static int tdav_speex_jitterbuffer_reset(tmedia_jitterbuffer_t* self)
{
tdav_speex_jitterbuffer_t *jitterbuffer = (tdav_speex_jitterbuffer_t *)self;
if(jitterbuffer->state){
jitter_buffer_reset(jitterbuffer->state);
}
return 0;
}
/** Destroy jitter buffer */
EXPORT void jitter_buffer_destroy(JitterBuffer *jitter)
{
if (jitter)
{
jitter_buffer_reset(jitter);
speex_free(jitter);
}
}
static int tdav_speex_jitterbuffer_reset(tmedia_jitterbuffer_t* self)
{
tdav_speex_jitterbuffer_t *jb = (tdav_speex_jitterbuffer_t *)self;
if (jb->state) {
jitter_buffer_reset(jb->state);
}
jb->num_pkt_in = 0;
jb->num_pkt_miss = 0;
return 0;
}
/** Initialise jitter buffer */
JitterBuffer *jitter_buffer_init(int tick)
{
JitterBuffer *jitter = (JitterBuffer*)speex_alloc(sizeof(JitterBuffer));
if (jitter)
{
int i;
for (i=0;i<SPEEX_JITTER_MAX_BUFFER_SIZE;i++)
jitter->buf[i]=NULL;
jitter->tick_size = tick;
jitter->buffer_margin = 1;
jitter_buffer_reset(jitter);
}
return jitter;
}
/** Initialise jitter buffer */
EXPORT JitterBuffer *jitter_buffer_init(int step_size)
{
JitterBuffer *jitter = (JitterBuffer*)speex_alloc(sizeof(JitterBuffer));
if (jitter)
{
int i;
spx_int32_t tmp;
for (i=0;i<SPEEX_JITTER_MAX_BUFFER_SIZE;i++)
jitter->packets[i].data=NULL;
jitter->delay_step = step_size;
jitter->concealment_size = step_size;
/*FIXME: Should this be 0 or 1?*/
jitter->buffer_margin = 0;
jitter->late_cutoff = 50;
jitter->destroy = NULL;
jitter->latency_tradeoff = 0;
jitter->auto_adjust = 1;
tmp = 4;
jitter_buffer_ctl(jitter, JITTER_BUFFER_SET_MAX_LATE_RATE, &tmp);
jitter_buffer_reset(jitter);
}
return jitter;
}
void jitterFill(JitterBuffer *jb) {
char buffer[65536];
JitterBufferPacket in, out;
int i;
out.data = buffer;
jitter_buffer_reset(jb);
for(i=0;i<100;++i) {
synthIn(&in, i, 1);
jitter_buffer_put(jb, &in);
out.len = 65536;
if (jitter_buffer_get(jb, &out, 10, NULL) != JITTER_BUFFER_OK) {
printf("Fill test failed iteration %d\n", i);
}
if (out.timestamp != i * 10) {
printf("Fill test expected %d got %d\n", i*10, out.timestamp);
}
jitter_buffer_tick(jb);
}
}
/** Put one packet into the jitter buffer */
void jitter_buffer_put(JitterBuffer *jitter, const JitterBufferPacket *packet)
{
int i,j;
spx_int32_t arrival_margin;
/*fprintf (stderr, "put packet %d %d\n", timestamp, span);*/
if (jitter->reset_state)
{
jitter->reset_state=0;
jitter->pointer_timestamp = packet->timestamp;
jitter->current_timestamp = packet->timestamp;
/*fprintf(stderr, "reset to %d\n", timestamp);*/
}
/* Cleanup buffer (remove old packets that weren't played) */
for (i=0;i<SPEEX_JITTER_MAX_BUFFER_SIZE;i++)
{
if (jitter->buf[i] && LE32(jitter->timestamp[i] + jitter->span[i], jitter->pointer_timestamp))
{
/*fprintf (stderr, "cleaned (not played)\n");*/
speex_free(jitter->buf[i]);
jitter->buf[i] = NULL;
}
}
/*Find an empty slot in the buffer*/
for (i=0;i<SPEEX_JITTER_MAX_BUFFER_SIZE;i++)
{
if (jitter->buf[i]==NULL)
break;
}
/*fprintf(stderr, "%d %d %f\n", timestamp, jitter->pointer_timestamp, jitter->drift_average);*/
/*No place left in the buffer*/
if (i==SPEEX_JITTER_MAX_BUFFER_SIZE)
{
int earliest=jitter->timestamp[0];
i=0;
for (j=1;j<SPEEX_JITTER_MAX_BUFFER_SIZE;j++)
{
if (!jitter->buf[i] || LT32(jitter->timestamp[j],earliest))
{
earliest = jitter->timestamp[j];
