int
main(int argc, char *argv[])
{
ps_decoder_t *ps;
cmd_ln_t *config;
FILE *fh;
char const *hyp, *uttid;
int16 buf[512];
int rv;
int32 score;
//int i;
config = cmd_ln_init(NULL, ps_args(), TRUE,
"-hmm", MODELDIR "/hmm/en_US/hub4wsj_sc_8k",
"-lm", MODELDIR "/lm/en/turtle.DMP",
"-dict", MODELDIR "/lm/en/turtle.dic",
NULL);
if (config == NULL)
return 1;
ps = ps_init(config);
if (ps == NULL)
return 1;
fh = fopen("goforward.raw", "rb");
if (fh == NULL) {
perror("Failed to open goforward.raw");
return 1;
}
rv = ps_decode_raw(ps, fh, "goforward", -1);
if (rv < 0)
return 1;
hyp = ps_get_hyp(ps, &score, &uttid);
if (hyp == NULL)
return 1;
printf("Recognized: %s\n", hyp);
fseek(fh, 0, SEEK_SET);
rv = ps_start_utt(ps, "goforward");
if (rv < 0)
return 1;
while (!feof(fh)) {
size_t nsamp;
nsamp = fread(buf, 2, 512, fh);
rv = ps_process_raw(ps, buf, nsamp, FALSE, FALSE);
}
rv = ps_end_utt(ps);
if (rv < 0)
return 1;
hyp = ps_get_hyp(ps, &score, &uttid);
if (hyp == NULL)
return 1;
printf("Recognized: %s\n", hyp);
fclose(fh);
ps_free(ps);
return 0;
}
void EventHandlerBasic::event(ps_decoder_t* decoder)
{
char const* message = ps_get_hyp( decoder, NULL );
if( mCb != nullptr && message != NULL && strlen( message ) > 0 )
mCb( std::string( message ) );
}
std::string PocketSphinxServer::decode(const PocketSphinxIce::sample& signal, const Ice::Current& c)
{
std::cout << "Decode\n";
int rv;
const char* hyp;
const char* uttid;
int32 score;
rv = ps_start_utt(ps);
if (rv < 0)
throw PocketSphinxIce::Error("Error in ps_start_utt");
ps_process_raw(ps, signal.data(), signal.size(), FALSE, FALSE);
rv = ps_end_utt(ps);
if (rv < 0)
throw PocketSphinxIce::Error("Error in ps_end_utt");
hyp = ps_get_hyp(ps, &score);
if (!hyp)
throw PocketSphinxIce::Error("ps_get_hyp returned NULL");
std::cout << "return:" << hyp << '\n';
return hyp;
}
int
main(int argc, char *argv[])
{
cmd_ln_t *config;
ps_decoder_t *ps;
FILE *rawfh;
char const *hyp;
char const *uttid;
int32 score;
TEST_ASSERT(config =
cmd_ln_init(NULL, ps_args(), TRUE,
"-hmm", DATADIR "/an4_ci_cont",
"-lm", MODELDIR "/lm/en/turtle.DMP",
"-dict", MODELDIR "/lm/en/turtle.dic",
"-mllr", DATADIR "/mllr_matrices",
"-samprate", "16000", NULL));
TEST_ASSERT(ps = ps_init(config));
TEST_ASSERT(rawfh = fopen(DATADIR "/goforward.raw", "rb"));
ps_decode_raw(ps, rawfh, "goforward", -1);
fclose(rawfh);
hyp = ps_get_hyp(ps, &score, &uttid);
printf("FWDFLAT (%s): %s (%d)\n", uttid, hyp, score);
ps_free(ps);
cmd_ln_free_r(config);
return 0;
}
int processRaw(const char *rawFile) {
const char *hyp, *uttid;
int16 buf[512];
int rv;
int32 score;
// Open the wav file passed from argument
printf("file: %s\n", rawFile);
fh = fopen(rawFile, "rb");
if (fh == NULL) {
fprintf(stderr, "Unable to open input file %s\n", rawFile);
return -1;
}
// Start utterance
rv = ps_start_utt(ps);
// Process buffer, 512 samples at a time
while (!feof(fh)) {
size_t nsamp;
nsamp = fread(buf, 2, 512, fh);
rv = ps_process_raw(ps, buf, nsamp, FALSE, FALSE);
}
// Recieve the recognized string
rv = ps_end_utt(ps);
hyp = ps_get_hyp(ps, &score);
printf("Recognized: |%s|\n", hyp);
fflush(stdout);
// Close file
fclose(fh);
return 0;
}
std::tuple<std::string, double> TwitchStreamChunk::process(std::string body){
int pos = _uri.find_last_of('/');
std::string fileName = _uri.substr(pos + 1);
std::ofstream file(fileName);
file << body;
file.flush();
file.close();
std::stringstream cmd;
std::string audioFile = boost::filesystem::unique_path().native();
audioFile.append(".wav");
