Research And Implementation On Algorithm Of Adaptive Digital Filters

Adaptive filter is one of the Most popular problems of today’s digital signal processing fields. With the development of computer science and the theory and method of signal processing, we need to deal with more and more data, at the same time, the demand of accuracy become more and more high across the signal processing. The basic idea of adaptive filter is to use the result of the last moment to change the coefficient of the next moment, in order to adjust the unknown input signals and obtain the best output. The aim of the adaptive filter is always the uncertain information process or system. Because of the strong abilities of tracking and learning, it was used for the processing of voice signal, system identification,linear prediction, adaptive equalization, noise elimination and so on since the 1960 s.The basic algorithms of adaptive filter are the least mean square(LMS) algorithm,recursive least squares(RLS) algorithm, and some improved algorithms, just as normalized least square(NLMS) algorithm, variable step size algorithm and so on. Different filter algorithm is suitable for different occasions, they have their own characteristics and advantages. The adaptive filter performance is the key of the research, because it related to the output result. The LMS algorithm is widely used because of the its small amount of calculation, the advantage of good stability. But the traditional algorithm has its disadvantage,the problem of slowly convergence. In this paper, we will put forward a kind of algorithm based on the traditional LMS algorithm that will change the variable step size based on the ACC(n) of error signal and input signal. It will change the adaptive size independently in order to track the input signals.The simulation results show that the improved algorithm of this paper is to achieve fast convergence, at the time, it also win the good steady-state misadjustment characteristic, and has perfect anti-interference ability to noise, compared with the traditional variable step size LMS algorithm, and at the end, it has more excellent performance.