kenmirrors/freeswitch
1
0
Fork 0
mirror of https://github.com/signalwire/freeswitch.git synced 2025-04-17 01:02:12 +00:00
Code Issues Projects Releases Wiki Activity

FS-11443 [core] reworked switch_vad.c and added voice_ms and silence_ms as parameters.

Browse Source
This commit is contained in:
Chris Rienzo 2018-10-05 19:43:34 +00:00 committed by Andrey Volk
parent eb84654383
commit 8534a4b86a
1 changed files with 63 additions and 76 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

127
src/switch_vad.c
View File

@ -37,20 +37,18 @@
#endif #endif
struct switch_vad_s { struct switch_vad_s {
int talking; // configs
int talked;
int talk_hits;
int listen_hits;
int hangover;
int hangover_len;
int divisor;
int thresh;
int channels; int channels;
int sample_rate; int sample_rate;
int debug; int debug;
int _hangover_len; int divisor;
int _thresh; int thresh;
int _listen_hits; int voice_samples_thresh;
int silence_samples_thresh;
// VAD state
int voice_samples;
int silence_samples;
switch_vad_state_t vad_state; switch_vad_state_t vad_state;
#ifdef SWITCH_HAVE_FVAD #ifdef SWITCH_HAVE_FVAD
Fvad *fvad; Fvad *fvad;
@ -82,9 +80,13 @@ SWITCH_DECLARE(switch_vad_t *) switch_vad_init(int sample_rate, int channels)
memset(vad, 0, sizeof(*vad)); memset(vad, 0, sizeof(*vad));
vad->sample_rate = sample_rate ? sample_rate : 8000; vad->sample_rate = sample_rate ? sample_rate : 8000;
vad->channels = channels; vad->channels = channels;
vad->_hangover_len = 25; vad->silence_samples_thresh = 500 * (vad->sample_rate / 1000);
vad->_thresh = 100; vad->voice_samples_thresh = 200 * (vad->sample_rate / 1000);
vad->_listen_hits = 10; vad->thresh = 100;
vad->divisor = vad->sample_rate / 8000;
if (vad->divisor <= 0) {
vad->divisor = 1;
}
switch_vad_reset(vad); switch_vad_reset(vad);
return vad; return vad;
@ -129,13 +131,29 @@ SWITCH_DECLARE(void) switch_vad_set_param(switch_vad_t *vad, const char *key, in
if (!key) return; if (!key) return;
if (!strcmp(key, "hangover_len")) { if (!strcmp(key, "hangover_len")) {
vad->hangover_len = vad->_hangover_len = val; /* convert old-style hits to samples assuming 20ms ptime */
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "hangover_len is deprecated, setting silence_ms to %d\n", 20 * val);
switch_vad_set_param(vad, "silence_ms", val * 20);
} else if (!strcmp(key, "silence_ms")) {
if (val > 0) {
vad->silence_samples_thresh = val * (vad->sample_rate / 1000);
} else {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "Ignoring invalid silence_ms of %d\n", val);
}
} else if (!strcmp(key, "thresh")) { } else if (!strcmp(key, "thresh")) {
vad->thresh = vad->_thresh = val; vad->thresh = val;
} else if (!strcmp(key, "debug")) { } else if (!strcmp(key, "debug")) {
vad->debug = val; vad->debug = val;
} else if (!strcmp(key, "voice_ms")) {
if (val > 0) {
vad->voice_samples_thresh = val * (vad->sample_rate / 1000);
} else {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "Ignoring invalid voice_ms of %d\n", val);
}
} else if (!strcmp(key, "listen_hits")) { } else if (!strcmp(key, "listen_hits")) {
vad->listen_hits = vad->_listen_hits = val; /* convert old-style hits to samples assuming 20ms ptime */
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "listen_hits is deprecated, setting voice_ms to %d\n", 20 * val);
switch_vad_set_param(vad, "voice_ms", 20 * val);
} }
} }
@ -144,34 +162,23 @@ SWITCH_DECLARE(void) switch_vad_reset(switch_vad_t *vad)
#ifdef SWITCH_HAVE_FVAD #ifdef SWITCH_HAVE_FVAD
if (vad->fvad) { if (vad->fvad) {
fvad_reset(vad->fvad); fvad_reset(vad->fvad);
return;
} }
#endif #endif
vad->talking = 0;
vad->talked = 0;
vad->talk_hits = 0;
vad->hangover = 0;
vad->listen_hits = vad->_listen_hits;
