bool pa_speex_ec_init(pa_core *c, pa_echo_canceller *ec,
pa_sample_spec *rec_ss, pa_channel_map *rec_map,
pa_sample_spec *play_ss, pa_channel_map *play_map,
pa_sample_spec *out_ss, pa_channel_map *out_map,
uint32_t *nframes, const char *args) {
int rate;
uint32_t frame_size_ms, filter_size_ms;
pa_modargs *ma;
if (!(ma = pa_modargs_new(args, valid_modargs))) {
pa_log("Failed to parse submodule arguments.");
goto fail;
}
filter_size_ms = DEFAULT_FILTER_SIZE_MS;
if (pa_modargs_get_value_u32(ma, "filter_size_ms", &filter_size_ms) < 0 || filter_size_ms < 1 || filter_size_ms > 2000) {
pa_log("Invalid filter_size_ms specification");
goto fail;
}
frame_size_ms = DEFAULT_FRAME_SIZE_MS;
if (pa_modargs_get_value_u32(ma, "frame_size_ms", &frame_size_ms) < 0 || frame_size_ms < 1 || frame_size_ms > 200) {
pa_log("Invalid frame_size_ms specification");
goto fail;
}
pa_speex_ec_fixate_spec(rec_ss, rec_map, play_ss, play_map, out_ss, out_map);
rate = out_ss->rate;
*nframes = pa_echo_canceller_blocksize_power2(rate, frame_size_ms);
pa_log_debug ("Using nframes %d, channels %d, rate %d", *nframes, out_ss->channels, out_ss->rate);
ec->params.priv.speex.state = speex_echo_state_init_mc(*nframes, (rate * filter_size_ms) / 1000, out_ss->channels, out_ss->channels);
if (!ec->params.priv.speex.state)
goto fail;
speex_echo_ctl(ec->params.priv.speex.state, SPEEX_ECHO_SET_SAMPLING_RATE, &rate);
if (!pa_speex_ec_preprocessor_init(ec, out_ss, *nframes, ma))
goto fail;
pa_modargs_free(ma);
return true;
fail:
if (ma)
pa_modargs_free(ma);
if (ec->params.priv.speex.pp_state) {
speex_preprocess_state_destroy(ec->params.priv.speex.pp_state);
ec->params.priv.speex.pp_state = NULL;
}
if (ec->params.priv.speex.state) {
speex_echo_state_destroy(ec->params.priv.speex.state);
ec->params.priv.speex.state = NULL;
}
return false;
}
pa_bool_t pa_speex_ec_init(pa_core *c, pa_echo_canceller *ec,
pa_sample_spec *source_ss, pa_channel_map *source_map,
pa_sample_spec *sink_ss, pa_channel_map *sink_map,
uint32_t *blocksize, const char *args)
{
int framelen, y, rate;
uint32_t frame_size_ms, filter_size_ms;
pa_modargs *ma;
if (!(ma = pa_modargs_new(args, valid_modargs))) {
pa_log("Failed to parse submodule arguments.");
goto fail;
}
filter_size_ms = DEFAULT_FILTER_SIZE_MS;
if (pa_modargs_get_value_u32(ma, "filter_size_ms", &filter_size_ms) < 0 || filter_size_ms < 1 || filter_size_ms > 2000) {
pa_log("Invalid filter_size_ms specification");
goto fail;
}
frame_size_ms = DEFAULT_FRAME_SIZE_MS;
if (pa_modargs_get_value_u32(ma, "frame_size_ms", &frame_size_ms) < 0 || frame_size_ms < 1 || frame_size_ms > 200) {
pa_log("Invalid frame_size_ms specification");
goto fail;
}
pa_speex_ec_fixate_spec(source_ss, source_map, sink_ss, sink_map);
rate = source_ss->rate;
framelen = (rate * frame_size_ms) / 1000;
/* framelen should be a power of 2, round down to nearest power of two */
y = 1 << ((8 * sizeof (int)) - 2);
while (y > framelen)
y >>= 1;
framelen = y;
*blocksize = framelen * pa_frame_size (source_ss);
pa_log_debug ("Using framelen %d, blocksize %u, channels %d, rate %d", framelen, *blocksize, source_ss->channels, source_ss->rate);
ec->params.priv.speex.state = speex_echo_state_init_mc (framelen, (rate * filter_size_ms) / 1000, source_ss->channels, source_ss->channels);
