Time-scale and pitch modifications of speech signals and resynthesis from the discrete short-time Fourier transform

Abstract

The modification methods described in this paper combine characteristics of PSOLA-based methods and algorithms that resynthesize speech from its short-time Fourier magnitude only. The starting point is a short-time Fourier representation of the signal. In the case of duration modification, portions, in voiced speech corresponding to pitch periods, are removed from or inserted in this representation. In the case of pitch modification, pitch periods are shortened or extended in this representation, and a number of pitch periods is inserted or removed, respectively. Since it is an important tool for both duration and pitch modification, the resynthesis-from-short-time-Fourier-magnitude-only method of Griffin and Lim (1984) and Griffin et al. (1984) is reviewed and adapted. Duration and pitch modification methods and their results are presented.