static VALUE t_to_mel(VALUE self, VALUE m) {
return rb_float_new(mel(NUM2DBL(m)));
}
void
whs_extractor_process (WhsExtractor *self, const gfloat *in, WhsFeatureVector *ret)
{
gdouble tmp;
// Calculate MFCC
// http://de.wikipedia.org/wiki/MFCC
// Copy to our temporary array and apply hamming window
gdouble *freqdata = self->priv->fft.freqdata;
for (gint i = 0; i < self->frame_length; i++) {
freqdata[i] = in[i] * self->priv->cos[i];
}
// Take FFT
rdft (self->frame_length, 1, freqdata, self->priv->fft.ip, self->priv->fft.w);
// Store magnitude spectrum in freqdata[0...n/2]
for (guint i = 0; i < self->frame_length; i += 2) {
gdouble cur;
if (i == 0) {
freqdata[0] = freqdata[0] * freqdata[0];
freqdata[1] = freqdata[1] * freqdata[1];
} else {
cur = freqdata[i / 2 + 1] = freqdata[i] * freqdata[i] + freqdata[i+1] * freqdata[i+1];
}
}
// Move freqdata[1] to the end, it's for the nyquist frequency!
tmp = freqdata[1];
g_memmove (&freqdata[1], &freqdata[2], sizeof (gdouble) * (self->frame_length / 2 - 1));
freqdata[self->frame_length / 2] = tmp;
// Take logarithms
for (gint i = 0; i < self->frame_length / 2 + 1; i++) {
if (freqdata[i] != 0.0)
freqdata[i] = CLAMP (log10 (sqrt (freqdata[i]/(self->frame_length*self->frame_length))), -500.0, G_MAXDOUBLE);
else
freqdata[i] = -500.0;
}
// Convert to mel spectrum
gdouble bins[32];
#if 0
{
gint bin = 0;
gint step = mel (self->sample_rate / 2) / 32; // 32 bins
for (bin = 0; bin < 32; bin++) {
gint i;
bins[bin] = 0.0;
// Fill bin from 'start' to 'stop' with triangular weighting
gdouble start = (bin > 0) ? bin * step - 0.5 * step : 0;
gdouble stop = (bin < 31) ? (bin + 1) * step + 0.5 * step : (bin + 1) * step;
for (i = 0; i < self->frame_length / 2 + 1; i++) {
gdouble f = (((gfloat) i) / ((gfloat) (self->frame_length))) * self->sample_rate;
bins[bin] += triangle (start, stop, mel (f)) * freqdata[i];
}
}
}
#else
{
#if 1
gint i, j = 0, bin = 0;
gint start_m = (self->min_freq > 0) ? mel (CLAMP (self->min_freq - (self->sample_rate / self->frame_length), 0, self->sample_rate / 2)) : 0;
gint stop_m = (self->max_freq > 0) ? mel (CLAMP (self->max_freq + (self->sample_rate / self->frame_length), 0, self->sample_rate / 2)) : mel (self->sample_rate / 2);
gint step = (stop_m - start_m) / 32; // 32 bins
#else
gint i, j = 0, bin = 0;
gint start_m = 0;
gint step = mel (self->sample_rate / 2) / 32; // 32 bins
#endif
for (bin = i = 0; bin < 32; bin++, i += j) {
bins[bin] = 0.0;
// Fill bin
for (j = 0; (i + j) <= self->frame_length / 2 && mel (((i + j) * (self->sample_rate / 2)) / (self->frame_length / 2)) <= start_m + step * (bin + 1); j++) {
bins[bin] += freqdata[i+j];
}
// Normalize bin
if (j != 0)
bins[bin] /= j;
}
}
#endif
// Calculate DCT
ddct (32, -1, bins, self->priv->dct.ip, self->priv->dct.w);
for (gint i = 0; i < 32; i++)
ret->mfcc[i] = bins[i];
}
