Add drum detection for channel A in live capture
- DrumDetector.cs: 3-band onset detection (kick/snare/brass) with adaptive thresholds, per-sub-tick timing within 100ms frames - Kick: low-band onset, 150ms deep thump - Snare: broadband (mid+high) onset, 50ms mid punch - Brass: high-only onset, 10ms buzzy - Priority: kick > snare > brass when multiple hit same sub-tick - Intensity = max (100) on hit, 0 on silence — user limiter scales down - MusicAnalyzer: add BuildMelodyFrame for chB only (pitch+energy) - LiveCapture: chA now from DrumDetector, chB unchanged (melody) - Sub-tick mapping: FFT window position → 0-3 within each 100ms tick
This commit is contained in:
+146
@@ -0,0 +1,146 @@
|
||||
namespace Substation;
|
||||
|
||||
public class DrumDetector
|
||||
{
|
||||
const int HistorySize = 50;
|
||||
|
||||
readonly double[] _lowFluxHistory = new double[HistorySize];
|
||||
readonly double[] _midFluxHistory = new double[HistorySize];
|
||||
readonly double[] _highFluxHistory = new double[HistorySize];
|
||||
int _historyIndex;
|
||||
|
||||
double[]? _prevLowMag;
|
||||
double[]? _prevMidMag;
|
||||
double[]? _prevHighMag;
|
||||
|
||||
readonly bool[] _subKick = new bool[4];
|
||||
readonly bool[] _subSnare = new bool[4];
|
||||
readonly bool[] _subBrass = new bool[4];
|
||||
|
||||
static readonly byte[] FreqKick = Freq.Compress4(new[] { 150, 150, 150, 150 });
|
||||
static readonly byte[] FreqSnare = Freq.Compress4(new[] { 50, 50, 50, 50 });
|
||||
static readonly byte[] FreqBrass = Freq.Compress4(new[] { 10, 10, 10, 10 });
|
||||
static readonly byte[] FreqSilent = Freq.Compress4(new[] { 10, 10, 10, 10 });
|
||||
|
||||
public void ProcessWindow(double[] magnitude, int sampleRate, int windowSize, int subTickIndex)
|
||||
{
|
||||
if (subTickIndex < 0 || subTickIndex > 3) return;
|
||||
|
||||
double binWidth = (double)sampleRate / windowSize;
|
||||
int lowLo = (int)(20 / binWidth), lowHi = (int)(150 / binWidth);
|
||||
int midLo = (int)(200 / binWidth), midHi = (int)(500 / binWidth);
|
||||
int highLo = (int)(5000 / binWidth), highHi = (int)(15000 / binWidth);
|
||||
|
||||
var lowMag = ExtractBand(magnitude, lowLo, lowHi);
|
||||
var midMag = ExtractBand(magnitude, midLo, midHi);
|
||||
var highMag = ExtractBand(magnitude, highLo, highHi);
|
||||
|
||||
double lowFlux = ComputeFlux(lowMag, _prevLowMag);
|
||||
double midFlux = ComputeFlux(midMag, _prevMidMag);
|
||||
double highFlux = ComputeFlux(highMag, _prevHighMag);
|
||||
|
||||
_prevLowMag = lowMag;
|
||||
_prevMidMag = midMag;
|
||||
_prevHighMag = highMag;
|
||||
|
||||
// Update history and compute adaptive thresholds
|
||||
_lowFluxHistory[_historyIndex] = lowFlux;
|
||||
_midFluxHistory[_historyIndex] = midFlux;
|
||||
_highFluxHistory[_historyIndex] = highFlux;
|
||||
_historyIndex = (_historyIndex + 1) % HistorySize;
|
||||
|
||||
double lowThresh = AdaptiveThreshold(_lowFluxHistory);
|
||||
double midThresh = AdaptiveThreshold(_midFluxHistory);
|
||||
double highThresh = AdaptiveThreshold(_highFluxHistory);
|
||||
|
||||
// Detect onsets
|
||||
bool kickOnset = lowFlux > lowThresh && lowFlux > 0.001;
|
||||
bool snareOnset = midFlux > midThresh && highFlux > highThresh && midFlux > 0.001;
|
||||
bool brassOnset = highFlux > highThresh && !snareOnset && highFlux > 0.001;
|
||||
|
||||
if (kickOnset) _subKick[subTickIndex] = true;
|
||||
if (snareOnset) _subSnare[subTickIndex] = true;
|
||||
if (brassOnset) _subBrass[subTickIndex] = true;
|
||||
}
|
||||
|
||||
public WaveFrame BuildFrame()
|
||||
{
|
||||
var intensity = new byte[4];
|
||||
var freqBytes = new byte[16]; // 4 sub-ticks × 4 bytes (but we use per-sub-tick freq)
|
||||
|
||||
for (int i = 0; i < 4; i++)
|
||||
{
|
||||
byte freq;
|
||||
if (_subKick[i])
|
||||
{
|
||||
intensity[i] = 100;
|
||||
freq = Freq.Compress(150);
|
||||
}
|
||||
else if (_subSnare[i])
|
||||
{
|
||||
intensity[i] = 100;
|
||||
freq = Freq.Compress(50);
|
||||
}
|
||||
else if (_subBrass[i])
|
||||
{
|
||||
intensity[i] = 100;
|
||||
freq = Freq.Compress(10);
|
||||
}
|
||||
else
|
||||
{
|
||||
intensity[i] = 0;
|
||||
freq = 10;
|
||||
}
|
||||
|
||||
freqBytes[i] = freq;
|
||||
}
|
||||
|
||||
// Reset for next tick
