Research on the De-Noising Algorithm

Signal processing is the most popular method for time series de-noising. But with this method, valued information with high frequency but high swing is also thrown away. This is because signal processing method just takes frequency into consideration. Furthermore, the selection of a filter and its input arguments is also difficult. To avoid the above shortages of signal processing method, de-noising algorithm based on an event is proposed. The algorithm utilizes curve features as extremum, slope and curvature to achieve de-noising. Advantages of the method are illustrated by the comparison between de-noising method based on Fourier transform and the new method through principle research and experimentations.