void CMFCC::init(float SampleF,char* mempool, int &mpidx,long DefFFTLen,long DefFrameLen,long DefSubBandNum, long DefCepstrumNum)
{
long M, SubBandNo, i, j;
float ms,pi_factor, mfnorm, a, melk, t;
MPidx = &mpidx;
m_pMemPool = mempool;
historyMPidx = mpidx;
FrameLen = DefFrameLen;
SubBandNum = DefSubBandNum;
CepstrumNum = DefCepstrumNum;
initFFT(m_pMemPool,mpidx,DefFFTLen);
FFT_LEN = GetFFTAnalyseLen();
// std::cout <<"FFT_LEN = "<<FFT_LEN<<std::endl;
if(((unsigned int)(m_pMemPool + mpidx))%4)
mpidx += 4 - ((unsigned int)(m_pMemPool + mpidx))%4;
cosTab = (float*)(m_pMemPool + mpidx);
mpidx += sizeof(float) * (SubBandNum+1)*(SubBandNum+1); //DCT变换系数
hamWin = (float*)(m_pMemPool + mpidx);
mpidx += sizeof(float) * FrameLen; //hamming 窗系数
cepWin = (float*)(m_pMemPool + mpidx);
mpidx += sizeof(float) * (SubBandNum+1);
MelBandBoundary = (float*)(m_pMemPool + mpidx);
mpidx += sizeof(float) * (SubBandNum+2);
SubBandWeight = (float*)(m_pMemPool + mpidx);
mpidx += sizeof(float) * FFT_LEN/2;
SubBandIndex = (long*)(m_pMemPool + mpidx);
mpidx += sizeof(long) * FFT_LEN/2;
SubBandEnergy = (float*)(m_pMemPool + mpidx);
mpidx += sizeof(float) * (SubBandNum+2); //for accumulating the corresponding
FFTFrame = (float*)(m_pMemPool + mpidx);
mpidx += sizeof(float) * (FFT_LEN); //for accumulating the corresponding
Cepstrum = (float*)(m_pMemPool + mpidx);
mpidx += sizeof(float) * (SubBandNum+1); //for accumulating the corresponding
fres = SampleF/(FFT_LEN*700.0f);
//caculating the mel scale sub-band boundary
//由于子带0的系数（直流分量）不参与MFCC特征计算，所以子带个数要多一个
M = SubBandNum+1;
ms = Mel(FFT_LEN/2); //计算1/2采样率所对应的MEL刻度
//Note that the sub-band 0 is not used for cepstrum caculating
for ( SubBandNo = 0; SubBandNo <= M; SubBandNo++ )
{ //计算每个子带的起始MEL刻度
MelBandBoundary[SubBandNo] = ( (float)SubBandNo/(float)M )*ms;
}
//mapping the FFT frequence component into the corresponding sub-band
for ( i=0,SubBandNo=1; i< FFT_LEN/2; i++)
{
melk = Mel(i);
while( MelBandBoundary[SubBandNo] < melk ) SubBandNo++;
SubBandIndex[i] = SubBandNo-1;
}
//caculating the weighting coefficients for each FFT frequence components
for(i=0; i< FFT_LEN/2; i++)
{ //以子带的起始MEL频率为中心，计算三角窗加权系数
SubBandNo = SubBandIndex[i];
SubBandWeight[i] = (MelBandBoundary[SubBandNo+1]-Mel(i))/(MelBandBoundary[SubBandNo+1]-MelBandBoundary[SubBandNo]);
}
pi_factor = (float)( asin(1.0)*2.0/(float)SubBandNum );
mfnorm = (float)sqrt(2.0f/(float)SubBandNum);
for( i=1; i<= CepstrumNum; i++ )
{
t = (float)i*pi_factor;
for(j=1; j<=SubBandNum; j++)
cosTab[i*(SubBandNum+1)+j] = (float)cos(t*(j-0.5f))*mfnorm;
}
a =(float)( asin(1.0)*4/(FrameLen-1) );
for(i=0;i<FrameLen;i++)
hamWin[i] = 0.54f - 0.46f * (float)cos(a*i);
for(i=1;i<=CepstrumNum;i++)
cepWin[i-1] = (float)i * (float)exp(-(float)i*2.0/(float)CepstrumNum);
}
CMFCC::CMFCC(float SampleF,long DefFFTLen,long DefFrameLen,long DefSubBandNum, long DefCepstrumNum):CFFTanalyser(DefFFTLen)
{
long M, SubBandNo, i, j;
float ms,pi_factor, mfnorm, a, melk, t;
FrameLen = DefFrameLen;
SubBandNum = DefSubBandNum;
CepstrumNum = DefCepstrumNum;
FFT_LEN = GetFFTAnalyseLen();
// std::cout <<"FFT_LEN = "<<FFT_LEN<<std::endl;
cosTab = new float[(SubBandNum+1)*(SubBandNum+1)]; //DCT变换系数
hamWin = new float[FrameLen]; //hamming 窗系数
cepWin = new float[SubBandNum+1];
MelBandBoundary = new float[SubBandNum+2];
SubBandWeight = new float[FFT_LEN/2]; //weighting coefficients for
//energy of each FFT frequence component
SubBandIndex = new long[FFT_LEN/2]; //mapping of the frequence to sub-band No.
