IoT-powered wearable assist device for visually impaired

Area of Science:

• Assistive Technology
• Computer Vision
• Internet of Things (IoT)

Background:

• Visual impairments significantly impact mobility and environmental interaction.
• Existing assistive devices often lack real-time environmental perception and adaptive navigation capabilities.
• There is a need for intelligent, wearable solutions to enhance independence for the visually impaired.

Purpose of the Study:

• To develop and evaluate an Internet of Things (IoT)-enabled wearable assistive device for improving obstacle perception, mobility, and environmental interaction for visually impaired individuals.
• To create a unified framework integrating real-time object detection, facial recognition, and obstacle sensing for autonomous navigation.
• To assess the device's effectiveness in facilitating secure and independent navigation.

Main Methods:

• A prototype was developed using an ESP32-CAM module, Arduino microcontroller, ultrasonic sensor, and text-to-speech feedback.
• Employed deep learning (ResNet-50 CNN, YOLOv4) and machine learning (SVM) algorithms for facial and object detection.
• Evaluated system performance using accuracy, precision, recall, and F1-score under diverse environmental conditions.

Main Results:

• The system achieved over 90% recognition accuracy with minimal computational latency, suitable for real-time use.
• The hybrid SVM-YOLOv4 model demonstrated robust multi-object recognition across various lighting and environmental conditions.
• Quantitative results validated the wearable device's effectiveness and adaptability for assistive purposes.

Conclusions:

• The IoT-enabled wearable device shows significant potential to improve mobility, awareness, and independence for the visually impaired.
• Integration of computer vision and IoT enhances practical rehabilitative applications.
• Future work includes user validation to optimize usability and comfort, with potential for broader assistive technology applications.