Deterministic signal processing of acoustic array outputs

Abstract

Acoustic signal processing currently relies heavily upon probabilistic assumptions concerning data and upon statistical properties of data to develop efficient processing principles. In contrast, deterministic signal processing uses only the actual sensor output amplitudes over a finite time (a data set) to estimate signal waveform and input power for that data set, employing no probabilistic assumptions or statistically based techniques. The major result presented is a demonstration of the improvement in output S/N and in signal waveform and power estimation of deterministic processing versus probabilitistic averaging under conditions of a priori unknown signal waveform and input power, very low input S/N, short processing time and small number of sensors. For example, with input S/N in the region of −15 to −30 dB, order of 3 to 10 sensors, and 5 to 20 time sampling points, improvements in output S/N on the order of 5 to 15 dB are achieved. These results are preceded by a summary of relevant deterministic properties of data sets and derivation of the deterministic algorithm used. Areas of further necessary work and implementation concepts are discussed. [Work supported by NUSC.]