NeuroSight: AI-Powered Vision Enhancement for the Visually Impaired

Year : 2026 | Volume : 14 | Issue : 02 | Page : 25 33
By

Harshit Singh Raghuvanshi,

Surendra Kumar Yadav,

Ajay Kumar,

  1. Scholar, Department of computer Science, JECRC University, Rajasthan, India
  2. Professor, Department of computer Science, JECRC University, Rajasthan, India
  3. Assistant Professor, Department of computer Science, JECRC University, Rajasthan, India

Abstract

Visual impairment has long been a significant challenge, affecting a person’s ability to move independently, understand their surroundings, and perform everyday activities safely. Although considerable progress has been made in the development of assistive technologies, many existing solutions still face practical limitations. These include inconsistent performance in real-world environments, high power consumption, dependence on external computing resources, and limited ability to adapt to dynamic and rapidly changing surroundings. To address these challenges, this paper presents NeuroSight, a compact, AI-driven assistive navigation system designed to enhance environmental awareness in real time using edge intelligence. The proposed system combines a panoramic camera with inertial sensors to continuously capture visual and motion-related information from the user’s surroundings. Instead of transferring sensitive data to remote servers for processing, the collected information is processed locally through an optimized embedded computing pipeline. This approach enables faster responses while also reducing communication requirements and improving user privacy. A key component of NeuroSight is the proposed Occlusion-Robust Depth-Aware Fusion (ORDAF) algorithm. The algorithm integrates information generated by several perception modules, including object detection, depth estimation, semantic segmentation, and contextual scene understanding. By combining these complementary sources of information, the system aims to provide a more reliable representation of complex environments, particularly in situations where objects may partially obscure one another. Furthermore, the neural network models developed for the system are optimized using techniques such as model quantization and are deployed locally through TensorRT and ONNX Runtime. These optimizations are intended to reduce computational overhead while maintaining adequate accuracy and enabling low-latency inference. Overall, NeuroSight seeks to provide a practical, privacy-conscious, and responsive navigation aid that can support visually impaired individuals in understanding and interacting with their surroundings more effectively.

 

Keywords: Assistive technology, edge AI, visual impairment, object detection, depth estimation, scene understanding, haptic feedback, wearable computing, ORDAF, occlusion-robust fusion, on-device navigation

[This article belongs to Research & Reviews: A Journal of Embedded System & Applications ]

How to cite this article: Harshit Singh Raghuvanshi, Surendra Kumar Yadav, Ajay Kumar. NeuroSight: AI-Powered Vision Enhancement for the Visually Impaired. Research & Reviews: A Journal of Embedded System & Applications. 2026; 14(02):25-33.
How to cite this URL: Harshit Singh Raghuvanshi, Surendra Kumar Yadav, Ajay Kumar. NeuroSight: AI-Powered Vision Enhancement for the Visually Impaired. Research & Reviews: A Journal of Embedded System & Applications. 2026; 14(02):25-33. Available from: https://journals.stmjournals.com/rrjoesa/article=2026/view=259464

