static void
phone_transition(phone_loop_search_t *pls, int frame_idx)
{
int32 thresh = pls->best_score + pls->pbeam;
int nf = frame_idx + 1;
int i;
/* Now transition out of phones whose last states are inside the
* phone transition beam. */
for (i = 0; i < pls->n_phones; ++i) {
hmm_t *hmm = (hmm_t *)&pls->phones[i];
int32 newphone_score;
int j;
if (hmm_frame(hmm) != nf)
continue;
newphone_score = hmm_out_score(hmm) + pls->pip;
if (newphone_score BETTER_THAN thresh) {
/* Transition into all phones using the usual Viterbi rule. */
for (j = 0; j < pls->n_phones; ++j) {
hmm_t *nhmm = (hmm_t *)&pls->phones[j];
if (hmm_frame(nhmm) < frame_idx
|| newphone_score BETTER_THAN hmm_in_score(nhmm)) {
hmm_enter(nhmm, newphone_score, hmm_out_history(hmm), nf);
}
}
}
}
}
static void
fwdflat_renormalize_scores(ngram_search_t *ngs, int frame_idx, int32 norm)
{
root_chan_t *rhmm;
chan_t *hmm;
int32 i, cf, w, *awl;
cf = frame_idx;
/* Renormalize individual word channels */
i = ngs->n_active_word[cf & 0x1];
awl = ngs->active_word_list[cf & 0x1];
for (w = *(awl++); i > 0; --i, w = *(awl++)) {
rhmm = (root_chan_t *) ngs->word_chan[w];
if (hmm_frame(&rhmm->hmm) == cf) {
hmm_normalize(&rhmm->hmm, norm);
}
for (hmm = rhmm->next; hmm; hmm = hmm->next) {
if (hmm_frame(&hmm->hmm) == cf) {
hmm_normalize(&hmm->hmm, norm);
}
}
}
ngs->renormalized = TRUE;
}
static void
compute_fwdflat_sen_active(ngram_search_t *ngs, int frame_idx)
{
int32 i, w;
int32 *awl;
root_chan_t *rhmm;
chan_t *hmm;
acmod_clear_active(ps_search_acmod(ngs));
i = ngs->n_active_word[frame_idx & 0x1];
awl = ngs->active_word_list[frame_idx & 0x1];
for (w = *(awl++); i > 0; --i, w = *(awl++)) {
rhmm = (root_chan_t *)ngs->word_chan[w];
if (hmm_frame(&rhmm->hmm) == frame_idx) {
acmod_activate_hmm(ps_search_acmod(ngs), &rhmm->hmm);
}
for (hmm = rhmm->next; hmm; hmm = hmm->next) {
if (hmm_frame(&hmm->hmm) == frame_idx) {
acmod_activate_hmm(ps_search_acmod(ngs), &hmm->hmm);
}
}
}
}
static void
prune_hmms(state_align_search_t *sas, int frame_idx)
{
int nf = frame_idx + 1;
int i;
/* Check all phones to see if they remain active in the next frame. */
for (i = 0; i < sas->n_phones; ++i) {
hmm_t *hmm = sas->hmms + i;
if (hmm_frame(hmm) < frame_idx)
continue;
hmm_frame(hmm) = nf;
}
}
void
hmm_enter(hmm_t *h, int32 score, int32 histid, int frame)
{
hmm_in_score(h) = score;
hmm_in_history(h) = histid;
hmm_frame(h) = frame;
}
/** Evaluate active PHMMs */
static int32
phmm_eval_all(allphone_search_t * allphs, const int16 * senscr)
{
s3cipid_t ci;
phmm_t *p;
int32 best;
bin_mdef_t *mdef;
phmm_t **ci_phmm;
mdef = ((ps_search_t *) allphs)->acmod->mdef;
ci_phmm = allphs->ci_phmm;
best = WORST_SCORE;
hmm_context_set_senscore(allphs->hmmctx, senscr);
for (ci = 0; ci < mdef->n_ciphone; ci++) {
