void AudioEchoWidget::paintEvent(QPaintEvent *) {
QPainter paint(this);
paint.scale(width(), height());
paint.fillRect(rect(), Qt::black);
AudioInputPtr ai = g.ai;
if (! ai || ! ai->sesEcho)
return;
ai->qmSpeex.lock();
spx_int32_t sz;
speex_echo_ctl(ai->sesEcho, SPEEX_ECHO_GET_IMPULSE_RESPONSE_SIZE, &sz);
STACKVAR(spx_int32_t, w, sz);
STACKVAR(float, W, sz);
speex_echo_ctl(ai->sesEcho, SPEEX_ECHO_GET_IMPULSE_RESPONSE, w);
ai->qmSpeex.unlock();
int N = 160;
int n = 2 * N;
int M = sz / n;
drft_lookup d;
mumble_drft_init(&d, n);
for (int j=0;j<M;j++) {
for (int i=0;i<n;i++)
W[j*n+i] = static_cast<float>(w[j*n+i]) / static_cast<float>(n);
mumble_drft_forward(&d, & W[j*n]);
}
mumble_drft_clear(&d);
float xscale = 1.0f / static_cast<float>(N);
float yscale = 1.0f / static_cast<float>(M);
for (int j = 0; j < M; j++) {
for (int i=1;i < N; i++) {
float xa = static_cast<float>(i) * xscale;
float ya = static_cast<float>(j) * yscale;
float xb = xa + xscale;
float yb = ya + yscale;
const QColor &c = mapEchoToColor(sqrtf(W[j*n+2*i]*W[j*n+2*i]+W[j*n+2*i-1]*W[j*n+2*i-1]) / 65536.f);
paint.fillRect(QRectF(QPointF(xa, ya), QPointF(xb, yb)), c);
}
}
QPolygonF poly;
xscale = 1.0f / (2.0f * static_cast<float>(n));
yscale = 1.0f / (200.0f * 32767.0f);
for (int i = 0; i < 2 * n; i++) {
poly << QPointF(static_cast<float>(i) * xscale, 0.5f + static_cast<float>(w[i]) * yscale);
}
paint.setPen(QColor::fromRgbF(1.0f, 0.0f, 1.0f));
paint.drawPolyline(poly);
}
void AudioEchoWidget::paintGL() {
AudioInputPtr ai = g.ai;
if (! ai || ! ai->sesEcho)
return;
ai->qmSpeex.lock();
spx_int32_t sz;
speex_echo_ctl(ai->sesEcho, SPEEX_ECHO_GET_IMPULSE_RESPONSE_SIZE, &sz);
STACKVAR(spx_int32_t, w, sz);
STACKVAR(float, W, sz);
speex_echo_ctl(ai->sesEcho, SPEEX_ECHO_GET_IMPULSE_RESPONSE, w);
ai->qmSpeex.unlock();
int N = 160;
int n = 2 * N;
int M = sz / n;
drft_lookup d;
mumble_drft_init(&d, n);
for (int j=0;j<M;j++) {
for (int i=0;i<n;i++)
W[j*n+i] = static_cast<float>(w[j*n+i]) / static_cast<float>(n);
mumble_drft_forward(&d, & W[j*n]);
}
mumble_drft_clear(&d);
float xscale = 1.0f / static_cast<float>(N);
float yscale = 1.0f / static_cast<float>(M);
glBegin(GL_QUADS);
for (int j = 0; j < M; j++) {
for (int i=1;i < N; i++) {
float xa = static_cast<float>(i) * xscale;
float ya = static_cast<float>(j) * yscale;
float xb = xa + xscale;
float yb = ya + yscale;
mapEchoToColor(sqrtf(W[j*n+2*i]*W[j*n+2*i]+W[j*n+2*i-1]*W[j*n+2*i-1]) / 65536.f);
glVertex2f(xa, ya);
glVertex2f(xb, ya);
glVertex2f(xb, yb);
glVertex2f(xa, yb);
}
}
glEnd();
glBegin(GL_LINE_STRIP);
glColor3f(1.0f, 0.0f, 1.0f);
xscale = 1.0f / (2.0f*static_cast<float>(n));
yscale = 1.0f / (200.0f * 32767.0f);
for (int i=0;i<2*n;i++) {
glVertex2f(static_cast<float>(i)*xscale, 0.5f + static_cast<float>(w[i]) * yscale);
}
glEnd();
}
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();
}
