Abstract： In order to accurately and quickly detect accidental falls of the elderly who live alone, and to minimize the risk of accidental deaths and injuries caused by accidental falls, the six human skeleton joints, including head, left shoulder, right shoulder, left hip, right  hip and hip center, were chosen using the human skeleton tracking algorithm based on Kinect sensor. By realtime compution of the spatial location, relative position, kinematic velocity and static time of the six joint points, the occurrence of human falls could be determined. Meanwhile, the movements like sitting, squatting, retrieving and other nonfall movements could be accurately discriminated. As a result, the misjudgment ratio was reduced. The experiment results showed that the automatic detection of human fall was realized, and the misjudgment rate of human fall was reduced to 7%.By using the skeleton tracking technology, the privacy of the elderly could be protected in the monitoring process. The detection system did not depend on visible light and could make real time detection at 24 hours. These advantages provide a guarantee for timely security assistance for the elderly who suffer from accidental falls.