Machine Learning Based Real Time Sign Language Detection

Abstract

In this paper, a real time ML based system was built for the Sign Language Detection using images that have been captured with the help of a PC camera. The main purpose of this project is to design a system for the differently abled people to communicate with others with ease. This model is one of the first models to detect signs irrespective of their sign language standards (i.e., the American Standard or the Indian Standard). The existing digital translators are very slow since every alphabet has to be gestured out and the amount of time it would take to just form a simple sentence would be a lot. This model, which was trained using the SSD ML Algorithm, overcomes the above problem by directly recognizing the signs as words instead of alphabets. This model was trained using a set of 20 images for a particular sign in different conditions such as different lighting, different skin tones, backgrounds, etc., in order to increase the accuracy of detecting the gesture. The system displayed a high accuracy for all the datasets when new test data, which had not been used in the training, were introduced. The results have shown a high accuracy of 85% for the sign detection.