Detection Of Abnormal Behavior And Posture Recognition Based On Characteristics Of Human Motion

Detection of abnormal behavior in human movement and human movement recognition is the hot issues in the computer vision. Human motion recognition is the use of some means of detecting, tracking the movement of the human body to obtain the body motion parameters, and from reconstruction of the human body structure and posture, and ultimately to the understanding of human movement and apply. This paper based on this correlation method and theory to study human motion in abnormal detection and gesture recognition algorithm.For the detection of abnormal behavior,In this article abnormal behavior defined refers to the movement of people in the squat or a sudden fall this situation, this is one of the work to detect this situation.The abnormal behavior detection of human based on motion feature, this paper using GMM to get foreground image of the sequence using the minimum bounding rectangle 's height, width, aspect ratio and target velocity as the motion characteristics of the human,combination the K-means cluster algorithm to detected the abnormal behavior in human movements,achieved satisfactory results.For the gesture recognition, the algorithm is taken out by extracting the prospect of human skeleton outline of the star's motion as a feature ,classification algorithm with HMM as the behavior of human body movement and posture classification and recognition,achieved good recognition effect.