void
run_acmod_test(acmod_t *acmod)
{
FILE *rawfh;
int16 *buf;
int16 const *bptr;
size_t nread, nsamps;
int nfr;
int frame_counter;
cmn_live_set(acmod->fcb->cmn_struct, cmninit);
nsamps = 2048;
frame_counter = 0;
buf = ckd_calloc(nsamps, sizeof(*buf));
TEST_ASSERT(rawfh = fopen(DATADIR "/goforward.raw", "rb"));
TEST_EQUAL(0, acmod_start_utt(acmod));
printf("Incremental(2048):\n");
while (!feof(rawfh)) {
nread = fread(buf, sizeof(*buf), nsamps, rawfh);
bptr = buf;
while ((nfr = acmod_process_raw(acmod, &bptr, &nread, FALSE)) > 0) {
int16 best_score;
int frame_idx = -1, best_senid;
while (acmod->n_feat_frame > 0) {
acmod_score(acmod, &frame_idx);
acmod_advance(acmod);
best_score = acmod_best_score(acmod, &best_senid);
printf("Frame %d best senone %d score %d\n",
frame_idx, best_senid, best_score);
TEST_EQUAL(frame_counter, frame_idx);
++frame_counter;
frame_idx = -1;
}
}
}
TEST_EQUAL(0, acmod_end_utt(acmod));
nread = 0;
{
int16 best_score;
int frame_idx = -1, best_senid;
while (acmod->n_feat_frame > 0) {
acmod_score(acmod, &frame_idx);
acmod_advance(acmod);
best_score = acmod_best_score(acmod, &best_senid);
printf("Frame %d best senone %d score %d\n",
frame_idx, best_senid, best_score);
TEST_EQUAL(frame_counter, frame_idx);
++frame_counter;
frame_idx = -1;
}
}
fclose(rawfh);
}
static int
phone_loop_search_step(ps_search_t *search, int frame_idx)
{
phone_loop_search_t *pls = (phone_loop_search_t *)search;
acmod_t *acmod = ps_search_acmod(search);
int16 const *senscr;
int i;
/* All CI senones are active all the time. */
if (!ps_search_acmod(pls)->compallsen)
for (i = 0; i < pls->n_phones; ++i)
acmod_activate_hmm(acmod, (hmm_t *)&pls->phones[i]);
/* Calculate senone scores for current frame. */
senscr = acmod_score(acmod, &frame_idx);
/* Renormalize, if necessary. */
if (pls->best_score + (2 * pls->beam) WORSE_THAN WORST_SCORE) {
E_INFO("Renormalizing Scores at frame %d, best score %d\n",
frame_idx, pls->best_score);
renormalize_hmms(pls, frame_idx, pls->best_score);
}
/* Evaluate phone HMMs for current frame. */
pls->best_score = evaluate_hmms(pls, senscr, frame_idx);
/* Prune phone HMMs. */
prune_hmms(pls, frame_idx);
/* Do phone transitions. */
phone_transition(pls, frame_idx);
return 0;
}
static int
consumer(sbthread_t *th)
{
acmod_t *acmod = sbthread_arg(th);
int frame_idx;
printf("Consumer %p started\n", acmod);
acmod_consumer_start_utt(acmod, -1);
while ((frame_idx = acmod_consumer_wait(acmod, -1)) >= 0) {
int senid, score;
/* Score a frame. */
acmod_score(acmod, frame_idx);
score = acmod_best_score(acmod, &senid);
printf("Consumer %p scored frame %d best %d score %d\n",
acmod, frame_idx, senid, score);
/* Release it. */
acmod_consumer_release(acmod, frame_idx);
}
TEST_ASSERT(acmod_eou(acmod));
acmod_consumer_end_utt(acmod);
printf("Consumer %p exiting\n", acmod);
return 0;
}
static int
state_align_search_step(ps_search_t *search, int frame_idx)
{
state_align_search_t *sas = (state_align_search_t *)search;
acmod_t *acmod = ps_search_acmod(search);
int16 const *senscr;
