static int
process_ctl_line(ps_decoder_t *ps, cmd_ln_t *config,
char const *file, char const *uttid, int32 sf, int32 ef)
{
FILE *infh;
char const *cepdir, *cepext;
char *infile;
if (ef != -1 && ef < sf) {
E_ERROR("End frame %d is < start frame %d\n", ef, sf);
return -1;
}
cepdir = cmd_ln_str_r(config, "-cepdir");
cepext = cmd_ln_str_r(config, "-cepext");
/* Build input filename. */
infile = string_join(cepdir ? cepdir : "",
"/", file,
cepext ? cepext : "", NULL);
if (uttid == NULL) uttid = file;
if ((infh = fopen(infile, "rb")) == NULL) {
E_ERROR_SYSTEM("Failed to open %s", infile);
ckd_free(infile);
return -1;
}
/* Build output directories. */
if (cmd_ln_boolean_r(config, "-build_outdirs"))
build_outdirs(config, uttid);
if (cmd_ln_boolean_r(config, "-senin")) {
/* start and end frames not supported. */
ps_decode_senscr(ps, infh, uttid);
}
else if (cmd_ln_boolean_r(config, "-adcin")) {
if (ef != -1) {
ef = (int32)((ef - sf)
* (cmd_ln_float32_r(config, "-samprate")
/ cmd_ln_int32_r(config, "-frate"))
+ (cmd_ln_float32_r(config, "-samprate")
* cmd_ln_float32_r(config, "-wlen")));
}
sf = (int32)(sf
* (cmd_ln_float32_r(config, "-samprate")
/ cmd_ln_int32_r(config, "-frate")));
fseek(infh, cmd_ln_int32_r(config, "-adchdr") + sf * sizeof(int16), SEEK_SET);
ps_decode_raw(ps, infh, uttid, ef);
}
else {
mfcc_t **mfcs;
int nfr;
if (NULL == (mfcs = read_mfc_file(infh, sf, ef, &nfr,
cmd_ln_int32_r(config, "-ceplen")))) {
E_ERROR("Failed to read MFCC from the file '%s'\n", infile);
fclose(infh);
ckd_free(infile);
return -1;
}
ps_start_utt(ps, uttid);
ps_process_cep(ps, mfcs, nfr, FALSE, TRUE);
ps_end_utt(ps);
ckd_free_2d(mfcs);
}
fclose(infh);
ckd_free(infile);
return 0;
}
int
ps_decoder_test(cmd_ln_t *config, char const *sname, char const *expected)
{
ps_decoder_t *ps;
mfcc_t **cepbuf;
FILE *rawfh;
int16 *buf;
int16 const *bptr;
size_t nread;
size_t nsamps;
int32 nfr, i, score, prob;
char const *hyp;
char const *uttid;
double n_speech, n_cpu, n_wall;
ps_seg_t *seg;
TEST_ASSERT(ps = ps_init(config));
/* Test it first with pocketsphinx_decode_raw() */
TEST_ASSERT(rawfh = fopen(DATADIR "/goforward.raw", "rb"));
ps_decode_raw(ps, rawfh, "goforward", -1);
hyp = ps_get_hyp(ps, &score, &uttid);
prob = ps_get_prob(ps, &uttid);
printf("%s (%s): %s (%d, %d)\n", sname, uttid, hyp, score, prob);
TEST_EQUAL(0, strcmp(hyp, expected));
TEST_ASSERT(prob <= 0);
ps_get_utt_time(ps, &n_speech, &n_cpu, &n_wall);
printf("%.2f seconds speech, %.2f seconds CPU, %.2f seconds wall\n",
n_speech, n_cpu, n_wall);
printf("%.2f xRT (CPU), %.2f xRT (elapsed)\n",
n_cpu / n_speech, n_wall / n_speech);
/* Test it with ps_process_raw() */
clearerr(rawfh);
fseek(rawfh, 0, SEEK_END);
nsamps = ftell(rawfh) / sizeof(*buf);
fseek(rawfh, 0, SEEK_SET);
TEST_EQUAL(0, ps_start_utt(ps, NULL));
nsamps = 2048;
buf = ckd_calloc(nsamps, sizeof(*buf));
while (!feof(rawfh)) {
nread = fread(buf, sizeof(*buf), nsamps, rawfh);
ps_process_raw(ps, buf, nread, FALSE, FALSE);
}
TEST_EQUAL(0, ps_end_utt(ps));
hyp = ps_get_hyp(ps, &score, &uttid);
prob = ps_get_prob(ps, &uttid);
printf("%s (%s): %s (%d, %d)\n", sname, uttid, hyp, score, prob);
TEST_EQUAL(0, strcmp(uttid, "000000000"));
TEST_EQUAL(0, strcmp(hyp, expected));
ps_get_utt_time(ps, &n_speech, &n_cpu, &n_wall);
printf("%.2f seconds speech, %.2f seconds CPU, %.2f seconds wall\n",
n_speech, n_cpu, n_wall);
printf("%.2f xRT (CPU), %.2f xRT (elapsed)\n",
n_cpu / n_speech, n_wall / n_speech);
/* Now read the whole file and produce an MFCC buffer. */
clearerr(rawfh);
fseek(rawfh, 0, SEEK_END);
nsamps = ftell(rawfh) / sizeof(*buf);
fseek(rawfh, 0, SEEK_SET);
bptr = buf = ckd_realloc(buf, nsamps * sizeof(*buf));
