Driver Behavior Profiling based on Efficient Recurrence Quantification Analysis

Abstract: Following the hypothesis that nonlinear dynamics provide a more sensitive description of driving variability compared to conventional linear metrics, we investigate whether nonlinear dynamic measures derived from recurrence quantification analysis (RQA) can effectively reveal and quantify recurrent driver behavior from inertial sensor data (accelerometer and gyroscope signals). Recurrence plots (RPs) of tri-axial accelerometer and gyroscope data facilitate recurrence quantification analysis (RQA) by using recurrence rate (RR), determinism (DET), entropy (ENTR), laminarity (LAM), trapping time (TT), and line-based metrics. The cross-recurrence quantification analysis (CRQA) is used to identify multivariate correlations among sensor data. These data could be used in intelligent transportation systems, driver assistance technology, and safety surveillance. The results hold potential applications in intelligent transportation systems, driver assistance technologies, and safety monitoring. The ability of CRQA to capture inter-signal dependencies further opens possibilities for multimodal driver monitoring frameworks.