Sign Detection Using an N-Gram Language Model and MobileNet

Typically a human being has no trouble communicating with each other through speech, gestures, body language, reading, and writing. However, people with speech impairments rely solely on sign language, which makes it harder for them to interact with the majority. This creates a need for sign language recognition systems that can understand and convert sign language into spoken or written language, and vice versa so that others can understand. Existing systems for this purpose are limited, expensive, and difficult to use and much work has not been done for Indian sign language. As a result, the main objective of this paper is to increase the accuracy of sign detection using language models for Indian sign language. For this N-gram sign model is used in MobileNet machine learning system. The simulation results showing the improvement of sign detection with accuracy of nearly 91% from 85% with the prediction model followed by machine learning method.