Audio time stretching with an adaptive multiresolution phase vocoder

Nicolas Juillerat; Béat Hirsbrunner

ICASSP 2017accepted0 citations

Audio time stretching with an adaptive multiresolution phase vocoder

Abstract

This paper presents a novel adaptive resolution technique for audio time stretching. The signal's transience is analyzed in both time and frequency, and controls the time-frequency resolution of the processing. Higher time resolution is used for transients and higher frequency resolution for steady sounds. To preserve phase coherence between components of different transience, only the magnitude is processed adaptively. This is done by first using a single resolution phase vocoder, and then by applying a multiresolution magnitude correction step. The proposed technique uses a decomposition of the signal in both time and frequency according to several degrees of transience. Yet the actual time stretching transformation is entirely based on the phase vocoder.

BibTeX

@inproceedings{icassp2017_audiotimestretch,
  title = {Audio time stretching with an adaptive multiresolution phase vocoder},
  author = {Nicolas Juillerat and Béat Hirsbrunner},
  booktitle = {ICASSP 2017},
  year = {2017}
}