Long Short-Term Memory (LSTM) model for Indian sign language recognition

• Published in: Journal of intelligent & fuzzy systems Vol. 46; no. 4; pp. 11185 - 11203
• Main Authors: Nihalani, Rahul, Chouhan, Siddharth Singh, Mittal, Devansh, Vadula, Jai, Thakur, Shwetank, Chakraborty, Sandeepan, Patel, Rajneesh Kumar, Singh, Uday Pratap, Ghosh, Rajdeep, Singh, Pritpal, Saxena, Akash
• Format: Journal Article
• Language: English
• Published: London Sage Publications Ltd 18.04.2024
• Subjects: Artificial intelligence; Computer vision; Deep learning; Machine learning; Recognition; Sign language
• ISSN: 1064-1246; 1875-8967
• DOI: 10.3233/JIFS-233250

The human-computer interaction process is a vital task in attaining artificial intelligence, especially for a person suffering from hearing or speaking disabilities. Recognizing actions more traditionally known as sign language is a common way for them to interact. Computer vision and Deep learning models are capable of understanding these actions and can simulate them to build up a sustainable learning process. This sign language mechanism will be helpful for both the persons with disabilities and the machines to unbound the gap to achieve intelligence. Therefore, in the proposed work, a real-time sign language system is introduced that is capable of identifying numbers ranging from 0 to 9. The database is acquired from the 8 different subjects respectively and processed to achieve approximately 200k amount of data. Further, a deep learning model named LSTM is used for sign recognition. The results were compared with different approaches and on distinct databases proving the supremacy of the proposed work with 91.50% accuracy. Collection of daily life useful signs and further improving the efficiency of the LSTM model is the research direction for future work. The code and data will be available at https://github.com/rahuln2002/Sign-Language-Recognition-using-LSTM-model.