cst_utterance *cart_intonation(cst_utterance *u)
{
cst_cart *accents, *tones;
cst_item *s;
const cst_val *v;
if (feat_present(u->features,"no_intonation_accent_model"))
return u; /* not all languages have intonation models */
accents = val_cart(feat_val(u->features,"int_cart_accents"));
tones = val_cart(feat_val(u->features,"int_cart_tones"));
for (s=relation_head(utt_relation(u,"Syllable")); s; s=item_next(s))
{
v = cart_interpret(s,accents);
if (!cst_streq("NONE",val_string(v)))
item_set_string(s,"accent",val_string(v));
v = cart_interpret(s,tones);
if (!cst_streq("NONE",val_string(v)))
item_set_string(s,"endtone",val_string(v));
DPRINTF(0,("word %s gpos %s stress %s ssyl_in %s ssyl_out %s accent %s endtone %s\n",
ffeature_string(s,"R:SylStructure.parent.name"),
ffeature_string(s,"R:SylStructure.parent.gpos"),
ffeature_string(s,"stress"),
ffeature_string(s,"ssyl_in"),
ffeature_string(s,"ssyl_out"),
ffeature_string(s,"accent"),
ffeature_string(s,"endtone")));
}
return u;
}
cst_utterance *cst_spamf0(cst_utterance *utt)
{
cst_track *spamf0_track = NULL;
cst_track *param_track = NULL;
cst_item *s;
cst_cg_db *cg_db;
const cst_cart *acc_tree, *phrase_tree;
float end, f0val, syldur;
int num_frames, f, i;
cg_db = val_cg_db(utt_feat_val(utt, "cg_db"));
spamf0_track = new_track();
cst_track_resize(spamf0_track,
(utt_feat_int(utt, "param_track_num_frames")), 1);
acc_tree = cg_db->spamf0_accent_tree;
phrase_tree = cg_db->spamf0_phrase_tree;
end = 0.0;
num_frames = 0;
for (s = utt_rel_head(utt, "Segment"); s; s = item_next(s))
{
end = ffeature_float(s, "end");
if (cst_streq("pau", ffeature_string(s, "name")))
{
f0val = 0;
}
else
{
f0val = val_float(cart_interpret(s, phrase_tree));
}
for (;
((num_frames * cg_db->frame_advance) <= end)
&& (num_frames < utt_feat_int(utt, "param_track_num_frames"));
num_frames++)
{
spamf0_track->frames[num_frames][0] = f0val;
}
}
for (s = utt_rel_head(utt, "Syllable"); s; s = item_next(s))
{
f = val_int(cart_interpret(s, acc_tree));
syldur = ffeature_float(s, "R:SylStructure.daughtern.R:Segment.end")
- ffeature_float(s, "R:SylStructure.daughter1.R:Segment.p.end");
cst_synthtilt(cg_db,
ffeature_float(s,
"R:SylStructure.daughter1.R:Segment.p.end"),
cg_db->spamf0_accent_vectors[f][0],
cg_db->spamf0_accent_vectors[f][2], syldur,
cg_db->spamf0_accent_vectors[f][6], spamf0_track);
}
param_track = val_track(utt_feat_val(utt, "param_track"));
for (i = 0; i < utt_feat_int(utt, "param_track_num_frames"); i++)
{
param_track->frames[i][0] = spamf0_track->frames[i][0];
}
delete_track(spamf0_track);
return utt;
}
cst_utterance *default_phrasing(cst_utterance *u)
{
cst_relation *r;
cst_item *w, *p, *lp=NULL;
const cst_val *v;
cst_cart *phrasing_cart;
r = utt_relation_create(u,"Phrase");
phrasing_cart = val_cart(feat_val(u->features,"phrasing_cart"));
for (p=NULL,w=relation_head(utt_relation(u,"Word")); w; w=item_next(w))
{
if (p == NULL)
{
p = relation_append(r,NULL);
lp = p;
#ifdef FLITE_PLUS_HTS_ENGINE
item_set_string(p,"name","BB");
#else
item_set_string(p,"name","B");
#endif /* FLITE_PLUS_HTS_ENGINE */
}
item_add_daughter(p,w);
v = cart_interpret(w,phrasing_cart);
if (cst_streq(val_string(v),"BB"))
p = NULL;
}
if (lp && item_prev(lp)) /* follow festival */
item_set_string(lp,"name","BB");
return u;
}
static float cg_state_duration(cst_item *s, cst_cg_db *cg_db)
{
float zdur, dur;
const char *n;
int i, x;
zdur = val_float(cart_interpret(s,cg_db->dur_cart));