int i;
/* Calculate senone scores. */
for (i = 0; i < sas->n_phones; ++i)
acmod_activate_hmm(acmod, sas->hmms + i);
senscr = acmod_score(acmod, &frame_idx);
/* Renormalize here if needed. */
/* FIXME: Make sure to (unit-)test this!!! */
if ((sas->best_score - 0x300000) WORSE_THAN WORST_SCORE) {
E_INFO("Renormalizing Scores at frame %d, best score %d\n",
frame_idx, sas->best_score);
renormalize_hmms(sas, frame_idx, sas->best_score);
}
/* Viterbi step. */
sas->best_score = evaluate_hmms(sas, senscr, frame_idx);
prune_hmms(sas, frame_idx);
/* Transition out of non-emitting states. */
phone_transition(sas, frame_idx);
/* Generate new tokens from best path results. */
record_transitions(sas, frame_idx);
/* Update frame counter */
sas->frame = frame_idx;
return 0;
}
int
main(int argc, char *argv[])
{
acmod_t *acmod;
logmath_t *lmath;
cmd_ln_t *config;
FILE *rawfh;
int16 *buf;
int16 const *bptr;
mfcc_t **cepbuf, **cptr;
size_t nread, nsamps;
int nfr;
int frame_counter;
int bestsen1[270];
lmath = logmath_init(1.0001, 0, 0);
config = cmd_ln_init(NULL, ps_args(), TRUE,
"-featparams", MODELDIR "/hmm/en_US/hub4wsj_sc_8k/feat.params",
"-mdef", MODELDIR "/hmm/en_US/hub4wsj_sc_8k/mdef",
"-mean", MODELDIR "/hmm/en_US/hub4wsj_sc_8k/means",
"-var", MODELDIR "/hmm/en_US/hub4wsj_sc_8k/variances",
"-tmat", MODELDIR "/hmm/en_US/hub4wsj_sc_8k/transition_matrices",
"-sendump", MODELDIR "/hmm/en_US/hub4wsj_sc_8k/sendump",
"-compallsen", "true",
"-cmn", "prior",
"-tmatfloor", "0.0001",
"-mixwfloor", "0.001",
"-varfloor", "0.0001",
"-mmap", "no",
"-topn", "4",
"-ds", "1",
"-input_endian", "little",
"-samprate", "16000", NULL);
TEST_ASSERT(config);
TEST_ASSERT(acmod = acmod_init(config, lmath, NULL, NULL));
cmn_prior_set(acmod->fcb->cmn_struct, prior);
nsamps = 2048;
frame_counter = 0;
buf = ckd_calloc(nsamps, sizeof(*buf));
TEST_ASSERT(rawfh = fopen(DATADIR "/goforward.raw", "rb"));
TEST_EQUAL(0, acmod_start_utt(acmod));
E_INFO("Incremental(2048):\n");
while (!feof(rawfh)) {
nread = fread(buf, sizeof(*buf), nsamps, rawfh);
bptr = buf;
while ((nfr = acmod_process_raw(acmod, &bptr, &nread, FALSE)) > 0 || nread > 0) {
int16 const *senscr;
int16 best_score;
int frame_idx = -1, best_senid;
while (acmod->n_feat_frame > 0) {
senscr = acmod_score(acmod, &frame_idx);
acmod_advance(acmod);
best_score = acmod_best_score(acmod, &best_senid);
E_INFO("Frame %d best senone %d score %d\n",
frame_idx, best_senid, best_score);
TEST_EQUAL(frame_counter, frame_idx);
if (frame_counter < 190)
bestsen1[frame_counter] = best_score;
++frame_counter;
frame_idx = -1;
}
}
}
TEST_EQUAL(0, acmod_end_utt(acmod));
nread = 0;
{
int16 const *senscr;
int16 best_score;
int frame_idx = -1, best_senid;
while (acmod->n_feat_frame > 0) {
senscr = acmod_score(acmod, &frame_idx);
acmod_advance(acmod);
best_score = acmod_best_score(acmod, &best_senid);
E_INFO("Frame %d best senone %d score %d\n",
frame_idx, best_senid, best_score);
if (frame_counter < 190)
bestsen1[frame_counter] = best_score;
TEST_EQUAL(frame_counter, frame_idx);
++frame_counter;
