An optimization method of signal de-noising in discrete wavelet transform based on generalized cross-validation

Xiaojing Chen,Ke Liu, Peng Ye

College of Physics and Electronic Engineering Information, KeyLaboratory of Low-voltage Apparatus Intellectual

Technology ofZhejiang, Wenzhou University, Chashan University Town,Wenzhou,Zhejiang Province, 325035, P.R. China

A method for automatically selecting the asymptotical optimal parameters is presented for signal de-noising in discrete wavelet transform. The parameters of wavelet de-noising were first encoded. A generalized cross-validation algorithm was then used to select these parameters automatically. The parameters that obtained the smallest generalized cross-validation were asymptotically optimal. Simulation signals with different features and range signal-to-noise ratios were used to demonstrate the optimality of the proposed method. In addition, the Raman spectrum of edible oil and nuclear magnetic resonance spectrumof quinine and Boc-protected prolinewere employed as real-world data to validate the proposed method. The proposed method achieved superior performances in both real-world data and in artificial simulation.