Convolutive Mixtures Of Speech Signals Research Articles

New efficient subband structures for blind source separation

Convolutive mixtures of signals generated by more than one source and acquired by a set of microphones are commonly found in acoustic signal processing applications. Subband methods have been proposed to reduce computational complexity and improve the convergence rate of adaptive algorithms developed for blind source separation of these mixtures, without significantly impairing steady-state performance. Oversampled discrete Fourier transform (DFT) filter bank is a usual option for the generation of subband signals in order to avoid harmful aliasing effects, thereby maintaining sufficient samples to estimate the subband signal statistics. In this paper, two new subband structures, composed of cosine-modulated filter banks (CMFB) with critical or oversampled sampling and low-order adaptive subfilters, are proposed for efficient blind source separation approach in convolutive mixtures of speech signals. A time-varying step-size procedure that provides better convergence rate for several reverberation characteristics is advanced. In addition, both computational complexity and steady-state performance of the proposed structures are compared to those of the standard fullband algorithm, to other subband structures with oversampling and critical sampling, and to frequency domain algorithm. The advanced solutions are capable of improving the source interference ratio (SIR) by more than 5 dB. Finally, two strategies are presented to minimize the effects of remaining aliasing in the proposed subband approaches, which obtain an additional gain of about 3 dB in the asymptotic SIR.

Complex Blind Source Separation

Blind source separation (BSS) techniques aim at recovering the original source signals from observed mixtures without a priori information. The bivariate empirical mode decomposition (BEMD) algorithm combined with complex independent component analysis by entropy bound minimization (ICA-EBM) technique is proposed as an alternative to separate convolutive mixtures of speech signals. The empirical mode decomposition (EMD) is a local self-adaptive decomposition method that allows analyzing data from non-stationary and/or nonlinear processes. Its principle is based on the sequential extraction of different amplitude and frequency modulation single-component contributions called intrinsic mode functions (IMFs). The BEMD is an extension of the EMD to complex-valued signals. First, the convolutive mixtures in the frequency domain are decomposed into a set of IMFs using the BEMD algorithm, and then, the complex ICA-EBM method is applied to extract the independent sources. The performance of the proposed approach is tested on real speech sounds chosen from available databases and compared to the results obtained via conventional frequency ICA and BEMD-ICA-based separation for convolutive mixtures. Simulation results show that the proposed method of BSS outperforms the BEMD-ICA separation technique for convolutive mixtures and conventional frequency ICA.

An ICA Algorithm for Separation of Convolutive Mixture of Speech Signals

An ICA Algorithm for Separation of Convolutive Mixture of Speech Signals

Enhancement of speech signals separated from their convolutive mixture by FDICA algorithm

This paper presents a novel method for the enhancement of independent components of mixed speech signal segregated by the frequency domain independent component analysis (FDICA) algorithm. The enhancement algorithm proposed here is based on maximum a posteriori (MAP) estimation of the speech spectral components using generalized Gaussian distribution (GGD) function as the statistical model for the time-frequency series of speech (TFSS) signal. The proposed MAP estimator has been used and evaluated as the post-processing stage for the separation of convolutive mixture of speech signals by the fixed-point FDICA algorithm. It has been found that the combination of separation algorithm with the proposed enhancement algorithm provides better separation performance under both the reverberant and non-reverberant conditions.

Multistage Blind Source Separation and Deconvolution for Convolutive Mixture of Speech Signals

Multistage Blind Source Separation and Deconvolution for Convolutive Mixture of Speech Signals

Linear State-Space Models for Blind Source Separation

We apply a type of generative modelling to the problem of blind source separation in which prior knowledge about the latent source signals, such as time-varying auto-correlation and quasi-periodicity, are incorporated into a linear state-space model. In simulations, we show that in terms of signal-to-error ratio, the sources are inferred more accurately as a result of the inclusion of strong prior knowledge. We explore different schemes of maximum-likelihood optimization for the purpose of learning the model parameters. The Expectation Maximization algorithm, which is often considered the standard optimization method in this context, results in slow convergence when the noise variance is small. In such scenarios, quasi-Newton optimization yields substantial improvements in a range of signal to noise ratios. We analyze the performance of the methods on convolutive mixtures of speech signals.

Permutation Correction in the Frequency Domain in Blind Separation of Speech Mixtures

This paper presents a method for blind separation of convolutive mixtures of speech signals, based on the joint diagonalization of the time varying spectral matrices of the observation records. The main and still largely open problem in a frequency domain approach is permutation ambiguity. In an earlier paper of the authors, the continuity of the frequency response of the unmixing filters is exploited, but it leaves some frequency permutation jumps. This paper therefore proposes a new method based on two assumptions. The frequency continuity of the unmixing filters is still used in the initialization of the diagonalization algorithm. Then, the paper introduces a new method based on the time-frequency representations of the sources. They are assumed to vary smoothly with frequency. This hypothesis of the continuity of the time variation of the source energy is exploited on a sliding frequency bandwidth. It allows us to detect the remaining frequency permutation jumps. The method is compared with other approaches and results on real world recordings demonstrate superior performances of the proposed algorithm.

Evaluation of separation and dereverberation performance in frequency domain blind source separation

In this paper, we propose a new method for evaluating the separation and dereverberation performance of a convolutive blind source separation (BSS) system, and investigate a separating system obtained by employing frequency domain BSS based on independent component analysis (ICA). As a result, we reveal the acoustical characteristics of the frequency domain BSS for convolutive mixture of speech signals. We show that the separating system removes the direct sound of a jammer signal even when the frame length is relatively short, and it also reduces the reverberation of the jammer according to the frame length. We also confirm that the reverberation of the target is not reduced. Moreover, we propose a technique, suggested by the experimental results, for improving the quality of the separated signals by removing pre-echo noise.

Convolutive blind source separation of speech signals based on amplitude modulation decorrelation

The problem of blind separation of a convolutive mixture of speech signals is considered. Signal separation is performed in the frequency domain. The amplitude modulation decorrelation algorithm for convolutive blind source separation of speech is described, the first ideas of which were presented by Anemüller at the joint ASA/EAA/DAGA meeting, Berlin, 1999. The algorithm is based on the fact that the frequency-specific envelopes of speech signals exhibit correlation across different frequencies. This feature can be used to solve the ‘‘permutation problem’’ of frequency-domain-based blind source separation algorithms. Furthermore, it leads to separation of good quality since it results in a high number of constraints which must be fulfilled by the unmixed signals. Results for the separation of speech signals are presented for different mixing scenarios, including real-room reverberant mixing.