Exploiting temporal coherence models to extend detection range

Abstract

Turbulent flow over acoustic sensors and other sources of noise can severely corrupt signals and reduce detection range. The ranges can be much greater when based on the temporal coherences of signals and noise instead of their magnitudes. The signals need not be perfectly coherent to be detected at longer ranges, just more coherent than noise. This is illustrated using a fluctuation‐based temporal coherence signal‐processing model. This model uses measured data to produce temporal coherence patterns that are random for noise, deterministic for signal, and a blend of the two for small, e.g., negative, signal‐to‐noise ratio (SNR). Fluctuation‐based processing techniques are presented for temporal coherence pattern generation, recognition, and automatic identification of signal presence or only noise in a frequency bin. It is shown for both underwater and atmosphere acoustic data that a signal’s influence in a frequency bin alters the usual random ‘‘noiselike’’ temporal coherence patterns by causing them to cluster. Hence, signal‐altered patterns can be recognized and the signals preserved. Corresponding ‘‘noise‐only’’ patterns, being random, can also be recognized and attenuated, or eliminated. Using these models results in increased SNR and automatic signal detection at longer ranges with greater confidence. [Work supported by ARDEC and SMDC.]