Detection of human presence in a vehicle by vibration analysis
Résumé
The aim of this study is to propose a complete instrumentation and signal processing method able to detect the presence of a person seated on the rear bench of a vehicle. The sensor is based on a piezoelectric film (EMFI sensor), designed to detect mechanical vibrations. In order to avoid confusion between humans and heavy objects or empty seats, the authors focused on the extraction of a biological signature from the acquired signals. This physiological pattern was extracted using an original wavelet denoising algorithm and was used further as a matched filter, in order to detect human presence in the vibration signals. Physiologically significant features were extracted from the output of the (on-line) filtering process and fed further-on into a classical Bayes-based classifier. After training, the proposed method yielded very promising results, the output of the classifier being almost error-free for different acquisition conditions (stopped and on-road vehicle, new and artificially aged sensor).