0 = atof(*++argv);
break;
case 'l':
gv_weight_lpf = atof(*++argv);
break;
default:
Error(1, "hts_engine: Invalid option '-j%c'.\n", *(*argv + 2));
}
--argc;
break;
case 'k':
fn_gv_switch = *++argv;
--argc;
break;
case 'z':
audio_buff_size = atoi(*++argv);
--argc;
break;
default:
Error(1, "hts_engine: Invalid option '-%c'.\n", *(*argv + 1));
}
} else {
guifn = *argv;
}
}
/* number of models,trees check */
if (num_interp != num_ts_dur || num_interp != num_ts_mgc ||
num_interp != num_ts_lf0 || num_interp != num_ms_dur ||
num_interp != num_ms_mgc || num_interp != num_ms_lf0) {
Error(1, "hts_engine: specify %d models(trees) for each parameter.\n",
num_interp);
}
if (num_ms_lpf > 0 || num_ts_lpf > 0) {
if (num_interp != num_ms_lpf || num_interp != num_ts_lpf) {
Error(1, "hts_engine: specify %d models(trees) for each parameter.\n",
num_interp);
}
}
/* initialize (stream[0] = spectrum, stream[1] = lf0, stream[2] = low-pass filter) */
if (num_ms_lpf > 0 || num_ts_lpf > 0) {
HTS_Engine_initialize(&engine, 3);
} else {
HTS_Engine_initialize(&engine, 2);
}
/* load duration model */
HTS_Engine_load_duration_from_fn(&engine, fn_ms_dur, fn_ts_dur, num_interp);
/* load stream[0] (spectrum model) */
HTS_Engine_load_parameter_from_fn(&engine, fn_ms_mgc, fn_ts_mgc, fn_ws_mgc,
0, FALSE, num_ws_mgc, num_interp);
/* load stream[1] (lf0 model) */
HTS_Engine_load_parameter_from_fn(&engine, fn_ms_lf0, fn_ts_lf0, fn_ws_lf0,
1, TRUE, num_ws_lf0, num_interp);
/* load stream[2] (low-pass filter model) */
if (num_ms_lpf > 0 || num_ts_lpf > 0)
HTS_Engine_load_parameter_from_fn(&engine, fn_ms_lpf, fn_ts_lpf,
fn_ws_lpf, 2, FALSE, num_ws_lpf,
num_interp);
/* load gv[0] (GV for spectrum) */
if (num_interp == num_ms_gvm) {
if (num_ms_gvm == num_ts_gvm)
HTS_Engine_load_gv_from_fn(&engine, fn_ms_gvm, fn_ts_gvm, 0,
num_interp);
else
HTS_Engine_load_gv_from_fn(&engine, fn_ms_gvm, NULL, 0, num_interp);
}
/* load gv[1] (GV for lf0) */
if (num_interp == num_ms_gvl) {
if (num_ms_gvl == num_ts_gvl)
HTS_Engine_load_gv_from_fn(&engine, fn_ms_gvl, fn_ts_gvl, 1,
num_interp);
else
HTS_Engine_load_gv_from_fn(&engine, fn_ms_gvl, NULL, 1, num_interp);
}
/* load gv[2] (GV for low-pass filter) */
if (num_interp == num_ms_gvf && (num_ms_lpf > 0 || num_ts_lpf > 0)) {
if (num_ms_gvf == num_ts_gvf)
HTS_Engine_load_gv_from_fn(&engine, fn_ms_gvf, fn_ts_gvf, 0,
num_interp);
else
HTS_Engine_load_gv_from_fn(&engine, fn_ms_gvf, NULL, 2, num_interp);
}
/* load GV switch */
if (fn_gv_switch != NULL)
HTS_Engine_load_gv_switch_from_fn(&engine, fn_gv_switch);
/* set parameter */
HTS_Engine_set_sampling_rate(&engine, sampling_rate);
HTS_Engine_set_fperiod(&engine, fperiod);
HTS_Engine_set_alpha(&engine, alpha);
HTS_Engine_set_gamma(&engine, stage);
