ADL Classification Based on Autocorrelation Function of Inertial Signals

Recognition of human activities is one of the most promising research areas in artificial intelligence. This has come along with the technological advancement in sensing technologies as well as the high demand for applications that are mobile, context-aware, and real-time. In this paper, we use a smart watch to collect sensory data for 14 ADL activities. We collect three types of sensory signals: acceleration, angular velocity, and rotation displacement; each is a tri-axial signal. From each given signal we compute the autocorrelation function up to certain lag and take these computed values as representative features of the given signal. We then feed these features to a random-forest-based classifier for training and prediction. We experiment with different combinations of sensory data. The joint use of acceleration with angular velocity has achieved the best performance in prediction accuracy which reaches about 80% for the whole set of 14 activities.