Research on the De-Noising Algorithm

Abstract

to confirm the useful information better, the changing magnitude of the value must be considered. Secondly, the effect of filter is related to the choice and the input parameters of filter, but it is so difficult to confirm the choice and the input parameters of filter due to the complicated wave of the real time series. So this de-noising method will be complicated to apply. According to this shortage, here proposes a kind of de-noising method based on event. This method directly utilizes the mathematical features of curve such as the extremum slope and curvature to de-noise. The principle of this method is simple and easy to operate. In order to prove the validity and advantage, Here first it researches on the de-noising method based on Fourier transform and makes experiment based on real time series of the stock with two methods. The result proves that the new method can de-noise effectively, at the same time it is easy to operate and can keep the original characters of the time series furthest. II、 DE-NOISING METHOD BASED ON FOURIER TRANSFORM Fourier transform is one of the basic and common-used signal processing method. If let {at :t=…,-1,0,1,…} denote one infinite series of real value variables, then Fourier transform can be defined as complex value function below: