Add live audio capture via WASAPI loopback
- LiveCapture.cs: real-time WASAPI loopback capture with rolling buffer and continuous FFT analysis, enqueues WaveFrames at ~10Hz - MusicAnalyzer.cs: refactored ExtractFeatures/MapPitchToPeriod/BuildWaveFrame as public reusable methods with sampleRate parameter - MainForm: audio device selector (defaults to system default), Live button starts/stops capture, Stop All stops live too - Adaptive normalization (running max with slow decay) for live volume changes
This commit is contained in:
+151
@@ -0,0 +1,151 @@
|
||||
using System.Numerics;
|
||||
using FftSharp;
|
||||
using NAudio.CoreAudioApi;
|
||||
using NAudio.Wave;
|
||||
|
||||
namespace Substation;
|
||||
|
||||
public class LiveCapture : IDisposable
|
||||
{
|
||||
readonly State _state;
|
||||
readonly MMDevice _device;
|
||||
WasapiLoopbackCapture? _capture;
|
||||
Thread? _processThread;
|
||||
volatile bool _running;
|
||||
|
||||
readonly Queue<double> _sampleBuffer = new();
|
||||
readonly object _bufferLock = new();
|
||||
|
||||
double[]? _prevRhythmMag;
|
||||
double _liveMaxFlux;
|
||||
double _liveMaxMelodyEnergy;
|
||||
int _sampleRate;
|
||||
int _fftsPerTick;
|
||||
int _fftIndexInTick;
|
||||
|
||||
public event Action? Stopped;
|
||||
|
||||
public LiveCapture(State state, MMDevice device)
|
||||
{
|
||||
_state = state;
|
||||
_device = device;
|
||||
}
|
||||
|
||||
public void Start()
|
||||
{
|
||||
_running = true;
|
||||
_capture = new WasapiLoopbackCapture(_device);
|
||||
_sampleRate = _capture.WaveFormat.SampleRate;
|
||||
_fftsPerTick = Math.Max(1, (int)Math.Round(MusicAnalyzer.TickDuration * _sampleRate / MusicAnalyzer.HopSize));
|
||||
|
||||
_capture.DataAvailable += OnDataAvailable;
|
||||
_capture.RecordingStopped += OnRecordingStopped;
|
||||
|
||||
_processThread = new Thread(ProcessLoop) { IsBackground = true, Name = "LiveCapture-FFT" };
|
||||
_processThread.Start();
|
||||
|
||||
_capture.StartRecording();
|
||||
Console.WriteLine($"[live] capture started: {_device.FriendlyName} ({_sampleRate}Hz, {_capture.WaveFormat.Channels}ch)");
|
||||
}
|
||||
|
||||
public void Stop()
|
||||
{
|
||||
_running = false;
|
||||
try { _capture?.StopRecording(); } catch { }
|
||||
}
|
||||
|
||||
void OnDataAvailable(object? sender, WaveInEventArgs e)
|
||||
{
|
||||
int channels = _capture!.WaveFormat.Channels;
|
||||
int bytesPerSample = _capture.WaveFormat.BitsPerSample / 8;
|
||||
int frameSize = channels * bytesPerSample;
|
||||
int sampleCount = e.BytesRecorded / frameSize;
|
||||
|
||||
lock (_bufferLock)
|
||||
{
|
||||
for (int i = 0; i < sampleCount; i++)
|
||||
{
|
||||
int offset = i * frameSize;
|
||||
float left = BitConverter.ToSingle(e.Buffer, offset);
|
||||
float right = channels >= 2
|
||||
? BitConverter.ToSingle(e.Buffer, offset + bytesPerSample)
|
||||
: left;
|
||||
_sampleBuffer.Enqueue((left + right) * 0.5);
|
||||
}
|
||||
|
||||
// Cap buffer size to prevent memory growth if processing falls behind
|
||||
while (_sampleBuffer.Count > _sampleRate * 2)
|
||||
_sampleBuffer.Dequeue();
|
||||
}
|
||||
}
|
||||
|
||||
void OnRecordingStopped(object? sender, StoppedEventArgs e)
|
||||
{
|
||||
_running = false;
|
||||
Stopped?.Invoke();
|
||||
Console.WriteLine("[live] capture stopped");
|
||||
}
|
||||
|
||||
void ProcessLoop()
|
||||
{
|
||||
var window = new FftSharp.Windows.Hanning();
|
||||
var buffer = new double[MusicAnalyzer.WindowSize];
|
||||
var tf = new MusicAnalyzer.TickFeature();
|
||||
|
||||
while (_running)
|
||||
{
|
||||
double[]? windowData = null;
|
||||
|
||||
lock (_bufferLock)
|
||||
{
|
||||
if (_sampleBuffer.Count >= MusicAnalyzer.WindowSize)
|
||||
{
|
||||
for (int i = 0; i < MusicAnalyzer.WindowSize; i++)
|
||||
buffer[i] = _sampleBuffer.Dequeue();
|
||||
windowData = buffer;
|
||||
}
|
||||
}
|
||||
|
||||
if (windowData == null)
|
||||
{
|
||||
Thread.Sleep(5);
|
||||
continue;
|
||||
}
|
||||
|
||||
window.ApplyInPlace(windowData);
|
||||
var spectrum = FFT.Forward(windowData);
|
||||
var mag = FFT.Magnitude(spectrum);
|
||||
|
||||
var (rhythmEnergy, rhythmFlux, melodyEnergy, melodyFreq) =
|
||||
MusicAnalyzer.ExtractFeatures(mag, _prevRhythmMag, _sampleRate);
|
||||
|
||||
_prevRhythmMag = mag;
|
||||
|
||||
// Adaptive normalization: running max with slow decay
|
||||
_liveMaxFlux = Math.Max(rhythmFlux, _liveMaxFlux * 0.999);
|
||||
_liveMaxMelodyEnergy = Math.Max(melodyEnergy, _liveMaxMelodyEnergy * 0.999);
|
||||
|
||||
tf.RhythmFlux = Math.Max(tf.RhythmFlux, rhythmFlux);
|
||||
tf.MelodyEnergy += melodyEnergy;
|
||||
tf.MelodyFreqSamples.Add(melodyFreq);
|
||||
tf.MelodyCount++;
|
||||
_fftIndexInTick++;
|
||||
|
||||
if (_fftIndexInTick >= _fftsPerTick)
|
||||
{
|
||||
var (frameA, frameB) = MusicAnalyzer.BuildWaveFrame(tf, _liveMaxFlux, _liveMaxMelodyEnergy);
|
||||
_state.EnqueueStream('A', new[] { frameA });
|
||||
_state.EnqueueStream('B', new[] { frameB });
|
||||
|
||||
tf = new MusicAnalyzer.TickFeature();
|
||||
_fftIndexInTick = 0;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
public void Dispose()
|
||||
{
|
||||
_running = false;
|
||||
try { _capture?.Dispose(); } catch { }
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user