VAD code with optional libfvad

Collaboration diagram for VAD code with optional libfvad:

Functions
SWITCH_BEGIN_EXTERN_C switch_vad_t *	switch_vad_init (int sample_rate, int channels)
int	switch_vad_set_mode (switch_vad_t *vad, int mode)
void	switch_vad_set_param (switch_vad_t *vad, const char *key, int val)
switch_vad_state_t	switch_vad_process (switch_vad_t *vad, int16_t *data, unsigned int samples)
switch_vad_state_t	switch_vad_get_state (switch_vad_t *vad)
void	switch_vad_reset (switch_vad_t *vad)
void	switch_vad_destroy (switch_vad_t **vad)
const char *	switch_vad_state2str (switch_vad_state_t state)

Detailed Description

Function Documentation

switch_vad_destroy()

void switch_vad_destroy

(

switch_vad_t **

vad

)

Definition at line 254 of file switch_vad.c.

{
        if (*vad) {

#ifdef SWITCH_HAVE_FVAD
                if ((*vad)->fvad) fvad_free ((*vad)->fvad);
#endif

                free(*vad);
                *vad = NULL;
        }
}

switch_vad_get_state()

switch_vad_state_t switch_vad_get_state

(

switch_vad_t *

vad

)

Definition at line 248 of file switch_vad.c.

References switch_vad_s::vad_state.

{

        return vad->vad_state;
}

switch_vad_init()

SWITCH_BEGIN_EXTERN_C switch_vad_t* switch_vad_init	(	int	sample_rate,
		int	channels
	)

Definition at line 75 of file switch_vad.c.

References switch_vad_s::channels, switch_vad_s::divisor, memset(), switch_vad_s::sample_rate, switch_vad_s::silence_samples_thresh, switch_vad_reset(), switch_vad_s::thresh, and switch_vad_s::voice_samples_thresh.

{
        switch_vad_t *vad = malloc(sizeof(switch_vad_t));

        if (!vad) return NULL;

        memset(vad, 0, sizeof(*vad));
        vad->sample_rate = sample_rate ? sample_rate : 8000;
        vad->channels = channels;
        vad->silence_samples_thresh = 500 * (vad->sample_rate / 1000);
        vad->voice_samples_thresh = 200 * (vad->sample_rate / 1000);
        vad->thresh = 100;
        vad->divisor = vad->sample_rate / 8000;
        if (vad->divisor <= 0) {
                vad->divisor = 1;
        }
        switch_vad_reset(vad);

        return vad;
}

switch_vad_process()

switch_vad_state_t switch_vad_process	(	switch_vad_t *	vad,
		int16_t *	data,
		unsigned int	samples
	)

Definition at line 181 of file switch_vad.c.

References switch_vad_s::channels, count, switch_vad_s::debug, switch_vad_s::divisor, switch_vad_s::silence_samples, switch_vad_s::silence_samples_thresh, SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, SWITCH_LOG_INFO, SWITCH_LOG_NOTICE, switch_log_printf(), switch_vad_state2str(), SWITCH_VAD_STATE_NONE, SWITCH_VAD_STATE_START_TALKING, SWITCH_VAD_STATE_STOP_TALKING, SWITCH_VAD_STATE_TALKING, switch_vad_s::thresh, switch_vad_s::vad_state, switch_vad_s::voice_samples, and switch_vad_s::voice_samples_thresh.

{
        int score = 0;

        // Each frame has 2 possible outcomes- voice or not voice.
        // The VAD has 2 real states- talking / not talking with
        // begin talking and stop talking as events to mark transitions


        // determine if this is a voice or non-voice frame
#ifdef SWITCH_HAVE_FVAD
        if (vad->fvad) {
                // fvad returns -1, 0, or 1
                // -1: error
                //  0: non-voice frame
                //  1: voice frame
                int ret = fvad_process(vad->fvad, data, samples);

                // if voice frame set score > threshold
                score = ret > 0 ? vad->thresh + 100 : 0;
        } else {
#endif
                int energy = 0, j = 0, count = 0;
                for (energy = 0, j = 0, count = 0; count < samples; count++) {
                        energy += abs(data[j]);
                        j += vad->channels;
                }

                score = (uint32_t) (energy / (samples / vad->divisor));
#ifdef SWITCH_HAVE_FVAD
        }
#endif

        if (vad->debug > 9) {
                switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "score: %d\n", score);
        }