SubBandEnergy = new float[SubBandNum+2]; //for accumulating the corresponding
FFTFrame = new float[FFT_LEN];
Cepstrum = new float[SubBandNum+1];
fres = SampleF/(FFT_LEN*700.0f);
//caculating the mel scale sub-band boundary
//由于子带0的系数（直流分量）不参与MFCC特征计算，所以子带个数要多一个
M = SubBandNum+1;
ms = Mel(FFT_LEN/2); //计算1/2采样率所对应的MEL刻度
//Note that the sub-band 0 is not used for cepstrum caculating
for ( SubBandNo = 0; SubBandNo <= M; SubBandNo++ )
{ //计算每个子带的起始MEL刻度
MelBandBoundary[SubBandNo] = ( (float)SubBandNo/(float)M )*ms;
}
//mapping the FFT frequence component into the corresponding sub-band
for ( i=0,SubBandNo=1; i< FFT_LEN/2; i++)
{
melk = Mel(i);
while( MelBandBoundary[SubBandNo] < melk ) SubBandNo++;
SubBandIndex[i] = SubBandNo-1;
}
//caculating the weighting coefficients for each FFT frequence components
for(i=0; i< FFT_LEN/2; i++)
{ //以子带的起始MEL频率为中心，计算三角窗加权系数
SubBandNo = SubBandIndex[i];
SubBandWeight[i] = (MelBandBoundary[SubBandNo+1]-Mel(i))/(MelBandBoundary[SubBandNo+1]-MelBandBoundary[SubBandNo]);
}
pi_factor = (float)( asin(1.0)*2.0/(float)SubBandNum );
mfnorm = (float)sqrt(2.0f/(float)SubBandNum);
for( i=1; i<= CepstrumNum; i++ )
{
t = (float)i*pi_factor;
for(j=1; j<=SubBandNum; j++)
cosTab[i*(SubBandNum+1)+j] = (float)cos(t*(j-0.5f))*mfnorm;
}
a =(float)( asin(1.0)*4/(FrameLen-1) );
for(i=0;i<FrameLen;i++)
hamWin[i] = 0.54f - 0.46f * (float)cos(a*i);
for(i=1;i<=CepstrumNum;i++)
cepWin[i-1] = (float)i * (float)exp(-(float)i*2.0/(float)CepstrumNum);
}
/**
* Build filterbank information and generate tables for MFCC comptutation.
*
* @param w [i/o] MFCC calculation work area
* @param para [in] configuration parameters
*
* @return the generated filterbank information.
*/
boolean
InitFBank(MFCCWork *w, Value *para)
{
float mlo, mhi, ms, melk;
int k, chan, maxChan, nv2;
/* Calculate FFT size */
w->fb.fftN = 2; w->fb.n = 1;
while(para->framesize > w->fb.fftN){
w->fb.fftN *= 2; w->fb.n++;
}
nv2 = w->fb.fftN / 2;
w->fb.fres = 1.0E7 / (para->smp_period * w->fb.fftN * 700.0);
maxChan = para->fbank_num + 1;
w->fb.klo = 2; w->fb.khi = nv2;
mlo = 0; mhi = Mel(nv2 + 1, w->fb.fres);
/* lo pass filter */
if (para->lopass >= 0) {
mlo = 1127*log(1+(float)para->lopass/700.0);
w->fb.klo = ((float)para->lopass * para->smp_period * 1.0e-7 * w->fb.fftN) + 2.5;
if (w->fb.klo<2) w->fb.klo = 2;
}
/* hi pass filter */
if (para->hipass >= 0) {
mhi = 1127*log(1+(float)para->hipass/700.0);
w->fb.khi = ((float)para->hipass * para->smp_period * 1.0e-7 * w->fb.fftN) + 0.5;
if (w->fb.khi>nv2) w->fb.khi = nv2;
}
/* Create vector of fbank centre frequencies */
w->fb.cf = (float *)mymalloc((maxChan + 1) * sizeof(float));
ms = mhi - mlo;
for (chan = 1; chan <= maxChan; chan++)
w->fb.cf[chan] = ((float)chan / maxChan)*ms + mlo;
if (para->vtln_alpha != 1.0) {
/* Modify fbank center frequencies for VTLN */
if (VTLN_recreate_fbank_cf(w->fb.cf, para, mlo, mhi, maxChan) == FALSE) {
return FALSE;
}
}
/* Create loChan map, loChan[fftindex] -> lower channel index */
w->fb.loChan = (short *)mymalloc((nv2 + 1) * sizeof(short));
for(k = 1, chan = 1; k <= nv2; k++){
if (k < w->fb.klo || k > w->fb.khi) w->fb.loChan[k] = -1;
else {
melk = Mel(k, w->fb.fres);
while (w->fb.cf[chan] < melk && chan <= maxChan) ++chan;
