c@250: #include <stdio.h>
c@250: #include <stdlib.h>
c@250: #include <string.h>
c@250: #include <math.h>
c@250: 
c@250: #include "mfcc.h"
c@250: #include "SBFFT.h"
c@250: #include "windows.h"
c@250: 
c@250: /*
c@250:  * 
c@250:  *  Initialise the MFCC structure and return a pointer to a 
c@250:  *  feature vector 
c@250:  * 
c@250:  */
c@250: 
c@250: extern mfcc_t* init_mfcc(int fftSize, int nceps , int samplingRate, int WANT_C0){
c@250: 
c@250:   int i,j;
c@250:   /* Calculate at startup */
c@250:   double *freqs, *lower, *center, *upper, *triangleHeight, *fftFreqs;
c@250:   
c@250:   /* Allocate space for the structure */
c@250:   mfcc_t* mfcc_p = (mfcc_t*)malloc(sizeof(mfcc_t));
c@250:   
c@250:   mfcc_p->lowestFrequency   = 66.6666666;
c@250:   mfcc_p->linearFilters     = 13;
c@250:   mfcc_p->linearSpacing     = 66.66666666;
c@250:   mfcc_p->logFilters        = 27;
c@250:   mfcc_p->logSpacing        = 1.0711703;
c@250:   
c@250:   /* FFT and analysis window sizes */
c@250:   mfcc_p->fftSize           = fftSize;
c@250: 
c@250:   mfcc_p->totalFilters      = mfcc_p->linearFilters + mfcc_p->logFilters;
c@250:   
c@250:   mfcc_p->samplingRate      = samplingRate;
c@250:   
c@250:   /* The number of cepstral componenents */
c@250:   mfcc_p->nceps             = nceps;
c@250: 
c@250:   /* Set if user want C0 */
c@250:   mfcc_p->WANT_C0           = WANT_C0;
c@250:   
c@250:   /* Allocate space for feature vector */
c@250:   if (mfcc_p->WANT_C0==1) {
c@250:     mfcc_p->ceps              = (double*)calloc(nceps+1,sizeof(double));
c@250:   } else {
c@250:     mfcc_p->ceps              = (double*)calloc(nceps,sizeof(double));
c@250:   }
c@250:  
c@250:   /* Allocate space for local vectors */
c@250:   mfcc_p->mfccDCTMatrix     = (double**)calloc(mfcc_p->nceps+1, sizeof(double*));
c@250:   for (i=0;i<mfcc_p->nceps+1; i++) {
c@250:     mfcc_p->mfccDCTMatrix[i]= (double*)calloc(mfcc_p->totalFilters, sizeof(double)); 
c@250:   }
c@250: 
c@250:   mfcc_p->mfccFilterWeights = (double**)calloc(mfcc_p->totalFilters, sizeof(double*));
c@250:   for (i=0;i<mfcc_p->totalFilters; i++) {
c@250:     mfcc_p->mfccFilterWeights[i] = (double*)calloc(mfcc_p->fftSize, sizeof(double)); 
c@250:   }
c@250: 
c@250:   freqs  = (double*)calloc(mfcc_p->totalFilters+2,sizeof(double));
c@250:   
c@250:   lower  = (double*)calloc(mfcc_p->totalFilters,sizeof(double));
c@250:   center = (double*)calloc(mfcc_p->totalFilters,sizeof(double));
c@250:   upper  = (double*)calloc(mfcc_p->totalFilters,sizeof(double));
c@250: 
c@250:   triangleHeight = (double*)calloc(mfcc_p->totalFilters,sizeof(double));
c@250:   fftFreqs       = (double*)calloc(mfcc_p->fftSize,sizeof(double));
c@250:   
c@250:   for (i=0;i<mfcc_p->linearFilters;i++) {
c@250:     freqs[i] = mfcc_p->lowestFrequency + ((double)i) * mfcc_p->linearSpacing;
c@250:   }
c@250:   
c@250:   for (i=mfcc_p->linearFilters; i<mfcc_p->totalFilters+2; i++) {
c@250:     freqs[i] = freqs[mfcc_p->linearFilters-1] * 
c@250:                pow(mfcc_p->logSpacing, (double)(i-mfcc_p->linearFilters+1));
c@250:   }
c@250:   
c@250:   /* Define lower, center and upper */
c@250:   memcpy(lower,  freqs,mfcc_p->totalFilters*sizeof(double));
c@250:   memcpy(center, &freqs[1],mfcc_p->totalFilters*sizeof(double));
c@250:   memcpy(upper,  &freqs[2],mfcc_p->totalFilters*sizeof(double));
c@250:     
c@250:   for (i=0;i<mfcc_p->totalFilters;i++){
c@250:       triangleHeight[i] = 2./(upper[i]-lower[i]);
c@250:   }
c@250:   
c@250:   for (i=0;i<mfcc_p->fftSize;i++){
c@250:       fftFreqs[i] = ((double) i / ((double) mfcc_p->fftSize ) * 
