2026-05-19T18:30:06.584Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:3:9817 2024-12-03T20:10:47Z 3

Speeding Classification by a Deep Learning Audio Analysis System Optimized by the Reptile Search Algorithm, Chapter in AIS Algorithms for Intelligent Systems: IJCACI 2023: International Joint Conference on Advances in Computational Intelligence, Springer 2024 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/3/9817 https://link.springer.com/chapter/10.1007/978-981-97-0180-3_7 T. Dogandzic A. Petrovic L. Jovanovic N. Bacanin A. Jovanovic M. Zivkovic Speeding is a dangerous traffic transgression as it causes a large portion of casualties in traffic. Traditional approaches based on radar and lidar are effective to some extent, but the room for improvement is noticeable. There is also a factor of cost with this type of equipment. With advancements in artificial intelligence (AI) techniques, the capabilities of autonomous, precise, and robust systems are becoming achievable. The role of AI is to analyze the input from video or audio recorders. For this work, the focus was on audio recordings. This research proposes a deep neural network (DNN) approach for vehicle speed optimization along with a metaheuristic approach for the optimization of the DNN hyperparameters. A swarm-based algorithm was chosen for hyperparameter optimization that has been regarded as an efficient approach for solving non-deterministic polynomial time (NP) hard problems. Chosen algorithm is the reptile search algorithm. The problems of hyperparameter optimization and the real-world problem of speeding classification belong to this group of problems. The model was compared to other DNN-metaheuristic high-performing solutions from which it was deduced that the proposed approach is highly promising. bookPart Springer, Singapore 73 88 10.1007/978-981-97-0180-3_7 AIS Algorithms for Intelligent Systems: IJCACI 2023: Proceedings of International Joint Conference on Advances in Computational Intelligence