for (p = ci_phmm[(unsigned) ci]; p; p = p->next) {
if (hmm_frame(&(p->hmm)) == allphs->frame) {
int32 score;
allphs->n_hmm_eval++;
score = hmm_vit_eval((hmm_t *) p);
if (score > best)
best = score;
}
}
}
return best;
}
static void
record_transitions(state_align_search_t *sas, int frame_idx)
{
state_align_hist_t *tokens;
int i;
/* Push another frame of tokens on the stack. */
extend_tokenstack(sas, frame_idx);
tokens = sas->tokens + frame_idx * sas->n_emit_state;
/* Scan all active HMMs */
for (i = 0; i < sas->n_phones; ++i) {
hmm_t *hmm = sas->hmms + i;
int j;
if (hmm_frame(hmm) < frame_idx)
continue;
for (j = 0; j < sas->hmmctx->n_emit_state; ++j) {
int state_idx = i * sas->hmmctx->n_emit_state + j;
/* Record their backpointers on the token stack. */
tokens[state_idx].id = hmm_history(hmm, j);
tokens[state_idx].score = hmm_score(hmm, j);
/* Update backpointer fields with state index. */
hmm_history(hmm, j) = state_idx;
}
}
}
void
hmm_enter_obj(hmm_t *h, int32 score, void *histobj, int32 frame)
{
hmm_in_score(h) = score;
hmm_in_histobj(h) = histobj;
hmm_frame(h) = frame;
}
static void
prune_hmms(phone_loop_search_t *pls, int frame_idx)
{
int32 thresh = pls->best_score + pls->beam;
int nf = frame_idx + 1;
int i;
/* Check all phones to see if they remain active in the next frame. */
for (i = 0; i < pls->n_phones; ++i) {
hmm_t *hmm = (hmm_t *)&pls->phones[i];
if (hmm_frame(hmm) < frame_idx)
continue;
/* Retain if score better than threshold. */
if (hmm_bestscore(hmm) BETTER_THAN thresh) {
hmm_frame(hmm) = nf;
}
else
hmm_clear_scores(hmm);
}
}
static void
phone_transition(state_align_search_t *sas, int frame_idx)
{
int nf = frame_idx + 1;
int i;
for (i = 0; i < sas->n_phones - 1; ++i) {
hmm_t *hmm, *nhmm;
int32 newphone_score;
hmm = sas->hmms + i;
if (hmm_frame(hmm) != nf)
continue;
newphone_score = hmm_out_score(hmm);
/* Transition into next phone using the usual Viterbi rule. */
nhmm = hmm + 1;
if (hmm_frame(nhmm) < frame_idx
|| newphone_score BETTER_THAN hmm_in_score(nhmm)) {
hmm_enter(nhmm, newphone_score, hmm_out_history(hmm), nf);
}
}
}
static void
fwdflat_eval_chan(ngram_search_t *ngs, int frame_idx)
{
int32 i, w, nw, bestscore;
int32 *awl;
root_chan_t *rhmm;
chan_t *hmm;
nw = ngs->n_active_word[frame_idx & 0x1];
awl = ngs->active_word_list[frame_idx & 0x1];
bestscore = WORST_SCORE;
ngs->st.n_fwdflat_words += nw;
/* Scan all active words. */
for (i = 0; i < nw; i++) {
w = *(awl++);
rhmm = (root_chan_t *) ngs->word_chan[w];
if (hmm_frame(&rhmm->hmm) == frame_idx) {
int32 score = chan_v_eval(rhmm);
if ((score BETTER_THAN bestscore) && (w != ps_search_finish_wid(ngs)))
bestscore = score;
ngs->st.n_fwdflat_chan++;
}
for (hmm = rhmm->next; hmm; hmm = hmm->next) {
if (hmm_frame(&hmm->hmm) == frame_idx) {
int32 score = chan_v_eval(hmm);
