Improvement of wavelet threshold and application on thin-film thickness wideband monitoring system

During the course of noise reduction, wavelet threshold filtering algorithm can inhibite noise effectively, however Donoho threshold function hasn´t self-adaption between dimensions and results in signal details´ losing. Based on output signal of the thin-film thickness wideband monitoring system, it is processed by the improved Donoho threshold function whose threshold is added by micro-alignment factor, this factor makes wavelet coefficient threshold of the signal decrease, whose amplitude is similar to or less than niose one, this is beneficial to preserve wavelet coefficients of real signals; On the other hand, this factor makes that of noise increase, this is favorable to filter out noise. By experiment, it represents that both rejecting true probability and false declaration probability are reduced, the random noise is filtered well and the signal details are reserved perfectly in use of this algorithm, the peak error of signal is 0.7%~1.0%,the peak location one is 0.1%~0.3%,the system accuracy is improved.