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 int
state_align_search_start(ps_search_t *search)
{
state_align_search_t *sas = (state_align_search_t *)search;
/* Activate the initial state. */
hmm_enter(sas->hmms, 0, 0, 0);
return 0;
}
static int
phone_loop_search_start(ps_search_t *search)
{
phone_loop_search_t *pls = (phone_loop_search_t *)search;
int i;
/* Reset and enter all phone HMMs. */
for (i = 0; i < pls->n_phones; ++i) {
hmm_t *hmm = (hmm_t *)&pls->phones[i];
hmm_clear(hmm);
hmm_enter(hmm, 0, -1, 0);
}
phone_loop_search_free_renorm(pls);
pls->best_score = 0;
return 0;
}
void
ngram_fwdflat_start(ngram_search_t *ngs)
{
root_chan_t *rhmm;
int i;
ptmr_reset(&ngs->fwdflat_perf);
ptmr_start(&ngs->fwdflat_perf);
build_fwdflat_wordlist(ngs);
build_fwdflat_chan(ngs);
ngs->bpidx = 0;
ngs->bss_head = 0;
for (i = 0; i < ps_search_n_words(ngs); i++)
ngs->word_lat_idx[i] = NO_BP;
/* Reset the permanently allocated single-phone words, since they
* may have junk left over in them from previous searches. */
for (i = 0; i < ngs->n_1ph_words; i++) {
int32 w = ngs->single_phone_wid[i];
rhmm = (root_chan_t *) ngs->word_chan[w];
hmm_clear(&rhmm->hmm);
}
/* Start search with <s>; word_chan[<s>] is permanently allocated */
rhmm = (root_chan_t *) ngs->word_chan[ps_search_start_wid(ngs)];
hmm_enter(&rhmm->hmm, 0, NO_BP, 0);
ngs->active_word_list[0][0] = ps_search_start_wid(ngs);
ngs->n_active_word[0] = 1;
ngs->best_score = 0;
ngs->renormalized = FALSE;
for (i = 0; i < ps_search_n_words(ngs); i++)
ngs->last_ltrans[i].sf = -1;
if (!ngs->fwdtree)
ngs->n_frame = 0;
ngs->st.n_fwdflat_chan = 0;
ngs->st.n_fwdflat_words = 0;
ngs->st.n_fwdflat_word_transition = 0;
ngs->st.n_senone_active_utt = 0;
}
static int
phone_loop_search_start(ps_search_t *search)
{
phone_loop_search_t *pls = (phone_loop_search_t *)search;
int i;
/* Reset and enter all phone HMMs. */
for (i = 0; i < pls->n_phones; ++i) {
hmm_t *hmm = (hmm_t *)&pls->hmms[i];
hmm_clear(hmm);
hmm_enter(hmm, 0, -1, 0);
}
memset(pls->penalties, 0, pls->n_phones * sizeof(*pls->penalties));
for (i = 0; i < pls->window; i++)
memset(pls->pen_buf[i], 0, pls->n_phones * sizeof(*pls->pen_buf[i]));
phone_loop_search_free_renorm(pls);
pls->best_score = 0;
pls->pen_buf_ptr = 0;
return 0;
}
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_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);
}
}
}
}
}
