Abstract:

A wavelet transform and a wavelet neural network are discussed. The features of frequency specturm analysis at the different resolutions of the dyadic wavelet transform are extracted by using the properties of its better multiresolution analysis. These features are fed into the wavelet neural networks as the input patterns for training and classifying. The experiments are conducted to demonstrate the feasibility of the proposed method.

Key words: feature extraction, neural- net work, composite materials, damage-detection