Inkjet-printed Dry Flexible Electromyography Electrode Array for Classifying Finger Movements with AI Algorithms

Electromyography (EMG) signals are very useful for the development of active prosthetic devices and for detecting various muscle abnormalities now-a-days. Many modern technologies are being employed for this purpose, utilizing EMG data to evaluate muscle activity. The existing research on EMG-based movement classification faces the challenges of inaccurate classification, insufficient generalization ability, and weak robustness. To address these problems, this paper proposes a novel EMG signal acquisition system using an Inkjet-printed (IJP) electrode array. IJP EMG electrodes were fabricated on flexible polyimide substrates using silver nanoparticle inks for the conductive layer and provided the flexibility of the shape and size of electrodes. In this work, a classification method with different machine learning (ML) models was performed on the datasets collected from our designed EMG circuit system. We have extracted 145 features from the dataset to properly train and test our ML models. The Random Forest (RF) model yielded a maximum accuracy of 98% for the suggested system. To validate the result, we classified the datasets by using the K-nearest Neighbor Algorithm (90%), Decision Tree (94%), and Artificial Neural Network (97%) and compared the results with the RF model. This proposed method is highly accurate for actuating prosthetics, or it has the ability to maximize medical resources and serve as a clinical auxiliary diagnostic tool.