TEST_EQUAL(nsamps, fread(buf, sizeof(*buf), nsamps, rawfh));
fe_process_frames(ps->acmod->fe, &bptr, &nsamps, NULL, &nfr);
cepbuf = ckd_calloc_2d(nfr + 1,
fe_get_output_size(ps->acmod->fe),
sizeof(**cepbuf));
fe_start_utt(ps->acmod->fe);
fe_process_frames(ps->acmod->fe, &bptr, &nsamps, cepbuf, &nfr);
fe_end_utt(ps->acmod->fe, cepbuf[nfr], &i);
/* Decode it with process_cep() */
TEST_EQUAL(0, ps_start_utt(ps, NULL));
for (i = 0; i < nfr; ++i) {
ps_process_cep(ps, cepbuf + i, 1, FALSE, FALSE);
}
TEST_EQUAL(0, ps_end_utt(ps));
hyp = ps_get_hyp(ps, &score, &uttid);
prob = ps_get_prob(ps, &uttid);
printf("%s (%s): %s (%d, %d)\n", sname, uttid, hyp, score, prob);
TEST_EQUAL(0, strcmp(uttid, "000000001"));
TEST_EQUAL(0, strcmp(hyp, expected));
TEST_ASSERT(prob <= 0);
for (seg = ps_seg_iter(ps, &score); seg;
seg = ps_seg_next(seg)) {
char const *word;
int sf, ef;
int32 post, lscr, ascr, lback;
word = ps_seg_word(seg);
ps_seg_frames(seg, &sf, &ef);
post = ps_seg_prob(seg, &ascr, &lscr, &lback);
printf("%s (%d:%d) P(w|o) = %f ascr = %d lscr = %d lback = %d\n", word, sf, ef,
logmath_exp(ps_get_logmath(ps), post), ascr, lscr, lback);
TEST_ASSERT(post <= 2); // Due to numerical errors with float it sometimes could go out of 0
}
ps_get_utt_time(ps, &n_speech, &n_cpu, &n_wall);
printf("%.2f seconds speech, %.2f seconds CPU, %.2f seconds wall\n",
n_speech, n_cpu, n_wall);
printf("%.2f xRT (CPU), %.2f xRT (elapsed)\n",
n_cpu / n_speech, n_wall / n_speech);
ps_get_all_time(ps, &n_speech, &n_cpu, &n_wall);
printf("TOTAL: %.2f seconds speech, %.2f seconds CPU, %.2f seconds wall\n",
n_speech, n_cpu, n_wall);
printf("TOTAL: %.2f xRT (CPU), %.2f xRT (elapsed)\n",
n_cpu / n_speech, n_wall / n_speech);
fclose(rawfh);
ps_free(ps);
cmd_ln_free_r(config);
ckd_free_2d(cepbuf);
ckd_free(buf);
return 0;
}
int
main(int argc, char *argv[])
{
cmd_ln_t *config;
ps_decoder_t *ps;
int32 out_score;
const char *input_file_path;
const char *cfg;
const char *utt_id;
const char *hyp;
FILE *input_file;
int16 buf[PCM_BUF_LEN];
int k;
config = cmd_ln_parse_r(NULL, cont_args_def, argc, argv, TRUE);
/* Handle argument file as -argfile. */
if (config && (cfg = cmd_ln_str_r(config, "-argfile")) != NULL) {
config = cmd_ln_parse_file_r(config, cont_args_def, cfg, FALSE);
}
if (config == NULL)
return 1;
if (cmd_ln_str_r(config, "-kws") == NULL && cmd_ln_str_r(config, "-keyphrase") == NULL) {
E_ERROR("Keyword is missing. Use -keyphrase <keyphrase> or -kws <kws_file> to specify the phrase to look for.");
return 1;
}
input_file_path = cmd_ln_str_r(config, "-infile");
if (input_file_path == NULL) {
E_ERROR("Input file is missing. Use -infile <input_file> to specify the file to look in.\n");
return 1;
}
ps_default_search_args(config);
ps = ps_init(config);
if (ps == NULL) {
E_ERROR("Failed to create the decoder\n");
return 1;
}
input_file = fopen(input_file_path, "rb");
if (input_file == NULL) {
E_FATAL_SYSTEM("Failed to open input file '%s'", input_file_path);
}
ps_start_utt(ps, NULL);
if (cmd_ln_boolean_r(config, "-adcin")) {
fread(buf, 1, 44, input_file);
while ((k = fread(buf, sizeof(int16), PCM_BUF_LEN, input_file)) > 0) {
ps_process_raw(ps, buf, k, FALSE, FALSE);
}
} else {
mfcc_t **mfcs;
int nfr;
if (NULL == (mfcs = read_mfc_file(input_file, &nfr, cmd_ln_int32_r(config, "-ceplen")))) {
E_ERROR("Failed to read MFCC from the file '%s'\n", input_file_path);
fclose(input_file);
return -1;
}
ps_process_cep(ps, mfcs, nfr, FALSE, TRUE);
ckd_free_2d(mfcs);
}
ps_end_utt(ps);
hyp = ps_get_hyp(ps, &out_score, &utt_id);
printf("hypothesis: %s\n", hyp);
fflush(stdout);
fclose(input_file);
ps_free(ps);
cmd_ln_free_r(config);
return 0;
}
