As one of the home services by a mobile manipulator system, we are aiming at the realization of the motion assist for elderly people. The purpose of the work is to realize a safe assist by feedback control based on the information from a high speed tactile sensor and Kinect. This paper describes the skeleton data accuracy estimated by Kinect and the classification of standing-up motion phases using Kinect. A method for classification from Kinect skeleton data is proposed. The method is implemented and experimental results show the availability of Kinect for the classification.