Hand Gesture Recognition in Complex Background Based on Convolutional Pose Machine and Fuzzy Gaussian Mixture Models

Hand gesture is one of the most intuitive and natural ways for human to communicate with computers, and it has been widely adopted in many human-computer interaction applications. However, it is still a challenging problem when confronted with complex background, illumination variation and occlusion in real-world scenarios. In this paper, a two-stage hand gesture recognition method is proposed to tackle these problems. At the first stage, hand pose estimation is developed to locate the hand keypoints using the convolutional pose machine, which can effectively localize hand keypoints even in a complex background. At the second stage, the Fuzzy Gaussian mixture models (FGMMs) are tailored to reject the nongesture patterns and classify the gestures based on the estimated hand keypoints. Extensive experiments are conducted to evaluate the performance of the proposed method, and the result demonstrates that the proposed algorithm is effective, robust, and satisfactory in real-time scenarios.