static void check_and_change_tone(SHTSEngineMESynthUttProc105 *HTSsynth,
const SUtterance *utt, s_erc *error)
{
const SVoice *voice;
const SObject *tmp;
float var;
S_CLR_ERR(error);
/* get voice */
voice = SUtteranceVoice(utt, error);
if (S_CHK_ERR(error, S_CONTERR,
"check_and_change_tone",
"Call to \"SUtteranceVoice\" failed"))
return;
if (voice == NULL)
return;
/* get half-tone feature */
tmp = SVoiceGetFeature(voice, "half-tone", error);
if (S_CHK_ERR(error, S_CONTERR,
"check_and_change_tone",
"Call to \"SVoiceGetFeature\" failed"))
return;
if (tmp == NULL)
return;
var = SObjectGetFloat(tmp, error);
if (S_CHK_ERR(error, S_CONTERR,
"check_and_change_tone",
"Call to \"SObjectGetFloat\" failed"))
return;
if (var != 0.0)
{
int i;
double f;
for (i = 0; i < HTS_SStreamSet_get_total_state(&(HTSsynth->engine.sss)); i++)
{
f = HTS_SStreamSet_get_mean(&(HTSsynth->engine.sss), 1, i, 0); /* logf0 is stream 1 */
f += var * log(2.0) / 12;
if (f < log(10.0))
f = log(10.0);
HTS_SStreamSet_set_mean(&(HTSsynth->engine.sss), 1, i, 0, f);
}
}
}
/* HTS_Engine_synthesize_from_strings: synthesize speech from strings */
HTS_Boolean HTS_Engine_synthesize_from_strings(HTS_Engine * engine, char **lines, size_t num_lines)
{
size_t i;
double f;
HTS_Engine_refresh(engine);
// Generate state sequence
if (HTS_SStreamSet_create(&engine->sss, &engine->ms, lines, num_lines, engine->condition.speed) != TRUE) {
HTS_Engine_refresh(engine);
return FALSE;
}
if (engine->condition.additional_half_tone != 0.0) {
for (i = 0; i < HTS_SStreamSet_get_total_state(&engine->sss); i++) {
f = HTS_SStreamSet_get_mean(&engine->sss, 1, i, 0);
f += engine->condition.additional_half_tone * HALF_TONE;
if (f < MIN_LF0)
f = MIN_LF0;
else if (f > MAX_LF0)
f = MAX_LF0;
HTS_SStreamSet_set_mean(&engine->sss, 1, i, 0, f);
}
}
// Generate parameter sequence
if (HTS_PStreamSet_create(&engine->pss, &engine->sss, engine->condition.msd_threshold) != TRUE) {
HTS_Engine_refresh(engine);
return FALSE;
}
// Free state sequence
HTS_SStreamSet_clear(&engine->sss);
// Generate sound sample sequence
if (HTS_GStreamSet_create(&engine->gss, &engine->pss, engine->condition.sampling_frequency,
engine->condition.fperiod, engine->condition.alpha,
engine->condition.beta) != TRUE) {
HTS_Engine_refresh(engine);
return FALSE;
}
return TRUE;
}
/* HTS_Engine_get_state_mean: get mean value of state */
double HTS_Engine_get_state_mean(HTS_Engine * engine, size_t stream_index, size_t state_index, size_t vector_index)
{
return HTS_SStreamSet_get_mean(&engine->sss, stream_index, state_index, vector_index);
}
/* HTS_PStreamSet_create: parameter generation using GV weight */
void HTS_PStreamSet_create(HTS_PStreamSet * pss, HTS_SStreamSet * sss,
double *msd_threshold, double *gv_weight)
{
int i, j, k, l, m;
int frame, msd_frame, state;
HTS_PStream *pst;
HTS_Boolean not_bound;
if (pss->nstream)
HTS_error(1, "HTS_PstreamSet_create: HTS_PStreamSet should be clear.\n");
/* initialize */
pss->nstream = HTS_SStreamSet_get_nstream(sss);
pss->pstream = (HTS_PStream *) HTS_calloc(pss->nstream, sizeof(HTS_PStream));
pss->total_frame = HTS_SStreamSet_get_total_frame(sss);
/* create */
for (i = 0; i < pss->nstream; i++) {
pst = &pss->pstream[i];
if (HTS_SStreamSet_is_msd(sss, i)) { /* for MSD */
pst->length = 0;
for (state = 0; state < HTS_SStreamSet_get_total_state(sss); state++)
if (HTS_SStreamSet_get_msd(sss, i, state) > msd_threshold[i])
pst->length += HTS_SStreamSet_get_duration(sss, state);
pst->msd_flag =
(HTS_Boolean *) HTS_calloc(pss->total_frame, sizeof(HTS_Boolean));
for (state = 0, frame = 0; state < HTS_SStreamSet_get_total_state(sss);
state++)
if (HTS_SStreamSet_get_msd(sss, i, state) > msd_threshold[i])
for (j = 0; j < HTS_SStreamSet_get_duration(sss, state); j++) {
pst->msd_flag[frame] = TRUE;
frame++;
} else
for (j = 0; j < HTS_SStreamSet_get_duration(sss, state); j++) {
pst->msd_flag[frame] = FALSE;
frame++;
}
} else { /* for non MSD */
pst->length = pss->total_frame;
pst->msd_flag = NULL;
}
pst->vector_length = HTS_SStreamSet_get_vector_length(sss, i);
pst->width = HTS_SStreamSet_get_window_max_width(sss, i) * 2 + 1; /* band width of R */
pst->win_size = HTS_SStreamSet_get_window_size(sss, i);
