NEXT GENERATION ALGORITHMS FOR UNATTENDED GROUND SENSORS

摘要

Non-linear signal processing algorithms, developed originally by R.A. Wagstaff for ocean acoustics and named "AWSUM(K)," have been applied in this laboratory to atmospheric acoustic signals measured in the presence of severe intermittent noise. It has been found that the AWSUM(K) processors, which filter out strong signals and pass weak signals, can provide dramatic gains in the signal-to-noise ratio for a steady sinusoidal signal strongly degraded by intermittent manmade noise and intermittent wind noise. Further, applications of the processors to field data have shown a number of systematic behaviors that so far have not been explained or understood quantitatively. This article presents a theoretical analysis of the AWSUM(K) processors for a steady sinusoidal signal in the presence of exponential (Rayleigh) noise and intermittent (non-Rayleigh) noise. The theory quantitatively explains the observed behaviors of the AWSUM(K) processors. In particular, it is shown that in the limit of large sample number, the AWSUM(K) gain in the (signal+noise)-to-noise ratio is independent of the sample number and processor order (for K≥2). For a steady sinusoidal signal, the gain is determined solely by the shape of the noise distribution function near zero. For Rayleigh noise, for example, the gain is given by exp(SNR)/(SNR+1), where SNR is the usual linear signal-to-noise ratio (e.g., SNR = 1 corresponds to a signal-to-noise ratio of 0 dB). For intermittent manmade noise and intermittent wind noise, the measured noise distribution function is strongly peaked near zero, so that gains in approaching 20 dB are predicted, even for small values of SNR. The predictions are in accord with field data from an atmospheric sound propagation experiment.