void CSpeexEC::Init(int frame_size, int filter_length, int sampling_rate)
{
Reset();
if (frame_size<=0 || filter_length<=0 || sampling_rate<=0)
{
m_nFrameSize =160;
m_nFilterLen = 160*8;
m_nSampleRate = 8000;
}
else
{
m_nFrameSize =frame_size;
m_nFilterLen = filter_length;
m_nSampleRate = sampling_rate;
}
m_pState = speex_echo_state_init(m_nFrameSize, m_nFilterLen);
m_pPreprocessorState = speex_preprocess_state_init(m_nFrameSize, m_nSampleRate);
m_pfNoise = new int[m_nFrameSize+1];
m_bHasInit = true;
speex_echo_ctl(m_pState, SPEEX_ECHO_SET_SAMPLING_RATE, &m_nSampleRate);
speex_preprocess_ctl(m_pPreprocessorState, SPEEX_PREPROCESS_SET_ECHO_STATE, m_pState);
}
static void speex_ec_preprocess(MSFilter *f){
SpeexECState *s=(SpeexECState*)f->data;
int delay_samples=0;
mblk_t *m;
s->echostarted=FALSE;
s->filterlength=(s->tail_length_ms*s->samplerate)/1000;
s->framesize=adjust_framesize(s->framesize_at_8000,s->samplerate);
delay_samples=s->delay_ms*s->samplerate/1000;
ms_message("Initializing speex echo canceler with framesize=%i, filterlength=%i, delay_samples=%i",
s->framesize,s->filterlength,delay_samples);
s->ecstate=speex_echo_state_init(s->framesize,s->filterlength);
s->den = speex_preprocess_state_init(s->framesize, s->samplerate);
speex_echo_ctl(s->ecstate, SPEEX_ECHO_SET_SAMPLING_RATE, &s->samplerate);
speex_preprocess_ctl(s->den, SPEEX_PREPROCESS_SET_ECHO_STATE, s->ecstate);
/* fill with zeroes for the time of the delay*/
m=allocb(delay_samples*2,0);
m->b_wptr+=delay_samples*2;
ms_bufferizer_put (&s->delayed_ref,m);
s->min_ref_samples=-1;
s->nominal_ref_samples=delay_samples;
audio_flow_controller_init(&s->afc);
s->flow_control_time = f->ticker->time;
#ifdef SPEEX_ECHO_GET_BLOB
apply_config(s);
#else
if (s->state_str) ms_warning("This version of speex doesn't support echo canceller restoration state. Rebuild speex and mediatreamer2 if you want to use this feature.");
#endif
}
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;
}
/*
* Create the echo canceller.
*/
pjs_echo_canceller::pjs_echo_canceller(pj_pool_t *pool_, unsigned clock_rate,
unsigned samples_per_frame_, unsigned tail_ms, unsigned latency_ms,
unsigned options) {
int sampling_rate = clock_rate;
unsigned ptime, lat_cnt;
unsigned delay_buf_opt = 0;
lat_ready = PJ_FALSE;
/* Create new pool and instantiate and init the EC */
pool = pj_pool_create(pool_->factory, "ec%p", 256, 256, NULL);
lock = new PPJ_SemaphoreLock(pool, NULL, 1, 1);
samples_per_frame = samples_per_frame_;
frm_buf = (pj_int16_t*) pj_pool_alloc(pool, samples_per_frame << 1);
state = speex_echo_state_init(samples_per_frame,
clock_rate * tail_ms / 1000);
speex_echo_ctl(state, SPEEX_ECHO_SET_SAMPLING_RATE,
&sampling_rate);
preprocess = speex_preprocess_state_init(samples_per_frame,
clock_rate);
tmp_frame = (pj_int16_t*) pj_pool_zalloc(pool, 2*samples_per_frame);
speex_preprocess_ctl(preprocess, SPEEX_PREPROCESS_SET_ECHO_STATE,
state);
pj_list_init(&lat_buf);
pj_list_init(&lat_free);
PJ_LOG(5, (THIS_FILE, "Creating echo canceler"));
/* Create latency buffers */
ptime = samples_per_frame * 1000 / clock_rate;
if (latency_ms < ptime) {
/* Give at least one frame delay to simplify programming */
latency_ms = ptime;
}
lat_cnt = latency_ms / ptime;
while (lat_cnt--) {
struct frame *frm;
frm = (struct frame*) pj_pool_alloc(pool,
(samples_per_frame << 1) + sizeof(struct frame));
pj_list_push_back(&lat_free, frm);
}
/* Create delay buffer to compensate drifts */
if (options & PJMEDIA_ECHO_USE_SIMPLE_FIFO)
delay_buf_opt |= PJMEDIA_DELAY_BUF_SIMPLE_FIFO;
pjmedia_delay_buf_create(pool, NULL, clock_rate, samples_per_frame,
1, (PJMEDIA_SOUND_BUFFER_COUNT + 1) * ptime, delay_buf_opt,
&delay_buf);
PJ_LOG(4, (THIS_FILE, "ECHO canceller created, clock_rate=%d, channel=%d, "
"samples per frame=%d, tail length=%d ms, "
"latency=%d ms", clock_rate, 1, samples_per_frame, tail_ms, latency_ms));
}
JNIEXPORT void JNICALL Java_com_pullmi_shanghai_TalkActivity_speex_1EchoCanceller_1open
