A Multiresolutional Analysis/Synthesis Approach for the Pitch Modification of Speech Signals

Authors: Osman EROĞLU, Nur SERİNKEN

Abstract: A new analysis/synthesis approach is introduced in this paper for the pitch modification of speech signals. Unlike the other pitch modification methods proposed in the literature, this algorithm modifies the wavelet coefficients of the speech signal in order to obtain a pitch-modified version of the original signal instead of modifying the speech waveform itself. In this method, the speech signal is first divided into subbands using an 11-level quadrature mirror filter bank (QMF) in order to obtain more accurately localized temporal and frequency information for the pitch modification algorithm. These subbands are then modified using the various pitch modification algorithms. Finally, the inverse wavelet transform is applied to these modified subband coefficients in order to reconstruct a pitch-modified version of the input signal. The performance of the multiresolutional analysis/synthesis algorithm is evaluated through a series of subjective listening tests which show the multiresolutional pitch modification (MPM) of speech signals, with the desired pitch modification factor, increases the quality of the pitch-modified speech over the well-known pitch modification algorithms while preserving the duration of the original signal.