Gradient Algorithms for Complex Non-Gaussian Independent Component/Vector Extraction, Question of Convergence

Abstract

We revise the problem of extracting one independent component from an instantaneous linear mixture of signals. The mixing matrix is parameterized by two vectors, one column of the mixing matrix and one row of the de-mixing matrix. The separation is based on the non-Gaussianity of the source of interest, while the other background signals are assumed to be Gaussian. Three gradient-based estimation algorithms are derived using the maximum likelihood principle and are compared with the Natural Gradient algorithm for Independent Component Analysis and with One-unit FastICA based on negentropy maximization. The ideas and algorithms are also generalized for the extraction of a vector component when the extraction proceeds jointly from a set of instantaneous mixtures. Throughout the paper, we address the problem of the size of the region of convergence for which the algorithms guarantee the extraction of the desired source. We show how that size is influenced by the ratio of powers of the sources within the mixture. Simulations confirm this observation where several algorithms are compared. They show various convergence behaviour in a scenario where the source of interest is dominant or weak. Here, our proposed modifications of the gradient methods taking into account the dominance/weakness of the source show improved global convergence property.

Adaptive Blind Audio Source Extraction Supervised By Dominant Speaker Identification Using X-Vectors

“…, x d k, ] T corresponds to the signals on microphones. Using the parameterization from [15], W k can be partitioned as…”

“…To construct the pilot P XVEC via (15), we shorten the context of the pooling layer within the x-vector DNN to Lc = 10 frames,…”

Overdetermined Independent Vector Analysis

“…IP is a class of block coordinate descent (BCD) methods specialized for optimizing the maximumlikelihood-based IVA, with the advantages of high computational efficiency and no hyperparameters. Recently proposed OverIVA [5] (IVA for the overdetermined case), which is a multi-target-source extension of independent component/vector extraction (ICE/IVE [15,16]), takes the advantages of IP, resulting in significantly improved computational cost of IVA. In addition to IP, OverIVA as well as ICE/IVE also relies on the orthogonality constraint (OC) that the sample correlation between the separated target and noise signals is to be zero.…”

Independent Vector Analysis with More Microphones Than Sources

“…Few techniques have been proposed for overdetermined IVA. The single source case, i.e., K = 1, known as independent vector extraction (IVE), has been tackled with a gradient ascent method [21].…”