if (!ec->params.priv.speex.state)
goto fail;
speex_echo_ctl(ec->params.priv.speex.state, SPEEX_ECHO_SET_SAMPLING_RATE, &rate);
pa_modargs_free(ma);
return TRUE;
fail:
if (ma)
pa_modargs_free(ma);
return FALSE;
}
void AudioInput::resetAudioProcessor() {
if (!bResetProcessor)
return;
int iArg;
if (sppPreprocess)
speex_preprocess_state_destroy(sppPreprocess);
if (sesEcho)
speex_echo_state_destroy(sesEcho);
sppPreprocess = speex_preprocess_state_init(iFrameSize, iSampleRate);
iArg = 1;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_VAD, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_DENOISE, &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 = iroundf(floorf(20.0f * log10f(v)));
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_MAX_GAIN, &iArg);
iArg = -60;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_DECREMENT, &iArg);
iArg = g.s.iNoiseSuppress;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_NOISE_SUPPRESS, &iArg);
if (iEchoChannels > 0) {
sesEcho = speex_echo_state_init_mc(iFrameSize, iFrameSize * 10, 1, bEchoMulti ? iEchoChannels : 1);
iArg = iSampleRate;
speex_echo_ctl(sesEcho, SPEEX_ECHO_SET_SAMPLING_RATE, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_ECHO_STATE, sesEcho);
qWarning("AudioInput: ECHO CANCELLER ACTIVE");
} else {
sesEcho = NULL;
}
bResetEncoder = true;
bResetProcessor = false;
}
pa_bool_t pa_speex_ec_init(pa_core *c, pa_echo_canceller *ec,
pa_sample_spec *source_ss, pa_channel_map *source_map,
pa_sample_spec *sink_ss, pa_channel_map *sink_map,
uint32_t *nframes, const char *args)
{
int rate;
uint32_t y, frame_size_ms, filter_size_ms;
pa_modargs *ma;
if (!(ma = pa_modargs_new(args, valid_modargs))) {
pa_log("Failed to parse submodule arguments.");
goto fail;
}
filter_size_ms = DEFAULT_FILTER_SIZE_MS;
if (pa_modargs_get_value_u32(ma, "filter_size_ms", &filter_size_ms) < 0 || filter_size_ms < 1 || filter_size_ms > 2000) {
pa_log("Invalid filter_size_ms specification");
goto fail;
}
frame_size_ms = DEFAULT_FRAME_SIZE_MS;
if (pa_modargs_get_value_u32(ma, "frame_size_ms", &frame_size_ms) < 0 || frame_size_ms < 1 || frame_size_ms > 200) {
pa_log("Invalid frame_size_ms specification");
goto fail;
}
pa_speex_ec_fixate_spec(source_ss, source_map, sink_ss, sink_map);
rate = source_ss->rate;
*nframes = (rate * frame_size_ms) / 1000;
/* nframes should be a power of 2, round down to nearest power of two */
y = 1 << ((8 * sizeof (uint32_t)) - 2);
while (y > *nframes)
y >>= 1;
*nframes = y;
pa_log_debug ("Using nframes %d, channels %d, rate %d", *nframes, source_ss->channels, source_ss->rate);
ec->params.priv.speex.state = speex_echo_state_init_mc (*nframes, (rate * filter_size_ms) / 1000, source_ss->channels, source_ss->channels);
if (!ec->params.priv.speex.state)
goto fail;
speex_echo_ctl(ec->params.priv.speex.state, SPEEX_ECHO_SET_SAMPLING_RATE, &rate);
if (!pa_speex_ec_preprocessor_init(ec, source_ss, *nframes, ma))
goto fail;
pa_modargs_free(ma);
return TRUE;
fail:
if (ma)
pa_modargs_free(ma);
if (ec->params.priv.speex.pp_state) {
speex_preprocess_state_destroy(ec->params.priv.speex.pp_state);
ec->params.priv.speex.pp_state = NULL;
}
if (ec->params.priv.speex.state) {
speex_echo_state_destroy(ec->params.priv.speex.state);
ec->params.priv.speex.state = NULL;
}
return FALSE;
}
/*
* Create the AEC.