|
||||
Array.Clear(_subKick, 0, 4);
|
||||
Array.Clear(_subSnare, 0, 4);
|
||||
Array.Clear(_subBrass, 0, 4);
|
||||
|
||||
return new WaveFrame(freqBytes, intensity);
|
||||
}
|
||||
|
||||
static double[] ExtractBand(double[] magnitude, int loBin, int hiBin)
|
||||
{
|
||||
if (hiBin <= loBin || hiBin >= magnitude.Length)
|
||||
return Array.Empty<double>();
|
||||
|
||||
var band = new double[hiBin - loBin + 1];
|
||||
Array.Copy(magnitude, loBin, band, 0, band.Length);
|
||||
return band;
|
||||
}
|
||||
|
||||
static double ComputeFlux(double[] current, double[]? previous)
|
||||
{
|
||||
if (previous == null || current.Length != previous.Length) return 0;
|
||||
|
||||
double flux = 0;
|
||||
for (int i = 0; i < current.Length; i++)
|
||||
{
|
||||
double diff = current[i] - previous[i];
|
||||
if (diff > 0) flux += diff;
|
||||
}
|
||||
return flux;
|
||||
}
|
||||
|
||||
static double AdaptiveThreshold(double[] history)
|
||||
{
|
||||
double sum = 0, sumSq = 0;
|
||||
int count = 0;
|
||||
foreach (var v in history)
|
||||
{
|
||||
if (v <= 0) continue;
|
||||
sum += v;
|
||||
sumSq += v * v;
|
||||
count++;
|
||||
}
|
||||
if (count == 0) return double.MaxValue;
|
||||
double mean = sum / count;
|
||||
double variance = sumSq / count - mean * mean;
|
||||
double stddev = variance > 0 ? Math.Sqrt(variance) : 0;
|
||||
return mean + 2 * stddev;
|
||||
}
|
||||
}
|
||||
+9
-2
@@ -22,6 +22,7 @@ public class LiveCapture : IDisposable
|
||||
int _sampleRate;
|
||||
int _fftsPerTick;
|
||||
int _fftIndexInTick;
|
||||
readonly DrumDetector _drums = new();
|
||||
|
||||
public event Action? Stopped;
|
||||
|
||||
@@ -139,7 +140,12 @@ public class LiveCapture : IDisposable
|
||||
_liveMaxFlux = Math.Max(rhythmFlux, _liveMaxFlux * 0.999);
|
||||
_liveMaxMelodyEnergy = Math.Max(melodyEnergy, _liveMaxMelodyEnergy * 0.999);
|
||||
|
||||
tf.RhythmFlux = Math.Max(tf.RhythmFlux, rhythmFlux);
|
||||
// Drum detection: map this FFT window to a sub-tick (0-3)
|
||||
int subTick = _fftsPerTick > 0 ? _fftIndexInTick * 4 / _fftsPerTick : 0;
|
||||
if (subTick > 3) subTick = 3;
|
||||
_drums.ProcessWindow(mag, _sampleRate, MusicAnalyzer.WindowSize, subTick);
|
||||
|
||||
// Melody accumulation for chB
|
||||
tf.MelodyEnergy += melodyEnergy;
|
||||
tf.MelodyFreqSamples.Add(melodyFreq);
|
||||
tf.MelodyCount++;
|
||||
@@ -147,7 +153,8 @@ public class LiveCapture : IDisposable
|
||||
|
||||
if (_fftIndexInTick >= _fftsPerTick)
|
||||
{
|
||||
var (frameA, frameB) = MusicAnalyzer.BuildWaveFrame(tf, _liveMaxFlux, _liveMaxMelodyEnergy);
|
||||
var frameA = _drums.BuildFrame();
|
||||
var frameB = MusicAnalyzer.BuildMelodyFrame(tf, _liveMaxMelodyEnergy);
|
||||
_state.EnqueueStream('A', new[] { frameA });
|
||||
_state.EnqueueStream('B', new[] { frameB });
|
||||
|
||||
|
||||
@@ -79,6 +79,33 @@ public static class MusicAnalyzer
|
||||
public readonly List<double> MelodyFreqSamples = new();
|
||||
}
|
||||
|
||||
public static WaveFrame BuildMelodyFrame(TickFeature tf, double maxMelodyEnergy)
|
||||
{
|
||||
double avgEnergy = tf.MelodyCount > 0 ? tf.MelodyEnergy / tf.MelodyCount : 0;
|
||||
double normalizedEnergy = maxMelodyEnergy > 0 ? avgEnergy / maxMelodyEnergy : 0;
|
||||
int melodyIntensity = (int)Math.Round(normalizedEnergy * 80);
|
||||
melodyIntensity = Math.Clamp(melodyIntensity, 0, 100);
|
||||
|
||||
double weightedFreq = 0;
|
||||
double totalWeight = 0;
|
||||
foreach (var f in tf.MelodyFreqSamples)
|
||||
{
|
||||
weightedFreq += f * f;
|
||||
totalWeight += f;
|
||||
}
|
||||
double avgMelodyHz = totalWeight > 0 ? weightedFreq / totalWeight : 500;
|
||||
int estimsMs = MapPitchToPeriod(avgMelodyHz);
|
||||
|
||||
var intB = new[]
|
||||
{
|
||||
(byte)melodyIntensity, (byte)melodyIntensity,
|
||||
(byte)melodyIntensity, (byte)melodyIntensity
|
||||
};
|
||||
var freqB = Freq.Compress4(new[] { estimsMs, estimsMs, estimsMs, estimsMs });
|
||||
|
||||
return new WaveFrame(freqB, intB);
|
||||
}
|
||||
|
||||
public static (WaveFrame chA, WaveFrame chB) BuildWaveFrame(TickFeature tf, double maxFlux, double maxMelodyEnergy)
|
||||
{
|
||||
const int rhythmFreqMs = 150;
|
||||
|
||||
Reference in New Issue
Block a user