frame_idx = -1;
}
}
/* Now try to process the whole thing at once. */
E_INFO("Whole utterance:\n");
cmn_prior_set(acmod->fcb->cmn_struct, prior);
nsamps = ftell(rawfh) / sizeof(*buf);
clearerr(rawfh);
fseek(rawfh, 0, SEEK_SET);
buf = ckd_realloc(buf, nsamps * sizeof(*buf));
TEST_EQUAL(nsamps, fread(buf, sizeof(*buf), nsamps, rawfh));
bptr = buf;
TEST_EQUAL(0, acmod_start_utt(acmod));
acmod_process_raw(acmod, &bptr, &nsamps, TRUE);
TEST_EQUAL(0, acmod_end_utt(acmod));
{
int16 const *senscr;
int16 best_score;
int frame_idx = -1, best_senid;
frame_counter = 0;
while (acmod->n_feat_frame > 0) {
senscr = acmod_score(acmod, &frame_idx);
acmod_advance(acmod);
best_score = acmod_best_score(acmod, &best_senid);
E_INFO("Frame %d best senone %d score %d\n",
frame_idx, best_senid, best_score);
if (frame_counter < 190)
TEST_EQUAL_LOG(best_score, bestsen1[frame_counter]);
TEST_EQUAL(frame_counter, frame_idx);
++frame_counter;
frame_idx = -1;
}
}
/* Now process MFCCs and make sure we get the same results. */
cepbuf = ckd_calloc_2d(frame_counter,
fe_get_output_size(acmod->fe),
sizeof(**cepbuf));
fe_start_utt(acmod->fe);
nsamps = ftell(rawfh) / sizeof(*buf);
bptr = buf;
nfr = frame_counter;
fe_process_frames(acmod->fe, &bptr, &nsamps, cepbuf, &nfr);
fe_end_utt(acmod->fe, cepbuf[frame_counter-1], &nfr);
E_INFO("Incremental(MFCC):\n");
cmn_prior_set(acmod->fcb->cmn_struct, prior);
TEST_EQUAL(0, acmod_start_utt(acmod));
cptr = cepbuf;
nfr = frame_counter;
frame_counter = 0;
while ((acmod_process_cep(acmod, &cptr, &nfr, FALSE)) > 0) {
int16 const *senscr;
int16 best_score;
int frame_idx = -1, best_senid;
while (acmod->n_feat_frame > 0) {
senscr = acmod_score(acmod, &frame_idx);
acmod_advance(acmod);
best_score = acmod_best_score(acmod, &best_senid);
E_INFO("Frame %d best senone %d score %d\n",
frame_idx, best_senid, best_score);
TEST_EQUAL(frame_counter, frame_idx);
if (frame_counter < 190)
TEST_EQUAL_LOG(best_score, bestsen1[frame_counter]);
++frame_counter;
frame_idx = -1;
}
}
TEST_EQUAL(0, acmod_end_utt(acmod));
nfr = 0;
acmod_process_cep(acmod, &cptr, &nfr, FALSE);
{
int16 const *senscr;
int16 best_score;
int frame_idx = -1, best_senid;
while (acmod->n_feat_frame > 0) {
senscr = acmod_score(acmod, &frame_idx);
acmod_advance(acmod);
best_score = acmod_best_score(acmod, &best_senid);
E_INFO("Frame %d best senone %d score %d\n",
frame_idx, best_senid, best_score);
TEST_EQUAL(frame_counter, frame_idx);
if (frame_counter < 190)
TEST_EQUAL_LOG(best_score, bestsen1[frame_counter]);
++frame_counter;
frame_idx = -1;
}
}
/* Note that we have to process the whole thing again because
* !#@$@ s2mfc2feat modifies its argument (not for long) */
fe_start_utt(acmod->fe);
nsamps = ftell(rawfh) / sizeof(*buf);
bptr = buf;
nfr = frame_counter;
fe_process_frames(acmod->fe, &bptr, &nsamps, cepbuf, &nfr);
fe_end_utt(acmod->fe, cepbuf[frame_counter-1], &nfr);
E_INFO("Whole utterance (MFCC):\n");
cmn_prior_set(acmod->fcb->cmn_struct, prior);
TEST_EQUAL(0, acmod_start_utt(acmod));
cptr = cepbuf;
nfr = frame_counter;
