A signal analysis expert system for signal noise reduction

A signal-analysis expert system (SAES) has been developed to address the problem of noise-editing pulse-position modulation (PPM) signals in preparation for analysis. These signals have historically required manual noise editing, or a combination of manual and semiautomatic editing techniques. Due to the nature of the signal, standard frequency-domain noise-reduction techniques are not applicable. The objectives for SAES were to provide a substantially improved tool for automatically noise-editing the signal and to provide intelligent assistance to the signal analyst. To meet these goals, SAES uses a combination of statistical pattern recognition, image processing, and expert system technologies. The result is a tool which reduces the noise editing workload on the human analyst by approximately 80% over existing methods while retaining accuracy comparable to a human editor working manually. Implementation on a parallel-processing transputer system provides real-time processing rates