// peeks at the audio data and fills 'data' with the requested kind
// (currently kind=0 returns mono 16 bit PCM data, and kind=1 returns
// 256 point FFT data). Return value is number of samples returned,
// which may be 0.
/*static*/ int MediaPlayer::snoop(short* data, int len, int kind) {
sp<IMemory> p;
const sp<IMediaPlayerService>& service = getMediaPlayerService();
if (service != 0) {
// Take a peek at the waveform. The returned data consists of 16 bit mono PCM data.
p = service->snoop();
if (p == NULL) {
return 0;
}
if (kind == 0) { // return waveform data
int plen = p->size();
len *= 2; // number of shorts -> number of bytes
short *src = (short*) p->pointer();
if (plen > len) {
plen = len;
}
memcpy(data, src, plen);
return plen / sizeof(short); // return number of samples
} else if (kind == 1) {
// TODO: use a more efficient FFT
// Right now this uses the speex library, which is compiled to do a float FFT
if (!ffttable) ffttable = spx_fft_init(512);
short *usrc = (short*) p->pointer();
float fsrc[512];
for (int i=0;i<512;i++)
fsrc[i] = usrc[i];
float fdst[512];
spx_fft_float(ffttable, fsrc, fdst);
if (len > 512) {
len = 512;
}
len /= 2; // only half the output data is valid
for (int i=0; i < len; i++)
data[i] = fdst[i];
return len;
}
} else {
LOGE("Unable to locate media service");
}
return 0;
}
EXPORT SpeexEchoState *speex_echo_state_init_mc(int frame_size, int filter_length, int nb_mic, int nb_speakers)
{
int i,N,M, C, K;
SpeexEchoState *st = (SpeexEchoState *)speex_alloc(sizeof(SpeexEchoState));
st->K = nb_speakers;
st->C = nb_mic;
C=st->C;
K=st->K;
#ifdef DUMP_ECHO_CANCEL_DATA
if (rFile || pFile || oFile)
speex_fatal("Opening dump files twice");
rFile = fopen("aec_rec.sw", "wb");
pFile = fopen("aec_play.sw", "wb");
oFile = fopen("aec_out.sw", "wb");
#endif
st->frame_size = frame_size;
st->window_size = 2*frame_size;
N = st->window_size;
M = st->M = (filter_length+st->frame_size-1)/frame_size;
st->cancel_count=0;
st->sum_adapt = 0;
st->saturated = 0;
st->screwed_up = 0;
/* This is the default sampling rate */
st->sampling_rate = 8000;
st->spec_average = DIV32_16(SHL32(EXTEND32(st->frame_size), 15), st->sampling_rate);
#ifdef FIXED_POINT
st->beta0 = DIV32_16(SHL32(EXTEND32(st->frame_size), 16), st->sampling_rate);
st->beta_max = DIV32_16(SHL32(EXTEND32(st->frame_size), 14), st->sampling_rate);
#else
st->beta0 = (2.0f*st->frame_size)/st->sampling_rate;
st->beta_max = (.5f*st->frame_size)/st->sampling_rate;
#endif
st->leak_estimate = 0;
st->fft_table = spx_fft_init(N);
st->e = (spx_word16_t*)speex_alloc(C*N*sizeof(spx_word16_t));
st->x = (spx_word16_t*)speex_alloc(K*N*sizeof(spx_word16_t));
st->input = (spx_word16_t*)speex_alloc(C*st->frame_size*sizeof(spx_word16_t));
st->y = (spx_word16_t*)speex_alloc(C*N*sizeof(spx_word16_t));
