Recognizing bicycling states with HMM based on accelerometer and magnetometer data

In this paper, we design and implement an sBike (Sensorized Bike) prototype to support cyclists by recognizing various bicycling states including going straight, turning right or left, meandering, and stopping. An Android phone, which is integrated with an accelerometer, a magnetometer, and a GPS receiver, is mounted on the handle of bicycle to collect necessary data for analysis. Hidden Markov model (HMM) is adopted to recognize the bicycling states from raw sensor data. The experimental results show that the accuracy of recognition is as high as 98%. By knowing the bicycling states of cyclists, road conditions can be inferred and shared amongst users.