SIGNify – A mobile solution for Indian sign language using MobileNet architecture
Sajid Siddiq, S. Roopashree, Musfirah Suha, MRS Ruthvik, Kuruva Divyasree · 2021 2nd Global Conference for Advancement in Technology (GCAT) · 2021
Sign language assists in visual communication for the vocal and hearing-impaired population. It benefits people with other disabilities such as autism, down syndrome etc. The work recommends an intelligent system specifically to recognize the Indian Sign Language (ISL). It is an interesting and challenging problem, as the solution brings a leap in both social and technological aspects. A signer independent methodology based on different techniques of deep learning, is a real-time recognition system for Indian sign language developed to avoid the isolation of disabling groups from the rest of the society. In our work, we build and compare two pre-trained CNN models, MobileNet and InceptionV3 architectures. The suggested approach using the MobileNet model showed an accuracy of 99% on a custom-built dataset. The working CNN model can perform real-time recognition of ISL alphabets, numbers and a few selected gestures integrated into an Android mobile platform called SIGNify using React-Native, for better accessibility and user-friendly access. The study highlights a small step involved in human-computer interaction.