Navy SBIR Topic No. N92-081 solicits proposals for exploratory development of methods for processing active sonar returns which will permit earlier classification of submarines. In our opinion, the most promising approach for rapid advancement in this area will focus on the application of improved decision making and signal interpretation algorithms to currently available signal processing technology. This proposal is for a program of basic research and exploratory development to develop an improved signal interpretation algorithm for classifying a digitally processed acoustic signal based on an approach known as linear programming (LP) discriminant analysis. It is proposed the LP discriminant analysis, as a pattern classification technique, be applied to the pattern vector obtained by digitizing the spectral shape of a sonar signal. LP discriminant analysis offers several advantages over classical discriminant analysis including freedom from the classical statistical assumptions and significantly enhanced classification power. These properties make the technique well suited to pattern recognition applications including the training of neural networks.