cmd << "ffmpeg -i " << fileName << " -vn -ac 1 " << audioFile << " > /dev/null 2>&1";
system(cmd.str().c_str());
FILE *aFile = fopen(audioFile.c_str(), "r");
ps_decode_raw(getDecoder(), aFile, -1);
fclose(aFile);
auto logarithm = ps_get_logmath(getDecoder());
int confidence = 1;
const char * result = ps_get_hyp(getDecoder(), &confidence);
double tmp = logmath_exp(logarithm, confidence);
std::remove(fileName.c_str());
std::remove(audioFile.c_str());
return std::make_tuple(result == nullptr ? "" : std::string(result), tmp);
}
int ofApp::engineClose()
{
char const *uttid;
int rv;
int32 score;
rv = ps_end_utt(ps);
if (rv < 0)
{
return 1;
}
hyp = ps_get_hyp(ps, &score);
if (hyp == NULL)
{
return 1;
}
printf("NEWLINE_________\n\n\nRecognized: %s\n", hyp);
sentence = hyp;
process_result();
}
int
ps_end_utt(ps_decoder_t *ps)
{
int rv, i;
acmod_end_utt(ps->acmod);
/* Search any remaining frames. */
if ((rv = ps_search_forward(ps)) < 0) {
ptmr_stop(&ps->perf);
return rv;
}
/* Finish phone loop search. */
if (ps->phone_loop) {
if ((rv = ps_search_finish(ps->phone_loop)) < 0) {
ptmr_stop(&ps->perf);
return rv;
}
}
/* Search any frames remaining in the lookahead window. */
for (i = ps->acmod->output_frame - ps->pl_window;
i < ps->acmod->output_frame; ++i)
ps_search_step(ps->search, i);
/* Finish main search. */
if ((rv = ps_search_finish(ps->search)) < 0) {
ptmr_stop(&ps->perf);
return rv;
}
ptmr_stop(&ps->perf);
/* Log a backtrace if requested. */
if (cmd_ln_boolean_r(ps->config, "-backtrace")) {
char const *uttid, *hyp;
ps_seg_t *seg;
int32 score;
hyp = ps_get_hyp(ps, &score, &uttid);
if (hyp != NULL) {
E_INFO("%s: %s (%d)\n", uttid, hyp, score);
E_INFO_NOFN("%-20s %-5s %-5s %-5s %-10s %-10s %-3s\n",
"word", "start", "end", "pprob", "ascr", "lscr", "lback");
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);
E_INFO_NOFN("%-20s %-5d %-5d %-1.3f %-10d %-10d %-3d\n",
word, sf, ef, logmath_exp(ps_get_logmath(ps), post),
ascr, lscr, lback);
}
}
}
return rv;
}
/*! function to feed audio to the ASR */
static switch_status_t pocketsphinx_asr_feed(switch_asr_handle_t *ah, void *data, unsigned int len, switch_asr_flag_t *flags)
{
pocketsphinx_t *ps = (pocketsphinx_t *) ah->private_info;
int rv = 0;
if (switch_test_flag(ah, SWITCH_ASR_FLAG_CLOSED))
return SWITCH_STATUS_BREAK;
if (!switch_test_flag(ps, PSFLAG_HAS_TEXT) && switch_test_flag(ps, PSFLAG_READY)) {
if (stop_detect(ps, (int16_t *) data, len / 2)) {
char const *hyp;
switch_mutex_lock(ps->flag_mutex);
if ((hyp = ps_get_hyp(ps->ps, &ps->score, &ps->uttid))) {
if (!zstr(hyp)) {
ps_end_utt(ps->ps);
switch_clear_flag(ps, PSFLAG_READY);
if ((hyp = ps_get_hyp(ps->ps, &ps->score, &ps->uttid))) {
if (zstr(hyp)) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "Lost the text, never mind....\n");
ps_start_utt(ps->ps, NULL);
switch_set_flag(ps, PSFLAG_READY);
} else {
ps->hyp = switch_core_strdup(ah->memory_pool, hyp);
switch_set_flag(ps, PSFLAG_HAS_TEXT);
}
}
}
}
switch_mutex_unlock(ps->flag_mutex);
}
/* only feed ps_process_raw when we are listening */
if (ps->listening) {
switch_mutex_lock(ps->flag_mutex);
rv = ps_process_raw(ps->ps, (int16 *) data, len / 2, FALSE, FALSE);
switch_mutex_unlock(ps->flag_mutex);
}
if (rv < 0) {
return SWITCH_STATUS_FALSE;
}
}
return SWITCH_STATUS_SUCCESS;
}
ReturnType Recognizer::process(const std::vector<int16_t>& buffer) {
if ((decoder == NULL) || (!is_recording)) return BAD_STATE;
if (buffer.size() == 0)
return RUNTIME_ERROR;
ps_process_raw(decoder, (short int *) &buffer[0], buffer.size(), 0, 0);
const char* h = ps_get_hyp(decoder, &score, &sentence_id);
