Collection and Classification of Gait Patterns through Wireless Sensor and Video Data

Principal Investigator: Maria Regina Justina Estuar
Other Researchers: Kardi Teknomo, Nadia M. Leetian, Maricor Soriano (UP), Kenneth Ng (UP General Hospital), Josephine R. Bundoc (Physicians for Peace), Phoebe Gallanosa (UP)

PedLab, in cooperation with University of the Philippines, Physicians for Peace, and University of the Philippines, Philippine General Hospital.

For unilateral transtibial amputees, numerous factors affect long term prosthesis use including factors related to utility or the patient himself. Utility factors such as ill-fitting socket, incorrect alignment of prosthetic parts, appearance, weight, sounds as well as access to maintenance and repair affect long term prosthesis use. Patient related factors include general medical condition, residual stump health, cognitive status, and psychological state. In most cases, early detection of differences in gait pattern can determine which utility factors affect short term prosthesis use. This paper addresses the issue of the need to provide a simple and inexpensive remote wireless sensor device that can collect and classify gait patterns. In this study, we use two methods to collect gait data. The first method is a programmed application which uses tri-axial accelerometer, magnetometer and pedometer sensors embedded in a standard smart phone to collect gait data. The second method is captured motion in video format for video analysis. We used a 2 x 3 x 2 x 2 experimental design categorizing by type of person (normal, non-normal), type of sensor (accelerometer, magnetometer, pedometer), type of capture (sensor, video) and type of gait (walking, non-walking). We will perform time series analysis of sensor data after being smoothed using exponential moving average to detect gait to differentiate walking and non-walking patterns.