Activity recognition using a hierarchical model

In this paper, we propose a human daily activity recognition method that is used for Ambient Assisted Living. The proposed system is able to learn a user's activities using the data from motion and door sensors. We extract low level features from the sensor data and feed the features to a model that combines support vector machines (SVMs) and conditional random fields (CRFs) to give accurate recognition results. We propose to combine SVM and CRF classifiers in a hierarchical model which results in better accuracies and can also make use of high level features. We conducted experiments and presented the effectiveness and accuracies of the proposed method.