Multivariate calibration model from overlapping voltammetric signals employing wavelet neural networks

This work presents the use of a Wavelet Neural Network (WNN) to build the model for multianalyte quantification in an overlapped-signal voltammetric application. The Wavelet Neural Network is implemented with a feedforward multilayer perceptron architecture, in which the activation function in hidden layer neurons is substituted for the first derivative of a Gaussian function, used as a mother wavelet. The neural network is trained using a backpropagation algorithm, and the connection weights along with the network parameters are adjusted during this process. The principle is applied to the simultaneous quantification of three oxidizable compounds namely ascorbic acid, 4-aminophenol and paracetamol, that present overlapping voltammograms. The theory supporting this tool is presented and the results are compared to the more classical tool that uses the wavelet transform for feature extraction and an artificial neural network for modeling; results are of special interest in the work with voltammetric electronic tongues.