DWT-Based Pattern Recognition Systems

摘要

In this paper the discrete wavelet transform (DWT) is combined with eigenspace projection and Fisher discriminant analysis to form pattern recognition (PR) systems for the classification of 1-D signals. In the first system, the DWT is combined with eigenspace projection while in the second system the DWT is combined with Fisherspace projection. The performances of both systems are evaluated using 4 classes of 1-D ammunition Doppler radar signals with each class consisting of 5 signals. Though the systems are demonstrated on 1-D signals, they are equally suitable for use in the PR of 2-D images in computer vision and image processing applications.