i=j;
}
}
speex_free(jitter->buf[i]);
jitter->buf[i]=NULL;
if (jitter->lost_count>20)
{
jitter_buffer_reset(jitter);
}
/*fprintf (stderr, "Buffer is full, discarding earliest frame %d (currently at %d)\n", timestamp, jitter->pointer_timestamp);*/
}
/* Copy packet in buffer */
jitter->buf[i]=(char*)speex_alloc(packet->len);
for (j=0;j<(int)packet->len;j++)
jitter->buf[i][j]=packet->data[j];
jitter->timestamp[i]=packet->timestamp;
jitter->span[i]=packet->span;
jitter->len[i]=packet->len;
/* Adjust the buffer size depending on network conditions */
arrival_margin = (packet->timestamp - jitter->current_timestamp) - jitter->buffer_margin*jitter->tick_size;
if (arrival_margin >= -LATE_BINS*jitter->tick_size)
{
spx_int32_t int_margin;
for (i=0;i<MAX_MARGIN;i++)
{
jitter->shortterm_margin[i] *= .98;
jitter->longterm_margin[i] *= .995;
}
int_margin = LATE_BINS + arrival_margin/jitter->tick_size;
if (int_margin>MAX_MARGIN-1)
int_margin = MAX_MARGIN-1;
if (int_margin>=0)
{
jitter->shortterm_margin[int_margin] += .02;
jitter->longterm_margin[int_margin] += .005;
}
} else {
/*fprintf (stderr, "way too late = %d\n", arrival_margin);*/
if (jitter->lost_count>20)
{
jitter_buffer_reset(jitter);
}
}
#if 0 /* Enable to check how much is being buffered */
if (rand()%1000==0)
{
int count = 0;
for (j=0;j<SPEEX_JITTER_MAX_BUFFER_SIZE;j++)
{
if (jitter->buf[j])
count++;
}
fprintf (stderr, "buffer_size = %d\n", count);
}
#endif
}
/** Put one packet into the jitter buffer */
EXPORT void jitter_buffer_put(JitterBuffer * jitter,
const JitterBufferPacket * packet)
{
int i, j;
int late;
/*fprintf (stderr, "put packet %d %d\n", timestamp, span); */
/* Cleanup buffer (remove old packets that weren't played) */
if (!jitter->reset_state) {
for (i = 0; i < SPEEX_JITTER_MAX_BUFFER_SIZE; i++) {
/* Make sure we don't discard a "just-late" packet in case we want to play it next (if we interpolate). */
if (jitter->packets[i].data
&& LE32(jitter->packets[i].timestamp +
jitter->packets[i].span,
jitter->pointer_timestamp)) {
/*fprintf (stderr, "cleaned (not played)\n"); */
if (jitter->destroy)
jitter->destroy(jitter->packets[i].
data);
else
speex_free(jitter->packets[i].data);
jitter->packets[i].data = NULL;
}
}
}
/*fprintf(stderr, "arrival: %d %d %d\n", packet->timestamp, jitter->next_stop, jitter->pointer_timestamp); */
/* Check if packet is late (could still be useful though) */
if (!jitter->reset_state && LT32(packet->timestamp, jitter->next_stop)) {
update_timings(jitter,
((spx_int32_t) packet->timestamp) -
((spx_int32_t) jitter->next_stop) -
jitter->buffer_margin);
late = 1;
} else {
late = 0;
}
/* For some reason, the consumer has failed the last 20 fetches. Make sure this packet is
* used to resync. */
if (jitter->lost_count > 20) {
jitter_buffer_reset(jitter);
}
/* Only insert the packet if it's not hopelessly late (i.e. totally useless) */
if (jitter->reset_state
|| GE32(packet->timestamp + packet->span + jitter->delay_step,
jitter->pointer_timestamp)) {
/*Find an empty slot in the buffer */
for (i = 0; i < SPEEX_JITTER_MAX_BUFFER_SIZE; i++) {
if (jitter->packets[i].data == NULL)
break;
}
/*No place left in the buffer, need to make room for it by discarding the oldest packet */
if (i == SPEEX_JITTER_MAX_BUFFER_SIZE) {