HTS_Engine_set_log_gain(&engine, use_log_gain);
HTS_Engine_set_beta(&engine, beta);
HTS_Engine_set_audio_buff_size(&engine, audio_buff_size);
HTS_Engine_set_msd_threshold(&engine, 1, uv_threshold); /* set voiced/unvoiced threshold for stream[1] */
HTS_Engine_set_gv_weight(&engine, 0, gv_weight_mgc);
HTS_Engine_set_gv_weight(&engine, 1, gv_weight_lf0);
if (num_ms_lpf > 0 || num_ts_lpf > 0)
HTS_Engine_set_gv_weight(&engine, 2, gv_weight_lpf);
for (i = 0; i < num_interp; i++) {
HTS_Engine_set_duration_interpolation_weight(&engine, i, rate_interp[i]);
HTS_Engine_set_parameter_interpolation_weight(&engine, 0, i,
rate_interp[i]);
HTS_Engine_set_parameter_interpolation_weight(&engine, 1, i,
rate_interp[i]);
if (num_ms_lpf > 0 || num_ts_lpf > 0)
HTS_Engine_set_parameter_interpolation_weight(&engine, 2, i,
rate_interp[i]);
}
if (num_interp == num_ms_gvm)
for (i = 0; i < num_interp; i++)
HTS_Engine_set_gv_interpolation_weight(&engine, 0, i, rate_interp[i]);
if (num_interp == num_ms_gvl)
for (i = 0; i < num_interp; i++)
HTS_Engine_set_gv_interpolation_weight(&engine, 1, i, rate_interp[i]);
if (num_interp == num_ms_gvf && (num_ms_lpf > 0 || num_ts_lpf > 0))
for (i = 0; i < num_interp; i++)
HTS_Engine_set_gv_interpolation_weight(&engine, 2, i, rate_interp[i]);
/* synthesis */
guifp=Getfp(guifn, "r");
while (getline(&labfn,&labfn_size,guifp) != -1)
{
chomp(labfn);
HTS_Engine_load_label_from_fn(&engine, labfn); /* load label file */
if (phoneme_alignment) /* modify label */
HTS_Label_set_frame_specified_flag(&engine.label, TRUE);
if (speech_speed != 1.0) /* modify label */
HTS_Label_set_speech_speed(&engine.label, speech_speed);
HTS_Engine_create_sstream(&engine); /* parse label and determine state duration */
if (half_tone != 0.0) { /* modify f0 */
for (i = 0; i < HTS_SStreamSet_get_total_state(&engine.sss); i++) {
f = HTS_SStreamSet_get_mean(&engine.sss, 1, i, 0);
f += half_tone * log(2.0) / 12;
if (f < log(10.0))
f = log(10.0);
HTS_SStreamSet_set_mean(&engine.sss, 1, i, 0, f);
}
}
HTS_Engine_create_pstream(&engine); /* generate speech parameter vector sequence */
HTS_Engine_create_gstream(&engine); /* synthesize speech */
/* output */
if (rawfp)
HTS_Engine_save_generated_speech(&engine, rawfp);
/* free */
HTS_Engine_refresh(&engine);
}
if (guifp != NULL)
fclose(guifp);
free(labfn);
/* free memory */
HTS_Engine_clear(&engine);
free(rate_interp);
free(fn_ws_mgc);
free(fn_ws_lf0);
free(fn_ws_lpf);
free(fn_ms_mgc);
free(fn_ms_lf0);
free(fn_ms_lpf);
free(fn_ms_dur);
free(fn_ts_mgc);
free(fn_ts_lf0);
free(fn_ts_lpf);
free(fn_ts_dur);
free(fn_ms_gvm);
free(fn_ms_gvl);
free(fn_ms_gvf);
free(fn_ts_gvm);
free(fn_ts_gvl);
free(fn_ts_gvf);
/* close files */
if (rawfp != NULL)
fclose(rawfp);
return 0;
}