*/
PJ_DEF(pj_status_t) speex_aec_create(pj_pool_t *pool,
unsigned clock_rate,
unsigned channel_count,
unsigned samples_per_frame,
unsigned tail_ms,
unsigned options,
void **p_echo )
{
speex_ec *echo;
int sampling_rate;
*p_echo = NULL;
echo = PJ_POOL_ZALLOC_T(pool, speex_ec);
PJ_ASSERT_RETURN(echo != NULL, PJ_ENOMEM);
echo->samples_per_frame = samples_per_frame;
echo->options = options;
#if 0
echo->state = speex_echo_state_init_mc(echo->samples_per_frame,
clock_rate * tail_ms / 1000,
channel_count, channel_count);
#else
if (channel_count != 1) {
PJ_LOG(2,("echo_speex.c", "Multichannel EC is not supported by this "
"echo canceller. It may not work."));
}
echo->state = speex_echo_state_init(echo->samples_per_frame,
clock_rate * tail_ms / 1000);
#endif
if (echo->state == NULL) {
return PJ_ENOMEM;
}
/* Set sampling rate */
sampling_rate = clock_rate;
speex_echo_ctl(echo->state, SPEEX_ECHO_SET_SAMPLING_RATE,
&sampling_rate);
echo->preprocess = speex_preprocess_state_init(echo->samples_per_frame,
clock_rate);
if (echo->preprocess == NULL) {
speex_echo_state_destroy(echo->state);
return PJ_ENOMEM;
}
/* Disable all preprocessing, we only want echo cancellation */
#if 0
disabled = 0;
enabled = 1;
speex_preprocess_ctl(echo->preprocess, SPEEX_PREPROCESS_SET_DENOISE,
&enabled);
speex_preprocess_ctl(echo->preprocess, SPEEX_PREPROCESS_SET_AGC,
&disabled);
speex_preprocess_ctl(echo->preprocess, SPEEX_PREPROCESS_SET_VAD,
&disabled);
speex_preprocess_ctl(echo->preprocess, SPEEX_PREPROCESS_SET_DEREVERB,
&enabled);
#endif
/* Control echo cancellation in the preprocessor */
speex_preprocess_ctl(echo->preprocess, SPEEX_PREPROCESS_SET_ECHO_STATE,
echo->state);
/* Create temporary frame for echo cancellation */
echo->tmp_frame = (pj_int16_t*) pj_pool_zalloc(pool, 2*samples_per_frame);
PJ_ASSERT_RETURN(echo->tmp_frame != NULL, PJ_ENOMEM);
/* Done */
*p_echo = echo;
return PJ_SUCCESS;
}
void AudioInput::encodeAudioFrame() {
int iArg;
//ClientUser *p=ClientUser::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]);
iLevel = sqrtf(sum / static_cast<float>(iFrameSize)) * 9/32768.0f;
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 (psSpeaker && (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;
}*/
MutexLocker l(&qmSpeex);
bResetProcessor = false;
if (bResetProcessor) {
if (sppPreprocess)
speex_preprocess_state_destroy(sppPreprocess);
if (sesEcho)
speex_echo_state_destroy(sesEcho);
sppPreprocess = speex_preprocess_state_init(iFrameSize, iSampleRate);
iArg = 1;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_VAD, &iArg);
//speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_DENOISE, &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_struct.s.iMinLoudness);
iArg = (floorf(20.0f * log10f(v)));
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_MAX_GAIN, &iArg);
