ExamVox: A Proposed AI-Powered Voice-Based Examination Writing Device for Blind and Hand-Disabled Students

Abstract

Students who are blind or cannot use their hands often depend on human scribes during written examinations, which creates problems of scribe availability, privacy, consistency and independence. This paper presents ExamVox, a proposed dedicated, sealed and offline-capable voice-driven examination writing device. The system reads questions aloud, captures spoken answers, converts speech to editable text through automatic speech recognition (ASR), and lets the student review answers through read-back before submission. The proposed architecture combines a noise-adaptive audio front end, an exam-intent router that separates answers from commands, a text formatter, speech-to-LaTeX for mathematical expressions, voice-to-SVG for simple diagrams, encrypted auto-save and a signed PDF or printed answer sheet. The hardware concept is based on a Raspberry Pi 5 with a microphone, bone-conduction earpiece, display, battery backup and a sealed enclosure, with an estimated basic cost of INR 18,000–30,000. This is a design-stage contribution: no prototype results are claimed, and the paper sets out a plan for measuring recognition accuracy, command accuracy, latency and usability.

Introduction

The paper proposes ExamVox, a voice-controlled examination device designed to help blind students and students with upper-limb disabilities take written examinations independently, without depending on a human scribe.

Traditional examinations create difficulties for these students because finding a suitable scribe can be challenging, the quality of assistance may vary, privacy can be reduced, and spoken answers may be recorded incorrectly.

ExamVox aims to solve these problems through a dedicated, secure and offline-capable device that allows students to listen to questions, speak their answers, correct recognized text and generate a verified answer document.

1. Related Work

Existing voice-based examination systems use speech recognition and text-to-speech technologies to help visually impaired students complete examinations. Previous research has demonstrated that voice commands and audio assistance can improve independent participation. However, most existing solutions operate on computers or smartphones and mainly focus on visually impaired students.

ExamVox addresses this gap by proposing a dedicated device that also supports students who cannot use their hands, spoken mathematics, simple diagrams and secure answer storage.

2. Problem Statement and Objectives

Students with visual or upper-limb disabilities often depend on human scribes, creating challenges related to availability, accuracy, privacy and independence.

The main objectives of ExamVox are:

• Develop a dedicated and tamper-resistant examination device.

• Capture students' spoken answers clearly.

• Convert speech into editable text.

• Support spoken punctuation, mathematics and simple diagrams.

• Read questions and answers aloud.

• Provide answer review, correction and time alerts.

• Securely store answers and generate verifiable documents.

• Evaluate accuracy, response time and usability.

3. Proposed System

ExamVox combines several technologies:

• Automatic Speech Recognition (ASR): Converts spoken answers into text.

• Text-to-Speech (TTS): Reads questions and answers aloud.

• Noise Reduction: Reduces background noise to improve speech recognition.

• Offline Computing: Allows local processing without internet connectivity.

• Encryption and Access Control: Protects examination data and student privacy.

Conclusion

This paper proposed ExamVox, a voice-driven examination writing device for blind students and students without hands. It combines question read-out, spoken answer capture, speech recognition, review, secure storage and verifiable output in one dedicated workflow. The concept responds to scribe availability, privacy and independence concerns, but its performance and suitability for real examinations have not yet been established. Future work is to build a working prototype, verify offline performance on the selected hardware, test recognition under examination-like noise, measure WER and command accuracy, evaluate usability with consented participants, and consult accessibility experts and examination authorities.

References

[1] Dept. of Empowerment of Persons with Disabilities, Govt. of India, “Guidelines for conducting written examination for persons with benchmark disabilities,” 2018. [Online]. Available: https://depwd.gov.in/en/guidelines/ [2] I. H. H. N. Dharmasena and J. A. D. C. A. Jayakody, “Voice-based online examination system for visually impaired students,” in Proc. 2nd Int. Conf. Advanced Research in Computing (ICARC), 2022, pp. 367–372, doi: 10.1109/ICARC54489.2022.9754191. [3] E. C. Budak, N. Yurtay, Y. Budak, and A. Gecer, “Voice-assisted online exam management and system usability analysis with visually impaired students,” Interactive Learning Environments, vol. 31, no. 9, pp. 5508–5522, 2023, doi: 10.1080/10494820.2021.2010219. [4] A. Radford, J. W. Kim, T. Xu, G. Brockman, C. McLeavey, and I. Sutskever, “Robust speech recognition via large-scale weak supervision,” in Proc. 40th Int. Conf. Machine Learning (ICML), PMLR vol. 202, 2023, pp. 28492–28518. [5] R. Raju, R. Aiswarya, S. S. Sambavi, and R. Kanakadas, “Comprehensive survey on voice-enabled semantic grading system for visually impaired students,” in Proc. Int. Conf. Artificial Intelligence and Secure Data Analytics (ICAISDA 2025), 2026, pp. 299–306, doi: 10.2991/978-94-6239-616-6_25. [6] F. N. Tubo, I. E. Onyenwe, and D. C. Asogwa, “Development of an NLP-driven computer-based test guide for visually impaired students,” arXiv:2401.12375, 2024. [7] P. Ramos, M. Zapata, K. Valencia, V. Vargas, and C. Ramos-Galarza, “Low-cost human–machine interface for computer control with facial landmark detection and voice commands,” Sensors, vol. 22, no. 23, Art. 9279, 2022, doi: 10.3390/s22239279. [8] S. Dutta, S. Chandupatla, and J. H. L. Hansen, “Adapting Whisper for lightweight and efficient automatic speech recognition of children for on-device edge applications,” arXiv:2507.14451, 2025.

Copyright

Copyright © 2026 Shagufta Himmat Chikode, Rehana Aslam Killedar, Bibifatima Rasul Bagwan, Ashiya Aslam Mulla. This 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.