Speech articulation rate change using recursive bandwidth scaling

Abstract

Speech articulation rate change is done by analyzing the speech signal into several frequency channels, scaling the unwrapped phase signal in each channel and synthesizing a new speech signal using the modified channel signals and their scaled center frequencies. It is shown that each channel signal can be modeled as the simultaneous amplitude and phase modulation of a carrier and that only scaling the phase modulating signal does not result in a proportional scaling of the bandwidth of the channel signals which results in the introduction of different types of distortions like frequency aliasing between channels when an increase in the articulation rate is attempted and reverberation when a rate reduction is attempted. It is proposed that the amplitude modulating signal bandwidth should also be scaled and a recursive method to do this is discussed.