        // clear the STOP/START TALKING events
        if (vad->vad_state == SWITCH_VAD_STATE_STOP_TALKING) {
                vad->vad_state = SWITCH_VAD_STATE_NONE;
        } else if (vad->vad_state == SWITCH_VAD_STATE_START_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 {
                vad->silence_samples += samples;
                vad->voice_samples = 0;
        }

        // check for state transitions
        if (vad->vad_state == SWITCH_VAD_STATE_TALKING && vad->silence_samples > vad->silence_samples_thresh) {
                vad->vad_state = SWITCH_VAD_STATE_STOP_TALKING;
                if (vad->debug) switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_NOTICE, "vad state STOP_TALKING\n");
        } 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) switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "vad state START_TALKING\n");
        }

        if (vad->debug > 9) switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "vad state %s\n", switch_vad_state2str(vad->vad_state));

        return vad->vad_state;
}

switch_vad_reset()

void switch_vad_reset

(

switch_vad_t *

vad

)

Definition at line 167 of file switch_vad.c.

References switch_vad_s::debug, switch_vad_s::silence_samples, SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, switch_log_printf(), SWITCH_VAD_STATE_NONE, switch_vad_s::vad_state, and switch_vad_s::voice_samples.

Referenced by switch_vad_init().

{
#ifdef SWITCH_HAVE_FVAD
        if (vad->fvad) {
                fvad_reset(vad->fvad);
        }
#endif
        vad->vad_state = SWITCH_VAD_STATE_NONE;
        vad->voice_samples = 0;
        vad->silence_samples = 0;

        if (vad->debug) switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "reset vad state\n");
}

switch_vad_set_mode()

int switch_vad_set_mode	(	switch_vad_t *	vad,
		int	mode
	)

Definition at line 96 of file switch_vad.c.

References switch_vad_s::debug, switch_vad_s::sample_rate, SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, switch_log_printf(), and SWITCH_LOG_WARNING.

{
#ifdef SWITCH_HAVE_FVAD
        int ret = 0;

        if (mode < 0) {
                if (vad->fvad) fvad_free(vad->fvad);

                vad->fvad = NULL;
                return ret;
        } else if (mode > 3) {
                mode = 3;
        }

        if (!vad->fvad) {
                vad->fvad = fvad_new();

                if (!vad->fvad) {
                        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "libfvad init error\n");
                }
        }

        if (vad->fvad) {
                ret = fvad_set_mode(vad->fvad, mode);
                fvad_set_sample_rate(vad->fvad, vad->sample_rate);
        }

        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "libfvad started, mode = %d\n", mode);
        return ret;
#else
        if (vad->debug) switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "set vad mode = %d\n", mode);

        return 0;
#endif
}

switch_vad_set_param()

void switch_vad_set_param	(	switch_vad_t *	vad,
		const char *	key,
		int	val
	)

Definition at line 132 of file switch_vad.c.

References switch_vad_s::debug, switch_vad_s::sample_rate, switch_vad_s::silence_samples_thresh, SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, switch_log_printf(), SWITCH_LOG_WARNING, switch_vad_s::thresh, and switch_vad_s::voice_samples_thresh.

{
        if (!key) return;

        if (!strcmp(key, "hangover_len")) {
                /* 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")) {
                vad->thresh = val;
        } else if (!strcmp(key, "debug")) {
                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")) {
                /* 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);
        }

        if (vad->debug) {
                switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "set %s to %d\n", key, val);
        }
}

switch_vad_state2str()

const char* switch_vad_state2str

(

switch_vad_state_t

state

)

Definition at line 59 of file switch_vad.c.

References SWITCH_VAD_STATE_NONE, SWITCH_VAD_STATE_START_TALKING, SWITCH_VAD_STATE_STOP_TALKING, and SWITCH_VAD_STATE_TALKING.

Referenced by switch_vad_process().

{
        switch(state) {
        case SWITCH_VAD_STATE_NONE:
                return "none";
        case SWITCH_VAD_STATE_START_TALKING:
                return "start_talking";
        case SWITCH_VAD_STATE_TALKING:
                return "talking";
        case SWITCH_VAD_STATE_STOP_TALKING:
                return "stop_talking";
        default:
                return "error";
        }
}