Eye-Blink-Controlled Wheelchairs: A Critical Review of Blink Detection, Command Encoding, and Safety in EOG-Based Prototypes

A blink is one of the few voluntary actions left to wheelchair users who cannot operate a joystick, and electrooculography (EOG) records it cheaply with skin electrodes. Yet people blink all day without meaning to, so a blinkcontrolled chair must tell a chosen blink from a reflex. This review reads 25 papers. Twelve of them are eye-controlled wheelchair studies, published between 2007 and 2024 where a year is shown, which are analyzed in detail together with the blink-based systems they describe; the other thirteen give background and related methods. For each study it notes the sensing arrangement, blink detection, the job given to the blink, safety provisions and test method. Six of the twelve use a blink or sustained eye closure as a command (toggle, stop, reverse or mode switch), one treats the blink purely as an artifact to be rejected, and the rest report no blink command. Chosen and reflex blinks are separated by peak amplitude, pulse duration, timing windows or, in one study, a pruned decision tree. Reported command success lies between 87 and 100 percent, but the figures come from incompatible protocols and small groups, and none of the studies reports false activations from involuntary blinks per unit time, tests with motor-impaired users or measures eye fatigue. The paper closes with proposals for a shared blink-command benchmark, fail-safe command assignment and layered safety design.