Spectral Signal Denoising Method Based on Improved Wavelet Threshold Function

Abstract: In the quantitative analysis of spectral data, the noise often affects the accuracy of analytic results. In order to improve the analytic precision of the spectral data, the spectral signal collected with CCD need to denoise. According to the characteristics of spectral signal and combining with the advantages of the hard threshold function and the soft threshold function, we propose an improved threshold function denoising algorithm(ITFDA), which can not only overcome the discontinuity of hard threshold, but also overcome the constant error of soft threshold. The simulation experiments conducted with the improved threshold function are compared with the hard and soft threshold in terms of Signal to Noise Ratio(SNR) and Mean Square Error(MSE). The results show that the proposed denoising algorithm with improved threshold function can remove noise from the spectral signal in a certain extent, retain the important characteristics spectrum - peak of spectral signal, and reconstruct the spectral signal very well.