A Review of Machine Learning-Based Recognition of Sign Language

Abstract

Some people in society have impaired cognitive senses like speech and hearing where they cannot behave like normal people. It is quite a complex task for abnormal people to understand as well as recognize the gestures of normal people. This initiates to delve into the study of review of Sign Language Recognition (SLR), in specific to, machine learning techniques. In this work, a review of machine learning techniques based on SLR were portrayed. Several studies related to ML papers have been collected and discussed with their merits and demerits. Thus, the observation dictates that recognition of hand gesture is still a challenging task. There are two sorts of gesture recognition, namely, static and dynamic gesture recognition. Static gesture recognition is developed from the dynamic gesture recognition. Almost, Convolutional Neural Networks (CNNs), Hidden Markov Models (HMM) and Histogram analysis were used as recognition classifiers for sign language. Dynamic gesture recognition process operates on tracking the centroid of hand gesture. It changes the visual information in time basis. Henceforth, study on dynamic gesture recognition needs to be more focused using Machine learning techniques. Comparative analysis is done in perspectives of significance of segmentation models, feature extraction and vision-based approaches.