Classification of User Postures with Capacitive Proximity Sensors in AAL-Environments

In Ambient Assisted Living (AAL), the context-dependent adaption of a system to a person's needs is of particular interest. In the living area, a fine-grained context may not only contain information about the occupancy of certain furniture, but also the posture of a user on the occupied furniture. This information is useful in the application area of home automation, where, for example, a lying user may effect a different system reaction than a sitting user. In this paper, we present an approach for determining contextual information from furniture, using capacitive proximity sensors. Moreover, we evaluate the performance of Naive Bayes classifiers, decision trees and radial basis function networks, regarding the classification of user postures. Therefore, we use our generic classification framework to visualize, train and evaluate postures with up to two persons on a couch. Based on a data set collected from multiple users, we show that this approach is robust and suitable for real-time classification.