An Empirical Evaluation of Artificial Intelligence Algorithm for Hand Posture Classification

During the past decade, an intensive growth of Human–Computer Interaction (HCI) has been evolved. It includes, but is not limited to, Virtual Reality, Augmented Reality, Voice Control Systems, EEG-based systems, etc. Primarily, HCI is the hybridization of Information & Communication Technology and Bio-metric. Besides, the biometric inputs like Human Face, Voice, Physical actions, EEG signals, etc. are cascaded with the computing module for robust and intelligent decision making. The hand postures are one of the most common biometric for system automation & feedback. In this study, exhaustive empirical research of the machine learning algorithm for hand posture classification has been established. In this connection, a recent dataset, ‘Mocap Hand Postures Data Set,’ has been opted to employ the different variants of the machine learning algorithm. To the best of the knowledge, the exhaustive comparative study on the said dataset is found to be deficient in the literature. Primarily the principal focus is to identify the best candidate for real-time hand posture classification. Moreover, the performance of each method has been rigorously measuring as a function of training accuracy, testing accuracy, prediction speed, and training time. Besides, the percentage recognition of each corresponding class is illustrated using the respective confusion matrix.