current_hyp = (h == NULL) ? "" : h;
return SUCCESS;
}
int main(int argc, char *argv[]) {
ps_decoder_t *ps;
cmd_ln_t *config;
FILE *fh;
char const *hyp, *uttid;
int16 buf[512];
int rv;
int32 score;
config = cmd_ln_init(NULL, ps_args(), TRUE,
"-hmm", MODELDIR "/en-us/en-us",
"-lm", MODELDIR "/en-us/en-us.lm.bin",
"-dict", MODELDIR "/en-us/cmudict-en-us.dict",
NULL);
if (config == NULL) {
fprintf(stderr, "Failed to create config object, see log for details\n");
return -1;
}
// Initialize pocketsphinx
ps = ps_init(config);
if (ps == NULL) {
fprintf(stderr, "Failed to create recognizer, see log for details\n");
return -1;
}
// Open the wav file passed from argument
printf("file: %s\n", argv[1]);
fh = fopen(argv[1], "rb");
if (fh == NULL) {
fprintf(stderr, "Unable to open input file %s\n", argv[1]);
return -1;
}
// Start utterance
rv = ps_start_utt(ps);
// Process buffer, 512 samples at a time
while (!feof(fh)) {
size_t nsamp;
nsamp = fread(buf, 2, 512, fh);
rv = ps_process_raw(ps, buf, nsamp, FALSE, FALSE);
}
// Recieve the recognized string
rv = ps_end_utt(ps);
hyp = ps_get_hyp(ps, &score);
printf("Recognized: |%s|\n", hyp);
// free memory
fclose(fh);
ps_free(ps);
cmd_ln_free_r(config);
return 0;
}
SWIGINTERN Hypothesis *ps_decoder_s_getHyp(struct ps_decoder_s *self){
char const *hyp, *uttid;
int32 best_score;
hyp = ps_get_hyp(self, &best_score, &uttid);
if (hyp == NULL)
return NULL;
else
return new_Hypothesis(hyp, uttid, best_score);
}
/**
* You must sucessfully call spInitListener
* once before using this function.
*
* Reads the next block of audio from the microphone
* up to SPLBUFSIZE number of samples
* (defined in splistener.h).
* If an utterance was completed in that block,
* the transcription is stored in the string words.
*
* When calling this function in a realtime loop, delay
* by some amount of time between calls keeping in mind
* your recording's sample rate and maximum samples read
* per call (ex. sleep the thread for 100 milliseconds)
* so that some audio can be recorded for the next call.
*
* @return true if a speech session was completed and
* transcribed this block, otherwise false.
*/
static bool spDecode() {
static bool uttered = false;
// lock pocketsphinx resources to make sure they
// don't get freed by main thread while in use
std::lock_guard<std::mutex> ps_lock(ps_mtx);
if(!mic || !ps)
return false;
int samples_read = ad_read(mic, buf, SPLBUFSIZE);
if (samples_read <= 0) {
spError("failed to read audio :(");
return false;
}
ps_process_raw(ps, buf, samples_read, FALSE, FALSE);
bool talking = ps_get_in_speech(ps);
// Just started talking
if (talking && !uttered) {
uttered = true;
return false;
}
// Stopped talking, so transcribe what was said
// and begin the next utterance
if (!talking && uttered) {
ps_end_utt(ps);
const char *trans = ps_get_hyp(ps, NULL);
if (ps_start_utt(ps) < 0) {
spError("failed to start utterance :(");
}
uttered = false;
int l = strlen(trans);
if (trans && l > 0) {
std::lock_guard<std::mutex> lock(words_mtx);
if (words && l + 1 > words_buf_size) {
delete words;
words = NULL;
}
if (!words) {
words = new char[l + 1];
words_buf_size = l + 1;
}
std::copy(trans, trans + l, words);
words[l] = '\0';
return true;
}
}
return false;
}
ReturnType Recognizer::stop() {
if ((decoder == NULL) || (!is_recording)) return BAD_STATE;
if (ps_end_utt(decoder) < 0) {
return RUNTIME_ERROR;
}
const char* h = ps_get_hyp(decoder, &score, &sentence_id);
current_hyp = (h == NULL) ? "" : h;
is_recording = false;
return SUCCESS;
}
int
