Comparative Analysis of Siren Classification Technique for Emergency Vehicles

Abstract

Emergency vehicle sirens greatly aid traffic control and public safety awareness. Improving emergency response systems requires accurate siren classification. This study aims to categorize emergency vehicles, particularly fire trucks, police cars, and ambulances, based on the features of their sirens. It thoroughly analyses various schemes for categorizing emergency vehicle sirens. Mel-Frequency Cepstral Coefficients (MFCC), Zero-Crossing Rate (ZCR), Spectral Centroid, and hybrid methods that combine MFCC with ZCR and Spectral Centroid were observed for comparison. The data set is sourced from the Google Audio Set Ontology, ensuring robust training and evaluation of the models. This methodology involves preprocessing audio data, extracting relevant features, and training classifiers. The proposed hybrid method combines MFCC with Spectral Centroid to leverage their complementary strengths. Through rigorous experimentation, this system evaluates the performance of different classifiers, aiming to provide insights for optimal siren classification. The findings contribute to advancing audio classification methodologies and have implications for developing more robust emergency response and traffic management systems.