Optimization Design Of Digital Filter Based On Evolutionary Algorithm

Currently, the filter is an indispensable part in digital signal processing. It has been widely used in communication technology, image recognition, military radar, aerospace field, medical facility and voice pronunciation, etc. Excellent ability of signal processing has a growing market application value.To make analog filter designed with much more accuracy or meeting several technological specifications, some difficulties may arise such as complex design process, numerous components and extended structure. What’s worse, it will not necessarily succeed. As the development of digital computer technology and large scale integrated circuit, digital filter can be applied with real-time implementation in computer software and large scale integrated digital hardware. Although its rapid development has provided the convenience to the design and application of digital filters, people have to work more on the functional improvement of digital filter because common design cannot support its increasingly complex structure. In a word, computers are more and more important in designing filters.In recent years, researchers has put forward various methods of calculation in the design of digital filters, including genetic algorithm, simulate anneal arithmetic, taboo search, ant colony optimization, neural network algorithm, immune clonal selection algorithm, immune clonal selection algorithm, etc. All these methods determines the transmission coefficient first, takes digital filter’s structure into consideration then, only promising the best coefficient itself, but not the best one in the whole process.This thesis mainly proposes a design method based on evolutionary algorithms. The final digital filter is mostly optimized by optimizing filter structures and filter coefficients with difference algorithm and step length variation algorithm. The method is able to be used to directly design digital filers according to given specifications without identifying the coefficients of the transfer function.The research is conducted as follows:(1) It reviews the design methods of digital filters and focuses on evolutionary design methods.(2) It designs digital filter based on genetic algorithm. Meanwhile, the paper probes into crossing-over rate and mutation rate, which decides the performance and efficiency of genetic algorithm, and concludes the parameter valves of genetic algorithm.(3) The research optimizes filter coefficients with difference algorithm and step length variation algorithm. The results show that this method can improve the stopband attenuation clearly compared with the optimized structure result.(4) Finally, the thesis summarizes the whole research and forecasts future development.It can be testified through experiment that designing digital filter with evolutionary algorithms can meet the specification. Direct design of filter structures in accordance with target property without limitation of filter order, transfer function coefficient, has a wide range of applicability and suitable in other optimizing cases.