Hand gesture recognition using K-means clustering and Support Vector Machine

Abstract

Human-Robot Interaction (HRI) requires media for communication which can be both understood by robot and easily done by human. Usually, human using oral language to communicate but there are some situations that require performing non-verbal activities such as deaf people, patient, and old people, therefore gesture recognition as communication media is needed to give order to Robot. This paper discusses hand gesture recognition as input command for Bioloid Premium Robot using two methods, K-Means clustering and Support Vector Machine (SVM) with directed acyclic graph (DAG) decision. Four gestures (forward, right, left and stop) were recognized using Kinect v2. The testing was done 6 peoples for three distances (2m, 3m, and 4m) and three slopes position (45, 0, −45). The SVM required 10ms recognition time with accuracy reached 95.15%, while K-Means needed 4.45ms recognition time with 77.42% accuracy. This study resulted in Multiclass SVM with DAG decision performs better than K-Means clustering method.