main(int argc, char *argv[]) {
ps_decoder_t *ps;
cmd_ln_t *config;
FILE *fh;
const char *filename = "goforward.raw";
const char *word = "goforward";
char const *hyp, *uttid;
int rv;
int32 score;
/* setup the sphinx config */
config = cmd_ln_init(NULL, ps_args(), TRUE,
"-hmm", MODELDIR "/hmm/en_US/hub4wsj_sc_8k",
"-lm", MODELDIR "/lm/en/turtle.DMP",
"-dict", MODELDIR "/lm/en/turtle.dic",
NULL);
if(config == NULL) {
EXIT_ERROR;
}
/* initialize the config */
ps = ps_init(config);
if(ps == NULL) {
EXIT_ERROR;
}
/* open the audio file (stream?) */
fh = fopen(filename, "rb");
if(fh == NULL) {
perror(filename);
exit(1);
}
/* decode the file */
rv = ps_decode_raw(ps, fh, word, -1);
if(rv < 0) {
EXIT_ERROR;
}
/* get hypothesis */
hyp = ps_get_hyp(ps, &score, &uttid);
if(hyp == NULL) {
EXIT_ERROR;
}
printf("Recognized: %s; score: %d; uttid: %s\n",
hyp, score, uttid);
/* clean up */
fclose(fh);
ps_free(ps);
return 0;
}
int
main(int argc, char *argv[])
{
ps_decoder_t *ps;
cmd_ln_t *config;
FILE *fh;
char const *hyp, *uttid;
int16 buf[512];
int rv;
int32 score;
config = cmd_ln_init(NULL, ps_args(), TRUE,
"-hmm", MODELDIR "/en-us/en-us",
"-keyphrase", "marieta",
"-dict", MODELDIR "/en-us/cmudict-en-us.dict",
"-kws_threshold", "1e-30",
NULL);
if (config == NULL) {
fprintf(stderr, "Failed to create config object, see log for details\n");
return -1;
}
ps = ps_init(config);
if (ps == NULL) {
fprintf(stderr, "Failed to create recognizer, see log for details\n");
return -1;
}
fh = fopen("data/marieta.raw", "rb");
if (fh == NULL) {
fprintf(stderr, "Unable to open input file goforward.raw\n");
return -1;
}
rv = ps_start_utt(ps);
while (!feof(fh)) {
size_t nsamp;
nsamp = fread(buf, 2, 512, fh);
rv = ps_process_raw(ps, buf, nsamp, FALSE, FALSE);
}
rv = ps_end_utt(ps);
hyp = ps_get_hyp(ps, &score);
printf("Recognized: %s\n", hyp);
fclose(fh);
ps_free(ps);
cmd_ln_free_r(config);
return 0;
}
static void gst_sphinx_sink_process_chunk (GstSphinxSink *sphinxsink)
{
int32 k;
int16 adbuf[REQUIRED_FRAME_SAMPLES];
k = cont_ad_read (sphinxsink->cont, adbuf, REQUIRED_FRAME_SAMPLES);
if (k == 0 && sphinxsink->last_ts == 0) {
return;
} else if (k == 0 && sphinxsink->cont->read_ts - sphinxsink->last_ts >
DEFAULT_SAMPLES_PER_SEC) {
int32 score;
const char *hyp;
char *stripped_hyp;
int i, j;
ps_end_utt (sphinxsink->decoder);
if ((hyp = ps_get_hyp (sphinxsink->decoder, &score, NULL)) == NULL) {
gst_sphinx_sink_send_message (sphinxsink, "message", "");
g_message("Not Recognized");
} else {
stripped_hyp =
g_malloc (strlen (hyp) + 1);
for (i=0, j=0; hyp[i] != 0; i++) {
if (hyp[i] != '(' && hyp[i] != ')' && (hyp[i] < '0' || hyp[i] > '9')) {
stripped_hyp[j++] = hyp[i];
}
}
stripped_hyp [j] = 0;
gst_sphinx_sink_send_message (sphinxsink, "message", stripped_hyp);
g_message("Recognized: %s", stripped_hyp);
}
sphinxsink->last_ts = 0;
sphinxsink->ad.listening = 0;
} else if (k != 0) {
if (sphinxsink->ad.listening == 0) {
ps_start_utt (sphinxsink->decoder, NULL);
sphinxsink->ad.listening = 1;
gst_sphinx_sink_send_message (sphinxsink, "listening", NULL);
}
ps_process_raw (sphinxsink->decoder, adbuf, k, 0, 0);
sphinxsink->last_ts = sphinxsink->cont->read_ts;
}
}
/*
* Main utterance processing loop:
* for (;;) {
* start utterance and wait for speech to process
* decoding till end-of-utterance silence will be detected
* print utterance result;
* }
*/
static void
recognize_from_microphone()
{
ad_rec_t *ad;
int16 adbuf[2048];
uint8 utt_started, in_speech;
int32 k;
char const *hyp;
if ((ad = ad_open_dev(cmd_ln_str_r(config, "-adcdev"),