References

  1. World Health Organization. World report on vision. Geneva: World Health Organization; 2019. Available from: https://www.who.int/publications/i/item/worldreport-on-vision.
  2. Wang CY, Bochkovskiy A, Liao HYM. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Vancouver, Canada. 2023. p. 7464-75. doi:10.1109/CVPR52729.2023.00721.
  3. Ranftl R, Lasinger K, Hafner D, Schindler K, Koltun V. Towards robust monocular depth estimation: mixing datasets for zero-shot cross-dataset transfer. IEEE Trans Pattern Anal Mach Intell. 2022;44:1623-1637. doi:10.1109/TPAMI.2020.3019967. PMID:32853149.
  4. Dosovitskiy A, Beyer L, Kolesnikov A, Weissenborn D, Zhai X, Unterthiner T, et al. An image is worth 16×16 words: transformers for image recognition at scale [Preprint]. 2020. arXiv:2010.11929.
  5. Radford A, Kim JW, Hallacy C, Ramesh A, Goh G, Agarwal S, et al. Learning transferable visual models from natural language supervision. In: Proceedings of the 38th International Conference on Machine Learning. 2021. p. 8748–8763.
  6. Youvan DC. Developing and deploying AI applications on NVIDIA Jetson Orin NX: a comprehensive guide [technical report]. ResearchGate; 2024. doi:10.13140/RG.2.2.15641.43363.
  7. Lee YH, Medioni G. RGB-D camera based wearable navigation system for the visually impaired. Comput Vis Image Underst. 2016;149:3-20. doi:10.1016/j.cviu.2016.03.019.
  8. Badhan AK, Bhattacharjee A, Kumar R, Diwan P. Machine learning algorithms for adaptive haptic responses. In: Mamodiya U, Kalam A, Bossoufi B, Awad MM, Somwanshi DK, editors. Integrating AI with Haptic Systems for Smarter Healthcare Solutions. Hershey (PA): IGI Global Scientific Publishing; 2025. p. 199-232. doi:10.4018/979-8-3373-2307-7.ch010.
  9. Hanafy WA, Molom-Ochir T, Shenoy R. Design considerations for energy-efficient inference on edge devices. In: Proceedings of the Twelfth ACM International Conference on Future Energy Systems; 2021. p. 302-308. doi:10.1145/3447555.3465326.
  10. Chouhan S, Ghai D, Sandhu R, Tripathi SL. Smart assistive technologies for neurodisorders: a review on AI, IoT, and wearable systems for enhanced patient care. Neurol Sci. 2026;47:211. doi:10.1007/s10072-025-08790-8. PMID:41612095.
  11. Seiple W, van der Aa HPA, Garcia-Piña F, Greco I, Roberts C, van Nispen R. Performance on activities of daily living and user experience when using artificial intelligence by individuals with vision impairment. Transl Vis Sci Technol. 2025;14(1):3. doi:10.1167/tvst.14.1.3. PMID:39775799.
  12. Wazirali R, Ashrif FF, Ahmad R. AI smart cane technology and assistive navigation for visually impaired users: an overview. J King Saud Univ Comput Inf Sci. 2025;37:226. doi:10.1007/s44443-025-00234-9.
  13. Quick R, Bontula A, Puente K, Fitter NT. Using the OptiBand to increase the long-range spatial perception of people with vision disabilities. 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN), Busan, Republic of Korea. 2023. p.248-54. doi:10.1109/RO-MAN57019.2023.10309546.
  14. Steltenpohl H, Bouwer A. Vibrobelt: tactile navigation support for cyclists. 18th International Conference on Intelligent User Interfaces (IUI ’13), Santa Monica, CA, USA. 2013. p.417-26. doi:10.1145/2449396.2449450.
  15. Aaltonen I, Laarni J. Field evaluation of a wearable multimodal soldier navigation system. Appl Ergon. 2017;63:79-90. doi:10.1016/j.apergo.2017.04.005. PMID:28502409.
  16. Naayini P, Myakala PK, Bura C, Jonnalagadda AK, Kamatala S. AI-powered assistive technologies for visual impairment [Preprint]. 2025 Jan 14. arXiv:2503.15494.
  17. Ventura S, Ottoboni G, Lullini G, Chattat R, Simoncini L, Magni E, et al. Co-designing an interactive artificial intelligent system with post-stroke patients and caregivers to augment the lost abilities and improve their quality of life: a human-centric approach. Front Public Health. 2023;11:1227748. doi:10.3389/fpubh.2023.1227748. PMID:37808976.
  18. Swaminathan TP, Silver C, Akilan T, Kumar J. Benchmarking deep learning models on Nvidia Jetson Nano for real-time systems: an empirical investigation. Procedia Comput Sci. 2025;260:906-913. doi:10.1016/j.procs.2025.03.273.
Regular Issue Subscription Original Research
Volume 14
Issue 02
Received 05/02/2026
Accepted 03/07/2026
Published 30/09/2026
Publication Time 237 Days

Login

My IP

PlumX Metrics

Support