Toward a visual pedometer

Dawei Liu, Qi Shan, Dan Wu

Abstract: Traditional inertial-sensor-based pedometers are not user-friendly because of the limited availability of inertial sensors in existing mobile computing devices. To remedy the drawback, we propose a visual pedometer scheme. It employs a body-mounted camera to capture images and measures the number of steps based on a time-series analysis of image features. We address several key issues raised in the measurement and demonstrate the applicability of the proposed visual pedometer through a four-second experiment. Compared with traditional inertial-sensor-based pedometers, the proposed visual pedometer could be more user-friendly because of the extensive use of camera phones and digital cameras.