Automatic detection and localization of croaker’s fish calls using beamforming

Abstract

Many kinds of fish, including croaker, produce species-specific low-frequency sounds associated with courtship and spawning. A recording system was used to monitor underwater fish call sounds. We have proposed the method to detect croaker calls from data measured by a single hydrophone. However, it was difficult to detect the desired calls at a high detection rate because of a low signal-to-noise ratio. We proposed the method using beamforming to improve the detection rate. At first step, fish calls are detected from data measured on one hydrophone using the previous method, which detects calls automatically using the acoustic features, that is, duration and periodicity. At second step, additional calls are detected by beamforming the four-channel data. At third step, detected calls are localized by using the time differences of arrivals. It was clarified that the detection rate using the proposed beamforming method was higher than that under the previous method with a single data channel. In addition, it was shown that fish calls could be localized from the acoustic data measured during several weeks. Therefore, this method could monitor sounds from croaker in the ocean.