int earliest = jitter->packets[0].timestamp;
i = 0;
for (j = 1; j < SPEEX_JITTER_MAX_BUFFER_SIZE; j++) {
if (!jitter->packets[i].data
|| LT32(jitter->packets[j].timestamp,
earliest)) {
earliest = jitter->packets[j].timestamp;
i = j;
}
}
if (jitter->destroy)
jitter->destroy(jitter->packets[i].data);
else
speex_free(jitter->packets[i].data);
jitter->packets[i].data = NULL;
/*fprintf (stderr, "Buffer is full, discarding earliest frame %d (currently at %d)\n", timestamp, jitter->pointer_timestamp); */
}
/* Copy packet in buffer */
if (jitter->destroy) {
jitter->packets[i].data = packet->data;
} else {
jitter->packets[i].data =
(char *)speex_alloc(packet->len);
for (j = 0; j < (int)packet->len; j++)
jitter->packets[i].data[j] = packet->data[j];
}
jitter->packets[i].timestamp = packet->timestamp;
jitter->packets[i].span = packet->span;
jitter->packets[i].len = packet->len;
jitter->packets[i].sequence = packet->sequence;
jitter->packets[i].user_data = packet->user_data;
if (jitter->reset_state || late)
jitter->arrival[i] = 0;
else
jitter->arrival[i] = jitter->next_stop;
}
}
/** Destroy jitter buffer */
void jitter_buffer_destroy(JitterBuffer *jitter)
{
jitter_buffer_reset(jitter);
free(jitter);
}
void AudioInput::encodeAudioFrame() {
int iArg;
ClientPlayer *p=ClientPlayer::get(g.uiSession);
int i;
float sum;
short max;
short *psSource;
iFrameCounter++;
if (! bRunning) {
return;
}
sum=1.0f;
for (i=0;i<iFrameSize;i++)
sum += static_cast<float>(psMic[i] * psMic[i]);
dPeakMic=20.0f*log10f(sqrtf(sum / static_cast<float>(iFrameSize)) / 32768.0f);
if (dPeakMic < -96.0f)
dPeakMic = -96.0f;
max = 1;
for (i=0;i<iFrameSize;i++)
max = static_cast<short>(abs(psMic[i]) > max ? abs(psMic[i]) : max);
dMaxMic = max;
if (g.bEchoTest) {
STACKVAR(float, fft, iFrameSize);
STACKVAR(float, power, iFrameSize);
float scale = 1.f / static_cast<float>(iFrameSize);
for (i=0;i<iFrameSize;i++)
fft[i] = static_cast<float>(psMic[i]) * scale;
mumble_drft_forward(&fftTable, fft);
float mp = 0.0f;
int bin = 0;
power[0]=power[1]=0.0f;
for (i=2;i < iFrameSize / 2;i++) {
power[i] = sqrtf(fft[2*i]*fft[2*i]+fft[2*i-1]*fft[2*i-1]);
if (power[i] > mp) {
bin = i;
mp = power[i];
}
}
for (i=2;i< iFrameSize / 2;i++) {
if (power[i] * 2 > mp) {
if (i != bin)
bin = 0;
}
}
iBestBin = bin * 2;
}
if (iEchoChannels > 0) {
sum=1.0f;
for (i=0;i<iFrameSize;i++)
sum += static_cast<float>(psSpeaker[i] * psSpeaker[i]);
dPeakSpeaker=20.0f*log10f(sqrtf(sum / static_cast<float>(iFrameSize)) / 32768.0f);
if (dPeakSpeaker < -96.0f)
dPeakSpeaker = -96.0f;
} else {
dPeakSpeaker = 0.0;
}
QMutexLocker l(&qmSpeex);
if (bResetProcessor) {
if (sppPreprocess)
speex_preprocess_state_destroy(sppPreprocess);
if (sesEcho)
speex_echo_state_destroy(sesEcho);
sppPreprocess = speex_preprocess_state_init(iFrameSize, SAMPLE_RATE);
iArg = 1;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_VAD, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_DENOISE, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_DEREVERB, &iArg);
iArg = 30000;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_TARGET, &iArg);
float v = 30000.0f / static_cast<float>(g.s.iMinLoudness);
iArg = lroundf(floorf(20.0f * log10f(v)));
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_MAX_GAIN, &iArg);
