[Application of a modified method of wavelet noise removing to noisy ICP-AES spectra]

A new method for noise removal from signal by the wavelet transform was developed. Compared with analytical signal, noise has higher frequency and smaller amplitude. By the new wavelet filtering method, the high frequency components were first removed, and then the small ones in the remaining transformed vectors were discarded. The proposed approach was evaluated by the processing of simulated and experimental noisy ICP-AES spectra. Different amounts of noise were added to a Gaussian peak to obtain a series of noisy ICP spectra. The simulated noisy spectra with R (signal to noise ratio) = 6 and N (data number) = 51, and with R = 6 and N = 17 were used to illustrate the feasibility of the proposed method. The performances of noise removal by the wavelet smoothing, the wavelet denoising and the proposed technique were compared. It was found that using the new approach, the relative errors of peak height would be no more than 5% for spectra with normal sampling points and R > or = 2. Moreover, the baseline could be easily defined, which was helpful to the accurate measurement of peak height. Experimental spectra of Al and V at low concentrations were processed by the proposed method. Intense noises were efficiently removed and the spectra became smoother without underestimating the analytical signal. The distortion of V 303.310 nm line was substantially rectified. The linear correlation coefficients between the peak heights in the reconstructed spectra and the concentrations were found to be 0.9953 for Al and 0.9836 for V, respectively.