Human behavior recognition via top-view vision for intelligent space

Intelligent Space (IS), a kind of intelligent home, is one of the popular of the applying RT into our daily life. In intelligent space environment, human behavior is one of the meaningful information for interpretation of our intention and needs. In this paper, we proposed the human posture recognition approach which focuses on a top-view vision. A top-view vision enables our system to observe the whole area with single camera and also avoid horizontal obstacle. Moreover, a human posture feature which based on DFT (Discrete-time Fourier Transform) coefficients of log-polar histogram has been proposed. We have categorized human behavior to two types. The first one is a stationary behavior which consists of standing, sitting and lying. The second one is a moving behavior.