A Recognition Method of Misjudgment Gesture Based on Convolutional Neural Network

Based on the Kinect 2.0, 17 kinds of static gesture libraries were established and trained by Convolutional Neural Network. A lot of statistical experiments have been done on the classification of each gesture. During the experiment, we found a phenomenon that several gestures in the 17 gestures were easily confused. And for the sake of description, we call these gestures as similarity gestures. It is assumed that the test result of convolutional neural network model satisfies the large number theorem from the angle of large data. Therefore, For misjudgment gestures, this paper presents a recognition method based on probability statistics.