BandSplitRNNEnhancer<T>

Models & Types · AiDotNet.Audio.Enhancement

Band-Split RNN speech enhancement model (Luo and Yu, 2023).

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

Implements: IAudioEnhancer<T>, IGeneratedParameterRegistrar<T>

For Beginners

Imagine multiple specialized listeners, each focused on a different pitch range (bass, midrange, treble). Each cleans up their range independently, then they combine results. This divide-and-conquer approach works well because different types of noise affect different frequency ranges differently.

Usage:

How It Works

Band-Split RNN splits the spectrogram into non-overlapping frequency bands, processes each band independently with a shared RNN, then fuses across bands. Originally designed for music source separation, it also excels at speech enhancement by treating noise as a source to separate.

Constructors

Constructor Summary
BandSplitRNNEnhancer(NeuralNetworkArchitecture<T> architecture, BandSplitRNNEnhancerOptions options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer)
BandSplitRNNEnhancer(NeuralNetworkArchitecture<T> architecture, string modelPath, BandSplitRNNEnhancerOptions options)

Properties

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

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()
Task<Tensor<T>> EnhanceAsync(Tensor<T> noisyAudio, CancellationToken cancellationToken)
Tensor<T> Enhance(Tensor<T> noisyAudio)
Tensor<T> EnhanceWithReference(Tensor<T> audio, Tensor<T> reference)
Tensor<T> ProcessChunk(Tensor<T> audioChunk)
void EstimateNoiseProfile(Tensor<T> noiseOnlyAudio)