The audio signal is converted to the frequency domain and partitioned according to critical voice components. The voice component analysis block considers temporal and correlative properties of speech and noise to develop a predictive model of the speech components.

The extracted voice components are recombined and converted into the time domain. The voice extractor then modifies each frequency component according to the voice and noise model, taking advantage of psycho-acoustics principles to minimise noise floor artifacts and perceived voice distortion.