Unsupervised learninig and detection of pulse signals

This Paper describes an adaptive learning system for the estimation and detection of a pulse type signal of known duration but unknown shape or epoch, randomely occuring in white Gaussuion noise. A combination of energy detectio and cross correlation is used to build up estimates of the signal. Result are presented for different signal shapes showing that error rate approaches that of the ideal matched filter. the system is intended for on-line real-time implementation. the system has potential applications to medical signal processing.