Efficient Implementation and Performance Evaluation of the Second Order Volterra Filter Based on the MMD Approximation

摘要

Considered as a prototype of nonlinear system models the second order Volterra filter (FV_2) is characterized by an increased complexity in comparison with a linear filter. This complexity is given by the large number of filter coefficients, as well as by the filter operations and it requires an increased computational power in technical applications. The filter based on the multi memory decomposition (MMD) structure represents a good approximation of the FV_2 and significantly reduces the number of filter operations. In this paper we propose an efficient implementation of the MMD filter studied in a typical nonlinear identification problem. The simulations show the very good performance of our proposed MMD structure in comparison with the results obtained by using a second order LMS Volterra filter. We have evaluated the performance of our method based on the system response to different input signals.