Proposing Hand Gesture Recognition System Using MediaPipe Holistic and LSTM

The paper proposes a method of recognizing and tracking human hand gestures that can be applied to virtual reality applications, medical monitoring, and smart device control. The proposed method estimates hand gestures using the MediaPipe algorithm to track and long short-term memory (LSTM) to classify these gestures. Nine suggested gestures include punching, cutting, swiping right, swiping left, swiping up, swiping down, raising, waving, and clapping hands. The output of the system is to extract the human upper body skeleton and the corresponding labels for the gestures. The results show that the average accuracy of the proposed solution is 93.3% which is capable to be applied in medical monitoring and controlling of electronic devices in smart homes or virtual reality systems.