Time-scale modification of complex acoustic signals

Abstract

A new approach is introduced for time-scale modification of short-duration complex acoustic signals. The method preserves an approximate time-scaled temporal envelope of the signal, thus capitalizing on the perceptual importance of the signal’s temporal structure, while also maintaining its spectral character. The basis for the approach is a subband signal representation, derived from a filterbank analysis/synthesis, whose channel phases are controlled to shape the temporal envelope of the time-scaled signal. To account for a stochastic element in the signal, channel amplitudes and filterbank inputs are selected to shape the spectrum and correlation of the time-scaled noise component. The phase, amplitude, and input control are derived from locations of events which occur within filterbank outputs. A frame-based generalization of the method imposes phase consistency and noise continuity across consecutive synthesis frames. The approach is applied to synthetic and actual complex acoustic signals consisting of closely spaced sequential time components and is applicable to music, biological, and underwater sound processing.