if (score BETTER_THAN bestscore)
bestscore = score;
ngs->st.n_fwdflat_chan++;
}
}
}
ngs->best_score = bestscore;
}
static void
evaluate_hmms(phone_loop_search_t *pls, int16 const *senscr, int frame_idx)
{
int32 bs = WORST_SCORE;
int i;
hmm_context_set_senscore(pls->hmmctx, senscr);
for (i = 0; i < pls->n_phones; ++i) {
hmm_t *hmm = (hmm_t *)&pls->hmms[i];
int32 score;
if (hmm_frame(hmm) < frame_idx)
continue;
score = hmm_vit_eval(hmm);
if (score BETTER_THAN bs) {
bs = score;
}
}
pls->best_score = bs;
}
static int32
evaluate_hmms(state_align_search_t *sas, int16 const *senscr, int frame_idx)
{
int32 bs = WORST_SCORE;
int i;
hmm_context_set_senscore(sas->hmmctx, senscr);
for (i = 0; i < sas->n_phones; ++i) {
hmm_t *hmm = sas->hmms + i;
int32 score;
if (hmm_frame(hmm) < frame_idx)
continue;
score = hmm_vit_eval(hmm);
if (score BETTER_THAN bs) {
bs = score;
}
}
return bs;
}
static void
phmm_exit(allphone_search_t * allphs, int32 best)
{
s3cipid_t ci;
phmm_t *p;
int32 th, nf;
history_t *h;
blkarray_list_t *history;
bin_mdef_t *mdef;
int32 curfrm;
phmm_t **ci_phmm;
int32 *ci2lmwid;
th = best + allphs->pbeam;
history = allphs->history;
mdef = ps_search_acmod(allphs)->mdef;
curfrm = allphs->frame;
ci_phmm = allphs->ci_phmm;
ci2lmwid = allphs->ci2lmwid;
nf = curfrm + 1;
for (ci = 0; ci < mdef->n_ciphone; ci++) {
for (p = ci_phmm[(unsigned) ci]; p; p = p->next) {
if (hmm_frame(&(p->hmm)) == curfrm) {
if (hmm_bestscore(&(p->hmm)) >= th) {
h = (history_t *) ckd_calloc(1, sizeof(*h));
h->ef = curfrm;
h->phmm = p;
h->hist = hmm_out_history(&(p->hmm));
h->score = hmm_out_score(&(p->hmm));
if (!allphs->lm) {
h->tscore = allphs->inspen;
}
else {
if (h->hist > 0) {
int32 n_used;
history_t *pred =
blkarray_list_get(history, h->hist);
if (pred->hist > 0) {
history_t *pred_pred =
blkarray_list_get(history,
h->hist);
h->tscore =
ngram_tg_score(allphs->lm,
ci2lmwid
[pred_pred->phmm->ci],
ci2lmwid[pred->
phmm->ci],
ci2lmwid[p->ci],
&n_used) >>
SENSCR_SHIFT;
}
else {
h->tscore =
ngram_bg_score(allphs->lm,
ci2lmwid
[pred->phmm->ci],
ci2lmwid[p->ci],
&n_used) >>
SENSCR_SHIFT;
}
}
else {
/*
* This is the beginning SIL and in srch_allphone_begin()
* it's inscore is set to 0.
*/
h->tscore = 0;
}
}
static void
fwdflat_word_transition(ngram_search_t *ngs, int frame_idx)
{
int32 cf, nf, b, thresh, pip, i, w, newscore;
int32 best_silrc_score = 0, best_silrc_bp = 0; /* FIXME: good defaults? */
bptbl_t *bp;
int32 *rcss;
root_chan_t *rhmm;
int32 *awl;
float32 lwf;
dict_t *dict = ps_search_dict(ngs);
dict2pid_t *d2p = ps_search_dict2pid(ngs);
cf = frame_idx;
nf = cf + 1;
thresh = ngs->best_score + ngs->fwdflatbeam;
pip = ngs->pip;
best_silrc_score = WORST_SCORE;
lwf = ngs->fwdflat_fwdtree_lw_ratio;
/* Search for all words starting within a window of this frame.
