A Kind Of Auto-suitable Blind Separation Algorithm Containning Noise Based On Infomax Critertion

Blind singnal processing is new field that is quickly developed in recently years. It is a important development direction in the questi–on of signal restoration and also is a hot direction in intelligent computation. The development in signal processing field is promoted by it. It plays an important role in many fields such as Biology, Medical,Science, Economics, Engineering and so on. At present,one of research difficulties is blind signal separation containing noise, until recently, the result are still less.This paper mainly disscuss this kind of separation which is added a special additional gaussian white noise. The whole algorithm can be seperated in two steps: whitening the noise separattion of the composited non-gauss signal.In the second chapter, a objective function is established that can optimize the algori- thm based on the information theory related to the knowledge of the information transmission maximization criterion and discuss the concrete setup procedure.In the third chapter,the paper introduces firstly the survey of several kind of albinism denoising algorithm.then proposes an auto-adaptive algorithm including the information item of the noise and proves its stability, and indicates its effect through the simulation testing.In the fourth chapter, the paper designs an kind of auto-adaptive algorithm that it can be able to effectively separates the mixed situation of sub and super Gauss signals based on the criterion of the natural gradient's information maximization (Infomax) in order to improve the insufficiency of certain algorithms that can only separate one kind of situation before, and analies its stable condition and finally gives the simulation testing.