n = item_feat_string(s,"name");
for (x=i=0; cg_db->dur_stats[i]; i++)
{
if (cst_streq(cg_db->dur_stats[i]->phone,n))
{
x=i;
break;
}
}
if (!cg_db->dur_stats[i]) /* unknown type name */
x = 0;
dur = (zdur*cg_db->dur_stats[x]->stddev)+cg_db->dur_stats[x]->mean;
/* dur = 1.2 * (float)exp((float)dur); */
return dur;
}
cst_utterance *default_phrasing(cst_utterance *u)
{
cst_relation *r;
cst_item *w, *p;
const cst_val *v;
cst_cart *phrasing_cart;
r = utt_relation_create(u,"Phrase");
phrasing_cart = val_cart(feat_val(u->features,"phrasing_cart"));
for (p=NULL,w=relation_head(utt_relation(u,"Word")); w; w=item_next(w))
{
if (p == NULL)
{
p = relation_append(r,NULL);
item_set_string(p,"name","BB");
}
item_add_daughter(p,w);
v = cart_interpret(w,phrasing_cart);
if (cst_streq(val_string(v),"BB"))
p = NULL;
}
return u;
}
static float cg_state_duration(cst_item *s, cst_cg_db *cg_db)
{
float zdur, dur;
const char *n;
int i, x;
#ifdef STATICANALYZEDEPENDENCIES
#define __clang_analyzer__ 1
#endif
#if !defined(__clang_analyzer__) || defined(STATICANALYZEDEPENDENCIES)
#undef __clang_analyzer__
zdur = val_float(cart_interpret(s,cg_db->dur_cart));
n = item_feat_string(s,"name");
for (x=i=0; cg_db->dur_stats[i]; i++)
{
if (cst_streq(cg_db->dur_stats[i]->phone,n))
{
x=i;
break;
}
}
if (!cg_db->dur_stats[i]) /* unknown type name */
x = 0;
dur = (zdur*cg_db->dur_stats[x]->stddev)+cg_db->dur_stats[x]->mean;
/* dur = 1.2 * (float)exp((float)dur); */
return dur;
}
cst_utterance *russian_postlex_function(cst_utterance *u)
{
const cst_item *word,*seg;
const char *answer,*name,*pair;
for(word=relation_head(utt_relation(u,"Transcription"));word;word=item_next(word))
{
if(item_feat_present(word,"no_pl")||item_feat_present(word,"no_vr"))
continue;
for(seg=item_daughter(word);seg;seg=item_next(seg))
{
name=item_feat_string(seg,"name");
if(cst_member_string(name,unstressed_vowels))
{
answer=val_string(cart_interpret(item_as(seg,"Segment"),&ru_vowel_reduction_cart));
if(!cst_streq(answer,"N"))
item_set_string(seg,"name",answer);
}
else
{
if(cst_streq(name,"ii")&&
cst_streq(ffeature_string(seg,"R:Segment.p.ph_csoft"),"-")&&
!(cst_streq(item_feat_string(word,"name"),"и")&&
cst_streq(ffeature_string(word,"gpos"),"content")))
{
item_set_string(seg,"name","yy");
}
}
}
}
for(word=relation_tail(utt_relation(u,"Transcription"));word;word=item_prev(word))
{
if(item_feat_present(word,"no_pl"))
continue;
for(seg=item_last_daughter(word);seg;seg=item_prev(seg))
{
name=item_feat_string(seg,"name");
pair=russian_vpair(name);
if(pair!=NULL)
{
answer=val_string(cart_interpret(item_as(seg,"Segment"),&ru_vpair_cart));
if(cst_streq(answer,"Y"))
item_set_string(seg,"name",pair);
}
}
}
return u;
}
cst_utterance *cart_duration(cst_utterance *u)
{
cst_cart *dur_tree;
cst_item *s;
float zdur, dur_stretch, local_dur_stretch, dur;
float end;
dur_stats *ds;
const dur_stat *dur_stat;
end = 0;
if (feat_present(u->features,"no_segment_duration_model"))
return u; /* not all methods need segment durations */
dur_tree = val_cart(feat_val(u->features,"dur_cart"));
dur_stretch = get_param_float(u->features,"duration_stretch", 1.0);
ds = val_dur_stats(feat_val(u->features,"dur_stats"));
for (s=relation_head(utt_relation(u,"Segment")); s; s=item_next(s))
{
zdur = val_float(cart_interpret(s,dur_tree));
dur_stat = phone_dur_stat(ds,item_name(s));
local_dur_stretch = ffeature_float(s, "R:SylStructure.parent.parent."
"R:Token.parent.local_duration_stretch");
if (local_dur_stretch)
local_dur_stretch *= dur_stretch;
else
local_dur_stretch = dur_stretch;
dur = local_dur_stretch * ((zdur*dur_stat->stddev)+dur_stat->mean);
DPRINTF(0,("phone %s accent %s stress %s pdur %f stretch %f mean %f std %f dur %f\n",
item_name(s),
ffeature_string(s,"R:SylStructure.parent.accented"),
ffeature_string(s,"R:SylStructure.parent.stress"),
zdur, local_dur_stretch, dur_stat->mean,
dur_stat->stddev, dur));
end += dur;
item_set_float(s,"end",end);
}
return u;
}
cst_utterance *default_pos_tagger(cst_utterance *u)
{
cst_item *word;
const cst_val *p;
const cst_cart *tagger;
p = get_param_val(u->features,"pos_tagger_cart",NULL);
if (p == NULL)
return u;
tagger = val_cart(p);
for (word=relation_head(utt_relation(u,"Word"));
word; word=item_next(word))
{
p = cart_interpret(word,tagger);
item_set_string(word,"pos",val_string(p));
}
return u;
}
static cst_utterance *cg_predict_params(cst_utterance *utt)
{
cst_cg_db *cg_db;
cst_track *param_track;
cst_track *str_track = NULL;
cst_item *mcep;
const cst_cart *mcep_tree, *f0_tree;
int i,j,f,p,fd,o;
const char *mname;
float f0_val;
int fff;
int extra_feats = 0;
cg_db = val_cg_db(utt_feat_val(utt,"cg_db"));
param_track = new_track();
if (cg_db->do_mlpg) /* which should be the default */
fff = 1; /* copy details with stddevs */
else
fff = 2; /* copy details without stddevs */
extra_feats = 1; /* voicing */
if (cg_db->mixed_excitation)
{
extra_feats += 5;
str_track = new_track();
cst_track_resize(str_track,
utt_feat_int(utt,"param_track_num_frames"),
5);
}
cst_track_resize(param_track,
utt_feat_int(utt,"param_track_num_frames"),
(cg_db->num_channels0/fff)-
(2 * extra_feats));/* no voicing or str */
for (i=0,mcep=utt_rel_head(utt,"mcep"); mcep; i++,mcep=item_next(mcep))
{
mname = item_feat_string(mcep,"name");
for (p=0; cg_db->types[p]; p++)
if (cst_streq(mname,cg_db->types[p]))
break;
if (cg_db->types[0] == NULL)
p=0; /* if there isn't a matching tree, use the first one */
/* Predict F0 */
f0_tree = cg_db->f0_trees[p];
f0_val = val_float(cart_interpret(mcep,f0_tree));
param_track->frames[i][0] = f0_val;
/* what about stddev ? */
if (cg_db->multimodel)
{ /* MULTI model */
f = val_int(cart_interpret(mcep,cg_db->param_trees0[p]));
fd = val_int(cart_interpret(mcep,cg_db->param_trees1[p]));
item_set_int(mcep,"clustergen_param_frame",f);
param_track->frames[i][0] =
(param_track->frames[i][0]+
CG_MODEL_VECTOR(cg_db,model_vectors0,f,0)+
CG_MODEL_VECTOR(cg_db,model_vectors1,fd,0))/3.0;
for (j=2; j<param_track->num_channels; j++)
param_track->frames[i][j] =
(CG_MODEL_VECTOR(cg_db,model_vectors0,f,(j)*fff)+
CG_MODEL_VECTOR(cg_db,model_vectors1,fd,(j)*fff))/2.0;
if (cg_db->mixed_excitation)
{
o = j;
for (j=0; j<5; j++)
{
str_track->frames[i][j] =
(CG_MODEL_VECTOR(cg_db,model_vectors0,f,(o+(2*j))*fff)+
