QuailVad<T>

Models & Types · AiDotNet.Audio.VoiceActivity

Quail VAD - lightweight voice activity detection optimized for on-device deployment.

Inheritance: NeuralNetworkBase<T> → AudioNeuralNetworkBase<T> → QuailVad<T>

Implements: IVoiceActivityDetector<T>

For Beginners

Quail VAD is a lightweight “is someone talking?” detector that runs efficiently on phones and small devices. Despite being much smaller than models like Silero VAD, it achieves competitive accuracy by learning from larger, more powerful models during training.

Usage:

How It Works

Quail VAD (2024) is a compact CNN-RNN voice activity detector designed for edge devices. It uses knowledge distillation from larger teacher models to maintain high accuracy while keeping the model small enough for mobile phones and embedded systems. The model processes short audio frames and outputs speech probabilities with minimal latency.

Constructors

Constructor Summary
QuailVad(NeuralNetworkArchitecture<T> architecture, QuailVadOptions options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer) Creates a Quail VAD model in native training mode.
QuailVad(NeuralNetworkArchitecture<T> architecture, string modelPath, QuailVadOptions options) Creates a Quail VAD model in native training mode.

Properties

Property Summary
double Threshold { get; set; }
int FrameSize { get; }
int MinSilenceDurationMs { get; set; }
int MinSpeechDurationMs { get; set; }
int SampleRate { get; }

Methods

Method Summary
IReadOnlyList<ValueTuple<int, int>> DetectSpeechSegments(Tensor<T> audio)
ModelMetadata<T> GetModelMetadata()
T GetSpeechProbability(Tensor<T> audioFrame)
T[] GetFrameProbabilities(Tensor<T> audio)
ValueTuple<bool, T> ProcessChunk(Tensor<T> audioChunk)
bool DetectSpeech(Tensor<T> audioFrame)
void ResetState()