Abstract: A multivariate calibration method was used for the data analysis of trace elements determination, which is based on the combination of wavelet packet decomposition and neural network. Firstly,the differential pulse cathodic stripping voltammetry (DPCSV) signal was decomposed by wavelet packet, and the eigenvector was extracted based on the power features which are distributed in different frequency bands. Then a three layers BP network was used for training the features. The forecast results and calculated results of samples showed that the method has strong anti-jamming capability, high measure accuracy and practical values.

Key words: wavelet packet analysis, BP neural network, electrochemistry, simultaneous determination, trace element