Mohamed Anouar Ben Messaoud

Autocorrelation of the Speech Multi-Scale Product for Voicing Decision and Pitch Estimation

In this work, we present an algorithm for voiced/unvoiced decision and pitch estimation from speech signals. Our approach is based on classifying the peaks provided by the autocorrelation of the speech multi-scale product. The multi-scale product is based on making the product of the speech wavelet transform coefficients at three successive dyadic scales. The autocorrelation function of the multi-scale product is calculated over frames of a specific length. The experimental results show the robustness and the effectiveness of our approach. Besides, the proposed method outperforms some existing algorithms in a clean and noisy environment.

Spectral multi-scale product analysis for pitch estimation from noisy speech signal

In this work, we present an algorithm for estimating the fundamental frequency in speech signals. Our approach is based on the spectral multi-scale product analysis. It consists of operating a short Fourier transform on the speech multi-scale product. The multi-scale product is based on making the product of the speech wavelet transform coefficients at three successive dyadic scales. The wavelet used is the quadratic spline function with a support of 0.8 ms. We estimate the pitch for each time frame based on its multi-scale product harmonic structure. We evaluate our approach on the Keele database. Experimental results show the effectiveness of our method presenting a good performance surpassing other algorithms. Besides, the proposed approach is robust for noisy speech.

Single channel speech separation based on sinusoidal modeling

We present an approach of single channel speech separation based on sinusoidal modeling and multi-scale product analysis. To construct the sinusoidal model of speech, we need to determine three parameters which are namely frequency, amplitude and phase. For fundamental frequency determination, we apply an effective method based on multi-scale product analysis. For amplitude estimation, we adopt an iterative peak picking algorithm. This approach effectively estimates multi-pitch of the desired speaker in a multiple speaker context.

Pitch estimation of speech and music sound based on multi-scale product with auditory feature extraction

The pitch is a crucial parameter in speech and music signals. However, due to severe noisy conditions, missing harmonics, unsuitable physical vibration, the determination of pitch presents a great challenge when desiring to get a good accuracy. In this paper, we propose a method for pitch estimation of speech and music sounds. Our method is based on the fast Fourier transform (FFT) of the multi-scale product (MP) provided by a feature auditory model of the sound signals. The auditory model simulates the spectral behaviour of the cochlea by a gammachirp filter-bank, and the out/middle ear filtering by a low-pass filter. For the two output channels, the FFT function of the MP is computed over frames. The MP is based on constituting the product of the speech and music wavelet transform coefficients at three scales. The experimental results show that our method estimates the pitch with high accuracy. Besides, our proposed method outperforms several other pitch detection algorithms in clean and noisy environments.

A Robust pitch estimation approach for clean speech

In this work, we present an algorithm for estimating the fundamental frequency in speech signals. Our approach is based on the spectral compression by the autocorrelation of the speech multi-scale product analysis. It consists of operating the product of compressed copies of the original spectrum on the multi-scale product autocorrelation. The multi-scale product is based on making the product of the speech wavelet transform coefficients at three successive dyadic scales. The wavelet used is the quadratic spline function with a support of 0.8 ms. We estimate the pitch for each time frame based on its multi-scale product autocorrelation of the harmonic product spectrum structure. We evaluate our approach on the Keele database. Experimental results show the effectiveness of our method presenting a good performance surpassing the other algorithms.

Time-frequency masks for monaural speech separation: A comparative review

In this paper we present a comparative analysis of different time-frequency (T-F) masking techniques used for single channel speech separation (SCSS). We survey T-F masking concept and compare different types of masks in different criteria. The comparison is conduct theoretically by mathematical study and numerically by objective and subjective assessment. Also, we study the effect of the masking techniques on the perceptual quality of speech and their ability to separate a target speech from monaural mixing.