A 3D gesture posture estimation and interaction method

Aiming the needs of non-contact convenient interaction in applications, we proposed a 3D gesture pose estimation and interaction feedback method. The accuracy of gesture posture estimation is improved by a cascade gesture posture estimation network based on DeepPrior++ and the Kalman filter. We also design a 3D gesture interaction modal. The recognized gestures and estimated gesture poses are transformed into operation commands through a gesture understanding engine, and the interaction feedback is updated based on the operation commands. Thus, non-contact interaction can be achieved in a convenient manner, to provide convenient and practical 3D gesture interaction tools.