Microsoft Word - 508-182_budura.rtf

Nonlinear adaptive filtering techniques, based on the Volterra model, are widely used for the nonlinearities identification in many applications. This paper proposes a new implementation of the third order LMS Volterra filter. A third order nonlinear system with memory is identified using the new LMS algorithm implementation for the Volterra kernels estimation. The accuracy of the proposed algorithm is appreciated by comparing the estimated third order kernel with the real kernel. It is demonstrated that the adaptive algorithm gives satisfactory convergence in different conditions of noise. The extension of the LMS algorithm implementation to higher order Volterra filters is also possible and involves a few simple changes. Key-Words: Volterra kernels estimation, LMS algorithm, Nonlinear filter identification.