Classification Accuracy Comparison between Machine Learning Algorithms and a Deep Learning Algorithm in Predicting Hand Gestures

Abstract

In this paper four Machine Learning (ML) Algorithms have been implemented for the classification of four hand gestures using electromyography (EMG) dataset. The classifiers opted are Support Vector Machine (SVM), Random Forest (RF), Bagged tree and Extreme Gadient Boosting (XGBoost). The prediction accuracy of the machine learning algorithms were subsequently compared with Long Short-Term Memory (LSTM) which is a Deep learning based classification technique. Among the machine learning algorithms, XGBoost provided the highest accuracy of approximately 97% while LSTM provided a superior accuracy close to 99% which promises to provide the physiologically natural upper-limb movement control. In addition to the pursuit of improved accuracy in the research, the effect of removing the noisiest channel in the accuracy of the algorithms has been examined in order to decrease the volume of data processing.

Keywords

• electromyography (emg)
• myo gesture control
• prosthesis
• machine learning
• deep learning