Wavelet transform use for signal classification by self-organizing neural networks

Detection and classification of signal components belongs to very common problems in various engineering, economical and biomedical applications. To recognize groups of similar input vectors it is possible to use various methods of cluster analysis and mathematical models of neural networks studied in the paper as well. Characteristic signal features forming network patterns during the learning stage can be based either upon signal segments identification and modelling or its frequency components analysis. The paper is devoted to the use of the discrete wavelet transform (DWT) for their evaluation providing an alternative to the commonly used discrete Fourier transform (DFT). Fundamentals of wavelet analysis and signal decomposition with different resolution both in the time and frequency domains are presented at first. Resulting signal features are then used for signal classification. The method is verified for simulated signals and then applied for a given encephalogram (EEG) signal classification.