Abstract
This paper focuses on the problem of sensing vehicle's steer change via smartphones. Different from other works, we propose a framework that only use off-the-shelf smartphone's IMU sensors to detect and classify the vechicle's steer change during driving. First, we achieve a sensor fusion method based on Kalman Filter to get smartphone's attitude accurately in order to get the relationship between smartphone coordinate and vehicle coordinate. Step further, in the process of steer change classification, we propose a multi feature score method to classify lane change pattern and use K-means algorithm to differentiate left turn, right turn and U-turn pattern. From the experiment results in real driving condition, the classification accuracy is acceptable at around 90% in average.