Research On Hunman Behavior Recognition In Home Environment

The increasing development of robot techniques makes it possible to devise and develop service robots for civil applications. Human behavior recognition is a prerequisite to robots autonomous services. However, human behavior recognition is a challenging subject because of the variety of human behavior patterns, the complexity of the background and the robots' limited capacity of sensors and computing power. Thus, we build an assistant platform to help robots to complete service tasks. Intelligent space technology is used, which is expected to help service robots to work in complex environment more easily and effectively.Machine vision technology combined with intelligent space information is used in order to recognize human behavior and realize human intention in home environment. The main contributions are as follows:1. Firstly, we introduce the general framework for human behavior recognition and analysis of key technologies. Then, we introduct the assistant platform for robot and describe human behavior recognition plan.2. Human motion detecting is the primary stage of human behaviors recognition. Improve GMM-based motion detection algorithm as follows: the mutative learning rate was adopt to improve the accuracy of background model. The part background update strategy was introduced to slove the proplem that objects stop moving temporarily may be absorbed into the background model. The GMM motion detection algorithm and edge difference algorithm were combined to remove breaking and holes in the foreground.3. A novel method is proposed to detect human daily activities and fall over behavior based on features fusion. Actions are described by a set of postures and features of postures change velocity and position changes, then simple postures events and features are combined to express complex human activities events.4. A novel method is proposed in this paper to recognize human behavior based on scene knowledge in the platform of intelligent space. Using machine vision technology combined with intelligent space information, we can detect abnormal habits behavior and real-time unexpected accident, recognize human intention. First, we extract the location of the human body in the environment via motion detection. Then, due to different scene rules for different behaviors, we divide the environment into different states and introduce state-based matrix and state-time matrix to model those behaviors. At the same time, the state transition probability and its duration time distribution are gotten based on Markov model as the human daily behavior templates. Abnormal behaviors can be found by calculating the likehood of current acts with daily templates. This method requires low computation complexity thus can be widely used for abnormal behavior detection and human intention recognition.5. Human behavior recognition as a way for robot to access to environmental information have incomparable advantages over other sensors In the final chapter, human behavior information is used in robots navigating, which embodies information sharing concept in intelligent space system.