Two wheeler rash drive detection using smartphones

摘要

Two-wheeler accident rates have increased significantly year by year in India. Technological interventions such as closed-circuit cameras are limited in that they cannot monitor individual driving behaviors. Also, most solutions proposed by the research community have been implemented and tested on four wheelers. In this paper, we propose a portable method using just the built-in accelerometer of a smartphone to detect and identify abnormal driving behaviors such as weaving, lane changing, and sudden braking. The detection process is implemented on a server by applying dynamic time warping based pattern matching. Driving traces collected by multiple volunteers were used for the evaluation, resulting in an overall accuracy of 90%.