(JNIEnv *env, jobject jobj, jint jSampleRate, jint jBufSize, jint jTotalSize)
{
//init
int sampleRate = jSampleRate;
st = speex_echo_state_init(jBufSize, jTotalSize);
den = speex_preprocess_state_init(jBufSize, sampleRate);
speex_echo_ctl(st, SPEEX_ECHO_SET_SAMPLING_RATE, &sampleRate);
speex_preprocess_ctl(den, SPEEX_PREPROCESS_SET_ECHO_STATE, st);
}
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;
}
RTC::ReturnCode_t EchoCanceler::onActivated(RTC::UniqueId ec_id)
{
RTC_DEBUG(("onActivated start"));
is_active = true;
BufferClr();
if ( mp_sest == NULL ) {
int sampleRate = 16000;
mp_sest = speex_echo_state_init(ECHOLEN, sampleRate * 2);
speex_echo_ctl(mp_sest, SPEEX_ECHO_SET_SAMPLING_RATE, &sampleRate);
}
RTC_DEBUG(("onActivated finish"));
return RTC::RTC_OK;
}
static int aec_alloc(struct speex_st **stp, void **ctx, struct aufilt_prm *prm)
{
struct speex_st *st;
uint32_t sampc;
int err, tmp, fl;
if (!stp || !ctx || !prm)
return EINVAL;
if (*ctx) {
*stp = mem_ref(*ctx);
return 0;
}
st = mem_zalloc(sizeof(*st), speex_aec_destructor);
if (!st)
return ENOMEM;
sampc = prm->srate * prm->ch * prm->ptime / 1000;
st->out = mem_alloc(2 * sampc, NULL);
if (!st->out) {
err = ENOMEM;
goto out;
}
/* Echo canceller with 200 ms tail length */
fl = 10 * sampc;
st->state = speex_echo_state_init(sampc, fl);
if (!st->state) {
err = ENOMEM;
goto out;
}
tmp = prm->srate;
err = speex_echo_ctl(st->state, SPEEX_ECHO_SET_SAMPLING_RATE, &tmp);
if (err < 0) {
warning("speex_aec: speex_echo_ctl: err=%d\n", err);
}
info("speex_aec: Speex AEC loaded: srate = %uHz\n", prm->srate);
out:
if (err)
mem_deref(st);
else
*ctx = *stp = st;
return err;
}
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;
}
extern "C" JNIEXPORT void Java_com_haitou_xiaoyoupai_imservice_support_audio_Speex_initEcho(
JNIEnv *env, jobject jobj, jint frame_size, jint filter_length) {
if (aec_status == AEC_OPENED)
return;
aec_status = AEC_OPENED;
int frm_size;
int f_length;
frm_size = frame_size;
f_length = filter_length;
echoState = speex_echo_state_init(frame_size, filter_length);
den = speex_preprocess_state_init(frame_size, sampleRate);
speex_echo_ctl(echoState, SPEEX_ECHO_SET_SAMPLING_RATE, &sampleRate);
speex_preprocess_ctl(den, SPEEX_PREPROCESS_SET_ECHO_STATE, echoState);
}
static void apply_config(SpeexECState *s){
if (s->state_str!=NULL){
size_t buflen=strlen(s->state_str);
uint8_t *buffer=alloca(buflen);
SpeexEchoStateBlob *blob;
if ((buflen=b64_decode(s->state_str,strlen(s->state_str),buffer,buflen))<=0){
ms_error("Could not decode base64 %s",s->state_str);
return;
}
blob=speex_echo_state_blob_new_from_memory(buffer,buflen);
if (blob==NULL){
ms_error("Could not create blob from config string");
return;
}
if (speex_echo_ctl(s->ecstate, SPEEX_ECHO_SET_BLOB, blob)!=0){
ms_error("Could not apply speex echo blob !");
}
speex_echo_state_blob_free(blob);
ms_message("speex echo state restored.");
}
}
static void fetch_config(SpeexECState *s){
SpeexEchoStateBlob *blob=NULL;
char *txt;
size_t txt_len;
if (s->ecstate==NULL) return;
if (speex_echo_ctl(s->ecstate, SPEEX_ECHO_GET_BLOB, &blob)!=0){
ms_error("Could not retrieve speex echo blob !");
return;
}
txt_len=(speex_echo_state_blob_get_size(blob)*4)+1;
txt=ms_malloc0(txt_len);
if (b64_encode(speex_echo_state_blob_get_data(blob),speex_echo_state_blob_get_size(blob),
txt,txt_len)==0){
ms_error("Base64 encoding failed.");
ms_free(txt);
return;
}
speex_echo_state_blob_free(blob);
if (s->state_str) ms_free(s->state_str);
s->state_str=txt;
}
int main(int argc, char **argv)
{
FILE *echo_fd, *ref_fd, *e_fd;
short echo_buf[NN], ref_buf[NN], e_buf[NN];
SpeexEchoState *st;
SpeexPreprocessState *den;
int sampleRate = 8000;
if (argc != 4)
{
fprintf(stderr, "testecho mic_signal.sw speaker_signal.sw output.sw\n");
exit(1);
}
echo_fd = fopen(argv[2], "rb");
ref_fd = fopen(argv[1], "rb");
e_fd = fopen(argv[3], "wb");
st = speex_echo_state_init(NN, TAIL);
den = speex_preprocess_state_init(NN, sampleRate);
speex_echo_ctl(st, SPEEX_ECHO_SET_SAMPLING_RATE, &sampleRate);
