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{209302,
        author = {Dr MK Jayanthi Kannan and Manikeerthan and Thanush and Mohd Minhaz and Ansh Sonkar and Kartavya and Aadi Atharva},
        title = {SignLearn: Edge Computer Vision Innovation in Real-Time Sign Language Recognition, Kinematic Tracking, and Assistive Communication},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {5},
        pages = {1631-1645},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=209302},
        abstract = {The rapid evolution of computer vision and human-computer interaction highlights the necessity of intelligent assistive systems to bridge the communication gap between the Deaf and Hard-of-Hearing (DHH) community and the hearing world. Over 70 million individuals globally rely on sign language as their primary mode of communication, yet fewer than 5% of the hearing population understand basic sign gestures. Existing automated translation frameworks rely predominantly on cloud-hosted multimodal deep learning APIs, which introduce severe operational bottlenecks: elevated latency (800ms–2000ms), prohibitive per-call financial costs, strict API rate-limit quota exhaustion (HTTP 429 errors), and privacy vulnerabilities arising from continuous video streaming. This paper presents SignLearn, an edge-computing assistive platform that executes real-time sign recognition, dynamic kinematics tracking, and synthetic voice generation entirely client-side within the web browser. Leveraging Google MediaPipe Tasks Vision compiled to WebAssembly (WASM) with WebGL GPU acceleration, SignLearn extracts 21 three-dimensional anatomical hand landmarks at 30–60 frames per second. By introducing rotation-invariant 3D joint angle formulations, dynamic palm-span scale normalization, an 800ms temporal motion buffer with direction-reversal kinematics, and dual-hand spatial topology, the system resolves complex static, dynamic, body-directed, and compound two-handed signs. Recognized tokens are stabilized via rolling majority voting and synthesized into natural audible speech via the native Web Speech API. Empirical benchmarking reveals that SignLearn eliminates network latency (reducing response time from 1200ms to <25ms), operates at zero server cost, guarantees total client-side privacy, and achieves an average recognition accuracy of 95.1% across conversational vocabularies.},
        keywords = {Sign Language Recognition, Edge Computer Vision, Google MediaPipe, WebAssembly, Biomechanical Kinematics, Temporal Motion Tracking, Assistive Technology, Text-to-Speech (TTS), Human-Computer Interaction (HCI).},
        month = {October},
        }