(int) cmd_ln_float32_r(config,
"-samprate"))) == NULL)
E_FATAL("Failed to open audio device\n");
if (ad_start_rec(ad) < 0)
E_FATAL("Failed to start recording\n");
if (ps_start_utt(ps) < 0)
E_FATAL("Failed to start utterance\n");
utt_started = FALSE;
E_INFO("Ready....\n");
for (;;) {
if ((k = ad_read(ad, adbuf, 2048)) < 0)
E_FATAL("Failed to read audio\n");
ps_process_raw(ps, adbuf, k, FALSE, FALSE);
in_speech = ps_get_in_speech(ps);
if (in_speech && !utt_started) {
utt_started = TRUE;
E_INFO("Listening...\n");
}
if (!in_speech && utt_started) {
/* speech -> silence transition, time to start new utterance */
ps_end_utt(ps);
hyp = ps_get_hyp(ps, NULL );
if (hyp != NULL) {
printf("%s\n", hyp);
fflush(stdout);
}
if (ps_start_utt(ps) < 0)
E_FATAL("Failed to start utterance\n");
utt_started = FALSE;
E_INFO("Ready....\n");
}
sleep_msec(100);
}
ad_close(ad);
}
/*
* Continuous recognition from a file
*/
int
recognize_from_file()
{
int16 adbuf[2048];
const char *fname;
const char *hyp;
int32 k;
char str[1000]="";
uint8 utt_started, in_speech;
fname = "C:/Users/Reza/Documents/GitHub/speech_agent/presentation_samples/italy1_reza.wav";
rawfd = fopen(fname, "rb");
if (strlen(fname) > 4 && strcmp(fname + strlen(fname) - 4, ".wav") == 0) {
char waveheader[44];
fread(waveheader, 1, 44, rawfd);
}
ps_start_utt(ps);
utt_started = FALSE;
while ((k = fread(adbuf, sizeof(int16), 2048, rawfd)) > 0) {
ps_process_raw(ps, adbuf, k, FALSE, FALSE);
in_speech = ps_get_in_speech(ps);
if (in_speech && !utt_started) {
utt_started = TRUE;
}
if (!in_speech && utt_started) {
ps_end_utt(ps);
hyp = ps_get_hyp(ps, NULL);
if (hyp != NULL){
strncpy_s( str, hyp, strlen(hyp));
printf("%s\n", hyp);
listenCallback(str);
}
ps_start_utt(ps);
utt_started = FALSE;
}
}
ps_end_utt(ps);
fclose(rawfd);
return 0;
}
int
main(int argc, char *argv[])
{
ps_decoder_t *ps;
ps_nbest_t *nbest;
cmd_ln_t *config;
FILE *rawfh;
char const *hyp;
int32 score, n;
TEST_ASSERT(config =
cmd_ln_init(NULL, ps_args(), TRUE,
"-hmm", MODELDIR "/en-us/en-us",
"-lm", MODELDIR "/en-us/en-us.lm.bin",
"-dict", MODELDIR "/en-us/cmudict-en-us.dict",
"-fwdtree", "yes",
"-fwdflat", "yes",
"-bestpath", "yes",
"-input_endian", "little",
"-samprate", "16000", NULL));
TEST_ASSERT(ps = ps_init(config));
TEST_ASSERT(rawfh = fopen(DATADIR "/goforward.raw", "rb"));
ps_decode_raw(ps, rawfh, -1);
fclose(rawfh);
hyp = ps_get_hyp(ps, &score);
printf("BESTPATH: %s (%d)\n", hyp, score);
for (n = 1, nbest = ps_nbest(ps); nbest && n < 10; nbest = ps_nbest_next(nbest), n++) {
ps_seg_t *seg;
hyp = ps_nbest_hyp(nbest, &score);
printf("NBEST %d: %s (%d)\n", n, hyp, score);
for (seg = ps_nbest_seg(nbest); seg;
seg = ps_seg_next(seg)) {
char const *word;
int sf, ef;
word = ps_seg_word(seg);
ps_seg_frames(seg, &sf, &ef);
printf("%s %d %d\n", word, sf, ef);
}
}
if (nbest)
ps_nbest_free(nbest);
ps_free(ps);
cmd_ln_free_r(config);
return 0;
}
static void
recognize_from_microphone()
{
ad_rec_t *ad;
int16 adbuf[2048];
uint8 utt_started, in_speech;
int32 k;
char const *hyp;
if ((ad = ad_open_dev(AUDIO_DEVICE_NAME,
(int) SAMPLE_RATE )) == NULL) {
E_FATAL("Failed to open audio device\n");
}
if (ad_start_rec(ad) < 0) {
E_FATAL("Failed to start recording\n");
}
if (ps_start_utt(ps) < 0) {
E_FATAL("Failed to start utterance\n");
}
utt_started = FALSE;
printf("READY....\n");
for (;;) {
if ((k = ad_read(ad, adbuf, 2048)) < 0)
E_FATAL("Failed to read audio\n");
ps_process_raw(ps, adbuf, k, FALSE, FALSE);
in_speech = ps_get_in_speech(ps);
if (in_speech && !utt_started) {