speex_preprocess_ctl(den, SPEEX_PREPROCESS_SET_ECHO_STATE, st);
while (!feof(ref_fd) && !feof(echo_fd))
{
fread(ref_buf, sizeof(short), NN, ref_fd);
fread(echo_buf, sizeof(short), NN, echo_fd);
speex_echo_cancellation(st, ref_buf, echo_buf, e_buf);
speex_preprocess_run(den, e_buf);
fwrite(e_buf, sizeof(short), NN, e_fd);
}
speex_echo_state_destroy(st);
speex_preprocess_state_destroy(den);
fclose(e_fd);
fclose(echo_fd);
fclose(ref_fd);
return 0;
}
int main(int argc, char **argv)
{
int echo_fd, ref_fd, e_fd;
spx_int32_t noise[NN+1];
short echo_buf[NN], ref_buf[NN], e_buf[NN];
SpeexEchoState *st;
SpeexPreprocessState *den;
if (argc != 4)
{
fprintf (stderr, "testecho mic_signal.sw speaker_signal.sw output.sw\n");
exit(1);
}
echo_fd = open (argv[2], O_RDONLY);
ref_fd = open (argv[1], O_RDONLY);
e_fd = open (argv[3], O_WRONLY | O_CREAT | O_TRUNC, 0644);
st = speex_echo_state_init(NN, TAIL);
den = speex_preprocess_state_init(NN, 8000);
int tmp = 8000;
speex_echo_ctl(st, SPEEX_ECHO_SET_SAMPLING_RATE, &tmp);
while (read(ref_fd, ref_buf, NN*2))
{
read(echo_fd, echo_buf, NN*2);
speex_echo_cancel(st, ref_buf, echo_buf, e_buf, noise);
/*speex_preprocess(den, e_buf, noise);*/
write(e_fd, e_buf, NN*2);
}
speex_echo_state_destroy(st);
speex_preprocess_state_destroy(den);
close(e_fd);
close(echo_fd);
close(ref_fd);
return 0;
}
void KotiAEC_init(int16_t frame_size, int32_t sample_freq, AEC_CORE aec_core,
int speex_filter_length, int16_t agc_mode, int16_t compression_gain_db, uint8_t limiter_enable, float snd_amplification)
{
int32_t speex_agc = 0; float speex_agc_level = 24000, speex_agc_level_tmp = 3000;
int32_t speex_noise_suppress = -15;
aec_core_used = aec_core;
output_sound_amplification = snd_amplification;
switch(aec_core)
{
#ifdef WEBRTC_AEC_CORE_ENABLED
case WEBRTC_AEC:
if(webrtc_aec_pty.webrtc_aec == NULL)
{
webrtc_aec_pty.frame_size = frame_size;
webrtc_aec_pty.sample_freq = sample_freq;
webrtc_aec_pty.sndcard_sample_freq = sample_freq;
webrtc_aec_pty.sndcard_delay_ms = ((float)speex_filter_length/sample_freq) * 1000;
if( WebRtcAec_Create(&webrtc_aec_pty.webrtc_aec) == 0)
WebRtcAec_Init(webrtc_aec_pty.webrtc_aec, webrtc_aec_pty.sample_freq, webrtc_aec_pty.sndcard_sample_freq);
if( WebRtcNsx_Create((NsxHandle**)&webrtc_aec_pty.webrtc_ns) == 0)
{
WebRtcNsx_Init((NsxHandle*)webrtc_aec_pty.webrtc_ns, webrtc_aec_pty.sample_freq);
WebRtcNsx_set_policy((NsxHandle*)webrtc_aec_pty.webrtc_ns, 1);
}
if( WebRtcAgc_Create(&webrtc_aec_pty.webrtc_agc) == 0)
{
WebRtcAgc_Init(webrtc_aec_pty.webrtc_agc, 0, 255, agc_mode/*kAgcModeFixedDigital*/, webrtc_aec_pty.sample_freq);
WebRtcAgc_config_t conf = {0, compression_gain_db, limiter_enable};
WebRtcAgc_set_config(webrtc_aec_pty.webrtc_agc, conf);
}
}
else
{
WebRtcAec_Init(webrtc_aec_pty.webrtc_aec, webrtc_aec_pty.sample_freq, webrtc_aec_pty.sndcard_sample_freq);
}
break;
case WEBRTC_AECM:
if(webrtc_aecm_pty.webrtc_aec == NULL)
{
webrtc_aecm_pty.frame_size = frame_size;
webrtc_aecm_pty.sample_freq = sample_freq;
webrtc_aecm_pty.sndcard_sample_freq = sample_freq;
webrtc_aecm_pty.sndcard_delay_ms = ((float)speex_filter_length/sample_freq) * 1000;
// webrtc_aecm_pty.sndcard_delay_ms = speex_filter_length/frame_size;
if( WebRtcAecm_Create(&webrtc_aecm_pty.webrtc_aec) == 0)
WebRtcAecm_Init(webrtc_aecm_pty.webrtc_aec, webrtc_aecm_pty.sample_freq);
if( WebRtcNsx_Create((NsxHandle**)&webrtc_aecm_pty.webrtc_ns) == 0)
{
WebRtcNsx_Init((NsxHandle*)webrtc_aecm_pty.webrtc_ns, webrtc_aecm_pty.sample_freq);
WebRtcNsx_set_policy((NsxHandle*)webrtc_aecm_pty.webrtc_ns, 1);
}
if( WebRtcAgc_Create(&webrtc_aecm_pty.webrtc_agc) == 0)
{
WebRtcAgc_Init(webrtc_aecm_pty.webrtc_agc, 0, 255, agc_mode/*kAgcModeFixedDigital*/, webrtc_aecm_pty.sample_freq);
WebRtcAgc_config_t conf = {0, compression_gain_db, limiter_enable};
WebRtcAgc_set_config(webrtc_aecm_pty.webrtc_agc, conf);
}
}
break;
#endif
case SPEEX_AEC:
default:
#ifdef OLD_SPEEX_AEC
if(speex_aec_pty.speex_echo_state == NULL && speex_aec_pty.speex_preprocess_state == NULL)
