Classifying Human Walking Patterns using Accelerometer Data from Smartphone

This paper presents a study on identifying different individuals using accelerometer data from a smartphone presented on their walking patterns. The identifier algorithm was trained and evaluated in an experiment with twenty human subjects including 12 males and 8 females in real-world conditions. Various classifiers were tested using descriptive statistical features. Our model recognizes patterns and regularities in human walking movement using limited accelerometer data captured from a mobile device. The accelerometer data are decomposed into gravitational acceleration and body motion acceleration using a low pass filter, and then extracted features of those acceleration components is fed to multi-class classifiers. The proposed model is developed based on an informative and stable body acceleration feature set that gives rise to a high performance multi-class identification model. The results show that using the Decision tables as our classification method enables the identification to be made in overall accuracy rate of 98.45%.