Real-time static custom gestures recognition based on skeleton hand

Gesture recognition is one of the natural ways of human-computer interaction (HCI) that will positively affect their use. This paper presents an approach for real-time static gestures recognition based on the skeleton of a hand using a MediaPipe framework and Support Vector Machine(SVM) classification. The approach demonstrated high accuracy of recognition gestures 98.74% on a dataset of sign-digit-gestures as well as runtime 71 fps. Moreover, the approach is required only one camera sensor for recognition. The proposed approach can be extended for dynamic gesture recognition and used to control robots and computer devices.