{
speex_aec_pty.frame_size = frame_size;
speex_aec_pty.sample_freq = sample_freq;
speex_aec_pty.filter_length = speex_filter_length;
if(speex_aec_pty.filter_length < frame_size)
speex_aec_pty.filter_length = frame_size*FILTER_LENGTH_MULTIPLE_OF_FRAME_SIZE;
speex_aec_pty.speex_echo_state = speex_echo_state_init(frame_size, speex_aec_pty.filter_length);
speex_aec_pty.speex_preprocess_state = speex_preprocess_state_init(frame_size, sample_freq);
speex_echo_ctl((SpeexEchoState*)speex_aec_pty.speex_echo_state, SPEEX_ECHO_SET_SAMPLING_RATE, &speex_aec_pty.sample_freq);
speex_aec_pty.nosie = (int32_t*) malloc( (frame_size+1) * sizeof(int32_t) );
}
#else
if(speex_aec_pty.speex_echo_state == NULL && speex_aec_pty.speex_preprocess_state == NULL)
{
speex_aec_pty.frame_size = frame_size;
speex_aec_pty.sample_freq = sample_freq;
speex_aec_pty.filter_length = speex_filter_length;
if(speex_aec_pty.filter_length < frame_size)
speex_aec_pty.filter_length = frame_size*FILTER_LENGTH_MULTIPLE_OF_FRAME_SIZE;
speex_aec_pty.speex_echo_state = speex_echo_state_init(frame_size, speex_aec_pty.filter_length);
speex_aec_pty.speex_preprocess_state = speex_preprocess_state_init(frame_size, sample_freq);
speex_echo_ctl((SpeexEchoState*)speex_aec_pty.speex_echo_state, SPEEX_ECHO_SET_SAMPLING_RATE, &speex_aec_pty.sample_freq);
speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_SET_ECHO_STATE,
speex_aec_pty.speex_echo_state);
speex_aec_pty.nosie = (int32_t*) malloc( (frame_size+1) * sizeof(int32_t) );
speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_SET_AGC, &speex_agc);
speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_SET_AGC_LEVEL,
&speex_agc_level);
speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_SET_NOISE_SUPPRESS,
&speex_noise_suppress);
speex_aec_pty.speex_preprocess_state_tmp = speex_preprocess_state_init(frame_size, sample_freq);
speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state_tmp, SPEEX_PREPROCESS_SET_AGC, &speex_agc);
speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state_tmp, SPEEX_PREPROCESS_SET_AGC_LEVEL,
&speex_agc_level_tmp);
speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state_tmp, SPEEX_PREPROCESS_SET_NOISE_SUPPRESS,
&speex_noise_suppress);
// int32_t speex_vad = 1;
// int32_t speex_vad_prob_start = 80;
// int32_t speex_vad_prob_continue = 65;
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_SET_VAD, &speex_vad);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_SET_PROB_START,
// &speex_vad_prob_start);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_SET_PROB_CONTINUE,
// &speex_vad_prob_continue);
// int32_t tmp = -1; float tmp1 = 0.0f;
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_DENOISE, &tmp);
// printf("SPEEX_PREPROCESS_GET_DENOISE: %d\n", tmp);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_NOISE_SUPPRESS, &tmp);
// printf("SPEEX_PREPROCESS_GET_NOISE_SUPPRESS: %d\n", tmp);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_AGC, &tmp);
// printf("SPEEX_PREPROCESS_GET_AGC: %d\n", tmp);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_VAD, &tmp);
// printf("SPEEX_PREPROCESS_GET_VAD: %d\n", tmp);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_PROB_START, &tmp);
// printf("SPEEX_PREPROCESS_GET_PROB_START: %d\n", tmp);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_PROB_CONTINUE, &tmp);
// printf("SPEEX_PREPROCESS_GET_PROB_CONTINUE: %d\n", tmp);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_AGC_LEVEL, &tmp1);
// printf("SPEEX_PREPROCESS_GET_AGC_LEVEL: %f\n", tmp1);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_AGC_DECREMENT, &tmp);
// printf("SPEEX_PREPROCESS_GET_AGC_DECREMENT: %d\n", tmp);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_AGC_INCREMENT, &tmp);
// printf("SPEEX_PREPROCESS_GET_AGC_INCREMENT: %d\n", tmp);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_AGC_MAX_GAIN, &tmp);
// printf("SPEEX_PREPROCESS_GET_AGC_MAX_GAIN: %d\n", tmp);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_AGC_GAIN, &tmp);
