int
corpus_set_interval(uint32 n_skip,
uint32 run_len)
{
sv_n_skip = n_skip;
sv_run_len = run_len;
if (n_skip) {
E_INFO("skipping %d utts.\n", n_skip);
for (begin = 0; (n_skip > 0) && corpus_next_utt(); --n_skip, begin++);
E_INFO("Last utt skipped: %s\n", corpus_utt());
}
if (run_len != UNTIL_EOF)
n_run = run_len;
n_proc = 0;
return S3_SUCCESS;
}
int
agg_phn_seg(lexicon_t *lex,
acmod_set_t *acmod_set,
feat_t *fcb,
segdmp_type_t type)
{
uint16 *seg;
vector_t *mfcc;
vector_t **feat;
int32 n_frame;
uint32 tick_cnt;
acmod_id_t *phone;
uint32 *start;
uint32 *len;
uint32 n_phone;
uint32 s;
char *btw_mark;
char *trans;
char **word;
uint32 n_word;
int32 mfc_veclen = cmd_ln_int32("-ceplen");
uint32 n_stream;
uint32 *veclen;
tick_cnt = 0;
n_stream = feat_dimension1(fcb);
veclen = feat_stream_lengths(fcb);
while (corpus_next_utt()) {
if ((++tick_cnt % 500) == 0) {
E_INFOCONT("[%u] ", tick_cnt);
}
if (corpus_get_sent(&trans) != S3_SUCCESS) {
E_FATAL("Unable to read word transcript for %s\n", corpus_utt_brief_name());
}
if (corpus_get_seg(&seg, &n_frame) != S3_SUCCESS) {
E_FATAL("Unable to read Viterbi state segmentation for %s\n", corpus_utt_brief_name());
}
n_word = str2words(trans, NULL, 0);
word = ckd_calloc(n_word, sizeof(char*));
str2words(trans, word, n_word);
phone = mk_phone_list(&btw_mark, &n_phone, word, n_word, lex);
start = ckd_calloc(n_phone, sizeof(uint32));
len = ckd_calloc(n_phone, sizeof(uint32));
/* check to see whether the word transcript and dictionary entries
agree with the state segmentation */
if (ck_seg(acmod_set, phone, n_phone, seg, n_frame, corpus_utt()) != S3_SUCCESS) {
free(trans); /* alloc'ed using strdup, not ckd_*() */
free(seg); /* alloc'ed using malloc in areadshort(), not ckd_*() */
ckd_free(word);
ckd_free(phone);
E_ERROR("ck_seg failed");
continue;
}
if (cvt2triphone(acmod_set, phone, btw_mark, n_phone) != S3_SUCCESS) {
free(trans); /* alloc'ed using strdup, not ckd_*() */
free(seg); /* alloc'ed using malloc in areadshort(), not ckd_*() */
ckd_free(word);
ckd_free(phone);
E_ERROR("cvt2triphone failed");
continue;
}
ckd_free(btw_mark);
if (mk_seg(acmod_set,
seg,
n_frame,
phone,
start,
len,
n_phone) != S3_SUCCESS) {
free(trans);
free(seg);
ckd_free(word);
ckd_free(phone);
E_ERROR("mk_seg failed");
continue;
}
if (corpus_provides_mfcc()) {
if (corpus_get_generic_featurevec(&mfcc, &n_frame, mfc_veclen) < 0) {
E_FATAL("Can't read input features from %s\n", corpus_utt());
}
if (n_frame < 9) {
E_WARN("utt %s too short\n", corpus_utt());
if (mfcc) {
ckd_free(mfcc[0]);
ckd_free(mfcc);
mfcc = NULL;
}
continue;
}
feat = feat_array_alloc(fcb, n_frame + feat_window_size(fcb));
feat_s2mfc2feat_live(fcb, mfcc, &n_frame, TRUE, TRUE, feat);
for (s = 0; s < n_phone; s++) {
segdmp_add_feat(phone[s],
&feat[start[s]],
len[s]);
}
feat_array_free(feat);
free(&mfcc[0][0]);
ckd_free(mfcc);
}
else {
E_FATAL("No data type specified\n");
}
free(trans); /* alloc'ed using strdup, not ckd_*() */
free(seg); /* alloc'ed using malloc in areadshort(), not ckd_*() */
ckd_free(word);
ckd_free(phone);
ckd_free(start);
ckd_free(len);
}
return 0;
}
int
cnt_phn_seg(model_def_t *mdef,
lexicon_t *lex,
uint32 **out_n_seg,
uint32 ***out_n_frame_per)
{
uint32 seq_no = 0;
uint16 *seg;
uint32 n_frame;
uint32 i, j;
uint32 n_acmod;
uint32 *phone;
uint32 n_phone;
uint32 *n_seg;
uint32 **n_frame_per;
uint32 *start;
uint32 *len;
seg_len_t *cur;
seg_len_t *tmp;
seg_len_t *phn_hd;
seg_len_t *phn_tl;
n_acmod = acmod_set_n_acmod(mdef->acmod_set);
E_INFO("Counting # occ. for %u models\n", n_acmod);
n_seg = ckd_calloc(n_acmod, sizeof(uint32));
