Concentration Prediction from Oscillopolarographic Signals via Hidden Information Detector

A practical and systematic scheme of concentration prediction of chemical substance from oscillopolarographic signals is presented, and a hidden information detector (HID) — put forward in our previous work — is employed. The HID consists of two connected neural networks called cascade neural networks, in which the first one extracts main features of signals by employing a modified DOG (difference of Gaussian) wavelet basis, and the second one predicts concentration from the extracted features by using a simple back-propagation neural networks in which a dipole sigmoid basis function is utilized instead of the traditional one. The application in predicting concentration of two kinds of chemical substance indicates the good quality of our scheme, which can be easily generalized to deal with solution containing three or more kinds of chemical substance.