c@250:                      (double) mfcc_p->samplingRate);
c@250:   }
c@250: 
c@250:   /* Build now the mccFilterWeight matrix */
c@250:   for (i=0;i<mfcc_p->totalFilters;i++){
c@250: 
c@250:     for (j=0;j<mfcc_p->fftSize;j++) {
c@250:       
c@250:         if ((fftFreqs[j] > lower[i]) && (fftFreqs[j] <= center[i])) {
c@250:           
c@250:           mfcc_p->mfccFilterWeights[i][j] = triangleHeight[i] * 
c@250:                                             (fftFreqs[j]-lower[i]) / (center[i]-lower[i]); 
c@250:           
c@250:         }
c@250:         else
c@250:         {
c@250:           
c@250:           mfcc_p->mfccFilterWeights[i][j] = 0.0;
c@250:         
c@250:         }
c@250: 
c@250:         if ((fftFreqs[j]>center[i]) && (fftFreqs[j]<upper[i])) {
c@250: 
c@250:           mfcc_p->mfccFilterWeights[i][j] = mfcc_p->mfccFilterWeights[i][j] + triangleHeight[i] * (upper[i]-fftFreqs[j]) 
c@250:                                             / (upper[i]-center[i]);
c@250:         }
c@250:         else
c@250:         {
c@250: 
c@250:           mfcc_p->mfccFilterWeights[i][j] = mfcc_p->mfccFilterWeights[i][j] + 0.0;
c@250:             
c@250:         }
c@250:     }
c@250: 
c@250:   }
c@250: 
c@250: #ifndef PI    
c@250: #define PI 3.14159265358979323846264338327950288
c@250: #endif 
c@250: 
c@250:   /*
c@250:    * 
c@250:    * We calculate now mfccDCT matrix 
c@250:    * NB: +1 because of the DC component
c@250:    * 
c@250:    */
c@250:   
c@250:   for (i=0; i<nceps+1; i++) {
c@250:     for (j=0; j<mfcc_p->totalFilters; j++) {
c@250:         mfcc_p->mfccDCTMatrix[i][j] = (1./sqrt((double) mfcc_p->totalFilters / 2.))  
c@250:                                       * cos((double) i * ((double) j + 0.5) / (double) mfcc_p->totalFilters * PI);
c@250:     }
c@250:   }
c@250: 
c@250:   for (j=0;j<mfcc_p->totalFilters;j++){
c@250:     mfcc_p->mfccDCTMatrix[0][j]     =  (sqrt(2.)/2.) * mfcc_p->mfccDCTMatrix[0][j];
c@250:   }
c@250:    
c@250:   /* The analysis window */
c@250:   mfcc_p->window = hamming(mfcc_p->fftSize);
c@250: 
c@250:   /* Allocate memory for the FFT */
c@250:   mfcc_p->imagIn      = (double*)calloc(mfcc_p->fftSize,sizeof(double));
c@250:   mfcc_p->realOut     = (double*)calloc(mfcc_p->fftSize,sizeof(double));
c@250:   mfcc_p->imagOut     = (double*)calloc(mfcc_p->fftSize,sizeof(double));
c@250:   
c@250:   free(freqs);
c@250:   free(lower);
c@250:   free(center);
c@250:   free(upper);
c@250:   free(triangleHeight);
c@250:   free(fftFreqs);
c@250: 
c@250:   return mfcc_p;
c@250: 
c@250: }
c@250: 
c@250: /*
c@250:  *
c@250:  *  Free the memory that has been allocated 
c@250:  *
c@250:  */ 
c@250: 
c@250: extern void close_mfcc(mfcc_t* mfcc_p) {
c@250: 
c@250:   int i;
c@250:   
c@250:   /* Free the structure */
c@250:   for (i=0;i<mfcc_p->nceps+1;i++) {
c@250:     free(mfcc_p->mfccDCTMatrix[i]);
c@250:   }
c@250:   free(mfcc_p->mfccDCTMatrix);
c@250: 
c@250:   for (i=0;i<mfcc_p->totalFilters; i++) {
c@250:     free(mfcc_p->mfccFilterWeights[i]);
c@250:   }
c@250:   free(mfcc_p->mfccFilterWeights);
c@250: 
c@250:   /* Free the feature vector */
c@250:   free(mfcc_p->ceps);
c@250: 
c@250:   /* The analysis window */
c@250:   free(mfcc_p->window);
c@250: 
c@250:   /* Free the FFT */
c@250:   free(mfcc_p->imagIn);
c@250:   free(mfcc_p->realOut);
c@250:   free(mfcc_p->imagOut);
c@250: 
c@250:   /* Free the structure itself */
c@250:   free(mfcc_p);
c@250:   mfcc_p = NULL;
c@250: 
c@250: }
c@250: 
c@250: /*
c@250:  * 
c@250:  * Extract the MFCC on the input frame 
c@250:  * 
c@250:  */ 
c@250: 
c@250: 
c@250: // looks like we have to have length = mfcc_p->fftSize ??????
