Font Size: a A A

Research On Particle Swarm Optimization And Its Application To Fir Digital Filters

Posted on:2017-08-16Degree:DoctorType:Dissertation
Country:ChinaCandidate:P ShaoFull Text:PDF
GTID:1368330485465949Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
Intelligent algorithms are developed due to increasing problems of complexity, high-dimension, and non-linearity by traditional opitimization algorithms are difficult to be solved in life and production. Inspired by the social behaviors of creatures or natural phenomena in nature, intelligent algorithms are proposed. Particle swarm optimization (PSO), one of intelligent algorithms, is proposed on the basis of the forgaing of bird swarm. Thanks to its advantages such as few parameters, easy implementation and faster convergence speed in the former period of PSO, PSO is employed to solve various complex problems and has achieved better optimization performance. However, compared with other optimization algorithms, PSO has also some other disadvantages over optimization such as slower convergence speed in the latter period of PSO, easy trapping into the local optima and lower convergence accuracy.Digital filters, as impotant parts of digital signal processing system, have broader pratical application to image processing, remote sensing, and medical system. They can be classified into two categories, namely infinite impulse response filters (IIR) and finite impulse response filters (FIR) according to the length of unit impulse response. FIR digital filters have excellent characteristics of strict linear phase and arbitrary amplitude which makes the filters more stable. These characteristics make FIR digital filters easy to implement and to obtain wider practical application compared with IIR digital filters. However, for digital filter design, it is not so easy for accurate determination of the boundary frequency. Moreover, they tend to get struck at local minima, lower convergence accuracy and slower convergence speed. Therefore, intelligent algorithms are introduced to optimize and design digital filters so that some weakness of traditional methods can be overcome.In order to avoid disadvantages of PSO and enchance the optimization performance of PSO, trigonometric factor, opposition-based learning and extremal optimization are used to improve the optimization performance of PSO respectively. Meanwhile, serveral improved PSO are introduced into digital filters design to improve their performance. This dissertation is mainly focusing on:(1) For the purpose of easily falling into local optimum and causing lower solution accuracy of PSO algorithm, this dissertation puts forward an improved PSO algorithm based on the trigonometric factor (TFPSO). The algorithm introduces the periodic oscillations trigonometric factor in the position update equation of PSO as its parmameter so that each particle obtains periodic oscillations to expand the search space and more easy escape from local optima and find the optimal value. Experimental results and analysis show that the proposed algorithm is not only easy implementation and converge fast, but also helpful to improve the accuracy of solution.(2) One of the typical of weaknesses is that particle swarm optimization algorithm consists in its liability to easy to fall into local optimum, but the opposition-based learning strategy is effective in dealing with the weakness. However, to improve the global search ability by using the opposition-based strategy it is necessary that in the late algorithm other strategies are combined to opposition-based strategy. To improve this, this dissertation improves the opposite process of the opposition-based strategy according to the refraction principle of light, and proposes the unified model of opposition-based learning (UOBL) and the improved particle swarm optimization algorithm based on the opposition-based model of the principle of refraction (refrPSO). The experiment results show that the model improves the global search ability of the refrPSO algorithm more effectively compared with other particle swarm algorithm based on opposite learning and the diversity of the population. Thanks to these improvements, the refrPSO enhances the convergence speed and the optimization accuracy.(3) The key disadvantage of PSO is easy to fall into local optima for solving complex problems. In other words, the local search capability of PSO, also called the exploitation ability, is weaker, which may result in lower convergence accuracy and slower convergence speed. To battle these dismerits, extremal optimization, a great local search algorithm, is introduced to enchance the exploitation of PSO. Meanwhile, opposition-based learning, a great global search strategy, is introduced to enhance the exploration of PSO further. Hence this dissertation proposes an adaptive hybrid PSO based on combination of extremal optimization and opposition-based learing. The improved PSO takes advantage of both so as to balance the exploitation and exploration of PSO, which leads to performance enhancement of PSO such as convergence speed and accuracy. The experiment results show that the impoved PSO has the excellent optimization performance.(4) Because of some disadvantages of the traditional design methods for FIR digital filters such as accurate determination of the boundary frequency, liability to fall into the local optima, this dissertation introduces several improved PSO mentioned above into FIR digital filter design. Compared with the serveral methods for FIR digital filter design, the experiment results show that the improved PSO algorithms enjoy more excellent optimization performance, and further improve the performance of FIR digital filters.
Keywords/Search Tags:Intelligent Algorithms, Particle Swarm Optimization, Opposition-based Learning, Extremal Optimization, FIR Digital Filters
PDF Full Text Request
Related items