Manuel Rodríguez Álvarez

Separation of speech signals for nonlinear mixtures

This paper shows an approach to recover original speech signals from their nonlinear mixtures. Using a geometric method that makes a piecewise linear approximation of the nonlinear mixing space, and the fact that the speech distributions are Laplacian or Gamma type, a set of slopes is obtained as a set of linear mixtures.

Can ICA Help Classify Skin Cancer and Benign Lesions

Various neural network models for the identification and classification of different skin lesions from ALA-induced fluorescence images are presented. After different image preprocessing steps, eigenimages and independent base images are extracted using PCA and ICA, respectively. In order to extract local information in the images rather than global features, Generative Topographic Mapping is added to cluster patches of the images first and then extract local features by ICA (local ICA). These components are used to distinguish skin cancer from benign lesions. An average classification rate of 70% is obtained, which considerably exceeds the rate achieved by an experienced physician.

Evolutionary Algorithm Using Mutual Information for Independent Component Analysis

Independent Component Analysis (ICA) is a method for finding underlying factors from multidimensional statistical data. ICA differs from other similar methods in that it looks for components that are both statistically independent and nongaussian. Blind Source Separation (BSS) consists in recovering unobserved signals from a known set of mixtures. Thus, ICA and BSS are equivalent when the mixture is assumed to be linear up to possible permutations and invertible scalings. However, when the mixing model is nonlinear, additional constraints are needed to assure that independent components correspond to the original signals. In this paper, we propose a simple though effective method based on estimating the probability densities of the outputs for solving the BSS problem in linear and nonlinear mixtures making use of genetic algorithms. A post-nonlinear mixture model is assumed so that the solution space in the nonlinear case is restricted to signals equivalent to the original ones.

Adaptive and heuristic approaches for nonlinear source separation

This paper presents a new adaptive procedure for the linear and non-linear separation of signals with non-uniform, symmetrical probability distributions, based on both simulated annealing (SA) and competitive learning (CL) methods by means of a neural network, considering the properties of the vectorial spaces of sources and mixtures, and using a multiple linearization in the mixture space. Also, the paper proposes the fusion of two important paradigms, genetic algorithms and the blind separation of sources (GABSS) in nonlinear mixtures. From experimental results, this paper demonstrates the possible benefits offered by GAs in combination with BSS, such as robustness against local minima, the parallel search for various solutions, and a high degree of flexibility in the evaluation function. The main characteristics of the method are its simplicity and the rapid convergence experimentally validated by the separation of many kinds of signals, with different probability density functions.

A new ica method based on a lattice of the observation space

This work shows a new method for blind separation of sources, based on geometrical considerations concerning the observation space. This new method is applied to a mixture of several sources and it obtains the estimated coeflicients of the unknown mixture mafrix A and separates the unknown sources. The principles of the new method and a description of the algorithm followed by some speed enhancements are shown. Finally, we illustrate with simulations of several source disfributions how the algorithm performs.