A Neuromorphic Trigger for Efficient Audio Event Detection

This paper presents a neuromorphic trigger for audio event detection utilizing a spiking neural network (SNN) to efficiently process continuous audio streams. The lightweight SNN selectively gates input to downstream models, achieving a one-second segment-based F1 score of 0.97 for anomalous sound detection on the URBAN-SED dataset and demonstrating a 42.6× reduction in FLOPs for sound event detection when combined with the Dang classifier on the DCASE 2017 Challenge Task 2 dataset. This approach significantly enhances real-time processing efficiency and reduces computational costs, making it relevant for practitioners developing resource-constrained audio analysis systems.