A modified Eckart post-beamformer filter for improved detection using broadband features

The problem of signal detection using passive sonar signal processing methods to extract features is addressed. We propose a modified Eckart post-beamformer filter to obtain better estimates of the signal and in particular the broadband component of the signal. The physical processes motivating the study are broadband features which are usually ignored in conventional narrowband detection and classification processors. Broadband features can arise from various sources like transients, propulsion mechanisms, cavitation, and hydrodynamic flow and are often present even in the absence of narrowband features. We separate the narrowband and broadband components of the signal and extract the broadband features from the noise normalized signal components which has high energy and a wide frequency extent compared to the center frequency. Preliminary results suggest that broadband based detection and classification performance outperforms methods that use only narrowband features.