vad->hangover_len = vad->_hangover_len;
vad->divisor = vad->sample_rate / 8000;
vad->thresh = vad->_thresh;
vad->vad_state = SWITCH_VAD_STATE_NONE; vad->vad_state = SWITCH_VAD_STATE_NONE;
vad->voice_samples = 0;
vad->silence_samples = 0;
} }
SWITCH_DECLARE(switch_vad_state_t) switch_vad_process(switch_vad_t *vad, int16_t *data, unsigned int samples) SWITCH_DECLARE(switch_vad_state_t) switch_vad_process(switch_vad_t *vad, int16_t *data, unsigned int samples)
{ {
int energy = 0, j = 0, count = 0;
int score = 0; int score = 0;
if (vad->vad_state == SWITCH_VAD_STATE_STOP_TALKING) { // Each frame has 2 possible outcomes- voice or not voice.
vad->vad_state = SWITCH_VAD_STATE_NONE; // The VAD has 2 real states- talking / not talking with
} // begin talking and stop talking as events to mark transitions
if (vad->vad_state == SWITCH_VAD_STATE_START_TALKING) {
vad->vad_state = SWITCH_VAD_STATE_TALKING;
}
// determine if this is a voice or non-voice frame
#ifdef SWITCH_HAVE_FVAD #ifdef SWITCH_HAVE_FVAD
if (vad->fvad) { if (vad->fvad) {
int ret = fvad_process(vad->fvad, data, samples); int ret = fvad_process(vad->fvad, data, samples);
@ -181,58 +188,38 @@ SWITCH_DECLARE(switch_vad_state_t) switch_vad_process(switch_vad_t *vad, int16_t
score = vad->thresh + ret - 1; score = vad->thresh + ret - 1;
} else { } else {
#endif #endif
int energy = 0, j = 0, count = 0;
for (energy = 0, j = 0, count = 0; count < samples; count++) { for (energy = 0, j = 0, count = 0; count < samples; count++) {
energy += abs(data[j]); energy += abs(data[j]);
j += vad->channels; j += vad->channels;
} }
score = (uint32_t) (energy / (samples / vad->divisor)); score = (uint32_t) (energy / (samples / vad->divisor));
#ifdef SWITCH_HAVE_FVAD #ifdef SWITCH_HAVE_FVAD
} }
#endif #endif
//printf("%d ", score); fflush(stdout); // clear the STOP/START TALKING events
//printf("yay %d %d %d\n", score, vad->hangover, vad->talking); if (vad->vad_state == SWITCH_VAD_STATE_STOP_TALKING) {
vad->vad_state = SWITCH_VAD_STATE_NONE;
if (vad->talking && score < vad->thresh) { } else if (vad->vad_state == SWITCH_VAD_STATE_START_TALKING) {
if (vad->hangover > 0) {
vad->hangover--;
} else {// if (hangover <= 0) {
vad->talking = 0;
vad->talk_hits = 0;
vad->hangover = 0;
}
} else {
if (score >= vad->thresh) {
vad->vad_state = vad->talking ? SWITCH_VAD_STATE_TALKING : SWITCH_VAD_STATE_START_TALKING;
vad->talking = 1;
vad->hangover = vad->hangover_len;
}
}
// printf("WTF %d %d %d\n", score, vad->talked, vad->talking);
if (vad->talking) {
vad->talk_hits++;
// printf("WTF %d %d %d\n", vad->talking, vad->talk_hits, vad->talked);
if (vad->talk_hits > vad->listen_hits) {
vad->talked = 1;
vad->vad_state = SWITCH_VAD_STATE_TALKING; vad->vad_state = SWITCH_VAD_STATE_TALKING;
} }
// adjust voice/silence run length counters
if (score > vad->thresh) {
vad->silence_samples = 0;
vad->voice_samples += samples;
} else { } else {
vad->talk_hits = 0; vad->silence_samples += samples;
vad->voice_samples = 0;
} }
if ((vad->talked && !vad->talking)) { // check for state transitions
// printf("NOT TALKING ANYMORE\n"); if (vad->vad_state == SWITCH_VAD_STATE_TALKING && vad->silence_samples > vad->silence_samples_thresh) {
vad->talked = 0;
vad->vad_state = SWITCH_VAD_STATE_STOP_TALKING; vad->vad_state = SWITCH_VAD_STATE_STOP_TALKING;
} } else if (vad->vad_state == SWITCH_VAD_STATE_NONE && vad->voice_samples > vad->voice_samples_thresh) {
vad->vad_state = SWITCH_VAD_STATE_START_TALKING;
if (vad->debug > 0) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "VAD DEBUG energy: %d state %s\n", score, switch_vad_state2str(vad->vad_state));
} }
return vad->vad_state; return vad->vad_state;