utt_started = TRUE;
printf("Listening...\n");
}
if (!in_speech && utt_started) {
/* speech -> silence transition, time to start new utterance */
ps_end_utt(ps);
hyp = ps_get_hyp(ps, NULL );
if (hyp != NULL)
printf("%s\n", hyp);
if (ps_start_utt(ps) < 0)
E_FATAL("Failed to start utterance\n");
utt_started = FALSE;
printf("READY....\n");
}
sleep_msec(100);
}
ad_close(ad);
}
int
main(int argc, char *argv[])
{
ps_decoder_t *ps;
cmd_ln_t *config;
FILE *fh;
char const *hyp, *uttid;
int16 buf[512];
int rv;
int32 score;
config = cmd_ln_init(NULL, ps_args(), TRUE,
"-hmm", MODELDIR "/en-us/en-us",
"-lm", MODELDIR "/en-us/en-us.lm.dmp",
"-dict", MODELDIR "/en-us/cmudict-en-us.dict",
NULL);
if (config == NULL)
return 1;
ps = ps_init(config);
if (ps == NULL)
return 1;
fh = fopen("goforward.raw", "rb");
if (fh == NULL)
return -1;
rv = ps_start_utt(ps);
if (rv < 0)
return 1;
while (!feof(fh)) {
size_t nsamp;
nsamp = fread(buf, 2, 512, fh);
rv = ps_process_raw(ps, buf, nsamp, FALSE, FALSE);
}
rv = ps_end_utt(ps);
if (rv < 0)
return 1;
hyp = ps_get_hyp(ps, &score);
if (hyp == NULL)
return 1;
printf("Recognized: %s\n", hyp);
fclose(fh);
ps_free(ps);
cmd_ln_free_r(config);
return 0;
}
int printResult()
{
char const *hyp, *uttid;
int32 score;
hyp = ps_get_hyp(ps, &score, &uttid);
if (hyp == NULL) {
printf("Error getting result.\n");
return 0;
}
printf("Result => %s, uttid: %s, score: %d\n", hyp, uttid, score);
if (strcmp(hyp, "left") == 0)
return 0;
return 1;
}
uint32 FSpeechRecognitionWorker::Run() {
char const *hyp;
// attempt to open the default recording device
if ((ad = ad_open_dev(cmd_ln_str_r(config, "-adcdev"),
(int)cmd_ln_float32_r(config,
"-samprate"))) == NULL) {
ClientMessage(FString(TEXT("Failed to open audio device")));
return 1;
}
if (ad_start_rec(ad) < 0) {
ClientMessage(FString(TEXT("Failed to start recording")));
return 2;
}
if (ps_start_utt(ps) < 0) {
ClientMessage(FString(TEXT("Failed to start utterance")));
return 3;
}
while (StopTaskCounter.GetValue() == 0) {
if ((k = ad_read(ad, adbuf, 1024)) < 0)
ClientMessage(FString(TEXT("Failed to read audio")));
ps_process_raw(ps, adbuf, k, 0, 0);
in_speech = ps_get_in_speech(ps);
if (in_speech && !utt_started) {
utt_started = 1;
}
if (!in_speech && utt_started) {
/* speech -> silence transition, time to start new utterance */
ps_end_utt(ps);
hyp = ps_get_hyp(ps, NULL);
if (hyp != NULL)
Manager->WordSpoken_method(FString(hyp));
if (ps_start_utt(ps) < 0)
ClientMessage(FString(TEXT("Failed to start")));
utt_started = 0;
}
}
ad_close(ad);
return 0;
}
static GstFlowReturn
gst_pocketsphinx_chain(GstPad * pad, GstBuffer * buffer)
{
GstPocketSphinx *ps;
ps = GST_POCKETSPHINX(GST_OBJECT_PARENT(pad));
/* Start an utterance for the first buffer we get (i.e. we assume
* that the VADER is "leaky") */
if (!ps->listening) {
ps->listening = TRUE;
ps_start_utt(ps->ps, NULL);
}
ps_process_raw(ps->ps,
(short *)GST_BUFFER_DATA(buffer),
GST_BUFFER_SIZE(buffer) / sizeof(short),
FALSE, FALSE);
/* Get a partial result every now and then, see if it is different. */
if (ps->last_result_time == 0
/* Check every 100 milliseconds. */
|| (GST_BUFFER_TIMESTAMP(buffer) - ps->last_result_time) > 100*10*1000) {
int32 score;
char const *hyp;
char const *uttid;
hyp = ps_get_hyp(ps->ps, &score, &uttid);
ps->last_result_time = GST_BUFFER_TIMESTAMP(buffer);
if (hyp && strlen(hyp) > 0) {
if (ps->last_result == NULL || 0 != strcmp(ps->last_result, hyp)) {
g_free(ps->last_result);
ps->last_result = g_strdup(hyp);
/* Emit a signal for applications. */
