Automatic High-Frequency Click Detection Algorithm for Cetaceans
Journal
2025 IEEE Underwater Technology, UT 2025
Part Of
2025 IEEE Underwater Technology, UT 2025
Start Page
1
End Page
5
Date Issued
2025-03-02
Author(s)
Abstract
The expansion of offshore wind power facilities in Taiwan's western waters poses significant threats to marine cetaceans, particularly the finless porpoise (Neophocaena phocaenoides sunameri), which is the most frequently stranded cetacean in the region. Causes of stranding include fishery bycatch, environmental changes, vessel collisions, and marine engineering activities. Traditional visual surveys are ineffective for monitoring this species due to its lack of a prominent dorsal fin, limiting the collection of ecological data and the development of conservation strategies. To overcome these challenges, we developed an automated detection algorithm for high-frequency (HF) cetacean clicks using underwater passive acoustic monitoring (PAM). The first algorithm utilized a time-domain computational approach, applying a high-pass filter (>110 kHz) and a peak detection algorithm to identify click trains, incorporating the Peak Energy Algorithm (Step A). The second algorithm introduced a band-pass filter (60-192 kHz) and combined Step A with a Signal-to-Noise Ratio (SNR) calculation (Step B). Additionally, two filters were integrated into the algorithm, informed by studies conducted in the Bohai Sea, Zhangzhou, and Hainan, China, to refine click event detection. Analysis of 7.5 hours of acoustic data from the Yun-Chang Rise validated the algorithm. The first algorithm achieved a precision ratio (PR) of 78.09% and a false positive ratio (FPR) of 8.06%, with misclassifications attributed to broadband pulses or closely spaced clicks. The second algorithm, incorporating the SNR threshold, improved PR to 96.3% (+17.4%) and reduced FPR to 0.89% (-7.17%). This algorithm enables efficient HF cetacean click detection and can be integrated into firmware for real-time marine operations. It provides a promising solution to the limitations of visual surveys, enhancing acoustic pattern recognition, and supporting improved cetacean monitoring and conservation in Taiwan's waters.
Event(s)
2025 IEEE Underwater Technology, UT 2025, Taipei2 March 2025through 5 March 2025. Code 208171
Subjects
Automatic Detection Algorithm
Click detection
Finless porpoise
ICI (Inter-Click-Interval)
Passive Acoustic Monitoring (PAM)
Peak Detection Algorithm
Signal to Noise Ratio
Publisher
IEEE
Type
conference paper