c@250: 
c@250: extern int do_mfcc(mfcc_t* mfcc_p, double* frame, int length){
c@250: 
c@250:   int i,j;
c@250:   
c@250:   double *fftMag;
c@250:   double *earMag;
c@250: 
c@250:   double *inputData;
c@250:   
c@250:   double tmp;
c@250: 
c@250:   earMag    = (double*)calloc(mfcc_p->totalFilters, sizeof(double));
c@250:   inputData = (double*)calloc(mfcc_p->fftSize, sizeof(double)); 
c@250:   
c@250:   /* Zero-pad if needed */
c@250:   memcpy(inputData, frame, length*sizeof(double));
c@250:   
c@250:   /* Calculate the fft on the input frame */
c@250:   fft_process(mfcc_p->fftSize, 0, inputData, mfcc_p->imagIn, mfcc_p->realOut, mfcc_p->imagOut);
c@250: 
c@250:   /* Get the magnitude */
c@250:   fftMag = abs_fft(mfcc_p->realOut, mfcc_p->imagOut, mfcc_p->fftSize);
c@250: 
c@250:   /* Multiply by mfccFilterWeights */
c@250:   for (i=0;i<mfcc_p->totalFilters;i++) {
c@250:     tmp = 0.;
c@250:     for(j=0;j<mfcc_p->fftSize/2; j++) {
c@250:       tmp = tmp + (mfcc_p->mfccFilterWeights[i][j]*fftMag[j]);
c@250:     }
c@250:     if (tmp>0)
c@250: 		earMag[i] = log10(tmp);
c@250: 	else
c@250: 		earMag[i] = 0.0;
c@250:   }
c@250: 
c@250:   /*
c@250:    * 
c@250:    * Calculate now the ceptral coefficients 
c@250:    * with or without the DC component
c@250:    *
c@250:    */
c@250:   
c@250:    if (mfcc_p->WANT_C0==1) {
c@250:      
c@250:     for (i=0;i<mfcc_p->nceps+1;i++) {
c@250:       tmp = 0.;
c@250:       for (j=0;j<mfcc_p->totalFilters;j++){
c@250:         tmp = tmp + mfcc_p->mfccDCTMatrix[i][j]*earMag[j];
c@250:       }
c@250:       /* Send to workspace */
c@250:       mfcc_p->ceps[i] = tmp;
c@250:     }
c@250: 
c@250:    }
c@250:     else 
c@250:    {  
c@250:       for (i=1;i<mfcc_p->nceps+1;i++) {
c@250:         tmp = 0.;
c@250:         for (j=0;j<mfcc_p->totalFilters;j++){
c@250:           tmp = tmp + mfcc_p->mfccDCTMatrix[i][j]*earMag[j];
c@250:         }
c@250:         /* Send to workspace */
c@250:         mfcc_p->ceps[i-1] = tmp;
c@250:       }
c@250:   }
c@250:     
c@250:   free(fftMag);
c@250:   free(earMag);
c@250:   free(inputData);
c@250: 
c@250:   return mfcc_p->nceps;
c@250: 
c@250: }
c@250: 
c@250: 
c@250: 
c@250: