Enhanced Multiple Sound Event Detection and Classification Using Physical Signal Properties in Recurrent Spiking Neural Networks
Date
Supervisor
Item type
Journal Article
Degree name
Journal Title
Journal ISSN
Volume Title
Publisher
Institute of Electrical and Electronics Engineers (IEEE)
Abstract
Sound event detection and classification present significant challenges, particularly in noisy environments with multiple overlapping sources. This paper introduces an innovative architecture for multiple sound event detection and classification utilizing recurrent spiking neural networks (SNNs). Our method uniquely leverages temporal data to detect and classify multiple sound sources simultaneously, integrating the physical concept of signal power matching with neuronal output power and employing a binaural strategy to enhance detection accuracy in real-world scenarios. The architecture processes spatiotemporal data to dynamically update synaptic weights, enabling precise identification of sound event categories and their occurrences. Our simulations reveal substantial performance improvements, achieving the highest precision of 73% in classification tasks, including multilayer perceptrons (MLP), convolutional recurrent neural networks (CRNN), and recurrent neural networks (RNN). Statistical analysis indicates that these improvements are significant (p-value < 0.05). These findings suggest practical applications in various fields such as surveillance, autonomous vehicles, and smart home systems, where robust sound event detection is critical.Description
Source
IEEE Access, ISSN: 2169-3536 (Print); 2169-3536 (Online), Institute of Electrical and Electronics Engineers (IEEE), PP(99), 1-1. doi: 10.1109/access.2025.3563346
Publisher's version
Rights statement
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/