基于优化的LMS算法的去噪原型研究.docVIP

  A De-noise Prototype Research Based on Improved LMS Algorithm Chu Weiyan, Yang Wenchuan** 5 10 15 20 25 30 (School Of Information and Communication Engineering,Beijing University of Posts and Telecommunications，Beijing 100876) Abstract: How to solve the problems arouse by noises has become more and more important. During the practical telecommunication information processing, there are enormous noise brought by the ambience, which affect the quality of communication badly. Linear Prediction is the most effective method in the processing of information, while it is the most easily affected by noises. We adapt adaptive filter to intensify the valuable signal, analyze the traditional LMS algorithm and bring forward a new algorithm to solve the conflict between convergence rate and stability error by choosing a different fixed step. The experiment results show that the algorithm is stable and credible Key words: Linear Prediction; LMS algorithm; Convergence; Telecommunication 0 Introduction For the telecommunication information processing, noise data is usually mixed with important data by surroundings and transmission medium. Also communication equipments and even other lecturers disturb produce noise data. These disturb will make the information polluted by noise information. If suffered by strong background noise disturb, communication quality will get serious influence [1]. So, we must introduce some strengthening system of information in order to raise the communication quality of information, and try to maintain systematic performance[2]. The types of noise in reality may be approximately classified and listed as: periodicity noise; swash noise; broadband noise; information disturb. The major method that reduces broadband noise has three kinds. We often use nonlinear algorithm to handle and reduce spectrum and offsets adaptively. Main nonlinear algorithms include cutting wave or waveform alternate. Adaptive filter may divide into linear adaptive filter and