NNSP-3.1

Blind Separation of Temporomandibular Joint Sound Signals
Yinchao Guo (Time-Frequency Lab, c/o Chris Koh, Division of SEM, School of Mechanical and Production Engineering, Nanyang Technological University, Singapore 639798), Farook Sattar (Division of Information Engineering, School of EEE, Nanyang Technological University, Nanyang Technological University, Singapore 639798), Christopher Koh (School of Mechanical and Production Engineering, Nanyang Technological University, Singapore 639798)

In order to develop a cheap, efficient and reliable diagnostic tool for the detection of temporomandibular joint disorders (TMD), sounds from the temporomandibular joint (TMJ) are recorded using a pair of microphone inserted in the auditory canals. However, the TMJ sounds originating from one side of head can also be picked up by microphone at the other side. Blind source separation (BSS) is thus proposed as a method to recover the original sound. The authors propose to use non-casual filters for the separation of TMJ signals. The algorithm is based on information theory and is an extension of early work by Torkkola. The separation was successful and the output can now be used for subsequent analysis of TMJ sounds.

NNSP-3.2

GENERALIZED ANTI-HEBBIAN LEARNING FOR SOURCE SEPARATION
Hsiao-Chun Wu, Jose C. Principe (Computational NeuroEngineering Laboratory, Department of Electrical and Computer Engineering, University of Florida)

The information-theoretic framework for source separation is highly suitable. However the choice of the nonlinearity or the estimation of the multidimensional joint probability density function are nontrivial. We propose here a generalized Gaussian model to construct a generalized blind source separation network based on the minimum entropy principle. This new separation network can suppress the interference to a significant amount compared to the traditional LMS-echo-canceler. The simulation is given to show the disparity of the performance as a varies. Finally how to choose the appropriate a in our generalized anti-Hebbian rule is discussed.

NNSP-3.3

Complex Independent Component Analysis By Nonlinear Generalized Hebbian Learning with Rayleigh Nonlinearity
Eraldo Pomponi, Simone Fiori, Francesco Piazza (University of Ancona)

The aim of this paper is to present a non-linear Extension of the Sanger's Generalized Hebbian Algorithm to the processing of complex-valued data. A possible choice of the involved non-linearity is discussed recalling the Sudjianto-Hassoun nterpretation of the non-linear Hebbian learning. Extension of this interpretation to the complex case leads to a nonlinearity called Rayleigh function, which allows for separating mixed independent complex-valued source signals.

NNSP-3.4

Sparse Basis Selection, ICA, and Majorization: Towards a Unified Perspective
Ken Kreutz-Delgado, Bhaskar D Rao (Dept. ECE, University of California, San Diego)

Sparse solutions to the linear inverse problem Ax = y and the determination of an environmentally adapted overcomplete dictionary (the columns of A) depend upon the choice of a ``regularizing function'' d(x) in several recently proposed procedures. We discuss the interpretation of d(x) within a Bayesian framework, and the desirable properties that ``good'' (i.e., sparsity ensuring) regularizing functions, d(x) might have. These properties are: Schur-concavity (d(x) is consistent with majorization); concavity (d(x) has sparse minima); parameterizability (d(x) is drawn from a large, parameterizable class); and factorizability of the gradient of d(x) in a certain manner. The last property (which naturally leads one to consider separable regularizing functions) allows d(x) to be efficiently minimized subject to Ax=y using an Affine Scaling Transformation (AST)-like algorithm ``adapted'' to the choice of d(x). A Bayesian framework allows the algorithm to be interpreted as an Independent Component Analysis (ICA) procedure.

NNSP-3.5

Two Spatio-temporal Decorrelation Learning Algorithms and their Application to Multichannel Blind Deconvolution
Seungjin CHOI (School of Electrical and Electronics Engineering, Chungbuk National University, KOREA), Andrzej CICHOCKI, Shun-ichi AMARI (Brain-style Information Systems Research Group, BSI, RIKEN, JAPAN)

In this paper we present and compare two different spatio-temporal decorrelation learning algorithms for updating the weights of a linear feedforward network with FIR synapses (MIMO FIR filter). Both standard gradient and the natural gradient are employed to derive the spatio-temporal decorrelation algorithms. These two algorithms are applied to multichannel blind deconvolution task and their performance is compared. The rigorous derivation of algorithms and computer simulation results are presented.

NNSP-3.6

Adaptive Paraunitary Filter Banks for Principal and Minor Subspace Analysis
Scott C Douglas (Dept. of Electrical Engr., Southern Methodist Univ., Dallas, TX USA), Shun-ichi Amari (RIKEN Brain Science Institute, Saitama, JAPAN), S.-Y. Kung (Dept. of Electrical Engineering, Princeton Univ., Princeton, NJ USA)

Paraunitary filter banks are important for several signal processing tasks. In this paper, we consider the task of adapting the coefficients of a multichannel FIR paraunitary filter bank via gradient ascent or descent on a chosen cost function. The proposed generalized algorithms inherently adapt the system's parameters in the space of paraunitary filters. Modifications and simplifications of the techniques for spatio-temporal principal and minor subspace analysis are described. Simulations verify one algorithm's useful behavior in this task.

< NNSP-2 NNSP-4 >

Last Update: February 4, 1999 Ingo Höntsch