CG_MODEL_VECTOR(cg_db,model_vectors1,fd,(o+(2*j))*fff))/2.0;
}
}
}
else
{ /* SINGLE model */
/* Predict Spectral */
mcep_tree = cg_db->param_trees0[p];
f = val_int(cart_interpret(mcep,mcep_tree));
item_set_int(mcep,"clustergen_param_frame",f);
param_track->frames[i][0] =
(param_track->frames[i][0]+
CG_MODEL_VECTOR(cg_db,model_vectors0,f,0))/2.0;
for (j=2; j<param_track->num_channels; j++)
param_track->frames[i][j] =
CG_MODEL_VECTOR(cg_db,model_vectors0,f,(j)*fff);
if (cg_db->mixed_excitation)
{
o = j;
for (j=0; j<5; j++)
{
str_track->frames[i][j] =
CG_MODEL_VECTOR(cg_db,model_vectors0,f,(o+(2*j))*fff);
}
}
}
/* last coefficient is average voicing for cluster */
item_set_float(mcep,"voicing",
CG_MODEL_VECTOR(cg_db,model_vectors0,f,
cg_db->num_channels0-2));
param_track->times[i] = i * cg_db->frame_advance;
}
cg_smooth_F0(utt,cg_db,param_track);
utt_set_feat(utt,"param_track",track_val(param_track));
if (cg_db->mixed_excitation)
utt_set_feat(utt,"str_track",track_val(str_track));
return utt;
}
static void assign_stress(cst_item *word)
{
int numsyls=ffeature_int(word,"word_numsyls");
const char *gpos=ffeature_string(word,"gpos");
const cst_item *word_in_phrase=item_as(word,"Phrase");
const cst_item *syls=item_as(word,"SylStructure");
const cst_item *syl=item_daughter(syls);
const cst_item *transcription=item_as(word,"Transcription");
int stressed=FALSE;
int n=item_feat_present(word,"stressed_syl_num")?item_feat_int(word,"stressed_syl_num"):0;
const char *name=item_name(word);
const char *pname=ffeature_string(word,"R:Phrase.p.name");
const char *nname=ffeature_string(word,"R:Phrase.n.name");
if(cst_streq(ffeature_string(word,"R:Token.p.name"),"по")&&
(cst_streq(name,"моему")||cst_streq(name,"своему")||cst_streq(name,"твоему")))
{
item_set_string(syl,"stress","1");
return;
}
else if((cst_streq(name,"не")||cst_streq(name,"ни"))&&
(cst_streq(nname,"был")||cst_streq(nname,"были")||cst_streq(nname,"было")))
{
item_set_string(syl,"stress","1");
return;
}
else if((cst_streq(name,"был")||cst_streq(name,"были")||cst_streq(name,"было"))&&
(cst_streq(pname,"не")||cst_streq(pname,"ни")))
return;
else if(cst_streq(gpos,"enc")&&item_prev(word_in_phrase))
return;
else if(cst_streq(gpos,"proc")&&item_next(word_in_phrase))
return;
if(!vowel_seg_between(item_daughter(transcription),item_last_daughter(transcription)))
return;
if(numsyls==1)
{
item_set_string(syl,"stress","1");
return;
}
for(;syl;syl=item_next(syl))
{
if(is_stressed_vowel(item_feat_string(item_daughter(item_as(syl,"SylVowel")),"name")))
{
item_set_string(syl,"stress","1");
stressed=TRUE;
}
}
if(n==0)
{
if(stressed)
return;
n=val_int(cart_interpret(word,&ru_stress_cart));
if((numsyls+n) < 0)
{
if(numsyls <= 4)
n=-2;
else
if(numsyls <= 6)
n=-3;
else
n=-4;
}
}
item_set_string(item_nth_daughter(syls,(numsyls+n)),"stress","1");
}