// printf("SPEEX_PREPROCESS_GET_AGC_GAIN: %d\n", tmp);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_AGC_TARGET, &tmp);
// printf("SPEEX_PREPROCESS_GET_AGC_TARGET: %d\n", tmp);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_AGC_LOUDNESS, &tmp);
// printf("SPEEX_PREPROCESS_GET_AGC_LOUDNESS: %d\n", tmp);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_ECHO_SUPPRESS, &tmp);
// printf("SPEEX_PREPROCESS_GET_ECHO_SUPPRESS: %d\n", tmp);
// speex_preprocess_ctl((SpeexPreprocessState*)speex_aec_pty.speex_preprocess_state, SPEEX_PREPROCESS_GET_ECHO_SUPPRESS_ACTIVE, &tmp);
// printf("SPEEX_PREPROCESS_GET_ECHO_SUPPRESS_ACTIVE: %d\n", tmp);
}
#endif
break;
}
}
/*
* 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 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;
//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;
}
static int tdav_speex_denoise_open(tmedia_denoise_t* self, uint32_t frame_size, uint32_t sampling_rate)
{
tdav_speex_denoise_t *denoiser = (tdav_speex_denoise_t *)self;
float f;
int i;
if(!denoiser->echo_state && TMEDIA_DENOISE(denoiser)->echo_supp_enabled){
TSK_DEBUG_INFO("Init Aec frame_size[%u] filter_length[%u] SampleRate[%u]",
(uint32_t)(frame_size* sizeof(spx_int16_t)),TMEDIA_DENOISE(denoiser)->echo_tail*frame_size,sampling_rate);
if((denoiser->echo_state = speex_echo_state_init(frame_size, TMEDIA_DENOISE(denoiser)->echo_tail))){
speex_echo_ctl(denoiser->echo_state, SPEEX_ECHO_SET_SAMPLING_RATE, &sampling_rate);
}
}
if(!denoiser->preprocess_state_record && !denoiser->preprocess_state_playback){
denoiser->frame_size = frame_size;
if((denoiser->preprocess_state_record = speex_preprocess_state_init(frame_size, sampling_rate))
&& (denoiser->preprocess_state_playback = speex_preprocess_state_init(frame_size, sampling_rate))
){
// Echo suppression
if(denoiser->echo_state){
int echo_supp , echo_supp_active = 0;
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_SET_ECHO_STATE, denoiser->echo_state);
TSK_FREE(denoiser->echo_output_frame);
denoiser->echo_output_frame = tsk_calloc(denoiser->frame_size, sizeof(spx_int16_t));
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_GET_ECHO_SUPPRESS , &echo_supp );
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_GET_ECHO_SUPPRESS_ACTIVE , &echo_supp_active );
TSK_DEBUG_INFO("AEC echo_supp level [%d] echo_supp_active level[%d] ", echo_supp , echo_supp_active);
echo_supp = -60 ;
echo_supp_active = -60 ;
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_SET_ECHO_SUPPRESS , &echo_supp );
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_SET_ECHO_SUPPRESS_ACTIVE , &echo_supp_active );
// TRACES
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_GET_ECHO_SUPPRESS , &echo_supp );
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_GET_ECHO_SUPPRESS_ACTIVE , &echo_supp_active );
TSK_DEBUG_INFO("New aec echo_supp level [%d] echo_supp_active level[%d] ", echo_supp , echo_supp_active);
}
// Noise suppression
if(TMEDIA_DENOISE(denoiser)->noise_supp_enabled){
TSK_DEBUG_INFO("SpeexDSP: Noise supp enabled");
i = 1;
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_SET_DENOISE, &i);
speex_preprocess_ctl(denoiser->preprocess_state_playback, SPEEX_PREPROCESS_SET_DENOISE, &i);
i = TMEDIA_DENOISE(denoiser)->noise_supp_level;
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_SET_NOISE_SUPPRESS, &i);
speex_preprocess_ctl(denoiser->preprocess_state_playback, SPEEX_PREPROCESS_SET_NOISE_SUPPRESS, &i);
}
else{
i = 0;
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_SET_DENOISE, &i);
speex_preprocess_ctl(denoiser->preprocess_state_playback, SPEEX_PREPROCESS_SET_DENOISE, &i);
}
// Automatic gain control
if(TMEDIA_DENOISE(denoiser)->agc_enabled){
float agc_level = TMEDIA_DENOISE(denoiser)->agc_level;
TSK_DEBUG_INFO("SpeexDSP: AGC enabled");
i = 1;
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_SET_AGC, &i);
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_SET_AGC_LEVEL, &agc_level);
}
else{
i = 0, f = 8000.0f;
