Abstract :
This paper discusses the automatic processing of data containing multiple signals. In particular, filters which modify their structures in order to recognize initially unknown waveforms in Gaussian noise and an unknown signal environment are investigated experimentally. The general structure, derived from decision theory, processes the data in nonlinear fashion and effectively sets up a narrow decision region about the estimate of the signal. The results demonstrate the possibility of using filters of the new type for automatically processing data although the relevant waveforms may be unknown {em a priori}.
