SpikingFullSubNet<T>

Models & Types · AiDotNet.Audio.Enhancement

Spiking-FullSubNet speech enhancement model using spiking neural networks.

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

Implements: IAudioEnhancer<T>, IGeneratedParameterRegistrar<T>

For Beginners

Spiking-FullSubNet cleans up noisy audio using a brain-inspired approach. Instead of traditional neural networks that pass numbers between layers, it uses “spikes” (on/off signals) like real neurons. This means it can run much more efficiently on specialized hardware while still producing clean, clear speech.

Usage:

How It Works

Spiking-FullSubNet (2024) replaces conventional RNNs with spiking neural network (SNN) layers in the FullSubNet architecture. SNNs use biologically-inspired binary spike events instead of continuous activations, achieving comparable speech enhancement quality with significantly reduced energy consumption, making it suitable for neuromorphic hardware.

Constructors

Constructor Summary
SpikingFullSubNet(NeuralNetworkArchitecture<T> architecture, SpikingFullSubNetOptions options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer) Creates a Spiking-FullSubNet model in native training mode.
SpikingFullSubNet(NeuralNetworkArchitecture<T> architecture, string modelPath, SpikingFullSubNetOptions options) Creates a Spiking-FullSubNet model in native training mode.

Properties

Property Summary
double EnhancementStrength { get; set; }
int LatencySamples { get; }
int NumChannels { get; }

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()
Tensor<T> Enhance(Tensor<T> audio)
Tensor<T> EnhanceWithReference(Tensor<T> audio, Tensor<T> reference)
Tensor<T> ProcessChunk(Tensor<T> audioChunk)
void EstimateNoiseProfile(Tensor<T> noiseOnlyAudio)
void ResetState()