Detection and Prediction of Abnormal Behavior Based on User Profile in Ubiquitous Home Network Using Hierarchical Hidden Markov Model

In this paper, we model the multilevel statistical structure as Hierarchical Hidden Markov Models (HHMM) for the problem of predicting the state of human behavior based on user profile in a ubiquitous home network. The HHMM is an extension of the hidden Markov model (HMM) to include a hierarchy of the hidden states and composed of sub-HMMs in a hierarchical model, providing functionality beyond a hidden Markov models for modeling complex systems. We present efficient algorithms for detecting abnormal behaviors in a ubiquitous environment and learning both the parameters and the model structures. Algorithms to analyze the behavioral patterns of a user using the information provided by the user in a home network system. We propose the detecting of abnormal behavior algorithm, which builds profile based on the actions taken when the user enters a room. The main contributions of this paper lie in the application of the shared structure HHMM, the estimation of the state of a user's behavior, and the detection of abnormal behavior. The user behavior data from an experiment show that directly modeling shared structures improves the recognition efficiency and prediction accuracy for the state of a human's behavior when compared with a flat HMM.