Intelligent Gesture Recognition System for Translating Indian Sign Language to English


Anagha Kulkarni (Corresponding author), Yael Robert, Yashshree Nigudkar, Pranjali Barve, Namita Mutha
Department of Information Technology, Cummins College of Engineering for Women, Karvenagar, Pune, 411052, Maharashtra, India.


Sign languages involve a combination of hand movements and facial gestures. Alphabets and digits form static signs whereas dynamic signs consist of words and sentences. Based upon the cultural dif erences and regional variations, dif erent signs have evolved for a word in each sign language. In reality, every sign language has its own set of signs for each word. As a result, recognizing words and phrases in sign languages is dif icult. The recognition of spatial and time-distributed features of Indian Sign Language is the focus of this research. The main goal of this work is to identify gestures in Indian Sign Language using a multi-class classification technique. Various experiments have been conducted using Convolutional Neural Network, Long-Short Term Memory, and Gated Recurrent Units. Processing videos posed challenges. Various experiments and the methodology yielded an accuracy of 87.5% on unseen test data. The most significant advantage of this system is that it does not require any special device such as a depth-sensing camera, hand gloves, or special t-shirts.