Survey on implementation of TinyML for real-time sign language recognition using smart gloves

Abstract

Communication is a major part of life. For people who are deaf and dumb, sign language is imperative. All sign languages contain various hand movements and finger placement which in turn form different gestures. Learning any sign language can be challenging. With the current number of different sign languages available around the world, it makes it very hard for people to use sign languages for communication without the help of a translator. Out of all the sign languages available around the world American Sign Language is used by many. Engineers and researchers around the world have been working with different machine learning algorithms to build a better system which can either be only software or along with hardware. The smart gloves have so far been the most effective hardware with machine learning software developed till now as compared to a camera. These smart gloves can be developed using TinyML with the Convolutional Neural Network (CNN) algorithm for best accuracy with low footprint. This comprehensive literature review goes into the details of the different algorithms explored for sign language recognition, the existing systems, and the proposed system using Raspberry Pi Pico.