Signal separation processor based on second-order statistic algorithms

Author

Ndjountche, Tertulien ; Unbehauen, Rolf ; Luo, Fa-Long

Author_Institution

Lehrstuhl fur Allgemeine und Theor. Elektrotech., Erlangen-Nurnberg Univ., Germany

Volume

1

fYear

1999

fDate

6/21/1905 12:00:00 AM

Firstpage

322

Abstract

A processor architecture for the separation of mixed signals is proposed. It consists of a demixing circuit which is based on the structure of a finite impulse response (FIR) filter and a learning circuit which updates the FIR variable coefficients according to a stochastic gradient algorithm with variable step size in order to minimize the cost function defined as the cross-correlation of the output signals. Using theoretical predictions and simulations, the components used in the building blocks were sized to meet the precision requirement of 12 bits. As a result, the convergence of the resulting structure is fast and reliable

Keywords

CMOS analogue integrated circuits; FIR filters; analogue processing circuits; convergence of numerical methods; discrete time filters; gradient methods; learning systems; least mean squares methods; minimisation; signal resolution; statistical analysis; CMOS parameters; FIR filter; FIR variable coefficient updating; analog discrete-time circuit; cost function minimization; demixing circuit; finite impulse response filter; learning circuit; mixed signals; output signal cross-correlation; second-order statistic algorithms; signal separation processor; stochastic gradient algorithm; variable step size; Circuits; Convergence; Delay estimation; Electronic mail; Finite impulse response filter; Least squares approximation; Signal processing; Signal processing algorithms; Source separation; Statistics;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks, 1999. IJCNN '99. International Joint Conference on

ISSN

1098-7576

Print_ISBN

0-7803-5529-6

Type

conf

DOI

10.1109/IJCNN.1999.831511

Filename

831511