IRIS: A Real-Time Gaze-Controlled Wheelchair System for Individuals with Severe Motor Impairments Using Intensity-Based Iris Localization

Abstract

Millions of people worldwide suffer from motor impairments that have a significant impact on their independence and mobility. Standard wheelchairs need residual limb function and are not accessible to people who do not have it - for example, people with advanced-stage muscular dystrophy, quadriplegia, and ALS. In this paper, we describe IRIS, an affordable, non-invasive, real-time, gaze-controlled wheelchair navigation system to fill this void. This system combines intensity-based iris localization implemented by OpenCV with an L298N motor driver and an Arduino Uno microcontroller to convert gaze direction to the commands of a wheelchair. Under laboratory conditions indoors, the accuracy of gaze classification (five classes) is 87% on average, while the latency of the entire system is less than 200 milliseconds (ms) and the processing rate of the frame is 20-30 frames per second (fps). The three-sensor HC-SR04 ultrasonic obstacle detection module was able to ensure collision avoidance with $\mathbf{1 0 0 \%}$ reliability in a $\mathbf{3 0 ~ c m}$ safety range. The proposed system demonstrates the construction of an effective and low-cost mobility assistive platform without relying on computationintensive deep learning models, and provides a platform in which the embedded inference and advanced gaze estimation algorithms can be integrated in the future.