Final Year Project

Project Title

Car Crash Detection Using Acoustic Signature

Description

This project aims to enhance road safety and accelerate emergency response times by developing an automated car crash detection system based on acoustic signatures. Utilizing a machine learning approach, the system processes audio data such as collisions, skids, and ambient background noise and converts it into spectrogram images. An RNN-LSTM model is then trained to accurately classify these acoustic events, reliably distinguishing actual crashes from normal traffic sounds. To enable real-time, low-latency detection at the edge, the architecture is planned for hardware deployment on a Zynq SoC platform. By instantly identifying accidents the moment they occur, this system provides a critical solution to reduce emergency dispatch delays and improve outcomes in life-threatening situations.