iArg = g_struct.s.iNoiseSuppress;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_NOISE_SUPPRESS, &iArg);
if (iEchoChannels > 0) {
sesEcho = speex_echo_state_init_mc(iFrameSize, iFrameSize*10, 1, bEchoMulti ? iEchoChannels : 1);
iArg = iSampleRate;
speex_echo_ctl(sesEcho, SPEEX_ECHO_SET_SAMPLING_RATE, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_ECHO_STATE, sesEcho);
Trace("AudioInput: ECHO CANCELLER ACTIVE");
} else {
sesEcho = NULL;
}
bResetProcessor = false;
}
int iIsSpeech=1;
psSource = psMic;
/*
//回音消除和音质处理
if (bEcho && sesEcho && psSpeaker)
{
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_struct.s.vsVAD == Settings::SignalToNoise) ? fSpeechProb : (1.0f + dPeakMic / 96.0f);
if (level > g_struct.s.fVADmax)
iIsSpeech = 1;
else if (level > g_struct.s.fVADmin && bPreviousVoice)
iIsSpeech = 1;
else
iIsSpeech = 0;
if (! iIsSpeech) {
iHoldFrames++;
if (iHoldFrames < g_struct.s.iVoiceHold)
iIsSpeech=1;
} else {
iHoldFrames = 0;
}*/
//tIdle.restart();
/*
int r = celt_encoder_ctl(ceEncoder, CELT_SET_POST_MDCT_CALLBACK(celtBack, NULL));
qWarning() << "Set Callback" << r;
*/
//编码 speex或者CELT
unsigned char buffer[512];
int len;
if (umtType != MessageHandler::UDPVoiceSpeex) {
if (cCodec == NULL)
{
cCodec = new CELTCodec;
umtType = MessageHandler::UDPVoiceCELT;
ceEncoder = cCodec->encoderCreate();
}
else if (cCodec && ! bPreviousVoice) {
cCodec->encoder_ctl(ceEncoder, CELT_RESET_STATE);
}
cCodec->encoder_ctl(ceEncoder, CELT_SET_PREDICTION(0));
cCodec->encoder_ctl(ceEncoder,CELT_SET_BITRATE(iAudioQuality));
len = cCodec->encode(ceEncoder, psSource, SAMPLE_RATE / 100, buffer, 512);
iBitrate = len * 100 * 8;
}
else {
int vbr = 0;
speex_encoder_ctl(esSpeex, SPEEX_GET_VBR_MAX_BITRATE, &vbr);
if (vbr != iAudioQuality) {
vbr = iAudioQuality;
speex_encoder_ctl(esSpeex, SPEEX_SET_VBR_MAX_BITRATE, &vbr);
}
if (! bPreviousVoice)
speex_encoder_ctl(esSpeex, SPEEX_RESET_STATE, NULL);
speex_encode_int(esSpeex, psSource, &sbBits);
len = speex_bits_write(&sbBits, reinterpret_cast<char *>(buffer), 127);
iBitrate = len * 50 * 8;
speex_bits_reset(&sbBits);
}
QByteArray qba;
for(int i=0; i<len; i++)
{
qba.push_back(buffer[i]);
}
flushCheck(qba, false);
if (! iIsSpeech)
iBitrate = 0;
bPreviousVoice = iIsSpeech;
}
void AudioInput::encodeAudioFrame() {
int iArg;
//ClientUser *p=ClientUser::get(g.uiSession);
int i;
float sum;
short max;
short *psSource;
iFrameCounter++;
if (! bRunning) {
return;
}
MutexLocker l(&qmSpeex);
bResetProcessor = false;
if (bResetProcessor) {
if (sppPreprocess)
speex_preprocess_state_destroy(sppPreprocess);
if (sesEcho)
speex_echo_state_destroy(sesEcho);
sppPreprocess = speex_preprocess_state_init(iFrameSize, iSampleRate);
iArg = 1;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_VAD, &iArg);
//speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_DENOISE, &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_struct.s.iMinLoudness);
iArg = (floorf(20.0f * log10f(v)));
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_AGC_MAX_GAIN, &iArg);
iArg = g_struct.s.iNoiseSuppress;
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_NOISE_SUPPRESS, &iArg);
if (iEchoChannels > 0) {
sesEcho = speex_echo_state_init_mc(iFrameSize, iFrameSize*10, 1, bEchoMulti ? iEchoChannels : 1);
iArg = iSampleRate;
speex_echo_ctl(sesEcho, SPEEX_ECHO_SET_SAMPLING_RATE, &iArg);
speex_preprocess_ctl(sppPreprocess, SPEEX_PREPROCESS_SET_ECHO_STATE, sesEcho);
Trace("AudioInput: ECHO CANCELLER ACTIVE");
} else {
sesEcho = NULL;
}
bResetProcessor = false;
}
int iIsSpeech=1;
psSource = psMic;
/*
//回音消除和音质处理
if (bEcho && sesEcho && psSpeaker)
{
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_struct.s.vsVAD == Settings::SignalToNoise) ? fSpeechProb : (1.0f + dPeakMic / 96.0f);
if (level > g_struct.s.fVADmax)
iIsSpeech = 1;
else if (level > g_struct.s.fVADmin && bPreviousVoice)
iIsSpeech = 1;
else
iIsSpeech = 0;
if (! iIsSpeech) {
iHoldFrames++;
if (iHoldFrames < g_struct.s.iVoiceHold)
iIsSpeech=1;
} else {
iHoldFrames = 0;
}*/
//tIdle.restart();
/*
int r = celt_encoder_ctl(ceEncoder, CELT_SET_POST_MDCT_CALLBACK(celtBack, NULL));
qWarning() << "Set Callback" << r;
*/
//编码 speex或者CELT
unsigned char buffer[512];
int len;
if (umtType == MessageHandler::UDPVoiceCELT) {
if (cCodec == NULL)
{
cCodec = CELTCodec::instance();
ceEncoder = cCodec->encoderCreate();
}
else if (cCodec && ! bPreviousVoice) {
cCodec->encoder_ctl(ceEncoder, CELT_RESET_STATE);
}
cCodec->encoder_ctl(ceEncoder, CELT_SET_PREDICTION(0));
cCodec->encoder_ctl(ceEncoder,CELT_SET_BITRATE(iAudioQuality));
len = cCodec->encode(ceEncoder, psSource, SAMPLE_RATE / 50, buffer, 512);
iBitrate = len * 50 * 8;
/*////////////////////////////////////////////////////////////////////////
if (m_de_cdDecoder == NULL) {
m_de_cdDecoder = cCodec->decoderCreate();
}
celt_int16 fout2[2560]={0};
if (cCodec)
{
int len3 = cCodec->decode(m_de_cdDecoder, buffer, len, fout2, SAMPLE_RATE / 50);
len3++;
UINT dwDataWrote;
if( FAILED(g_pWaveFile.Write( SAMPLE_RATE / 50*2*2, (BYTE*)fout2,
&dwDataWrote ) ))
{
int a=0;
a++;
}
else
{
OutputDebugString(L"plushuwav g_pWaveFile.Write 3");
}
}
///////////////////////////////////////////////////////////////////////*/
}
else {
assert(0);
}
QByteArray qba;
for(int i=0; i<len; i++)
{
qba.push_back(buffer[i]);
}
flushCheck(qba, false);
if (! iIsSpeech)
iBitrate = 0;
bPreviousVoice = iIsSpeech;
}
/** Creates a new echo canceller state */
EXPORT SpeexEchoState *speex_echo_state_init(int frame_size, int filter_length)
{
return speex_echo_state_init_mc(frame_size, filter_length, 1, 1);
}