Automated Acoustic Detection and Recognition of Moving Vehicles
- Members: Naoko Evans, Dave Chesmore
- Start Date: October 2006
Automated acoustic vehicle recognition has been studied in relation to various practical application areas such as; surveillance for security purposes, traffic monitoring and/or control as well as road planning, and also automated driver assistance systems, etc. Although the future goal of the project would be to develop a vehicle recognition system that employs multiple sensors in order to achieve a high accuracy, so far the work at University of York has concentrated on using acoustic and seismic sensors. Target detection performed with acoustic or seismic sensors is advantageous mainly due to its cost effectiveness, low power consumption, flexibility as well as those known benefits of passive sensing. The current research is aiming to develop a system that detects and classifies approaching vehicles by using signal processing techniques such as advanced version of Time Domain Signal Coding (TDSC) and Co-Occurrence Matrix, combined with artificial intelligence for pattern recognition.
An example showing how TDSC performs feature extraction.
Examples of plotted graphs obtained with Co-Occurrence Matrix on both training (left) and test samples (right), which indicate the algorithm’s good potentials in discriminating between large vehicles and small vehicles.