* These are the successors for words exiting now. */
get_expand_wordlist(ngs, cf, ngs->max_sf_win);
/* Scan words exited in current frame */
for (b = ngs->bp_table_idx[cf]; b < ngs->bpidx; b++) {
xwdssid_t *rssid;
int32 silscore;
bp = ngs->bp_table + b;
ngs->word_lat_idx[bp->wid] = NO_BP;
if (bp->wid == ps_search_finish_wid(ngs))
continue;
/* DICT2PID location */
/* Get the mapping from right context phone ID to index in the
* right context table and the bscore_stack. */
rcss = ngs->bscore_stack + bp->s_idx;
if (bp->last2_phone == -1)
rssid = NULL;
else
rssid = dict2pid_rssid(d2p, bp->last_phone, bp->last2_phone);
/* Transition to all successor words. */
for (i = 0; ngs->expand_word_list[i] >= 0; i++) {
int32 n_used;
w = ngs->expand_word_list[i];
/* Get the exit score we recorded in save_bwd_ptr(), or
* something approximating it. */
if (rssid)
newscore = rcss[rssid->cimap[dict_first_phone(dict, w)]];
else
newscore = bp->score;
if (newscore == WORST_SCORE)
continue;
/* FIXME: Floating point... */
newscore += lwf
* (ngram_tg_score(ngs->lmset,
dict_basewid(dict, w),
bp->real_wid,
bp->prev_real_wid,
&n_used) >> SENSCR_SHIFT);
newscore += pip;
/* Enter the next word */
if (newscore BETTER_THAN thresh) {
rhmm = (root_chan_t *) ngs->word_chan[w];
if ((hmm_frame(&rhmm->hmm) < cf)
|| (newscore BETTER_THAN hmm_in_score(&rhmm->hmm))) {
hmm_enter(&rhmm->hmm, newscore, b, nf);
/* DICT2PID: This is where mpx ssids get introduced. */
/* Look up the ssid to use when entering this mpx triphone. */
hmm_mpx_ssid(&rhmm->hmm, 0) =
dict2pid_ldiph_lc(d2p, rhmm->ciphone, rhmm->ci2phone,
dict_last_phone(dict, bp->wid));
assert(IS_S3SSID(hmm_mpx_ssid(&rhmm->hmm, 0)));
E_DEBUG(6,("ssid %d(%d,%d) = %d\n",
rhmm->ciphone, dict_last_phone(dict, bp->wid), rhmm->ci2phone,
hmm_mpx_ssid(&rhmm->hmm, 0)));
bitvec_set(ngs->word_active, w);
}
}
}
/* Get the best exit into silence. */
if (rssid)
silscore = rcss[rssid->cimap[ps_search_acmod(ngs)->mdef->sil]];
else
silscore = bp->score;
if (silscore BETTER_THAN best_silrc_score) {
best_silrc_score = silscore;
best_silrc_bp = b;
}
}
/* Transition to <sil> */
newscore = best_silrc_score + ngs->silpen + pip;
if ((newscore BETTER_THAN thresh) && (newscore BETTER_THAN WORST_SCORE)) {
w = ps_search_silence_wid(ngs);
rhmm = (root_chan_t *) ngs->word_chan[w];
if ((hmm_frame(&rhmm->hmm) < cf)
|| (newscore BETTER_THAN hmm_in_score(&rhmm->hmm))) {
hmm_enter(&rhmm->hmm, newscore,
best_silrc_bp, nf);
bitvec_set(ngs->word_active, w);
}
}
/* Transition to noise words */
newscore = best_silrc_score + ngs->fillpen + pip;
if ((newscore BETTER_THAN thresh) && (newscore BETTER_THAN WORST_SCORE)) {
for (w = ps_search_silence_wid(ngs) + 1; w < ps_search_n_words(ngs); w++) {
rhmm = (root_chan_t *) ngs->word_chan[w];
/* Noise words that aren't a single phone will have NULL here. */
if (rhmm == NULL)
continue;
if ((hmm_frame(&rhmm->hmm) < cf)
|| (newscore BETTER_THAN hmm_in_score(&rhmm->hmm))) {
hmm_enter(&rhmm->hmm, newscore,
best_silrc_bp, nf);
bitvec_set(ngs->word_active, w);
}
}
}
/* Reset initial channels of words that have become inactive even after word trans. */
i = ngs->n_active_word[cf & 0x1];
awl = ngs->active_word_list[cf & 0x1];
for (w = *(awl++); i > 0; --i, w = *(awl++)) {
