Study on the emg-based grasp gesture classification using deep learning and application to active prosthetics

In this paper, we extract the learning and test data for the "hand gesture of grasping" through the sEMG sensor, execute the Deep Learning CNN (convolutional neural network) algorithm by appropriately modifying it, and classify typical hand gestures that catch objects with a classification success rate (accuracy) of approximately 93.8%. In addition, we have constructed a system that can operate robot hands in real time from these classified commands to make active prosthetics. © ICROS 2019.