A Hybrid Machine Learning Model for Hand Gesture Recognition

Abstract

Sign language, a gesture based nonverbal technique of communication, is used by a significant number of populations throughout the world. Sign language recognition emerges as one of the most challenging exercises when the person interpreting it lacks the previous knowledge of signs. This paper proposes a method to recognize sign language using computer vision. Five features from a binary image of a hand shape, namely the area of the closed contour, the area of the convex hull, number of convexity defects, maximum depth of the defects, and the sum of the depths of the defects are extracted after preprocessing the image. In this study, for recognizing sign languages, the recognition task was accomplished by employing two classification algorithms: knearest neighbor (k-NN) and Support Vector Machine (SVM) individually. Then the system will decide the class of the gesture acquired from the result of a Logical AND operation from both of the classifiers.