Autonomous Tracking and Counting of Footsteps by Mobile Phone Cameras

In this paper, we present an autonomous method to track and count footsteps by using the camera data from mobile phones or tablets. Many step counters, relying on accelerometer data, are widely available. However, accelerometer-based algorithms are prone to overcounting. In our proposed method, feature points are detected first. Then, in order to increase robustness and accuracy especially in the case of highly-textured ground and floor surfaces, Kalman filter based tracking is performed. The proposed method is compared with existing accelerometer-based step counters. Experiments are performed with multiple subjects carrying five mobile devices simultaneously, including smart phones and watches, at different locations on their body. A Samsung (TM) Galaxy S4 smartphone is used to capture the videos. The results show that the proposed camera-based footstep counting has the lowest average error rate for different users, and is more reliable compared to accelerometer-based counters. The average error rate for the proposed method is 2.68%, and the standard deviation of the error is 2.39%.