Intelligent pattern recognition by feature selection through combined model of DWT and ANN

Abstract

This paper presented a combined model of Discrete Wavelet Transform(DWT) and Self-Organizing Map(SOM) to select features from irregular insect’s movement patterns. In the proposed method, the DWT was implemented to characterize different movement patterns in order to detect behavioral changes of insects. The extracted parameters based on combined model of DWT and SOM were subsequently provided to artificial neural networks to be trained to represent different patterns of the movement tracks before and after treatments of the insecticide. Finally, the proposed combined model of DWT and SOM was able to point out the occurrence of characteristic movement patterns, and could be a method for automatically detecting irregular patterns for nonlinear movements.