Classification of Finger Movements Using Multi-channel EMG and Machine Learning

This paper describes an experimental study on decoding of finger movements using surface electromyography (EMG) signals obtained from Myo-armband and machine learning techniques. The study is set out to determine whether machine learning algorithms and EMG signals could be used to precisely decode finger movements. The paper includes descriptions of the EMG dataset used in the study, pre-processing steps, feature extraction techniques, and machine learning algorithm development. The proposed model recognized seven pre-defined finger movements, with an overall cross-validated AUC of 95.29%. The study’s results, which show that Myo-bands and a support vector machine algorithm can predict finger movements with impressive accuracy, could have a big impact on how prosthetics and other tools help people with disabilities are made.