Implementation of finite impulse response (FIR) filter as test application on general purpose evolutionary algorithm testbed system
Metadata[+] Show full item record
[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI AT AUTHOR'S REQUEST.] A 19th order Finite Impulse Response (FIR) filter is implemented as a test application for Gpeat, a general purpose evolutionary algorithm testbed. The filter is used to evaluate the correctness, usability and efficiency of its performance. The main objective of the project is to be able to implement a real time evolutionary algorithm based application on the Gpeat system that can aid in the analysis of the performance and functionality of the Gpeat system. The FIR filter is implemented on the same hardware platform (Xilinx® Spartan™ 3A FPGA development board) as the Gpeat system and tests its results in a real world environment. The FIR filter evaluation block evaluates the fitness value for the randomly generated chromosomes generated by Gpeat in a real world environment and sends the fitness value back to the Gpeat system for further analysis. The Gpeat system is configured using a GUI. The information entered through the GUI is used to generate VHDL code and that code is compiled into a Xilinx® bit file for downloading. This bit file is dumped onto the hardware block to realize the user supplied information in hardware. Based on the information the Gpeat system generates random members to create a population. Each member of the population acts as a set of coefficients for the FIR filter. The FIR filter is provided with a description of the ideal response that we would like to see from the filter. An input signal is fed into the system from the function generator and the FIR filter uses the coefficients generated by the Gpeat system to generate a filter output. The filter output is compared with the ideal response provided and the sum of square error is generated. The sum of square error is deducted from the maximum error to generate the fitness value. Once a satisfactory member of the population is found the process stops until the user initiates the search again by changing the expected ideal response from the filter. The FIR filter generated is capable of working as low pass, high pass, band pass and band reject filter or any pattern of gaps and bands. This thesis describes the development of the FIR filter, the FIR intrinsic fitness evaluation system, a comparison to an existing FIR filter system implemented in software and other related topics to the evaluation of Gpeat performance.
Access to files is limited to the University of Missouri--Columbia.