Lattice architectures for signal expansion by Gaussian set wavelets with applications to recognition

Three issues are discussed. The first issue is the feasibility of nonorthogonal basis functions (BFs) for the representation of signals and images in particular. Novel BFs are suggested for signal expansion which are based on Gaussian sets (GSs) and Gaussian set wavelets (GSWs). Even though GSs are nonorthogonal, they are found to be quite efficient in the exploitation of local redundancies of signals. The second issue concerns a novel method of expansion for recognition applying template-similar functions as BFs. The results show significant improvement over traditional recognition methods. The third issue deals with hardware implementation of the above methods using adaptive lattice architectures that exploit the central limit