Spiking Neural Network for cetacean survey

Cetacean surveys are important to assess the state of marine biodiversity, but are energy and time consuming. In the past years, Passive Acoustic Monitoring (PAM) became a key element of these surveys since it allows the obtention of information about cetaceans hidden deep beneath the water surface [1]. However, processing all the audio files acquired during the PAM campaigns remains time and energy consuming, and there is a need for an automated and low power process for cetacean sounds classification [2]. Spiking Neural Network (SNN) could be one of the possible solutions to this challenge [3]. We present a simple SNN architecture for transient marine mammal sounds classification, which, together with an analog artificial cochlea model, could be used to make an embedded neuromorphic ultra-low power sound classification device.

[1] Zimmer, W. M. (2011). Passive acoustic monitoring of cetaceans. Cambridge University Press.

[2] Bittle, M., & Duncan, A. (2013). A review of current marine mammal detection and classification algorithms for use in automated passive acoustic monitoring. In Proceedings of Acoustics (Vol. 2013, pp. 1-8). Victor Harbor, SA: Australian Acoustical Society.

[3] Baek, S., & Lee, J. (2024). Snn and sound: a comprehensive review of spiking neural networks in sound. Biomedical Engineering Letters, 14(5), 981-991.