g_signal_emit(ps, gst_pocketsphinx_signals[SIGNAL_PARTIAL_RESULT],
0, hyp, uttid);
}
}
}
gst_buffer_unref(buffer);
return GST_FLOW_OK;
}
int processCommands()
{
int32 samples;
int16 audioBuf[BUFFER_SIZE];
char const *uttid;
char const *hyp;
while(run)
{
printf("Waiting for utterance...\n");
samples = waitForNextUtterance();
if(samples < 0)
return -1;
if(ps_start_utt(psDecoder, NULL) < 0) {
fprintf(stderr, "Failed to start next utterance\n");
return -1;
}
ps_process_raw(psDecoder, audioBuf, samples, FALSE, FALSE);
printf("Recording...\n");
fflush(stdout);
record();
ad_stop_rec(audioDevice);
while(ad_read(audioDevice, audioBuf, BUFFER_SIZE) >= 0);
cont_ad_reset(continousAudoDevice);
ps_end_utt(psDecoder);
hyp = ps_get_hyp(psDecoder, NULL, &uttid);
printf("Heard: %s\n", hyp);
if (ad_start_rec(audioDevice) < 0) {
fprintf(stderr, "Failed to start audio device.\n");
return -1;
}
}
return 0;
}
static int filter_frame(AVFilterLink *inlink, AVFrame *in)
{
AVFilterContext *ctx = inlink->dst;
AVDictionary **metadata = &in->metadata;
ASRContext *s = ctx->priv;
int have_speech;
const char *speech;
ps_process_raw(s->ps, (const int16_t *)in->data[0], in->nb_samples, 0, 0);
have_speech = ps_get_in_speech(s->ps);
if (have_speech && !s->utt_started)
s->utt_started = 1;
if (!have_speech && s->utt_started) {
ps_end_utt(s->ps);
speech = ps_get_hyp(s->ps, NULL);
if (speech != NULL)
av_dict_set(metadata, "lavfi.asr.text", speech, 0);
ps_start_utt(s->ps);
s->utt_started = 0;
}
return ff_filter_frame(ctx->outputs[0], in);
}
static void fsg_processor(context_t *ctx,
srs_srec_utterance_t *utt,
srs_srec_candidate_t *cands,
srs_srec_candidate_t **sorted)
{
filter_buf_t *filtbuf;
decoder_set_t *decset;
decoder_t *dec;
logmath_t *lmath;
const char *uttid;
int32_t score;
double prob;
srs_srec_candidate_t *cand;
srs_srec_token_t *tkn;
ps_lattice_t *dag;
ps_latlink_t *lnk;
ps_latnode_t *nod;
const char *token;
int32_t frlen;
int32_t start, end;
int16 fef, lef;
if (!ctx || !(filtbuf = ctx->filtbuf) ||
!(decset = ctx->decset) || !(dec = decset->curdec))
return;
frlen = filtbuf->frlen;
lmath = ps_get_logmath(dec->ps);
ps_get_hyp(dec->ps, &score, &uttid);
prob = logmath_exp(lmath, score);
cand = cands;
cand->score = 1.0;
cand->ntoken = 0;
tkn = NULL;
if ((dag = ps_get_lattice(dec->ps))) {
if ((lnk = ps_lattice_traverse_edges(dag, NULL, NULL))) {
ps_latlink_nodes(lnk, &nod);
if (nod && (token = ps_latnode_word(dag, nod)) && *token != '<') {
tkn = cand->tokens + cand->ntoken++;
tkn->token = tknbase(token);
tkn->start = ps_latnode_times(nod, &fef, &lef) * frlen;
tkn->end = ((fef + lef) / 2) * frlen;
}
goto handle_destination_node;
while ((lnk = ps_lattice_traverse_next(dag, NULL))) {
handle_destination_node:
nod = ps_latlink_nodes(lnk, NULL);
if (nod && (token = ps_latnode_word(dag,nod)) && *token != '<')
{
start = ps_latnode_times(nod, &fef, &lef) * frlen;
end = fef * frlen;
if (tkn && start < (int32_t)tkn->end)
break; /* just take one candidate */
if (!tkn || !tkneq(token, tkn->token)) {
tkn = cand->tokens + cand->ntoken++;
tkn->token = tknbase(token);
tkn->start = start;
tkn->end = end + frlen;
}
}
}
}
}
sorted[0] = cands;
sorted[1] = NULL;
utt->id = uttid;
utt->score = prob < 0.00001 ? 0.00001 : prob;
//utt->length = dag ? ps_lattice_n_frames(dag) * frlen : 0;
utt->length = filtbuf->len;
utt->ncand = 1;
utt->cands = sorted;
}
static void acoustic_processor(context_t *ctx,
srs_srec_utterance_t *utt,
srs_srec_candidate_t *cands,
srs_srec_candidate_t **sorted)
{
filter_buf_t *filtbuf;
decoder_set_t *decset;
decoder_t *dec;
