Automatic detection of tropical fish calls recorded on moored acoustic recording platforms

Passive acoustic recording of biological sound production on coral reefs can help identify spatial and temporal differences among reefs; however, the contributions of individual fish calls to overall trends are often overlooked. Given that the diversity of fish call types may be indicative of fish species diversity on a reef, quantifying these call types could be used as a proxy measure for biodiversity. Accordingly, automatic fish call detectors are needed because long acoustic recorders deployments can generate large volumes of data. In this investigation, we report the development and performance of two detectors—an entropy detector, which identifies troughs in entropy (i.e., uneven distribution of entropy across the frequency band of interest, 100–1000 Hz), and an energy detector, which identifies peaks in root mean square sound pressure level. Performance of these algorithms is assessed against a human identification of fish sounds recorded on a coral reef in the US Virgin Islands in 2013. Results in...