Artificial Intelligence IoT based EEG Application using Deep Learning for Movement Classification

People with disabilities such as hand amputations have limited motor activity. Several robotic prosthetic arms were developed to help them. The challenge arises when the robot's control source comes from the user's wishes extracted from brain signals via electroencephalography (EEG) signals. This research develops a raspberry-based embedded system device that is connected to EEG electrodes and functions as an artificial intelligence internet of things (AIoT) so that it can be controlled via the internet in real-time. The deep learning model used is convolutional neural networks (CNN) and autonomous deep learning (ADL). The results of the training with 5-fold cross-validation achieved an accuracy of about 98% in the four classes. The results of real-time testing over the network produce a pretty good response time of about 1 second.