A combined hardware/software optimization framework for signal representation and recognition

Abstract

This paper describes a signal recognition system that is jointly optimized from mathematical representation, algorithm design and final implementation. The goal is to exploit signal properties to jointly optimize a computation, beginning with first principles (mathematical representation) and completed with implementation. We use a BestBasis algorithm to search a large collection of orthogonal transforms derived from the Walsh-Hadamard transform to find a series of transforms which best discriminate among signal classes.The implementation exploits the structure of these matrices to compress the matrix representation, and in the process of multiplying the signal by the transform, reuse the results of prior computation and parallelize the implementation in hardware. Through this joint optimization, this dynamic, data-driven system is able to yield much more highly optimized results than if the optimizations were performed statically and in isolation. We provide results taken from applying this system to real input signals of spoken digits, and perform the initial analyses to demonstrate the properties of the transform matrices lead to optimized solutions. © Springer-Verlag Berlin Heidelberg 2007.