Reduce the number of fragments, to get smoother output.

[pitch] / pitch.cpp

diff --git a/pitch.cpp b/pitch.cpp
index be540560e1f309488f95f1141d6f3dc78eb48c25..4e0e3eba09fda41c971c9289ebb831662a6faa58 100644 (file)
--- a/pitch.cpp
+++ b/pitch.cpp
@@ -27,7 +27,9 @@ void read_chunk(int fd, double *in, unsigned num_samples);
 void apply_window(double *in, double *out, unsigned num_samples);
 std::pair<double, double> find_peak(double *in, unsigned num_samples);
 void find_peak_magnitudes(std::complex<double> *in, double *out, unsigned num_samples);
+std::pair<double, double> adjust_for_overtones(std::pair<double, double> base, double *in, unsigned num_samples);
 double bin_to_freq(double bin, unsigned num_samples);
+double freq_to_bin(double freq, unsigned num_samples);
 std::string freq_to_tonename(double freq);
 std::pair<double, double> interpolate_peak(double ym1, double y0, double y1);
 void print_spectrogram(double freq, double amp);
@@ -58,6 +60,7 @@ int main()
                fftw_execute(p);
                find_peak_magnitudes(out, bins, FFT_LENGTH);
                std::pair<double, double> peak = find_peak(bins, FFT_LENGTH);
+               peak = adjust_for_overtones(peak, bins, FFT_LENGTH);
 
                if (peak.first < 50.0 || peak.second - log10(FFT_LENGTH) < 0.0) {
 #if TUNING == WELL_TEMPERED_GUITAR
@@ -87,9 +90,12 @@ int get_dsp_fd()
        int chan = 1;
        ioctl(fd, SOUND_PCM_WRITE_CHANNELS, &chan);
        
-       int rate = 22050;
+       int rate = SAMPLE_RATE;
        ioctl(fd, SOUND_PCM_WRITE_RATE, &rate);
 
+       int fragments = 0x00020002;
+        ioctl(fd, SNDCTL_DSP_SETFRAGMENT, &fragments);
+       
        ioctl(3, SNDCTL_DSP_SYNC, 0);
        
        return fd;
@@ -234,10 +240,66 @@ std::pair<double, double> find_peak(double *in, unsigned num_samples)
        return std::make_pair(bin_to_freq(double(best_bin) + peak.first, num_samples), peak.second);
 }
 
+// it's perhaps not ideal to _first_ find the peak and _then_ the harmonics --
+// ideally, one should find the set of all peaks and then determine the likely
+// base from that... something for later. :-)
+std::pair<double, double> adjust_for_overtones(std::pair<double, double> base, double *in, unsigned num_samples)
+{
+       double mu = base.first, var = 1.0 / (base.second * base.second);
+       printf("mu=%f, var=%f\n", mu, var);
+
+       for (unsigned i = 2; i < 10; ++i) {
+               unsigned middle = unsigned(floor(freq_to_bin(base.first, num_samples) * i + 0.5));
+               unsigned lower = middle - (i+1)/2, upper = middle + (i+1)/2;
+
+               if (upper >= num_samples)
+                       upper = num_samples - 2;
+
+               // printf("Searching in [%u,%u] = %f..%f\n", lower, upper, bin_to_freq(lower, num_samples), bin_to_freq(upper, num_samples));
+
+               // search for a peak in this interval
+               double best_harmonics_freq = -1.0;
+               double best_harmonics_amp = -1.0;
+               for (unsigned j = lower; j <= upper; ++j) {
+                       if (in[j] > in[j-1] && in[j] > in[j+1]) {
+                               std::pair<double, double> peak =
+                                       interpolate_peak(in[j - 1],
+                                               in[j],
+                                               in[j + 1]);
+                               
+                               if (peak.second > best_harmonics_amp) {
+                                       best_harmonics_freq = bin_to_freq(j + peak.first, num_samples);
+                                       best_harmonics_amp = peak.second;
+                               }
+                       }
+               }
+
+               if (best_harmonics_amp <= 0.0)
+                       continue;
+
+               //printf("Found overtone %u at %.2f (amp=%5.2fdB)\n", i, best_harmonics_freq,
+               //       best_harmonics_amp);
+
+               double this_mu = best_harmonics_freq / double(i);
+               double this_var = 1.0 / (best_harmonics_amp * best_harmonics_amp);
+
+               double k = var / (var + this_var);
+               mu = (1.0 - k) * mu + k * this_mu;
+               var *= (1.0 - k);
+
+               printf("mu=%f, var=%f\n", mu, var);
+       }
+       return std::make_pair(mu, base.second);
+}
+
 double bin_to_freq(double bin, unsigned num_samples)
 {
        return bin * SAMPLE_RATE / double(num_samples);
 }
+double freq_to_bin(double freq, unsigned num_samples)
+{
+       return freq * double(num_samples) / double(SAMPLE_RATE);
+}
 
 /*
  * Given three bins, find the interpolated real peak based