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_SET_AGC, &i);
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_SET_AGC_LEVEL, &f);
}
// Voice Activity detection
i = TMEDIA_DENOISE(denoiser)->vad_enabled ? 1 : 0;
speex_preprocess_ctl(denoiser->preprocess_state_record, SPEEX_PREPROCESS_SET_VAD, &i);
return 0;
}
else{
TSK_DEBUG_ERROR("Failed to create Speex preprocessor state");
return -2;
}
}
return 0;
}
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 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();
}
static void
AudioQualityImprovement_updatePreprocess(AudioQualityImprovement *aqi)
{
if (aqi->echo)
{
int frameSize = 0;
if ((aqi->echoFilterLengthInMillis > 0)
&& (aqi->sampleRate > 0)
&& speex_echo_ctl(
aqi->echo,
SPEEX_ECHO_GET_FRAME_SIZE, &frameSize))
frameSize = 0;
if (frameSize
&& (aqi->frameSize
== (frameSize * (16 /* sampleSizeInBits */ / 8))))
{
int echoFilterLength
= (int)
((aqi->sampleRate * aqi->echoFilterLengthInMillis)
/ 1000);
if (aqi->filterLengthOfEcho != echoFilterLength)
frameSize = 0;
}
else
frameSize = 0;
if (frameSize < 1)
{
if (aqi->preprocess)
{
speex_preprocess_ctl(
aqi->preprocess,
SPEEX_PREPROCESS_SET_ECHO_STATE, NULL);
}
speex_echo_state_destroy(aqi->echo);
aqi->echo = NULL;
}
}
if (aqi->preprocess
&& ((aqi->frameSize != aqi->frameSizeOfPreprocess)
|| (aqi->sampleRate != aqi->sampleRateOfPreprocess)))
{
speex_preprocess_state_destroy(aqi->preprocess);
aqi->preprocess = NULL;
}
if ((aqi->frameSize > 0) && (aqi->sampleRate > 0))
{
if (aqi->echoFilterLengthInMillis > 0)
{
if (!(aqi->echo))
{
int echoFrameSize
= aqi->frameSize / (16 /* sampleSizeInBits */ / 8);
int echoFilterLength
= (int)
((aqi->sampleRate * aqi->echoFilterLengthInMillis)
/ 1000);
aqi->echo
= speex_echo_state_init(echoFrameSize, echoFilterLength);
aqi->filterLengthOfEcho = echoFilterLength;
/*
* Since echo has just been (re)created, make sure that the
* delay in play will happen again taking into consideration the
* latest frameSize.
*/
if (aqi->play)
AudioQualityImprovement_updatePlayIsDelaying(aqi);
}
if (aqi->echo)
{
speex_echo_ctl(
aqi->echo,
SPEEX_ECHO_SET_SAMPLING_RATE, &(aqi->sampleRate));
}
}
if (aqi->denoise || aqi->echo)
{
if (!(aqi->preprocess))
{
aqi->preprocess
= speex_preprocess_state_init(
aqi->frameSize / (16 /* sampleSizeInBits */ / 8),
aqi->sampleRate);
aqi->frameSizeOfPreprocess = aqi->frameSize;
aqi->sampleRateOfPreprocess = aqi->sampleRate;
if (aqi->preprocess)
{
int on = 1;
speex_preprocess_ctl(
aqi->preprocess,
SPEEX_PREPROCESS_SET_DEREVERB, &on);
speex_preprocess_ctl(
aqi->preprocess,
SPEEX_PREPROCESS_SET_VAD, &on);
}
}
if (aqi->preprocess)
{
int denoise = (aqi->denoise == JNI_TRUE) ? 1 : 0;
speex_preprocess_ctl(
aqi->preprocess,
SPEEX_PREPROCESS_SET_DENOISE, &denoise);
if (aqi->echo)
{
speex_preprocess_ctl(
aqi->preprocess,
SPEEX_PREPROCESS_SET_ECHO_STATE, aqi->echo);
}
}
}
}
}
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;
}
int main(int argc, char **argv)
{
int sd, rc, n, tmp;
char msg[MAX_MSG];
int nfds;
int send_timestamp = 0;
int recv_started = 0;
struct pollfd *pfds;
struct alsa_dev *dev;
CELTEncoder *enc_state;
CELTDecoder *dec_state;
CELTMode *mode;
struct sched_param param;
JitterBuffer *jitter;
SpeexEchoState *echo_state;
char mac_own[6], mac_remote[6];
if (argc != 4)
panic("Usage %s plughw:0,0 <lmac in xx:xx:xx:xx:xx:xx> <rmac>\n", argv[0]);
register_signal(SIGINT, sighandler);
hack_mac(mac_own, argv[2], strlen(argv[2]));
hack_mac(mac_remote, argv[3], strlen(argv[3]));
sd = socket(AF_LANA, SOCK_RAW, 0);
if (sd < 0)
panic("%s: cannot open socket \n", argv[0]);
printf("If ready hit key!\n"); //user must do binding
getchar();
dev = alsa_open(argv[1], SAMPLING_RATE, CHANNELS, FRAME_SIZE);
mode = celt_mode_create(SAMPLING_RATE, FRAME_SIZE, NULL);
enc_state = celt_encoder_create(mode, CHANNELS, NULL);
dec_state = celt_decoder_create(mode, CHANNELS, NULL);
param.sched_priority = sched_get_priority_min(SCHED_FIFO);
if (sched_setscheduler(0, SCHED_FIFO, ¶m))
