On the Assumption of Spherical Symmetry and Sparseness for the Frequency-Domain Speech Model

• Published in: IEEE transactions on audio, speech, and language processing speech, and language processing, 2007-07, Vol.15 (5), p.1521-1528
• Main Authors: Lee, Intae, Lee, Te-Won
• Format: Article
• Language: English
• Subjects: Algorithms; Blind source separation; Blind source separation (BSS); cocktail party problem; convolutive mixture; entropy estimator; frequency domain; Frequency domain analysis; Independent component analysis; independent component analysis (ICA); independent vector analysis (IVA); Instruction sets; Multidimensional systems; Natural language processing; order statistics; permutation problem; Probability density function; Probability density functions; RADICAL; Running; Separation; Signal analysis; Signal processing algorithms; Source separation; Speech; Speech analysis; statistical signal processing; Symmetry; Vector analysis
• ISSN: 1558-7916; 2329-9290; 1558-7924; 2329-9304
• DOI: 10.1109/TASL.2007.899231

A new independent component analysis (ICA) formulation called independent vector analysis (IVA) was proposed in order to solve the permutation problem in convolutive blind source separation (BSS). Instead of running ICA in each frequency bin separately and correcting the disorder with an additional algorithmic scheme afterwards, IVA exploited the dependency among the frequency components of a source and dealt with them as a multivariate source by modeling it with sparse and spherically, or radially, symmetric joint probability density functions (pdfs). In this paper, we compare the speech separation performances of IVA by using a group of l p -norm-invariant sparse pdfs where the value of and the sparseness can be controlled. Also, we derive an IVA algorithm from a nonparametric perspective with the constraint of spherical symmetry and high dimensionality. Simulation results confirm the efficiency of assuming sparseness and spherical symmetry for the speech model in the frequency domain.