hd = ckd_calloc(n_acmod, sizeof(seg_len_t *));
tl = ckd_calloc(n_acmod, sizeof(seg_len_t *));
for (seq_no = corpus_get_begin(); corpus_next_utt(); seq_no++) {
if (!(seq_no % 250)) {
fprintf(stderr, " cnt[%u]", seq_no);
fflush(stderr);
}
corpus_get_seg(&seg, &n_frame);
phone = get_next_phnseq(mdef, lex, &n_phone);
ck_seg(mdef->acmod_set, phone, n_phone, seg, n_frame, corpus_utt());
start = ckd_calloc(n_phone, sizeof(uint32));
len = ckd_calloc(n_phone, sizeof(uint32));
mk_seg(mdef->acmod_set, seg, n_frame, phone, start, len, n_phone);
ckd_free(start);
ckd_free(seg);
ckd_free(phone);
for (i = 0; i < n_phone; i++) {
/* insert the len for list phone[i] */
phn_hd = hd[phone[i]];
phn_tl = tl[phone[i]];
cur = (seg_len_t *)ckd_calloc(1, sizeof(seg_len_t));
cur->len = len[i];
if (phn_tl == NULL) {
hd[phone[i]] = tl[phone[i]] = cur;
}
else {
phn_tl->nxt = cur;
tl[phone[i]] = cur;
}
}
ckd_free(len);
}
n_frame_per = (uint32 **)ckd_calloc(n_acmod, sizeof(uint32 *));
for (i = 0; i < n_acmod; i++) {
if (hd[i] == NULL) {
n_seg[i] = 0;
}
else {
for (cur = hd[i], j = 0; cur != NULL; j++, cur = cur->nxt);
n_seg[i] = j;
n_frame_per[i] = (uint32 *)ckd_calloc(n_seg[i], sizeof(uint32));
for (cur = hd[i], j = 0; cur != NULL; j++, cur = cur->nxt)
n_frame_per[i][j] = cur->len;
for (cur = hd[i]; cur != NULL; cur = tmp) {
tmp = cur->nxt;
ckd_free(cur);
}
E_INFO("phn= %s n_seg= %u\n",
acmod_set_id2name(mdef->acmod_set, i),
n_seg[i]);
}
}
ckd_free(hd);
ckd_free(tl);
*out_n_seg = n_seg;
*out_n_frame_per = n_frame_per;
return S3_SUCCESS;
}
int
agg_all_seg(feat_t *fcb,
segdmp_type_t type,
const char *fn,
uint32 stride)
{
uint32 seq_no;
vector_t *mfcc = NULL;
uint32 mfc_veclen = cmd_ln_int32("-ceplen");
uint32 n_frame;
uint32 n_out_frame;
uint32 blksz=0;
vector_t **feat = NULL;
uint32 i, j;
uint32 t;
uint32 n_stream;
const uint32 *veclen;
FILE *fp;
uint32 ignore = 0;
long start;
int32 no_retries=0;
n_stream = feat_dimension1(fcb);
veclen = feat_stream_lengths(fcb);
for (i = 0, blksz = 0; i < n_stream; i++)
blksz += veclen[i];
fp = open_dmp(fn);
start = ftell(fp);
if (s3write(&i, sizeof(uint32), 1, fp, &ignore) != 1) {
E_ERROR_SYSTEM("Unable to write to dmp file");
return S3_ERROR;
}
for (seq_no = corpus_get_begin(), j = 0, n_out_frame = 0;
corpus_next_utt(); seq_no++) {
if (mfcc) {
free(mfcc[0]);
ckd_free(mfcc);
mfcc = NULL;
}
/* get the MFCC data for the utterance */
if (corpus_get_generic_featurevec(&mfcc, &n_frame, mfc_veclen) < 0) {
E_FATAL("Can't read input features from %s\n", corpus_utt());
}
if ((seq_no % 1000) == 0) {
E_INFO("[%u]\n", seq_no);
}
if (feat) {
feat_array_free(feat);
feat = NULL;
}
if (n_frame < 9) {
E_WARN("utt %s too short\n", corpus_utt());
if (mfcc) {
ckd_free(mfcc[0]);
ckd_free(mfcc);
mfcc = NULL;
}
continue;
}
feat = feat_array_alloc(fcb, n_frame + feat_window_size(fcb));
feat_s2mfc2feat_live(fcb, mfcc, &n_frame, TRUE, TRUE, feat);
for (t = 0; t < n_frame; t++, j++) {
if ((j % stride) == 0) {
while (s3write(&feat[t][0][0],
sizeof(float32),
blksz,
fp, &ignore) != blksz) {
static int rpt = 0;
if (!rpt) {
E_ERROR_SYSTEM("Unable to write to dmp file");
E_INFO("sleeping...\n");
no_retries++;
}
sleep(3);
if(no_retries > 10){
E_FATAL("Failed to write to a dmp file after 10 retries of getting MFCC(about 30 seconds)\n ");
}
}
++n_out_frame;
}
}
}
if (fseek(fp, start, SEEK_SET) < 0) {
E_ERROR_SYSTEM("Unable to seek to begin of dmp");
return S3_ERROR;
}
E_INFO("Wrote %u frames to %s\n", n_out_frame, fn);
if (s3write((void *)&n_out_frame, sizeof(uint32), 1, fp, &ignore) != 1) {
E_ERROR_SYSTEM("Unable to write to dmp file");
return S3_ERROR;
}
return S3_SUCCESS;
}