Research on biochemical spectrum denoising based on a novel wavelet threshold function and an improved translation-invariance method

Abstract

In this paper, an improved wavelet threshold denoising with combined translation invariance(TI)method is adopted to remove noises existed in the bio-chemical spectrum. Meanwhile, a novel wavelet threshold function and an optimal threshold determination algorithm are proposed. The new function is continuous and high-order derivable, it can overcome the vibration phenomena generated by the classical threshold function and decrease the error of reconstructed spectrum. So, it is superior to the frequency-domain filtering methods, the soft- and hard-threshold function proposed by D.L. Donoho and the semisoft-threshold function proposed by Gao, etc. The experimental results show that the improved TI wavelet threshold(TI-WT) denoising method can availably eliminate the Pseudo-Gibbs phenomena generated by the traditional wavelet thresholding method. At the same time, the improved wavelet threshold function and the TI-WT method present lower root mean-square-error (RMSE) and higher signal-to-noise ratio(SNR) than the frequency-domain filtering, classical soft and hard-threshold denoising The SNR increasing from 17.3200 to 32.5609, the RMSE decreasing from 4.0244 to 0.6257. Otherwise, The improved denoising method not only makes the spectrum smooth, but also effectively preserves the edge characteristics of the original spectrum.