rhmm = (root_chan_t *) ngs->word_chan[w];
if (hmm_frame(&rhmm->hmm) == cf) {
hmm_clear_scores(&rhmm->hmm);
}
}
}
static void
fwdflat_prune_chan(ngram_search_t *ngs, int frame_idx)
{
int32 i, cf, nf, w, pip, newscore, thresh, wordthresh;
int32 *awl;
root_chan_t *rhmm;
chan_t *hmm, *nexthmm;
cf = frame_idx;
nf = cf + 1;
i = ngs->n_active_word[cf & 0x1];
awl = ngs->active_word_list[cf & 0x1];
bitvec_clear_all(ngs->word_active, ps_search_n_words(ngs));
thresh = ngs->best_score + ngs->fwdflatbeam;
wordthresh = ngs->best_score + ngs->fwdflatwbeam;
pip = ngs->pip;
E_DEBUG(3,("frame %d thresh %d wordthresh %d\n", frame_idx, thresh, wordthresh));
/* Scan all active words. */
for (w = *(awl++); i > 0; --i, w = *(awl++)) {
rhmm = (root_chan_t *) ngs->word_chan[w];
/* Propagate active root channels */
if (hmm_frame(&rhmm->hmm) == cf
&& hmm_bestscore(&rhmm->hmm) BETTER_THAN thresh) {
hmm_frame(&rhmm->hmm) = nf;
bitvec_set(ngs->word_active, w);
/* Transitions out of root channel */
newscore = hmm_out_score(&rhmm->hmm);
if (rhmm->next) {
assert(!dict_is_single_phone(ps_search_dict(ngs), w));
newscore += pip;
if (newscore BETTER_THAN thresh) {
hmm = rhmm->next;
/* Enter all right context phones */
if (hmm->info.rc_id >= 0) {
for (; hmm; hmm = hmm->next) {
if ((hmm_frame(&hmm->hmm) < cf)
|| (newscore BETTER_THAN hmm_in_score(&hmm->hmm))) {
hmm_enter(&hmm->hmm, newscore,
hmm_out_history(&rhmm->hmm), nf);
}
}
}
/* Just a normal word internal phone */
else {
if ((hmm_frame(&hmm->hmm) < cf)
|| (newscore BETTER_THAN hmm_in_score(&hmm->hmm))) {
hmm_enter(&hmm->hmm, newscore,
hmm_out_history(&rhmm->hmm), nf);
}
}
}
}
else {
assert(dict_is_single_phone(ps_search_dict(ngs), w));
/* Word exit for single-phone words (where did their
* whmms come from?) (either from
* ngram_search_fwdtree, or from
* ngram_fwdflat_allocate_1ph(), that's where) */
if (newscore BETTER_THAN wordthresh) {
ngram_search_save_bp(ngs, cf, w, newscore,
hmm_out_history(&rhmm->hmm), 0);
}
}
}
/* Transitions out of non-root channels. */
for (hmm = rhmm->next; hmm; hmm = hmm->next) {
if (hmm_frame(&hmm->hmm) >= cf) {
/* Propagate forward HMMs inside the beam. */
if (hmm_bestscore(&hmm->hmm) BETTER_THAN thresh) {
hmm_frame(&hmm->hmm) = nf;
bitvec_set(ngs->word_active, w);
newscore = hmm_out_score(&hmm->hmm);
/* Word-internal phones */
if (hmm->info.rc_id < 0) {
newscore += pip;
if (newscore BETTER_THAN thresh) {
nexthmm = hmm->next;
/* Enter all right-context phones. */
if (nexthmm->info.rc_id >= 0) {
for (; nexthmm; nexthmm = nexthmm->next) {
if ((hmm_frame(&nexthmm->hmm) < cf)
|| (newscore BETTER_THAN
hmm_in_score(&nexthmm->hmm))) {
hmm_enter(&nexthmm->hmm,
newscore,
hmm_out_history(&hmm->hmm),
nf);
}
}
}
/* Enter single word-internal phone. */
else {
if ((hmm_frame(&nexthmm->hmm) < cf)
|| (newscore BETTER_THAN
hmm_in_score(&nexthmm->hmm))) {
hmm_enter(&nexthmm->hmm, newscore,
hmm_out_history(&hmm->hmm), nf);
}
}
}
}
/* Right-context phones - apply word beam and exit. */
else {
if (newscore BETTER_THAN wordthresh) {
ngram_search_save_bp(ngs, cf, w, newscore,
hmm_out_history(&hmm->hmm),
hmm->info.rc_id);
}
}
}
/* Zero out inactive HMMs. */
else if (hmm_frame(&hmm->hmm) != nf) {
hmm_clear_scores(&hmm->hmm);
}
}
}
}
}