iArg = g.s.iNoiseSuppress;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_NOISE_SUPPRESS, &iArg);
if (iEchoChannels > 0) {
sesEcho = speex_echo_state_init(iFrameSize, iFrameSize*10);
iArg = SAMPLE_RATE;
speex_echo_ctl(sesEcho, SPEEX_SET_SAMPLING_RATE, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_ECHO_STATE, sesEcho);
jitter_buffer_reset(jb);
qWarning("AudioInput: ECHO CANCELLER ACTIVE");
} else {
sesEcho = NULL;
}
iFrames = 0;
speex_bits_reset(&sbBits);
bResetProcessor = false;
}
int iIsSpeech;
if (sesEcho) {
speex_echo_cancellation(sesEcho, psMic, psSpeaker, psClean);
iIsSpeech=speex_preprocess_run(sppPreprocess, psClean);
psSource = psClean;
} else {
iIsSpeech=speex_preprocess_run(sppPreprocess, psMic);
psSource = psMic;
}
sum=1.0f;
for (i=0;i<iFrameSize;i++)
sum += static_cast<float>(psSource[i] * psSource[i]);
float micLevel = sqrtf(sum / static_cast<float>(iFrameSize));
dPeakSignal=20.0f*log10f(micLevel / 32768.0f);
if (dPeakSignal < -96.0f)
dPeakSignal = -96.0f;
spx_int32_t prob = 0;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_GET_PROB, &prob);
fSpeechProb = static_cast<float>(prob) / 100.0f;
float level = (g.s.vsVAD == Settings::SignalToNoise) ? fSpeechProb : (1.0f + dPeakMic / 96.0f);
if (level > g.s.fVADmax)
iIsSpeech = 1;
else if (level > g.s.fVADmin && bPreviousVoice)
iIsSpeech = 1;
else
iIsSpeech = 0;
if (! iIsSpeech) {
iHoldFrames++;
if (iHoldFrames < g.s.iVoiceHold)
iIsSpeech=1;
} else {
iHoldFrames = 0;
}
if (g.s.atTransmit == Settings::Continous)
iIsSpeech = 1;
else if (g.s.atTransmit == Settings::PushToTalk)
iIsSpeech = g.s.uiDoublePush && ((g.uiDoublePush < g.s.uiDoublePush) || (g.tDoublePush.elapsed() < g.s.uiDoublePush));
iIsSpeech = iIsSpeech || (g.iPushToTalk > 0) || (g.iAltSpeak > 0);
if (g.s.bMute || ((g.s.lmLoopMode != Settings::Local) && p && p->bMute) || g.bPushToMute) {
iIsSpeech = 0;
}
if (iIsSpeech) {
iSilentFrames = 0;
} else {
iSilentFrames++;
if (iSilentFrames > 200)
iFrameCounter = 0;
}
if (p)
p->setTalking(iIsSpeech, (g.iAltSpeak > 0));
if (g.s.bPushClick && (g.s.atTransmit == Settings::PushToTalk)) {
AudioOutputPtr ao = g.ao;
if (iIsSpeech && ! bPreviousVoice && ao)
ao->playSine(400.0f,1200.0f,5);
else if (ao && !iIsSpeech && bPreviousVoice && ao)
ao->playSine(620.0f,-1200.0f,5);
}
if (! iIsSpeech && ! bPreviousVoice) {
iBitrate = 0;
if (g.s.iIdleTime && ! g.s.bMute && ((tIdle.elapsed() / 1000000ULL) > g.s.iIdleTime)) {
emit doMute();
tIdle.restart();
}
return;
}
bPreviousVoice = iIsSpeech;
tIdle.restart();
if (! iIsSpeech) {
memset(psMic, 0, sizeof(short) * iFrameSize);
}
if (g.s.bTransmitPosition && g.p && ! g.bCenterPosition && (iFrames == 0) && g.p->fetch()) {
QByteArray q;
QDataStream ds(&q, QIODevice::WriteOnly);
ds << g.p->fPosition[0];
ds << g.p->fPosition[1];
ds << g.p->fPosition[2];
speex_bits_pack(&sbBits, 13, 5);
speex_bits_pack(&sbBits, q.size(), 4);
const unsigned char *d=reinterpret_cast<const unsigned char*>(q.data());
for (i=0;i<q.size();i++) {
speex_bits_pack(&sbBits, d[i], 8);
}
}
speex_encode_int(esEncState, psSource, &sbBits);
iFrames++;
speex_encoder_ctl(esEncState, SPEEX_GET_BITRATE, &iBitrate);
flushCheck();
}
PJ_DEF(void) jitter_buffer_flush(Jitter_Buffer *jitter_buffer) {
jitter_buffer_reset(jitter_buffer);
}
