A variety of both passive and active diver detection sonars have been developed for harbor underwater security applications. Diver classification is one of the most challenging processing problems for such sonars. Quasi-periodic breathing sounds are known to be a reliable classification feature for SCUBA diver detection and is often utilized in passive sonars. This paper discusses the possibility of automatically classifying diver targets via breathing event features analysis in both active and passive sonars. We show that breathing event features can be extracted from active multibeam sonar beamformer output — ping-to-ping waterfall image. Next, we consider the generalized approach to the breathing events periodicity estimation in both passive and active sonars. Periodicity of these events estimated via 2D filtering and FFT image processing, using the spectrogram image in the passive sonar or the ping-to-ping waterfall image in the active sonar.
展开▼