st->last_y = (spx_word16_t*)speex_alloc(C*N*sizeof(spx_word16_t));
st->Yf = (spx_word32_t*)speex_alloc((st->frame_size+1)*sizeof(spx_word32_t));
st->Rf = (spx_word32_t*)speex_alloc((st->frame_size+1)*sizeof(spx_word32_t));
st->Xf = (spx_word32_t*)speex_alloc((st->frame_size+1)*sizeof(spx_word32_t));
st->Yh = (spx_word32_t*)speex_alloc((st->frame_size+1)*sizeof(spx_word32_t));
st->Eh = (spx_word32_t*)speex_alloc((st->frame_size+1)*sizeof(spx_word32_t));
st->X = (spx_word16_t*)speex_alloc(K*(M+1)*N*sizeof(spx_word16_t));
st->Y = (spx_word16_t*)speex_alloc(C*N*sizeof(spx_word16_t));
st->E = (spx_word16_t*)speex_alloc(C*N*sizeof(spx_word16_t));
st->W = (spx_word32_t*)speex_alloc(C*K*M*N*sizeof(spx_word32_t));
#ifdef TWO_PATH
st->foreground = (spx_word16_t*)speex_alloc(M*N*C*K*sizeof(spx_word16_t));
#endif
st->PHI = (spx_word32_t*)speex_alloc(N*sizeof(spx_word32_t));
st->power = (spx_word32_t*)speex_alloc((frame_size+1)*sizeof(spx_word32_t));
st->power_1 = (spx_float_t*)speex_alloc((frame_size+1)*sizeof(spx_float_t));
st->window = (spx_word16_t*)speex_alloc(N*sizeof(spx_word16_t));
st->prop = (spx_word16_t*)speex_alloc(M*sizeof(spx_word16_t));
st->wtmp = (spx_word16_t*)speex_alloc(N*sizeof(spx_word16_t));
#ifdef FIXED_POINT
st->wtmp2 = (spx_word16_t*)speex_alloc(N*sizeof(spx_word16_t));
for (i=0;i<N>>1;i++)
{
st->window[i] = (16383-SHL16(spx_cos(DIV32_16(MULT16_16(25736,i<<1),N)),1));
st->window[N-i-1] = st->window[i];
}
#else
for (i=0;i<N;i++)
st->window[i] = .5-.5*cos(2*M_PI*i/N);
#endif
for (i=0;i<=st->frame_size;i++)
st->power_1[i] = FLOAT_ONE;
for (i=0;i<N*M*K*C;i++)
st->W[i] = 0;
{
spx_word32_t sum = 0;
/* Ratio of ~10 between adaptation rate of first and last block */
spx_word16_t decay = SHR32(spx_exp(NEG16(DIV32_16(QCONST16(2.4,11),M))),1);
st->prop[0] = QCONST16(.7, 15);
sum = EXTEND32(st->prop[0]);
for (i=1;i<M;i++)
{
st->prop[i] = MULT16_16_Q15(st->prop[i-1], decay);
sum = ADD32(sum, EXTEND32(st->prop[i]));
}
for (i=M-1;i>=0;i--)
{
st->prop[i] = DIV32(MULT16_16(QCONST16(.8f,15), st->prop[i]),sum);
}
}
st->memX = (spx_word16_t*)speex_alloc(K*sizeof(spx_word16_t));
st->memD = (spx_word16_t*)speex_alloc(C*sizeof(spx_word16_t));
st->memE = (spx_word16_t*)speex_alloc(C*sizeof(spx_word16_t));
st->preemph = QCONST16(.9,15);
if (st->sampling_rate<12000)
st->notch_radius = QCONST16(.9, 15);
else if (st->sampling_rate<24000)
st->notch_radius = QCONST16(.982, 15);
else
st->notch_radius = QCONST16(.992, 15);
st->notch_mem = (spx_mem_t*)speex_alloc(2*C*sizeof(spx_mem_t));
st->adapted = 0;
st->Pey = st->Pyy = FLOAT_ONE;
#ifdef TWO_PATH
st->Davg1 = st->Davg2 = 0;
st->Dvar1 = st->Dvar2 = FLOAT_ZERO;
#endif
st->play_buf = (spx_int16_t*)speex_alloc(K*(PLAYBACK_DELAY+1)*st->frame_size*sizeof(spx_int16_t));
st->play_buf_pos = PLAYBACK_DELAY*st->frame_size;
