Machine Learning-based Hand Sign Recognition
Greeshma Pala, Jagruti Bhagwan Jethwani, Satish S. Kumbhar, Shruti Dilip Patil
Abstract
Sign Language is used by the deaf and voiceless community to be able to communicate with others, but the most commonly faced problem here is that everyone around may not be able to understand sign language. The main motive behind this system is to bridge the communication gap between the communities, therefore, establish the interaction between the speechless community to communicate with others. Hand gestures differ from one person to another person in shape and orientation, therefore, a problem of linearity arises. Recent systems have come up with various ways and algorithms to accomplish the problem and build this system. Algorithms such as KNearest neighbors (KNN), Multi-class Super Vector Machine (SVM), and experiments using hand gloves were used to decode the hand gesture movements before. In this paper, a comparison between KNN, SVM, and CNN algorithms is done to determine which algorithm would provide the best accuracy among all. Approximately 29,000 images were split into test and train data and preprocessed to fit into the KNN, SVM, and CNN models to obtain an accuracy of 93.83%, 88.89%, and 98.49% respectively.