Skip to main content

nisqa

Function nisqa 

Source
pub fn nisqa(signal: &[f32], weights: Option<&[u8]>) -> Result<f32, AudioError>
Expand description

Predict a NISQA speech-quality MOS for signal.

  • signal: mono degraded speech (no reference needed).
  • weights: the trained NISQA model weights. None (or, currently, any value) means the learned backend is unavailable.

ยงFails closed

Returns AudioError::BackendUnavailable when weights is None, and also while the inference backend is unwired for Some(_). It never returns a synthesised MOS.