Date of Award

4-13-2017

Publication Type

Master Thesis

Degree Name

M.A.Sc.

Department

Electrical and Computer Engineering

Keywords

DE, FIR, PSO

Supervisor

Kwan, H. K.

Rights

info:eu-repo/semantics/openAccess

Abstract

Digital filter plays a vital part in digital signal processing field. It has been used in control systems, aerospace, telecommunications, medical applications, speech processing and so on. Digital filters can be divided into infinite impulse response filter (IIF) and finite impulse response filter (FIR). The advantage of FIR is that it can be linear phase using symmetric or anti-symmetry coefficients. Besides traditional methods like windowing function and frequency sampling, optimization methods can be used to design FIR filters. A common method for FIR filter design is to use the Parks-McClellan algorithm. Meanwhile, evolutional algorithm such as Genetic Algorithm (GA), Particle Swarm Optimization (PSO) [2], and Differential Evolution (DE) have shown successes in solving multi-parameters optimization problems. This thesis reports a comparison work on the use of PSO, DE, and two modified DE algorithms from [18] and [19] for designing six types of linear phase FIR filters, consisting of type1 lowpass, highpass, bandpass, and bandstop filters, and type2 lowpass and bandpass filters. Although PSO has been applied in this field for some years, the results of some of the designs, especially for high-dimensional filters, are not good enough when comparing with those of the Parks-McClellan algorithm. DE algorithms use parallel search techniques to explore optimal solutions in a global range. What’s more, when facing higher dimensional filter design problems, through combining the knowledge acquired during the searching process, the DE algorithm shows obvious advantage in both frequency response and computational time.

Download

Share

COinS