Proposal for Identification Scheme of Driver and Impostor based on Acceleration Data

Abstract

This paper proposes a continuous identification scheme monitoring a driver's behavior and a detection scheme for an impostor. The proposed scheme uses Long Short-Term Memory (LSTM) to create a classifier of the driver's behavior according to acceleration data and classifies a driver. Additionally, it also detects an impostor according to statistical information of output from the classifier. Experimental results show that an acceleration sensor on a real vehicle is enough to classify 15 drivers according to real-time driver's behavior. Additionally, the proposed scheme can detect an impostor through the verification of real experimental data.