A wearable multimodal assistive interface combining EOG, EEG, and IMU signals achieved an average operation accuracy of 87.53 ± 4.92% and an information transfer rate of 62.04 ± 15.93 bits/min, demonstrating feasibility for supporting computer interaction in stroke survivors with upper-limb impairments.
Key Findings
Results
The wearable multimodal assistive interface achieved an average operation accuracy of 87.53 ± 4.92% across stroke patients performing computer tasks.
Thirty stroke patients with upper-limb impairments participated in the experiments.
Tasks included news reading, video playback, and character spelling.
Accuracy was measured across common computer interaction tasks.
The system used EOG for blink-based clicking, IMU for cursor movement, and frontal EEG for attention-based command verification.
Results
The system achieved an average operation time of 3.49 ± 0.49 seconds per interaction.
Operation time was measured across the full set of computer tasks including news reading, video playback, and character spelling.
The sample consisted of 30 stroke survivors with upper-limb impairments.
This metric reflects the speed at which users could complete individual cursor control operations.
Results
The spelling task yielded an information transfer rate of 62.04 ± 15.93 bits/min.
The information transfer rate (ITR) was specifically measured during the character spelling task.
ITR of 62.04 ± 15.93 bits/min reflects both speed and accuracy of communication through the system.
Thirty stroke patients participated in this evaluation.
This metric provides a combined measure of system efficiency relevant for assistive communication applications.
Results
Subjective workload during system use was rated as moderate, with a mean NASA-TLX score of 32.1 ± 5.4.
The NASA Task Load Index (NASA-TLX) was used to assess subjective workload experienced by users.
A score of 32.1 ± 5.4 was interpreted by the authors as indicating 'a moderate subjective workload during system use.'
The assessment was conducted with 30 stroke patients with upper-limb impairments.
This finding suggests the system did not impose a high cognitive or physical burden on users.
Methods
The system used a lightweight headband integrating EOG, EEG, and an inertial measurement unit (IMU) to enable multimodal computer interaction.
EOG signals were processed to detect voluntary blinks to generate mouse clicks.
Head movements captured via IMU were mapped to cursor movements.
Frontal EEG signals were used to estimate attention as an auxiliary mechanism for command verification.
A rapid user-specific calibration procedure was introduced to adapt blink-detection thresholds to individual EOG characteristics without requiring extensive training.
Methods
A rapid user-specific calibration procedure was developed to adapt blink-detection thresholds to individual EOG characteristics.
The calibration was designed to accommodate variability in EOG signal characteristics across individual users.
The procedure was described as not requiring extensive training, aiming to reduce setup burden for stroke patients.
This approach was intended to improve the practicality of deployment in a clinical or home setting.
What This Means
This research suggests that a wearable headband combining three types of biosignals — eye movement signals (EOG), brainwave signals (EEG), and head motion sensors (IMU) — can allow stroke survivors who have lost the use of their arms and hands to control a computer cursor. The headband detects intentional eye blinks to simulate mouse clicks, tracks head movements to move the cursor around the screen, and monitors attention levels to help confirm that commands were intentional. A quick, individualized setup process was built in so the system could adapt to each person's unique eye signal patterns without lengthy training sessions.
In a study of 30 stroke patients with upper-limb impairments, participants were able to perform everyday computer tasks — including reading news articles, watching videos, and spelling words — with an average accuracy of about 87.5% and completing each operation in roughly 3.5 seconds. The character spelling task specifically achieved an information transfer rate of about 62 bits per minute, which is a combined measure of speed and accuracy. Users also reported only moderate mental and physical effort while using the system, as measured by a standard workload questionnaire (NASA-TLX score of 32.1 out of 100).
This research suggests that a low-cost, non-invasive wearable device using everyday biological signals from the head could meaningfully restore computer access for people who have had a stroke and lost upper-limb function. Such a tool could support greater independence in daily digital activities — such as communication, entertainment, and information access — without requiring surgery or extensive technical training. Further work would be needed to test the system's performance over longer periods and in real-world home environments.
Liang Y, Zhang L, Jiang Y, Zhu J, Zhao Y, Qin P, et al.. (2026). A Wearable Multimodal Assistive Interface for Virtual Cursor Control in Stroke Survivors with Upper-Limb Impairment.. Sensors (Basel, Switzerland). https://doi.org/10.3390/s26175436