Recognition of American Sign Language Gestures Based On Electromyogram (EMG) Signal With XGBoost Machine Learning

This paper discusses a potential way to recognize different American Sign Language(ASL) with the help of machine learning. In order to recognize different ASL, we decided to collect the data by multi-channel surface electromyogram(EMG). All the signals are processed during the time domain. After that, we extract the features in the data. The feature extractions this paper uses are mean absolute value, standard deviation, variance, and skewness. The extracted features are going to be analyzed by a machine learning model. The machine learning model this paper uses is the XGBClassifier, and the overall accuracy is about 85%. The samples this method chooses are from different college students whose ages are between 19 to 25 years old. This method also included the muscle fatigue situation. There are two muscle tired statuses that can be distinguished in the measurement to increase the gesture recognition accuracy.