whine("sched_setscheduler error!\n");
/* Setup all file descriptors for poll()ing */
nfds = alsa_nfds(dev);
pfds = xmalloc(sizeof(*pfds) * (nfds + 1));
alsa_getfds(dev, pfds, nfds);
pfds[nfds].fd = sd;
pfds[nfds].events = POLLIN;
/* Setup jitter buffer using decoder */
jitter = jitter_buffer_init(FRAME_SIZE);
tmp = FRAME_SIZE;
jitter_buffer_ctl(jitter, JITTER_BUFFER_SET_MARGIN, &tmp);
/* Echo canceller with 200 ms tail length */
echo_state = speex_echo_state_init(FRAME_SIZE, 10 * FRAME_SIZE);
tmp = SAMPLING_RATE;
speex_echo_ctl(echo_state, SPEEX_ECHO_SET_SAMPLING_RATE, &tmp);
register_signal_f(SIGALRM, timer_elapsed, SA_SIGINFO);
alsa_start(dev);
printf("ALSA started!\n");
itimer.it_interval.tv_sec = 0;
itimer.it_interval.tv_usec = interval;
itimer.it_value.tv_sec = 0;
itimer.it_value.tv_usec = interval;
setitimer(ITIMER_REAL, &itimer, NULL);
while (!sigint) {
poll(pfds, nfds + 1, -1);
/* Received packets */
if (pfds[nfds].revents & POLLIN) {
memset(msg, 0, MAX_MSG);
n = recv(sd, msg, MAX_MSG, 0);
if (n <= 0)
goto do_alsa;
pkts_in++;
int recv_timestamp;
memcpy(&recv_timestamp, msg, sizeof(recv_timestamp));
JitterBufferPacket packet;
packet.data = msg+4/*+6+6+2*/;
packet.len = n-4/*-6-6-2*/;
packet.timestamp = recv_timestamp;
packet.span = FRAME_SIZE;
packet.sequence = 0;
/* Put content of the packet into the jitter buffer,
except for the pseudo-header */
jitter_buffer_put(jitter, &packet);
recv_started = 1;
}
do_alsa:
/* Ready to play a frame (playback) */
if (alsa_play_ready(dev, pfds, nfds)) {
short pcm[FRAME_SIZE * CHANNELS] = {0};
if (recv_started) {
JitterBufferPacket packet;
/* Get audio from the jitter buffer */
packet.data = msg;
packet.len = MAX_MSG;
jitter_buffer_tick(jitter);
jitter_buffer_get(jitter, &packet, FRAME_SIZE,
NULL);
if (packet.len == 0)
packet.data=NULL;
celt_decode(dec_state, (const unsigned char *)
packet.data, packet.len, pcm);
}
/* Playback the audio and reset the echo canceller
if we got an underrun */
alsa_write(dev, pcm, FRAME_SIZE);
// if (alsa_write(dev, pcm, FRAME_SIZE))
// speex_echo_state_reset(echo_state);
/* Put frame into playback buffer */
// speex_echo_playback(echo_state, pcm);
}
/* Audio available from the soundcard (capture) */
if (alsa_cap_ready(dev, pfds, nfds)) {
short pcm[FRAME_SIZE * CHANNELS];
//pcm2[FRAME_SIZE * CHANNELS];
char outpacket[MAX_MSG];
alsa_read(dev, pcm, FRAME_SIZE);
/* Perform echo cancellation */
// speex_echo_capture(echo_state, pcm, pcm2);
// for (i = 0; i < FRAME_SIZE * CHANNELS; ++i)
// pcm[i] = pcm2[i];
celt_encode(enc_state, pcm, NULL, (unsigned char *)
(outpacket+4+6+6+2), PACKETSIZE);
/* Pseudo header: four null bytes and a 32-bit
timestamp; XXX hack */
memcpy(outpacket,mac_remote,6);
memcpy(outpacket+6,mac_own,6);
outpacket[6+6] = (uint8_t) 0xac;
outpacket[6+6+1] = (uint8_t) 0xdc;
memcpy(outpacket+6+6+2, &send_timestamp, sizeof(send_timestamp));
send_timestamp += FRAME_SIZE;
rc = sendto(sd, outpacket, PACKETSIZE+4+6+6+2, 0, NULL, 0);
if (rc < 0)
panic("cannot send to socket");
pkts_out++;
}
}
itimer.it_interval.tv_sec = 0;
itimer.it_interval.tv_usec = 0;
itimer.it_value.tv_sec = 0;
itimer.it_value.tv_usec = 0;
setitimer(ITIMER_REAL, &itimer, NULL);
close(sd);
return 0;
}
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;
}
int main(int argc, char *argv[])
{
int sd, rc, n;
int i;
struct sockaddr_in cliAddr, remoteAddr;
char msg[MAX_MSG];
struct hostent *h;
int local_port, remote_port;
int nfds;
struct pollfd *pfds;
SpeexPreprocessState *preprocess;
AlsaDevice *audio_dev;
int tmp;
if (argc != 5)
{
fprintf(stderr, "wrong options\n");
exit(1);
}
h = gethostbyname(argv[2]);
if(h==NULL) {
fprintf(stderr, "%s: unknown host '%s' \n", argv[0], argv[1]);
exit(1);
}
local_port = atoi(argv[3]);
remote_port = atoi(argv[4]);
printf("%s: sending data to '%s' (IP : %s) \n", argv[0], h->h_name,
inet_ntoa(*(struct in_addr *)h->h_addr_list[0]));
{
remoteAddr.sin_family = h->h_addrtype;
memcpy((char *) &remoteAddr.sin_addr.s_addr,
h->h_addr_list[0], h->h_length);
remoteAddr.sin_port = htons(remote_port);
}
/* socket creation */
sd=socket(AF_INET, SOCK_DGRAM, 0);
if(sd<0) {
printf("%s: cannot open socket \n",argv[0]);
exit(1);
}
/* bind any port */
cliAddr.sin_family = AF_INET;
cliAddr.sin_addr.s_addr = htonl(INADDR_ANY);
cliAddr.sin_port = htons(local_port);
rc = bind(sd, (struct sockaddr *) &cliAddr, sizeof(cliAddr));
if(rc<0) {
printf("%s: cannot bind port\n", argv[0]);
exit(1);
}
/* Setup audio device */
audio_dev = alsa_device_open(argv[1], SAMPLING_RATE, 1, FRAME_SIZE);
/* Setup the encoder and decoder in wideband */
void *enc_state, *dec_state;
enc_state = speex_encoder_init(&speex_wb_mode);
tmp = 8;
speex_encoder_ctl(enc_state, SPEEX_SET_QUALITY, &tmp);
tmp = 2;
speex_encoder_ctl(enc_state, SPEEX_SET_COMPLEXITY, &tmp);
dec_state = speex_decoder_init(&speex_wb_mode);
tmp = 1;
speex_decoder_ctl(dec_state, SPEEX_SET_ENH, &tmp);
SpeexBits enc_bits, dec_bits;
speex_bits_init(&enc_bits);
speex_bits_init(&dec_bits);
struct sched_param param;
/*param.sched_priority = 40; */
param.sched_priority = sched_get_priority_min(SCHED_FIFO);
if (sched_setscheduler(0,SCHED_FIFO,¶m))
perror("sched_setscheduler");
int send_timestamp = 0;
int recv_started=0;
/* Setup all file descriptors for poll()ing */
nfds = alsa_device_nfds(audio_dev);
pfds = malloc(sizeof(*pfds)*(nfds+1));
alsa_device_getfds(audio_dev, pfds, nfds);
pfds[nfds].fd = sd;
pfds[nfds].events = POLLIN;
/* Setup jitter buffer using decoder */
SpeexJitter jitter;
speex_jitter_init(&jitter, dec_state, SAMPLING_RATE);
/* Echo canceller with 200 ms tail length */
SpeexEchoState *echo_state = speex_echo_state_init(FRAME_SIZE, 10*FRAME_SIZE);
tmp = SAMPLING_RATE;
speex_echo_ctl(echo_state, SPEEX_ECHO_SET_SAMPLING_RATE, &tmp);
/* Setup preprocessor and associate with echo canceller for residual echo suppression */
preprocess = speex_preprocess_state_init(FRAME_SIZE, SAMPLING_RATE);
speex_preprocess_ctl(preprocess, SPEEX_PREPROCESS_SET_ECHO_STATE, echo_state);
alsa_device_start(audio_dev);
/* Infinite loop on capture, playback and receiving packets */
while (1)
{
/* Wait for either 1) capture 2) playback 3) socket data */
poll(pfds, nfds+1, -1);
/* Received packets */
if (pfds[nfds].revents & POLLIN)
{
/*fprintf (stderr, "x");*/
n = recv(sd, msg, MAX_MSG, 0);
int recv_timestamp = ((int*)msg)[1];
int payload = ((int*)msg)[0];
if ((payload & 0x80000000) == 0)
{
/* Put content of the packet into the jitter buffer, except for the pseudo-header */
speex_jitter_put(&jitter, msg+8, n-8, recv_timestamp);
recv_started = 1;
}
}
/* Ready to play a frame (playback) */
if (alsa_device_playback_ready(audio_dev, pfds, nfds))
{
short pcm[FRAME_SIZE];
if (recv_started)
{
/* Get audio from the jitter buffer */
speex_jitter_get(&jitter, pcm, NULL);
} else {
for (i=0; i<FRAME_SIZE; i++)
pcm[i] = 0;
}
/* Playback the audio and reset the echo canceller if we got an underrun */
if (alsa_device_write(audio_dev, pcm, FRAME_SIZE))
speex_echo_state_reset(echo_state);
/* Put frame into playback buffer */
speex_echo_playback(echo_state, pcm);
}
/* Audio available from the soundcard (capture) */
if (alsa_device_capture_ready(audio_dev, pfds, nfds))
{
short pcm[FRAME_SIZE], pcm2[FRAME_SIZE];
char outpacket[MAX_MSG];
/* Get audio from the soundcard */
alsa_device_read(audio_dev, pcm, FRAME_SIZE);
/* Perform echo cancellation */
speex_echo_capture(echo_state, pcm, pcm2);
for (i=0; i<FRAME_SIZE; i++)
pcm[i] = pcm2[i];
speex_bits_reset(&enc_bits);
/* Apply noise/echo suppression */
speex_preprocess_run(preprocess, pcm);
/* Encode */
speex_encode_int(enc_state, pcm, &enc_bits);
int packetSize = speex_bits_write(&enc_bits, outpacket+8, MAX_MSG);
/* Pseudo header: four null bytes and a 32-bit timestamp */
((int*)outpacket)[0] = htonl(0);
((int*)outpacket)[1] = send_timestamp;
send_timestamp += FRAME_SIZE;
rc = sendto(sd, outpacket, packetSize+8, 0,
(struct sockaddr *) &remoteAddr,
sizeof(remoteAddr));
if(rc<0) {
printf("cannot send audio data\n");
close(sd);
exit(1);
}
}
}
return 0;
}
