Adaptive Recognition of Random Binary Signals

摘要

The problem of this study is the detection of signals that are neither periodic nor of constant waveshape in the presence of white noise. It is assumed that the signals are initially known, that they are of a binary form, that they are modified at random times by the complementing of one ran domly chosen bit, and that they occur at random times. An adaptive matched filter based upon the link connection is proposed. The link connection is selected because of the relative simplicity with which it can perform cross-correlations and deliver samples of its input, and because of the large bandwidth it can provide. Several methods for estimating the input-signal waveform from its samples, including finite averaging, finite summation, and RC averaging, are discussed. Miss and false-alarm probabilities for the system are computed, estimated, or bounded, and its general behavior, from initial matching to noise-induced failure, is described. A digital computer program for the simulation of the adaptive system is described with flow charts. (Author)