Adjustments in beat detection

source/SoundTouch/BPMDetect.cpp

@@ -117,23 +117,19 @@ BPMDetect::BPMDetect(int numChannels, int aSampleRate)
 
     envelopeAccu = 0;
 
-    // Initialize RMS volume accumulator to RMS level of 3000 (out of 32768) that's
+    // Initialize RMS volume accumulator to RMS level of 1500 (out of 32768) that's
-    // a typical RMS signal level value for song data. This value is then adapted
+    // safe initial RMS signal level value for song data. This value is then adapted
     // to the actual level during processing.
 #ifdef SOUNDTOUCH_INTEGER_SAMPLES
     // integer samples
-    RMSVolumeAccu = (3000 * 3000) / avgnorm;
+    RMSVolumeAccu = (1500 * 1500) / avgnorm;
 #else
     // float samples, scaled to range [-1..+1[
-    RMSVolumeAccu = (0.092f * 0.092f) / avgnorm;
+    RMSVolumeAccu = (0.045f * 0.045f) / avgnorm;
 #endif
 
-    cutCoeff = 1.75;
+    // choose decimation factor so that result is approx. 1000 Hz
-    aboveCutAccu = 0;
+    decimateBy = sampleRate / 1000;
-    totalAccu = 0;
-
-    // choose decimation factor so that result is approx. 500 Hz
-    decimateBy = sampleRate / 500;
     assert(decimateBy > 0);
     assert(INPUT_BLOCK_SAMPLES < decimateBy * DECIMATED_BLOCK_SAMPLES);
 
@@ -270,31 +266,11 @@ void BPMDetect::calcEnvelope(SAMPLETYPE *samples, int numsamples)
 
         // cut amplitudes that are below cutoff ~2 times RMS volume
         // (we're interested in peak values, not the silent moments)
-        val -= cutCoeff * sqrt(RMSVolumeAccu * avgnorm);
+        if (val < 0.5 * sqrt(RMSVolumeAccu * avgnorm))
-        if (val > 0)
-        {
-            aboveCutAccu += 1.0;  // sample above threshold
-        }
-        else
         {
             val = 0;
         }
 
-        totalAccu += 1.0;
-
-        // maintain sliding statistic what proportion of 'val' samples is
-        // above cutoff threshold
-        aboveCutAccu *= 0.99931;  // 2 sec time constant
-        totalAccu *= 0.99931;
-
-        if (totalAccu > 500)
-        {
-            // after initial settling, auto-adjust cutoff level so that ~8% of 
-            // values are above the threshold
-            double d = (aboveCutAccu / totalAccu) - 0.08;
-            cutCoeff += 0.001 * d;
-        }
-
         // smooth amplitude envelope
         envelopeAccu *= decay;
         envelopeAccu += val;
@@ -306,12 +282,6 @@ void BPMDetect::calcEnvelope(SAMPLETYPE *samples, int numsamples)
 #endif // SOUNDTOUCH_INTEGER_SAMPLES
         samples[i] = (SAMPLETYPE)out;
     }
-
-    // check that cutoff doesn't get too small - it can be just silent sequence!
-    if (cutCoeff < 1.5) 
-    {
-        cutCoeff = 1.5;
-    }
 }
 
 
@@ -355,6 +325,26 @@ void BPMDetect::inputSamples(const SAMPLETYPE *samples, int numSamples)
 
 
 
+void BPMDetect::removeBias()
+{
+    int i;
+    float minval = 1e12f;   // arbitrary large number
+
+    for (i = windowStart; i < windowLen; i ++)
+    {
+        if (xcorr[i] < minval)
+        {
+            minval = xcorr[i];
+        }
+    }
+
+    for (i = windowStart; i < windowLen; i ++)
+    {
+        xcorr[i] -= minval;
+    }
+}
+
+
 float BPMDetect::getBpm()
 {
     double peakPos;
@@ -366,6 +356,9 @@ float BPMDetect::getBpm()
     // save bpm debug analysis data if debug data enabled
     _SaveDebugData(xcorr, windowStart, windowLen, coeff);
 
+    // remove bias from xcorr data
+    removeBias();
+
     // find peak position
     peakPos = peakFinder.detectPeak(xcorr, windowStart, windowLen);
 

source/SoundTouch/PeakFinder.cpp

@@ -103,7 +103,7 @@ int PeakFinder::findGround(const float *data, int peakpos, int direction) const
 
     pos = peakpos;
 
-    while ((pos > minPos) && (pos < maxPos))
+    while ((pos > minPos+1) && (pos < maxPos-1))
     {
         int prevpos;
 
@@ -132,7 +132,7 @@ int PeakFinder::findGround(const float *data, int peakpos, int direction) const
         {
             // going uphill, increase climbing counter
             climb_count ++;
-            if (climb_count >= 10) break;    // we've been climbing too long => it's next uphill => quit
+            if (climb_count > 5) break;    // we've been climbing too long => it's next uphill => quit
         }
     }
     return lowpos;
@@ -192,12 +192,9 @@ double PeakFinder::getPeakCenter(const float *data, int peakpos) const
     gp1 = findGround(data, peakpos, -1);
     gp2 = findGround(data, peakpos, 1);
 
-    groundLevel = max(data[gp1], data[gp2]);
+    groundLevel = 0.5 * (data[gp1] + data[gp2]);
     peakLevel = data[peakpos];
 
-    if (groundLevel < 1e-9) return 0;                // ground level too small => detection failed
-    if ((peakLevel / groundLevel) < 1.3) return 0;   // peak less than 30% of the ground level => no good peak detected
-
     // calculate 70%-level of the peak
     cutLevel = 0.70f * peakLevel + 0.30f * groundLevel;
     // find mid-level crossings
@@ -269,7 +266,7 @@ double PeakFinder::detectPeak(const float *data, int aminPos, int amaxPos)
         // now compare to highest detected peak
         i1 = (int)(highPeak + 0.5);
         i2 = (int)(peaktmp + 0.5);
-        if (data[i2] >= 0.5*data[i1])
+        if (data[i2] >= 0.4*data[i1])
         {
             // The harmonic is at least half as high primary peak,
             // thus use the harmonic peak instead