Joseph Di Martino

Real-time pitch tracking using the eCATE algorithm

This paper deals with the determination in real-time of the fundamental frequency FO of speech signals using an enhanced version of the CATE algorithm. The voiced / unvoiced decision as well as pitch determination is realized in real-time. Our approach was tested with the Bagshaw database of the Center for Speech Technology Research at Edinburgh. The presented algorithm, besides being able to work in real time, provides comparable or better results than those obtained by eSRPD, CATE or ALS algorithms.

On the use of an iterative estimation of continuous probabilistic transforms for voice conversion

Voice conversion is a technique that modifies a source speakers speech to be perceived as if a target speaker had spoken it. Among the algorithms of conversion published in the literature, the techniques using GMM are nowadays the reference. In this paper, we focus on a new technique for estimating parameters of the conversion function. We show that it is possible to determinate these parameters without statistical estimation using the classical Expectation Minimization (EM) algorithm. We propose a new method for determining the conversion function parameters. The technique proposed is based on an iterative statistical refinement algorithm working directly from data. The consequence of this strategy is that the estimation of the conversion function is very fast.

An architectural comparison of signal reconstruction algorithms from short-time Fourier transform magnitude spectra

This paper presents a comparison of different spectral re-synthesis algorithms. This study describes more particularly the Di Martino and Pierron (D&P) and real-time iterative spectrogram inversion with look-ahead algorithms from an architectural point of view because they are dedicated to real-time process. We use Python (as simulation language) because it allows easily the comparison of performances of the all the algorithms studied according to some important algorithm parameters as the number of iterations or the number of look-ahead frames. This comparison confirms the advantage of using D&P for real-time process from an architectural point of view.

A corroborative study on improving pitch determination by time–frequency cepstrum decomposition using wavelets

A new wavelet-based method is presented in this work for estimating and tracking the pitch period. The main idea of the proposed new approach consists in extracting the cepstrum excitation signal and applying on it a wavelet transform whose resulting approximation coefficients are smoothed, for a better pitch determination. Although the principle of the algorithms proposed has already been considered previously, the novelty of our methods relies in the use of powerful wavelet transforms well adapted to pitch determination. The wavelet transforms considered in this article are the discrete wavelet transform and the dual tree complex wavelet transform. This article, by all the provided experimental results, corroborates the idea of decomposing the cepstrum excitation by using wavelet transforms for improving pitch detection. Another interesting point of this article relies in using a simple but efficient voicing decision (which actually improves a similar voicing criterion we proposed in a preceding published study) which on one hand respects the real-time process with low latency and on the other hand allows obtaining low classifications errors. The accuracy of the proposed pitch tracking algorithms has been evaluated using the international Bagshaw and the Keele databases which include male and female speakers. Our various experimental results demonstrate that the proposed methods provide important performance improvements when compared with previously published pitch determination algorithms.