Phase-spread voicing analysis in parametric speech coders

摘要

A novel technique for accurately measuring the voicing content of a speech signal for parametric speech coders is presented. This technique does not use signal periodicity but utilises the phase-spread information contained in the individual pitch cycles of the speech signal, giving greater time accuracy. This non-reliance on periodicity also brings increased robustness to small variations in pitch, which can affect the accuracy of the voicing analysis. Compared to traditional methods this voicing technique more accurately reflects the natural evolution of the speech signal over time.