Machine Learning based Sign language Detection

The review presented in this paper highlights the continuous evolution of Indian Sign Language (ISL) recognition systems and their growing importance in bridging the communication gap between the hearing and non-hearing communities. With the integration of artificial intelligence, computer vision, and deep learning, researchers and developers have been able to transform sign language into interpretable text and speech, making interactions more inclusive and accessible. The comparison between the Web-Based ISL Converter team and the SignFlow framework proposed by Geetha et al. (IEEE, 2025) provides valuable insights into both practical and research-oriented approaches. The JSCOE system demonstrates that even a lightweight CNN-based model can successfully translate static ISL gestures into text and multilingual speech in real time. Its simplicity, cost-effectiveness, and easy deployment through a web interface make it highly suitable for educational and public use. Meanwhile, the SignFlow model represents the research frontier, utilizing a hybrid CNN–Transformer architecture with Connectionist Temporal Classification (CTC) to achieve high accuracy in continuous gesture recognition. This advancement shows how deep learning models can handle real-time ISL translation across diverse backgrounds and signing speeds. The analysis also reveals that dataset quality, model complexity, and computational resources play a vital role in recognition performance. While the web-based system performs effectively for isolated gestures, its scalability is limited by smaller datasets and lack of dynamic gesture recognition. Conversely, the SignFlow model, despite being computationally demanding, demonstrates robust performance and generalization through large, domain-specific ISL datasets. Both approaches contribute to the shared goal of enabling communication accessibility for the deaf and mute community. In conclusion, the integration of AI-driven models in sign language interpretation marks a major step toward inclusive technology and social empowerment.

[1]      M. Geetha, N. Aloysius, D. Somasundaran, A. Raghunath, and P. Nedungadi, “Toward Real-Time Recognition of Continuous Indian Sign Language: A Multi-Modal Approach Using RGB and Pose,” IEEE Access, Vol. 13, 2025. [2]      S. Renjith and R. Manazhy, “Sign Language: A Systematic Review on Classification and Recognition,” Multimedia Tools and Applications, Vol. 83, 2024. [3]      Hayani, S., Benaddy, M., El Meslouhi, O., & Kardouchi, M., “Arabic Sign Language Recognition with Convolutional Neural Networks,” IEEE Access, 2023. [4]      Aloysius, N., and Geetha, M., “Continuous Sign Language Recognition with Adapted Conformer via Unsupervised Pretraining,” arXiv preprint, 2022. [5]      R. Zuo and B. Mak, “Consistency-Enhanced Continuous Sign Language Recognition,” IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021.