acmod_process_cep(acmod, &cptr, &nfr, TRUE);
TEST_EQUAL(0, acmod_end_utt(acmod));
{
int16 const *senscr;
int16 best_score;
int frame_idx = -1, best_senid;
frame_counter = 0;
while (acmod->n_feat_frame > 0) {
senscr = acmod_score(acmod, &frame_idx);
acmod_advance(acmod);
best_score = acmod_best_score(acmod, &best_senid);
E_INFO("Frame %d best senone %d score %d\n",
frame_idx, best_senid, best_score);
if (frame_counter < 190)
TEST_EQUAL_LOG(best_score, bestsen1[frame_counter]);
TEST_EQUAL(frame_counter, frame_idx);
++frame_counter;
frame_idx = -1;
}
}
E_INFO("Rewound (MFCC):\n");
TEST_EQUAL(0, acmod_rewind(acmod));
{
int16 const *senscr;
int16 best_score;
int frame_idx = -1, best_senid;
frame_counter = 0;
while (acmod->n_feat_frame > 0) {
senscr = acmod_score(acmod, &frame_idx);
acmod_advance(acmod);
best_score = acmod_best_score(acmod, &best_senid);
E_INFO("Frame %d best senone %d score %d\n",
frame_idx, best_senid, best_score);
if (frame_counter < 190)
TEST_EQUAL_LOG(best_score, bestsen1[frame_counter]);
TEST_EQUAL(frame_counter, frame_idx);
++frame_counter;
frame_idx = -1;
}
}
/* Clean up, go home. */
ckd_free_2d(cepbuf);
fclose(rawfh);
ckd_free(buf);
acmod_free(acmod);
logmath_free(lmath);
cmd_ln_free_r(config);
return 0;
}
int
ngram_fwdflat_search(ngram_search_t *ngs, int frame_idx)
{
int16 const *senscr;
int32 nf, i, j;
int32 *nawl;
/* Activate our HMMs for the current frame if need be. */
if (!ps_search_acmod(ngs)->compallsen)
compute_fwdflat_sen_active(ngs, frame_idx);
/* Compute GMM scores for the current frame. */
senscr = acmod_score(ps_search_acmod(ngs), &frame_idx);
ngs->st.n_senone_active_utt += ps_search_acmod(ngs)->n_senone_active;
/* Mark backpointer table for current frame. */
ngram_search_mark_bptable(ngs, frame_idx);
/* If the best score is equal to or worse than WORST_SCORE,
* recognition has failed, don't bother to keep trying. */
if (ngs->best_score == WORST_SCORE || ngs->best_score WORSE_THAN WORST_SCORE)
return 0;
/* Renormalize if necessary */
if (ngs->best_score + (2 * ngs->beam) WORSE_THAN WORST_SCORE) {
E_INFO("Renormalizing Scores at frame %d, best score %d\n",
frame_idx, ngs->best_score);
fwdflat_renormalize_scores(ngs, frame_idx, ngs->best_score);
}
ngs->best_score = WORST_SCORE;
hmm_context_set_senscore(ngs->hmmctx, senscr);
/* Evaluate HMMs */
fwdflat_eval_chan(ngs, frame_idx);
/* Prune HMMs and do phone transitions. */
fwdflat_prune_chan(ngs, frame_idx);
/* Do word transitions. */
fwdflat_word_transition(ngs, frame_idx);
/* Create next active word list */
nf = frame_idx + 1;
nawl = ngs->active_word_list[nf & 0x1];
for (i = 0, j = 0; ngs->fwdflat_wordlist[i] >= 0; i++) {
if (bitvec_is_set(ngs->word_active, ngs->fwdflat_wordlist[i])) {
*(nawl++) = ngs->fwdflat_wordlist[i];
j++;
}
}
for (i = ps_search_start_wid(ngs); i < ps_search_n_words(ngs); i++) {
if (bitvec_is_set(ngs->word_active, i)) {
*(nawl++) = i;
j++;
}
}
if (!ngs->fwdtree)
++ngs->n_frame;
ngs->n_active_word[nf & 0x1] = j;
/* Return the number of frames processed. */
return 1;
}