logmath_t *lmath;
const char *uttid;
const char *hyp;
int32 score;
double prob;
ps_nbest_t *nb;
ps_seg_t *seg;
int32_t frlen;
int32 start, end;
size_t ncand;
srs_srec_candidate_t *cand;
srs_srec_token_t *tkn;
int32_t length;
if (!ctx || !(filtbuf = ctx->filtbuf) ||
!(decset = ctx->decset) || !(dec = decset->curdec))
return;
frlen = filtbuf->frlen;
lmath = ps_get_logmath(dec->ps);
uttid = "<unknown>";
hyp = ps_get_hyp(dec->ps, &score, &uttid);
prob = logmath_exp(lmath, score);
length = 0;
if (prob < 0.00000001)
prob = 0.00000001;
for (nb = ps_nbest(dec->ps, 0,-1, NULL,NULL), ncand = 0;
nb != NULL;
nb = ps_nbest_next(nb))
{
if (ncand >= CANDIDATE_MAX-1) {
break;
ps_nbest_free(nb);
}
if ((seg = ps_nbest_seg(nb, &score))) {
while (seg && strcmp(ps_seg_word(seg), "<s>"))
seg = ps_seg_next(seg);
if (!seg)
continue;
ps_seg_frames(seg, &start, &end);
cand = cands + ncand;
cand->score = logmath_exp(lmath, score) / prob;
cand->ntoken = 0;
length = 0;
while ((seg = ps_seg_next(seg))) {
if ((hyp = ps_seg_word(seg))) {
if (!strcmp(hyp, "</s>") ||
cand->ntoken >= CANDIDATE_TOKEN_MAX)
{
ncand++;
//memset(cand+1, 0, sizeof(srs_srec_candidate_t));
ps_seg_frames(seg, &start, &end);
ps_seg_free(seg);
//printf("hyp=</s> ncand=%d\n", ncand);
length = (end + 1) * frlen;
break;
}
else if (!strcmp(hyp, "<sil>")) {
ps_seg_frames(seg, &start, &end);
//printf("hyp=<sil> skip it\n");
}
else {
tkn = cand->tokens + cand->ntoken++;
tkn->token = tknbase(hyp);
ps_seg_frames(seg, &start, &end);
tkn->start = start * frlen;
tkn->end = (end + 1) * frlen;
//printf("hyp=%s (%d, %d) tkn count %d\n",
// tkn->token, tkn->start,tkn->end, cand->ntoken);
}
}
} /* while seg */
if (!seg && cand->ntoken > 0) {
ncand++;
cand->score *= 0.9; /* some penalty */
//memset(cand+1, 0, sizeof(srs_srec_candidate_t));
}
if (!length) {
tkn = cand->tokens + (cand->ntoken - 1);
length = tkn->end;
}
}
} /* for nb */
memset(cand+1, 0, sizeof(srs_srec_candidate_t));
utt->id = uttid;
utt->score = prob;
//utt->length = length;
utt->length = filtbuf->len;
utt->ncand = candidate_sort(cands, sorted);
utt->cands = sorted;
}
/*
* Continuous recognition from a file
*/
static void
recognize_from_file()
{
int16 adbuf[2048];
const char *fname;
const char *hyp;
int32 k;
uint8 utt_started, in_speech;
int32 print_times = cmd_ln_boolean_r(config, "-time");
fname = cmd_ln_str_r(config, "-infile");
if ((rawfd = fopen(fname, "rb")) == NULL) {
E_FATAL_SYSTEM("Failed to open file '%s' for reading",
fname);
}
if (strlen(fname) > 4 && strcmp(fname + strlen(fname) - 4, ".wav") == 0) {
char waveheader[44];
fread(waveheader, 1, 44, rawfd);
if (!check_wav_header(waveheader, (int)cmd_ln_float32_r(config, "-samprate")))
E_FATAL("Failed to process file '%s' due to format mismatch.\n", fname);
}
if (strlen(fname) > 4 && strcmp(fname + strlen(fname) - 4, ".mp3") == 0) {
E_FATAL("Can not decode mp3 files, convert input file to WAV 16kHz 16-bit mono before decoding.\n");
}
ps_start_utt(ps);
utt_started = FALSE;
while ((k = fread(adbuf, sizeof(int16), 2048, rawfd)) > 0) {
ps_process_raw(ps, adbuf, k, FALSE, FALSE);
in_speech = ps_get_in_speech(ps);
if (in_speech && !utt_started) {
utt_started = TRUE;
}
if (!in_speech && utt_started) {
ps_end_utt(ps);
hyp = ps_get_hyp(ps, NULL);
if (hyp != NULL)
printf("%s\n", hyp);
if (print_times)
print_word_times();
fflush(stdout);
ps_start_utt(ps);
utt_started = FALSE;
}
}
ps_end_utt(ps);
if (utt_started) {
hyp = ps_get_hyp(ps, NULL);
if (hyp != NULL) {
printf("%s\n", hyp);
if (print_times) {
print_word_times();
}
}
}
fclose(rawfd);
}