pst->static_length = pst->vector_length / pst->win_size;
pst->sm.mean = HTS_alloc_matrix(pst->length, pst->vector_length);
pst->sm.ivar = HTS_alloc_matrix(pst->length, pst->vector_length);
pst->sm.wum = (double *) HTS_calloc(pst->length, sizeof(double));
pst->sm.wuw = HTS_alloc_matrix(pst->length, pst->width);
pst->sm.g = (double *) HTS_calloc(pst->length, sizeof(double));
pst->par = HTS_alloc_matrix(pst->length, pst->static_length);
/* copy dynamic window */
pst->win_l_width = (int *) HTS_calloc(pst->win_size, sizeof(int));
pst->win_r_width = (int *) HTS_calloc(pst->win_size, sizeof(int));
pst->win_coefficient =
(double **) HTS_calloc(pst->win_size, sizeof(double));
for (j = 0; j < pst->win_size; j++) {
pst->win_l_width[j] = HTS_SStreamSet_get_window_left_width(sss, i, j);
pst->win_r_width[j] = HTS_SStreamSet_get_window_right_width(sss, i, j);
if (pst->win_l_width[j] + pst->win_r_width[j] == 0)
pst->win_coefficient[j] = (double *)
HTS_calloc(-2 * pst->win_l_width[j] + 1, sizeof(double));
else
pst->win_coefficient[j] = (double *)
HTS_calloc(-2 * pst->win_l_width[j], sizeof(double));
pst->win_coefficient[j] -= pst->win_l_width[j];
for (k = pst->win_l_width[j]; k <= pst->win_r_width[j]; k++)
pst->win_coefficient[j][k] =
HTS_SStreamSet_get_window_coefficient(sss, i, j, k);
}
/* copy GV */
if (HTS_SStreamSet_use_gv(sss, i)) {
pst->gv_mean =
(double *) HTS_calloc(pst->static_length, sizeof(double));
pst->gv_vari =
(double *) HTS_calloc(pst->static_length, sizeof(double));
for (j = 0; j < pst->static_length; j++) {
pst->gv_mean[j] =
HTS_SStreamSet_get_gv_mean(sss, i, j) * gv_weight[i];
pst->gv_vari[j] = HTS_SStreamSet_get_gv_vari(sss, i, j);
}
pst->gv_switch =
(HTS_Boolean *) HTS_calloc(pst->length, sizeof(HTS_Boolean));
if (HTS_SStreamSet_is_msd(sss, i)) { /* for MSD */
for (state = 0, frame = 0, msd_frame = 0;
state < HTS_SStreamSet_get_total_state(sss); state++)
for (j = 0; j < HTS_SStreamSet_get_duration(sss, state);
j++, frame++)
if (pst->msd_flag[frame])
pst->gv_switch[msd_frame++] =
HTS_SStreamSet_get_gv_switch(sss, i, state);
} else { /* for non MSD */
for (state = 0, frame = 0;
state < HTS_SStreamSet_get_total_state(sss); state++)
for (j = 0; j < HTS_SStreamSet_get_duration(sss, state); j++)
pst->gv_switch[frame++] =
HTS_SStreamSet_get_gv_switch(sss, i, state);
}
for (j = 0, pst->gv_length = 0; j < pst->length; j++)
if (pst->gv_switch[j])
pst->gv_length++;
} else {
pst->gv_switch = NULL;
pst->gv_length = 0;
pst->gv_mean = NULL;
pst->gv_vari = NULL;
}
/* copy pdfs */
if (HTS_SStreamSet_is_msd(sss, i)) { /* for MSD */
for (state = 0, frame = 0, msd_frame = 0;
state < HTS_SStreamSet_get_total_state(sss); state++)
for (j = 0; j < HTS_SStreamSet_get_duration(sss, state); j++) {
if (pst->msd_flag[frame]) {
/* check current frame is MSD boundary or not */
for (k = 0; k < pst->win_size; k++) {
not_bound = TRUE;
for (l = pst->win_l_width[k]; l <= pst->win_r_width[k];
l++)
if (frame + l < 0 || pss->total_frame <= frame + l
|| !pst->msd_flag[frame + l]) {
not_bound = FALSE;
break;
}
for (l = 0; l < pst->static_length; l++) {
m = pst->static_length * k + l;
pst->sm.mean[msd_frame][m] =
HTS_SStreamSet_get_mean(sss, i, state, m);
if (not_bound || k == 0)
pst->sm.ivar[msd_frame][m] =
HTS_finv(HTS_SStreamSet_get_vari
(sss, i, state, m));
else
pst->sm.ivar[msd_frame][m] = 0.0;
}
}
msd_frame++;
}
frame++;
}
} else { /* for non MSD */
for (state = 0, frame = 0;
state < HTS_SStreamSet_get_total_state(sss); state++) {
for (j = 0; j < HTS_SStreamSet_get_duration(sss, state); j++) {
for (k = 0; k < pst->win_size; k++) {
not_bound = TRUE;
for (l = pst->win_l_width[k]; l <= pst->win_r_width[k]; l++)
if (frame + l < 0 || pss->total_frame <= frame + l) {
not_bound = FALSE;
break;
}
for (l = 0; l < pst->static_length; l++) {
m = pst->static_length * k + l;
pst->sm.mean[frame][m] =
HTS_SStreamSet_get_mean(sss, i, state, m);
if (not_bound || k == 0)
pst->sm.ivar[frame][m] =
HTS_finv(HTS_SStreamSet_get_vari(sss, i, state, m));
else
pst->sm.ivar[frame][m] = 0.0;
}
}
frame++;
}
}
}
/* parameter generation */
HTS_PStream_mlpg(pst);
}
}
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;
}
