Atmospheric Observation Data De-Noising Based on a New Wavelet Threshold Function

Atmospheric observation data are unavoidable mixed with noise and the performance of the data processing system will degrade remarkably. Wavelet threshold de-noising techniques is an important method to reduce noise in signal process. In this paper a new wavelet threshold function is proposed for removing noise in signal, which is based on the conventional soft and hard threshold functions. Both quantities of signal to noise radio (SNR) and mean square error (MSE) are introduced to judge the de-noising effectiveness. Some precipitable water (PW) time series data are processed by combined use of the new threshold function with Kolmogorov entropy and neural network prediction which are used to judge the de-noising effectiveness. The study results show that the new threshold function proposed can overcome the shortcoming of the conventional soft and hard threshold functions to a certain degree and is efficient for signal de-noising.