st->play_buf_started = 0;
return st;
}
EXPORT SpeexEchoState *speex_echo_state_init_mc(int frame_size, int filter_length, int nb_mic, int nb_speakers)
{
int i,N,M, C, K;
SpeexEchoState *st = (SpeexEchoState *)speex_alloc(sizeof(SpeexEchoState));
st->K = nb_speakers;
st->C = nb_mic;
C=st->C;
K=st->K;
#ifdef DUMP_ECHO_CANCEL_DATA
if (rFile || pFile || oFile)
speex_fatal("Opening dump files twice");
rFile = fopen("aec_rec.sw", "wb");
pFile = fopen("aec_play.sw", "wb");
oFile = fopen("aec_out.sw", "wb");
#endif
st->frame_size = frame_size;
st->window_size = 2*frame_size;
N = st->window_size;
M = st->M = (filter_length+st->frame_size-1)/frame_size;
st->cancel_count=0;
st->sum_adapt = 0;
st->saturated = 0;
st->screwed_up = 0;
/* This is the default sampling rate */
st->sampling_rate = 8000;
st->spec_average = DIV32_16(SHL32(EXTEND32(st->frame_size), 15), st->sampling_rate);
#ifdef FIXED_POINT
st->beta0 = DIV32_16(SHL32(EXTEND32(st->frame_size), 16), st->sampling_rate);
st->beta_max = DIV32_16(SHL32(EXTEND32(st->frame_size), 14), st->sampling_rate);
#else
st->beta0 = (2.0f*st->frame_size)/st->sampling_rate;
st->beta_max = (.5f*st->frame_size)/st->sampling_rate;
#endif
st->leak_estimate = 0;
st->fft_table = spx_fft_init(N);
st->e = (spx_word16_t*)speex_alloc(C*N*sizeof(spx_word16_t));
st->x = (spx_word16_t*)speex_alloc(K*N*sizeof(spx_word16_t));
st->input = (spx_word16_t*)speex_alloc(C*st->frame_size*sizeof(spx_word16_t));
st->y = (spx_word16_t*)speex_alloc(C*N*sizeof(spx_word16_t));
st->last_y = (spx_word16_t*)speex_alloc(C*N*sizeof(spx_word16_t));
st->Yf = (spx_word32_t*)speex_alloc((st->frame_size+1)*sizeof(spx_word32_t));
st->Rf = (spx_word32_t*)speex_alloc((st->frame_size+1)*sizeof(spx_word32_t));
st->Xf = (spx_word32_t*)speex_alloc((st->frame_size+1)*sizeof(spx_word32_t));
st->Yh = (spx_word32_t*)speex_alloc((st->frame_size+1)*sizeof(spx_word32_t));
st->Eh = (spx_word32_t*)speex_alloc((st->frame_size+1)*sizeof(spx_word32_t));
st->X = (spx_word16_t*)speex_alloc(K*(M+1)*N*sizeof(spx_word16_t));
st->Y = (spx_word16_t*)speex_alloc(C*N*sizeof(spx_word16_t));
st->E = (spx_word16_t*)speex_alloc(C*N*sizeof(spx_word16_t));
st->W = (spx_word32_t*)speex_alloc(C*K*M*N*sizeof(spx_word32_t));
#ifdef TWO_PATH
st->foreground = (spx_word16_t*)speex_alloc(M*N*C*K*sizeof(spx_word16_t));
#endif
st->PHI = (spx_word32_t*)speex_alloc(N*sizeof(spx_word32_t));
st->power = (spx_word32_t*)speex_alloc((frame_size+1)*sizeof(spx_word32_t));
st->power_1 = (spx_float_t*)speex_alloc((frame_size+1)*sizeof(spx_float_t));
st->window = (spx_word16_t*)speex_alloc(N*sizeof(spx_word16_t));
st->prop = (spx_word16_t*)speex_alloc(M*sizeof(spx_word16_t));
st->wtmp = (spx_word16_t*)speex_alloc(N*sizeof(spx_word16_t));
#ifdef FIXED_POINT
st->wtmp2 = (spx_word16_t*)speex_alloc