Abstract: Sign language recognition belongs to the research category of gesture recognition.The traditional method of sign language recognition based on data glove can not capture all the elements of sign language,as it can not recognize sign language movements cooperated by hand and limbs.For IMU (Inertial Measurement Unit),due to its small size,low cost,it has been applied to the motion capture project widely.Based on the knowledge of robot kinematics,we proposed an IMU based language recognition skeleton model,which is in line with the biological characteristics of human body.The first construction procedure of the model is to make a selection of the skeleton,and then tocalibrate the size of the skeleton.Finally,an experimental method was proposed to calibrate the model size,which can be solved by using the data of the action set obtained by IMU.

Key words: Skeleton model,POE,Forward kinematics,Model calibration