Real Time Implementation of Fir Filter Based on Time Delay Neural Network

Abstract

An FPGA real time Implementation of Time Delay Neural Network (TDNN) is presented in this paper. The design and all of the work are geared towards the implementation of the TDNN in a scalable fashion. The TDNN is an adaptive FIR filter with 18-bit input and 18-bit output resolution. In this paper, the filter adapts its tap weights using the Least-Mean-Square (LMS) algorithm and then stores them in FPGA memory cell. The LMS algorithm that is used for weight adaptation is off chip implemented. The input is processed through a digital tapped delay line. The FIR neural network is used for real time adaptive noise cancellation. When the filter order is 10, the filter consumes 1168 Spartan 3E FPGA logic elements.