Authors: Assistant Professor S P.Premchand Goud, Pooja Kurikala
Abstract: The rapid expansion of urban transportation networks has increased traffic congestion, making it difficult for emergency vehicles to reach their destinations without delay. Ensuring uninterrupted movement for ambulances, police vehicles, and fire engines is essential for improving emergency response efficiency and public safety. This study presents an intelligent siren recognition framework that employs Mel-Frequency Cepstral Coefficients (MFCCs) for extracting discriminative audio features and a Convolutional Neural Network (CNN) for multi-class classification of emergency vehicle sounds. Prior to feature extraction, audio recordings undergo preprocessing and noise suppression to minimize the influence of environmental disturbances commonly found in city traffic. The CNN model learns the unique spectral and temporal characteristics of ambulance, police, and fire truck sirens, enabling accurate classification with high reliability. Once an emergency siren is detected, the system can support adaptive traffic signal control by assigning signal priority to the approaching emergency vehicle, thereby reducing waiting time at intersections. Experimental evaluation demonstrates that the proposed approach achieves classification accuracy exceeding 90%, confirming its suitability for practical intelligent transportation applications. The framework provides an efficient and scalable solution for smart cities and can be further enhanced through integration with real-time traffic monitoring systems and Internet of Things (IoT) infrastructure.
