MAST: Myo Armband Sign-Language Translator for Human Hand Activity Classification

As Computer Science has grown into an encompassed field in various scientific areas, the need for developing a computer aided and artificially intelligent device has become more important especially in the medical field. Artificial Intelligence (AI) plays a vital role not only in accelerating and optimizing common tasks but also in performing tasks that humans are incapable of. This paper presents a Myo Armband Sign-Language Translator (MAST), which is a novel algorithm to translate a hand's gestures into medical sign language using a Myo armband sensor which collects muscles' electromyography signals and then to classify them using an enhanced version of a dynamic random forest. Our experimental results indicate that a systematic fine tuning of MAST parameters leads to an accuracy improvement of 13% over the state-of-the-art scheme such as SCIKIT's random forest. Other comparison results show an improvement of over 20% compared to a popular classification scheme such as Support Vector Machines (SVM) and a deep learning technique such as Convolutional Neural Network (CNN).