WPTER: wavelet packet transform for efficient pattern recognition of signals

In the present work, we propose a novel algorithm based on the Wavelet Pack et Transform (WPT) for pattern recognition of signals, which operates both feature selection and classification at the same time: Wavelet Packet Trans form for Efficient pattern Recognition of signals (WPTER). The distinctive characteristics of WPTER with respect to the previously proposed algorithms for the WPT-based classification of signals consist mainly of two aspects: (1) a Classification Ability criterion is introduced into the procedure fo r selection of the best discriminant basis; (2) the signals are reconstruct ed in the original domain by using only the selected wavelet coefficients, which 1 allow for chemical interpretation of the results. The algorithm was First tested on an artificial (simulated) set of signals, consisting of a number of subsequent peaks, partially overlapped to each o ther, with added noise and baseline drift, simulating a three-class system. Then, it was applied to a data set consisting of X-ray diffractograms on f ired tiles subjected to different firing cycles, aiming at discriminating t he different firing methods on the basis of the phase composition. In both cases, satisfactory classifications were achieved. (C) 2001 Elsevier Scienc e B.V. All rights reserved.