In order to achieve the recognition of dynamic finger gestures based on motion trajectory and promote the development of the cross-cutting areas of human-computer interaction and virtual reality, a recognition method of the key feature points′trajectory is described, which is about finger gesture in three-dimensional (3D) space on basic of Leap Motion. The hand skeleton is extracted by skeleton algorithms with the modeling hand information, which is obtained by the Leap motion controller. Then, the thinning algorithm is used to get straight finger schematic, and extracts feature points and motion trajectory. The basic dynamic 3D finger gestures are identified and evaluated by means of support vector machine (SVM). Experimental results show that the identification of dynamic 3D finger gestures can be achieved in a more complex 3D space or hand block by themselves, and has a high accuracy and robustness.