Dolphin sounds detected amid snapping shrimp noise in Singapore waters
Dolphin broadband transient vocalization detection in impulsive ambient noise
Sound
Summary
Dolphins make quick clicking sounds that are hard to hear in noisy tropical waters full of loud snapping shrimp. The authors created a simple tool called TRIDENT that can pick out dolphin sounds from these noisy recordings. They trained TRIDENT carefully so it learns the timing patterns of dolphin clicks, not just loud noises. This tool works better than older methods and shows dolphins are more active during the day around different islands near Singapore.
What this means in practice
- •For marine ecologists: Identify dolphin presence and daily activity patterns using acoustic data in noisy tropical environments with snapping shrimp.
- •For environmental monitoring teams: Improve automated detection of dolphin broadband clicks in long-term underwater recordings where impulsive noise is common.
Authors
Hari Vishnu, Jin Xinhui, Mandar Chitre, Simen Hexeberg, Koay Teong Beng
Abstract
Passive acoustic monitoring in tropical waters such as those in Southeast Asia is hampered by the persistent, high-level impulsive noise of snapping shrimp. This is a particularly daunting challenge for detecting broadband transient vocalizations of dolphins - clicks, buzzes and burst-pulses. These vocalizations, unlike frequency-modulated whistles, share the broadband, impulsive character of the shrimp snaps and cannot be easily separated from them by frequency-localized or energy-based means. We develop TRIDENT, a lightweight waveform-based detector for these vocalizations, and apply it to 54.2 recorder-months of data from ten island sites around Singapore. We show that the training-data pipeline and the architecture are both important: controlled signal-to-noise-ratio based mixing to generate training data, appropriate normalization and hard negative mining, together steer the detector toward learning transient temporal structure rather than energy shortcuts. A single joint detector handles all three signal types better than specialized models for each, and outperforms established baselines. Applied to long-term recordings, it reveals tightly clustered, predominantly daytime activity, motivating days-with-detections as a robust indicator of cetacean presence.