Temporal and spectral feature extraction from fish school using a broadband split-beam echosounder

Abstract

For sustainable fisheries, remote discrimination of fish species has been demanded recently. Features for fish discrimination should be extracted from schools. High spatial resolution of broadband split-beam echosounder system enabled tracking individual fish in a school. Temporal and spectral features such as waveform envelopes and target strength spectra were calculated individually. In addition, incident angles of sound were calculated by using positions and velocities of a fish, which were measured by the tracking process of the four channels split-beam system. Because the acoustical features from individual fish depended on the incident angles of sound, all feature parameters were sorted according to the angles to create temporal and spectral averaged pattern. In free ranging condition, echoes of Japanese jack mackerel (Trachurus japonicus), chub mackerel (Scomber japonicus), and red sea bream (Pagrus major) were measured. Results showed clear difference among temporal and spectral averaged patterns, which was consistent with test measurement obtained in an acoustic tank. Broadband split-beam system seemed to be appropriate to extract species specific feature in the ocean. [Work supported by the Research and Development Program for New Bio-industry Initiatives.]