EMGNET

Summary: The main objective of this project is the design and implementation of a convolutional neural network (CNN), named EMGNet, for the classification of human upper limb movements using the surface EMG signals. The methodology consists of two stages as described next: 1) The EMG data acquisition from the muscle sites along the upper limb using wireless EMG sensors, 2) development, training & validation of deep learning algorithms for the limb movement classification, and finally online testing of the DL algorithm on the test EMG data. The key idea with the proposed system is that the CNN trained on the EMG data taken from the training subjects can be used to learn and classify the movements of entirely new subjects’ hands. The EMG data will be captured from the hands of subjects using the EMG sensors. Once the EMGNet model is trained, it will be used for testing over live and completely unknown data. We will take into account various factors such as subject’s age, gender, health conditions, daily variations in the EMG activity, etc. that influence the CNN performance and develop a database and the corresponding trained network that provides a robust performance against these factors.