Activity Recognition is a typical classification problem. The goal is to detect and recognize everyday activities of a person. This paper presents our approach to measurements and classification of a person’s movements. This is done by using two 3-axis radio accelerometers attached to the person’s body and by reconstruction and interpretation of the user’s behavior. We compared two machine learning algorithms (J48 and Random Forest) and various attributes characterizing the user’s behavior in order to obtain accurate classification of the behavior into predefined classes - activities.
Activity Recognition, 3-axial Accelerometers, Activities, Instances, Machine Learning, Classifiers.