Flexible Computer Vision for Assistive Autonomy: A Low-Cost, Non-Invasive Edge AI Solution
DOI:
https://doi.org/10.5753/reic.2026.6703Keywords:
Assistive Technology, Computer Vision, Edge AI, Fine-TuningAbstract
In Brazil, millions of people face daily barriers to maintaining their independence due to motor limitations. Although Computer Vision and Artificial Intelligence (AI) have advanced the field of assistive technology, most solutions available on the market are high-cost, inaccessible, and lack standardization for individualized adaptation. Given this scenario, this work proposes the development and evaluation of a low-cost Flexible Assistive Device (DAF), utilizing Edge AI and Computer Vision. The DAF aims to promote the autonomy of individuals with severe motor limitations, allowing for the control of external devices through the detection and interpretation of gestures or facial expressions. The major methodological differential lies in the flexible architecture, which enables individualized customization through remote fine-tuning. The caregiver uses the DAF (based on the Sipeed MAix Bit/Maixduino board) to capture automatically labeled images. Subsequently, these images are sent to a Web Platform for a server to perform the fine-tuning of the pre-trained model (MobileNet 2.5). The new weights are then returned via email to be updated on the device. Compared to existing technologies, the DAF demonstrated significant advantages, being a non-invasive solution, low-cost, and one that does not require a host (auxiliary computer), facilitating its modularity and transport. The system presented a satisfactory response time (1 to 2 seconds), suitable for practical use. The work thus addresses the need for flexible, accessible, and adaptable devices for the specific needs of each user.
Downloads
References
Ariza, J. and Pearce, J. M. (2022). Low-cost assistive technologies for disabled people using open-source hardware and software: A systematic literature review. IEEE Access, 10:124894–124927. DOI: 10.1109/ACCESS.2022.3221449.
Domellöf, F. P. (2019). The extraocular muscles are selectively spared in als. In Hegde, M. L., editor, Amyotrophic Lateral Sclerosis - Recent Advances and Therapeutic Challenges, chapter 6. IntechOpen, London. DOI: 10.5772/intechopen.89504.
González-Cely, A. X., Callejas-Cuervo, M., and Bastos-Filho, T. (2022). Wheelchair prototype controlled by position, speed and orientation using head movement. HardwareX, 11:e00306. DOI: https://doi.org/10.1016/j.ohx.2022.e00306.
Instituto Brasileiro de Geografia e Estatística, . (2022). Censo Demográfico 2022. IBGE, Rio de Janeiro.
Liu, Y., Jiang, F., and Gowda, M. (2020). Application informed motion signal processing for finger motion tracking using wearable sensors. In ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 8334–8338. DOI: 10.1109/ICASSP40776.2020.9053466.
Majaranta, P. and Räihä, K.-J. (2002). Twenty years of eye typing: systems and design issues. In Proceedings of the 2002 Symposium on Eye Tracking Research & Applications, ETRA ’02, page 15–22, New York, NY, USA. Association for Computing Machinery. DOI: 10.1145/507072.507076.
Miller, R. B. (1968). Response time in man-computer conversational transactions. In Proceedings of the December 9-11, 1968, Fall Joint Computer Conference, Part I, AFIPS ’68 (Fall, part I), page 267–277, New York, NY, USA. Association for Computing Machinery. DOI: 10.1145/1476589.1476628.
Naves, E., Rocha, L., and Pino, P. (2012). Alternative communication system for people with severe motor disabilities using myoelectric signal control. In 2012 ISSNIP Biosignals and Biorobotics Conference: Biosignals and Robotics for Better and Safer Living (BRC), pages 1–4. DOI: 10.1109/BRC.2012.6222176.
Pandey, M., Chaudhari, K., Kumar, R., Shinde, A., Totla, D., and Mali, N. (2018). Assistance for paralyzed patient using eye motion detection. In 2018 Fourth International Conference on Computing Communication Control and Automation (ICCUBEA), pages 1–5. DOI: 10.1109/ICCUBEA.2018.8697455.
Rabhi, Y., Mrabet, M., and Fnaiech, F. (2018). A facial expression controlled wheelchair for people with disabilities. Computer Methods and Programs in Biomedicine, 165:89–105. DOI: https://doi.org/10.1016/j.cmpb.2018.08.013.
Rafal, T., Grzegorz, W., Piotr, G., Nikodem, C., and Sebastian, L. (2023). Edge devices inference performance comparison. Journal of Computing Science and Engineering, 17(2):51–59. DOI: 10.5626/jcse.2023.17.2.51.
Rosero-Montalvo, P. D., Tözün, P., and Hernandez, W. (2024). Optimized cnn architectures benchmarking in hardware-constrained edge devices in iot environments. IEEE Internet of Things Journal, 11(11):20357–20366. DOI: 10.1109/JIOT.2024.3369607.
Torres-Sánchez, E., Alastruey-Benedé, J., and Torres-Moreno, E. (2020). Developing an ai iot application with open software on a risc-v soc. In 2020 XXXV Conference on Design of Circuits and Integrated Systems (DCIS), pages 1–6. DOI: 10.1109/DCIS51330.2020.9268645.
Wang, M.-Y., Kogkas, A. A., Darzi, A., and Mylonas, G. P. (2018). Free-view, 3d gaze-guided, assistive robotic system for activities of daily living.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 The authors
This work is licensed under a Creative Commons Attribution 4.0 International License.
