Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{203483,
        author = {YASH JITENDRA SAVDEKAR and Gunjan Nandkishor Rane and Ketki Praful Gokakkar and Sayali Rohidas Navale and Prof. Mrs. Priyanka Deshpande},
        title = {Bridging Communication Barriers: A Review and Conceptual Framework for Real-Time Indian Sign Language Translation Using Deep Learning},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {12672-12679},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=203483},
        abstract = {For the millions of hearings and speech impaired individuals across India whose daily communication depends on Indian Sign Language (ISL), the absence of practical, automated interpretation tools constitutes a persistent accessibility gap. Prevailing computational approaches to ISL recognition are largely frame-centric in design, carry substantial processing overhead, and fall short of the responsiveness demanded by consumer-grade hardware. This paper offers a structured critical review of sign language recognition research spanning 2019 to 2025, charting the field’s trajectory from handcrafted feature descriptors through convolutional architectures, hybrid CNN LSTM formulations, and attention-based transformer models. Within this analytical context, we introduce Sign verse, a compact and privacy-conscious framework for real-time ISL recognition built upon structured skeletal landmark sequences produced by MediaPipe Holistic. Rather than processing raw pixel data, the system models gesture dynamics through Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks operating on ordered keypoint vectors, substantially reducing inference cost while retaining temporal discriminability. A custom dataset covering eleven ISL sign classes (exceeding 100 samples per class) stored exclusively as landmark arrays underpins model training and assessment. Recognized sign tokens are subsequently routed to an NLP-based sentence framing stage that assembles grammatically well-formed natural language sentences. Systematic comparison across paradigms demonstrates that the landmark-sequential approach yields meaningful improvements in inference latency, memory consumption, resilience to background variation, and alignment with privacy requirements. Sign verse is framed not as a performance benchmark but as a practical, modular foundation for real-world ISL assistive communication.},
        keywords = {Indian Sign Language, MediaPipe Holistic, LSTM, GRU, Sequential Landmark Learning, Gesture Recognition, NLP Sentence Framing, Assistive Technology, Real-Time Recognition, Deep Learning},
        month = {May},
        }