M.A. Thesis:

Human activity recognition in smart home

The number of people aged who need help in their daily routine is increasing rapidly. Smart homes can enhance the elderly's lifestyle with their privacy and allow them to live longer in their home than to keep them in care homes or hospitals. The purpose of the smart home is to increase the comfort of the residents, guarantee safety, security and provide services to the elderly who live alone at home. In the smart home surveillance system, the behavior of the elderly person can be monitored and behaviors of the elderly person are being diagnosed. In this research, we investigate the detection of behavior in the smart home. In the proposed method, Bayesian-supervised algorithms and KNN and Kmedoide-unsupervised algorithms are used to detect behavior. The results showed that the bayesian algorithm had the highest detection accuracy than knn and kmedoide algorithms and knn's algorithm was more accurate than kmedoide in detecting behavior. The accuracy of the behavior detection in the Bayesian algorithm is %96.96, in the knn algorithm is %81.94 and in the kmedoide algorithm is %69.95. The proposed method has also been able to significantly improve the accuracy of behavior detection compared with previous methods.

Keywords: Keywords:Smart home ,Activity recognition, Machins lerning