Interactive Design of 3D Dynamic Gesture Based on SVM-LSTM Model

Abstract

Visual hand gesture interaction is one of the main ways of human-computer interaction, and provides users more interactive degrees of freedom and more realistic interactive experience. Authors present a hybrid model based on SVM-LSTM, and design a three-dimensional dynamic gesture interaction system. The system uses Leap Motion to capture gesture information, combined with SVM powerful static gesture classification ability and LSTM powerful variable-length time series gesture processing ability, enabling real-time recognition of user gestures. The gesture interaction method can automatically define the start and end of gestures, recognition accuracy reached 96.4%, greatly reducing the cost of learning. Experiments have shown that the gesture interaction method proposed by authors is effective. In the simulated mobile environment, the average gesture prediction only takes 0.15 seconds, and ordinary users can quickly grasp this method.