Add managed codec DSP and initial control server

This commit is contained in:
2026-09-15 22:51:33 +02:00
parent 2df79cdd4c
commit 4067bab7c2
52 changed files with 2503 additions and 20 deletions
@@ -0,0 +1,48 @@
namespace VoiceCat.Dsp;
public sealed class EnergyVadProcessor
{
private readonly TimeProvider timeProvider;
private long lastVoiceTimestamp;
private bool hasVoice;
private float threshold;
public float Threshold
{
get => Volatile.Read(ref threshold);
set
{
if (!float.IsFinite(value) || value is < 0 or > 1) throw new ArgumentOutOfRangeException(nameof(value));
Volatile.Write(ref threshold, value);
}
}
public TimeSpan HangTime { get; }
public EnergyVadProcessor(float threshold = 0.02f, TimeSpan? hangTime = null, TimeProvider? timeProvider = null)
{
Threshold = threshold;
HangTime = hangTime ?? TimeSpan.FromMilliseconds(300);
if (HangTime < TimeSpan.Zero) throw new ArgumentOutOfRangeException(nameof(hangTime));
this.timeProvider = timeProvider ?? TimeProvider.System;
}
public bool Process(ReadOnlySpan<short> pcm)
{
long now = timeProvider.GetTimestamp();
if (!pcm.IsEmpty)
{
double sum = 0;
foreach (short sample in pcm)
{
double normalized = sample / 32768.0;
sum += normalized * normalized;
}
if (Math.Sqrt(sum / pcm.Length) >= Threshold)
{
lastVoiceTimestamp = now;
hasVoice = true;
}
}
return hasVoice && timeProvider.GetElapsedTime(lastVoiceTimestamp, now) < HangTime;
}
}
@@ -0,0 +1,53 @@
using System.Runtime.InteropServices;
using Microsoft.Win32.SafeHandles;
namespace VoiceCat.Dsp;
public sealed unsafe partial class RnnoiseProcessor : IDisposable
{
public const int SampleRate = 48000;
public const int FrameSamples = 480;
private readonly RnnoiseHandle handle;
private readonly float[] input = new float[FrameSamples];
private readonly float[] output = new float[FrameSamples];
public RnnoiseProcessor()
{
handle = new(Create());
if (handle.IsInvalid) { handle.Dispose(); throw new OutOfMemoryException(); }
}
public void Process(Span<short> pcm, int sampleRate = SampleRate)
{
ObjectDisposedException.ThrowIf(handle.IsClosed, this);
if (sampleRate != SampleRate) return;
if (pcm.Length % FrameSamples != 0) throw new ArgumentException("RNNoise requires complete 480-sample mono chunks.", nameof(pcm));
fixed (float* source = input)
fixed (float* destination = output)
{
for (int offset = 0; offset < pcm.Length; offset += FrameSamples)
{
for (int i = 0; i < FrameSamples; i++) input[i] = pcm[offset + i];
ProcessFrame(handle, destination, source);
for (int i = 0; i < FrameSamples; i++)
pcm[offset + i] = (short)Math.Clamp(MathF.Round(output[i], MidpointRounding.AwayFromZero), short.MinValue, short.MaxValue);
}
}
}
public void Dispose() => handle.Dispose();
[LibraryImport("voicecat_media", EntryPoint = "vcm_rnnoise_create")]
private static partial nint Create();
[LibraryImport("voicecat_media", EntryPoint = "vcm_rnnoise_destroy")]
private static partial void Destroy(nint state);
[LibraryImport("voicecat_media", EntryPoint = "vcm_rnnoise_process")]
private static partial float ProcessFrame(RnnoiseHandle state, float* output, float* input);
private sealed class RnnoiseHandle : SafeHandleZeroOrMinusOneIsInvalid
{
public RnnoiseHandle() : base(true) { }
internal RnnoiseHandle(nint value) : this() => SetHandle(value);
protected override bool ReleaseHandle() { Destroy(handle); return true; }
}
}
@@ -0,0 +1,5 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<AllowUnsafeBlocks>true</AllowUnsafeBlocks>
</PropertyGroup>
</Project>
@@ -0,0 +1,6 @@
{
"version": 1,
"dependencies": {
"net10.0": {}
}
}