[Simultaneous determination of Sm( III) and Y(III) by spectrophotometry with a wavelet packet transform latent...

A wavelet packet transform latent variable regression (WPLVR) method was developed to perform simultaneous quantitative analysis of Sm(III) and Y(III). The quality of the noise removal was improved by combining wavelet packet transform with latent variable regression (VLR). Wavelet packet representations of signals provided a local time-frequency description, thus in the wavelet domain, the quality of the noise removal can be improved. The latent variables were made by projecting the wavelet packet processed signals onto orthogonal basis eigenvectors. The latent variable is expressible in term of linear combination of the original signals. By this method one can obtain highly selective information from unselective full-spectrum data. Through optimization, the wavelet function and wavelet packet decomposition levels (L) were selected. Two programs, PWPLVR and PFTLVR, were designed to perform WPLVR and Fourier transform latent variable regression (FTLVR) calculations. Experimental results showed that both methods were successful, but the WPLVR methed was better than FTLVR.