Coherent Signal Power Detector Using Higher-Order Statistics

摘要

A signal detection system divides a data sampling run into blocks and performs afast Fourier transform on each block, sorting or segregating results by frequency. Then certain combinations of results of the fast Fourier transform corresponding to each frequency are processed to derive a test statistic which is unbiased by Gaussian noise. An index look-up table is created by scanning through possible combinations of addresses and excluding those combinations of results of the fast Fourier transform which would be redundant over other combinations and would result in a contribution to biasing by noise. To allow detection for an accelerating or constant change of frequency with time (chirp) source or target, a lag index is derived for each combination of addresses in the index look-up table and the accumulation and averaging is done while sorting in accordance with the lag index. No knowledge of the characteristics of the signal to be detected is required and substantial information concerning the frequency behavior of the signal is derived in the course of detection. Detection is accomplished with increased sensitivity for a constant frequency and only slightly reduced sensitivity for a chirped signal as compared to known: spectrum analysis techniques which require such knowledge. Higher order embodiments allow increased numbers of channels to be correlated and different types of frequency variation with time.