A method for identity recognition is proposed, which fit skeletal motion features using Bézier curves. First, the OpenPose algorithm extracted key skeletal points of the human body. Limb gait feature triangles were constructed, and their variation patterns were fitted using Bézier curves, resulting in limb gait curves with strong discriminative ability. The optimal feature subset was selected by a greedy algorithm combined with an incremental feature fusion strategy. A gait feature vector was constructed, and a similarity measurement mechanism based on Euclidean distance was designed for precise comparison of limb gait features. Finally, the Bayesian inference framework was employed to integrate upper- and lower-limb feature similarities, producing the final identity recognition result. Experimental results demonstrate that this method significantly improves recognition accuracy and efficiency compared to existing methods, highlighting its broad application potential in practical scenarios.