2026-05-13T16:14:48.896Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:3:11402 2025-05-08T17:28:13Z 3

Exploring the Prospect of Hybridizing Mel Spectrograms and Neural Networks for Acoustic Speed Violation Detection, Chapter in LNNS Lecture Notes in Networks and Systems: SoCPaR 2023: International Conference on Soft Computing and Pattern Recognition, Springer, volume 1246 2025 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/3/11402 https://link.springer.com/chapter/10.1007/978-3-031-88992-9_3 M. Stankovic L. Jovanovic A. Bozovic N. Budimirovic M. Zivkovic N. Bacanin Enforcement of the vehicle’s acceleration limitations is vital for ensuring safer roads. This study delves into an avant-garde methodology that synergizes signal processing techniques with convolutional neural networks (CNN) for identification of the speeding breaches, tackling a vital facet of the traffic management and control. Despite the existence of the traditional approaches that are efficient, this kind of hybrid approach that combines signal processing with AI models is still largely unexamined. This study tries to overcome this gap by utilizing Mel spectrograms gathered from audio recordings of vehicles, containing intricate acoustic attributes. These spectrograms are fed to the input of a customized CNN structure, carefully developed for recognizing patterns in audio recordings associated with speeding. The introduced approach targets discrimination among regular driving sounds and cases of speed limit violations. This hybrid approach delivered very favorable initial outcomes, showcasing the potential for accurate identification of speed violation cases. The contributions of this research enhance traffic safety and road management, and at the same time blaze a trail for integration of signal processing with AI models, implying possible extension to wider audio analysis areas. bookPart Springer, Cham 26 38 10.1007/978-3-031-88992-9_3 LNNS Lecture Notes in Networks and Systems: SoCPaR 2023: Proceedings of the 15th International Conference on Soft Computing and Pattern Recognition, volume 1246