w->fb.loChan[k] = chan - 1;
}
}
/* Create vector of lower channel weights */
w->fb.loWt = (float *)mymalloc((nv2 + 1) * sizeof(float));
for(k = 1; k <= nv2; k++) {
chan = w->fb.loChan[k];
if (k < w->fb.klo || k > w->fb.khi) w->fb.loWt[k] = 0.0;
else {
if (chan > 0)
w->fb.loWt[k] = (w->fb.cf[chan + 1] - Mel(k, w->fb.fres)) / (w->fb.cf[chan + 1] - w->fb.cf[chan]);
else
w->fb.loWt[k] = (w->fb.cf[1] - Mel(k, w->fb.fres)) / (w->fb.cf[1] - mlo);
}
}
/* Create workspace for fft */
w->fb.Re = (float *)mymalloc((w->fb.fftN + 1) * sizeof(float));
w->fb.Im = (float *)mymalloc((w->fb.fftN + 1) * sizeof(float));
w->sqrt2var = sqrt(2.0 / para->fbank_num);
return TRUE;
}
/* EXPORT->InitFBank: Initialise an FBankInfo record */
FBankInfo InitFBank(MemHeap *x, int frameSize, long sampPeriod, int numChans,
float lopass, float hipass, Boolean usePower, Boolean takeLogs,
Boolean doubleFFT,
float alpha, float warpLowCut, float warpUpCut)
{
FBankInfo fb;
float mlo,mhi,ms,melk;
int k,chan,maxChan,Nby2;
/* Save sizes to cross-check subsequent usage */
fb.frameSize = frameSize; fb.numChans = numChans;
fb.sampPeriod = sampPeriod;
fb.usePower = usePower; fb.takeLogs = takeLogs;
/* Calculate required FFT size */
fb.fftN = 2;
while (frameSize>fb.fftN) fb.fftN *= 2;
if (doubleFFT)
fb.fftN *= 2;
Nby2 = fb.fftN / 2;
fb.fres = 1.0E7/(sampPeriod * fb.fftN * 700.0);
maxChan = numChans+1;
/* set lo and hi pass cut offs if any */
fb.klo = 2; fb.khi = Nby2; /* apply lo/hi pass filtering */
mlo = 0; mhi = Mel(Nby2+1,fb.fres);
if (lopass>=0.0) {
mlo = 1127*log(1+lopass/700.0);
fb.klo = (int) ((lopass * sampPeriod * 1.0e-7 * fb.fftN) + 2.5);
if (fb.klo<2) fb.klo = 2;
}
if (hipass>=0.0) {
mhi = 1127*log(1+hipass/700.0);
fb.khi = (int) ((hipass * sampPeriod * 1.0e-7 * fb.fftN) + 0.5);
if (fb.khi>Nby2) fb.khi = Nby2;
}
if (trace&T_MEL){
printf("FFT passband %d to %d out of 1 to %d\n",fb.klo,fb.khi,Nby2);
printf("Mel passband %f to %f\n",mlo,mhi);
}
/* Create vector of fbank centre frequencies */
fb.cf = CreateVector(x,maxChan);
ms = mhi - mlo;
for (chan=1; chan <= maxChan; chan++) {
if (alpha == 1.0) {
fb.cf[chan] = ((float)chan/(float)maxChan)*ms + mlo;
}
else {
/* scale assuming scaling starts at lopass */
float minFreq = 700.0 * (exp (mlo / 1127.0) - 1.0 );
float maxFreq = 700.0 * (exp (mhi / 1127.0) - 1.0 );
float cf = ((float)chan / (float) maxChan) * ms + mlo;
cf = 700 * (exp (cf / 1127.0) - 1.0);
fb.cf[chan] = 1127.0 * log (1.0 + WarpFreq (warpLowCut, warpUpCut, cf, minFreq, maxFreq, alpha) / 700.0);
}
}
/* Create loChan map, loChan[fftindex] -> lower channel index */
fb.loChan = CreateShortVec(x,Nby2);
for (k=1,chan=1; k<=Nby2; k++){
melk = Mel(k,fb.fres);
if (k<fb.klo || k>fb.khi) fb.loChan[k]=-1;
else {
while (fb.cf[chan] < melk && chan<=maxChan) ++chan;
fb.loChan[k] = chan-1;
}
}
/* Create vector of lower channel weights */
fb.loWt = CreateVector(x,Nby2);
for (k=1; k<=Nby2; k++) {
chan = fb.loChan[k];
if (k<fb.klo || k>fb.khi) fb.loWt[k]=0.0;
else {
if (chan>0)
fb.loWt[k] = ((fb.cf[chan+1] - Mel(k,fb.fres)) /
(fb.cf[chan+1] - fb.cf[chan]));
else
fb.loWt[k] = (fb.cf[1]-Mel(k,fb.fres))/(fb.cf[1] - mlo);
}
}
/* Create workspace for fft */
fb.x = CreateVector(x,fb.fftN);
return fb;
}