Emergency Vehicle Detection using Vehicle Sound Classification: A Deep Learning Approach

Abstract

Emergency vehicles are equipped with audio and visual warning devices, which are meant to help them navigate through traffic by demanding manoeuvring space from other vehicles. Notable deaths happen due to the delay in reaching their destinations by ambulance and fire engine vehicles. Depending on local legislation, vehicles on the road may be compelled to cede the right of way to emergency responders utilizing their warning devices. Emergency vehicles happen to wait at signalized intersections with fixed cycle timing. Though there are DL-based vehicle classification techniques that support intelligent traffic light systems, this study discusses the Emergency vehicle sound detection model based on Deep Learning techniques as additional prop data to improve the accuracy of existing vehicle detection. The Convolutional Neural Network (CNN) model was trained based on short audio signals. The sound was processed using the Mel-frequency Cepstral Coefficients (MFCC) feature extraction technique to transform into an image. The model successfully reached 93% accuracy.