ScaleInvariantSignalToDistortionRatio<T>

Models & Types · AiDotNet.Metrics

Scale-Invariant Signal-to-Distortion Ratio (SI-SDR) metric for source separation evaluation.

How It Works

SI-SDR is the standard metric for evaluating source separation quality. It measures how well the estimated signal matches the target signal, ignoring scale differences.

Formula: SI-SDR = 10 * log10(||s_target||^2 / ||e_noise||^2) where s_target is the projection of the estimate onto the target, and e_noise is the residual.

Higher values indicate better separation quality. Typical values:

  • 15 dB: Excellent separation

  • 10-15 dB: Good separation
  • 5-10 dB: Fair separation
  • <5 dB: Poor separation

Constructors

Constructor Summary
ScaleInvariantSignalToDistortionRatio() Initializes a new instance of SI-SDR calculator.

Methods

Method Summary
T Compute(Tensor<T> estimated, Tensor<T> target) Computes SI-SDR between an estimated signal and the target signal.
T ComputeImprovement(Tensor<T> estimated, Tensor<T> target, Tensor<T> baseline) Computes SI-SDR improvement relative to a baseline (typically the mixture).