Automatic grunt detector and recognizer for Atlantic cod (Gadus morhua)

Northwest Atlantic cod (Gadus morhua) have been heavily overfished in recent years and have not yet recovered. Passive acoustic technology offers a new approach to identify the spatial location of spawning fish, as well as their seasonal and long term persistence in an area. To date, the lack of a species-specific detector has made searching for Atlantic cod grunts in large amounts of passive acoustic data cumbersome. To address this problem, an automatic grunt detection and recognition algorithm that processes yearlong passive acoustic data recordings was designed. The proposed technique is a two-stage hypothesis testing algorithm that includes detecting and recognizing all grunt-like sounds. Test results demonstrated that the algorithm provided a detection probability of 0.93 for grunts with a signal-to-noise ratio (SNR) higher than 10 dB, and a detection probability of 0.8 for grunts with the SNR ranging from 3 to 10 dB. This detector is being used to identify cod in current and historical data from U.S. waters. Its use has significantly reduced the time required to find and validate the presence of cod grunts.