static int _ms_audio_diff_one_chunk(int16_t *s1, int16_t *s2, int nsamples, int max_shift_samples, int nchannels, double *ret, int64_t *s1_energy, ProgressContext *pctx){
int xcorr_size;
int max_index_r;
int max_index_l;
int max_pos;
ProgressContext local_pctx;
int64_t er, el;
xcorr_size=max_shift_samples*2;
if (nchannels == 2){
float *xcorr_r = ms_new0(float, xcorr_size);
float *xcorr_l = ms_new0(float, xcorr_size);
double max = 0;
double max_r, max_l;
int i;
progress_context_push(pctx, &local_pctx, 0.5);
max_index_r = compute_cross_correlation(s1, nsamples, s2, xcorr_r, xcorr_size, &local_pctx, 2, &er);
max_r = xcorr_r[max_index_r];
progress_context_pop(pctx, &local_pctx);
progress_context_push(pctx, &local_pctx, 0.5);
max_index_l = compute_cross_correlation(s1 + 1, nsamples, s2 + 1, xcorr_l, xcorr_size, &local_pctx, 2, &el);
max_l = xcorr_l[max_index_l];
progress_context_pop(pctx, &local_pctx);
max_pos = 0;
/*sum the square of r and l xcorr signals to determine the global maximum*/
for (i = 0; i < max_shift_samples; ++i){
xcorr_r[i] = xcorr_r[i]*xcorr_r[i] + xcorr_l[i]*xcorr_l[i];
if (xcorr_r[i] > max){
max = xcorr_r[i];
max_pos = i;
}
}
max = sqrt(max/2);
ms_message("chunk - max stereo cross-correlation obtained at position [%i,%i], similarity factor=%g,%g",
max_index_r-max_shift_samples,max_index_l-max_shift_samples,max_r, max_l);
max_pos = max_pos - max_shift_samples;
ms_message("chunk - max stereo overall cross-correlation obtained at position [%i], similarity factor=[%g]", max_pos, max);
*ret = max;
if (s1_energy) *s1_energy = (er + el)/2;
ms_free(xcorr_r);
ms_free(xcorr_l);
}else{
int ms_audio_diff(const char *file1, const char *file2, double *ret, int max_shift_percent, MSAudioDiffProgressNotify func, void *user_data){
FileInfo *fi1,*fi2;
int64_t *xcorr;
int xcorr_size;
int max_shift_samples;
int max_index_r;
int max_index_l;
double max_r, max_l;
*ret=0;
fi1=file_info_new(file1);
if (fi1==NULL) return 0;
fi2=file_info_new(file2);
if (fi2==NULL){
file_info_destroy(fi1);
return -1;
}
if (fi1->rate!=fi2->rate){
ms_error("Comparing files of different sampling rates is not supported (%d vs %d)", fi1->rate, fi2->rate);
return -1;
}
if (fi1->nchannels!=fi2->nchannels){
ms_error("Comparing files with different number of channels is not supported (%d vs %d)", fi1->nchannels, fi2->nchannels);
return -1;
}
file_info_compute_energy(fi1);
file_info_compute_energy(fi2);
if (fi1->energy_r==0 || fi2->energy_r==0){
/*avoid division by zero*/
ms_error("One of the two files is pure silence.");
return -1;
}
max_shift_samples = MIN(fi1->nsamples, fi2->nsamples) * MIN(MAX(1, max_shift_percent), 100) / 100;
xcorr_size=max_shift_samples*2;
xcorr=ms_new0(int64_t,xcorr_size);
if (fi1->nchannels == 2){
max_index_r=compute_cross_correlation_interleaved(fi1->buffer,fi1->nsamples,fi2->buffer,fi2->nsamples,xcorr,xcorr_size, func, user_data, 0, max_shift_samples);
max_r=xcorr[max_index_r];
ms_message("max_r=%g", (double)max_r);
max_r/=sqrt((double)fi1->energy_r*(double)fi2->energy_r);
max_index_l=compute_cross_correlation_interleaved(fi1->buffer,fi1->nsamples,fi2->buffer,fi2->nsamples,xcorr,xcorr_size, func, user_data, 1, max_shift_samples);
max_l=xcorr[max_index_l];
ms_message("max_l=%g", (double)max_l);
max_l/=sqrt((double)fi1->energy_l*(double)fi2->energy_l);
ms_message("Max stereo cross-correlation obtained at position [%i,%i], similarity factor=%g,%g",
max_index_r-max_shift_samples,max_index_l-max_shift_samples,max_r, max_l);
*ret = 0.5 * (fabs(max_r) + fabs(max_l)) * (1 - (double)abs(max_index_r-max_index_l)/(double)xcorr_size);
}else{
max_index_r=compute_cross_correlation(fi1->buffer,fi1->nsamples,fi2->buffer,fi2->nsamples,xcorr,xcorr_size, func, user_data, max_shift_samples);
max_r=xcorr[max_index_r];
max_r/=(sqrt(fi1->energy_r)*sqrt(fi2->energy_r));
*ret=max_r;
ms_message("Max cross-correlation obtained at position [%i], similarity factor=%g",max_index_r-max_shift_samples,*ret);
}
ms_free(xcorr);
file_info_destroy(fi1);
file_info_destroy(fi2);
return 0;
}
