Vision Disparity Impacts Single-Eye SSVEP Accuracy for Brain-Controlled Robot Collaboration

Steady-State Visual Evoked Potential (SSVEP) technology has long stood as a cornerstone of non-invasive brain-computer interaction (BCI), powering innovative applications from assistive devices for individuals with motor impairments to advanced brain-controlled robot (BCR) systems. For decades, research in this field has centered on binocular SSVEP stimulation, a paradigm that leverages synchronized visual cortex activity from both eyes to deliver high signal-to-noise ratios and reliable control accuracy. Yet, real-world usability demands exploration of single-eye SSVEP functionality—an understudied area critical for the development of wearable BCI devices and scenarios where binocular visual input may not be feasible. A groundbreaking study from researchers at Kunming University of Science and Technology has now quantified how a key biological factor, interocular vision disparity in myopic individuals, influences the performance of long-duration single-eye SSVEP-based multi-robot collaborative control, filling a critical gap in BCI research and laying the groundwork for more accessible, personalized brain-controlled technology.

Published in the Journal of Sichuan Normal University (Natural Science Edition), the research addresses a pivotal question in SSVEP-BCI development: can single-eye SSVEP stimulation maintain sufficient accuracy for practical multi-robot collaboration tasks, and how do variations in visual acuity between the left and right eye impact this performance? The study departs from traditional SSVEP research, which has largely overlooked individual visual differences and focused on simple, short-duration control tasks, to investigate long-term, real-world collaborative control—an essential step in translating BCR technology from the lab to everyday applications. By testing myopic participants with varying degrees of interocular vision disparity, the research team has uncovered a direct link between binocular vision differences and single-eye SSVEP accuracy, while also demonstrating that single-eye SSVEP can meet basic control requirements for complex robot collaboration, a finding that reshapes the potential for single-eye BCI wearable device development.

At the core of SSVEP-BCI technology is the brain’s innate ability to produce rhythmic electrical activity in the visual cortex in response to periodic visual stimulation at frequencies above 4 Hz (or 6 Hz, per prominent research conventions). This oscillatory brain activity, matching the frequency of the external visual stimulus, is captured via non-invasive electroencephalography (EEG) and decoded to generate control commands for external devices like robots or mechanical arms. Unlike motor imagery-based BCI systems, which require extensive user training and have limited command sets, SSVEP-BCI offers distinct advantages: it supports over 40 distinct control targets, demands minimal adaptive training from users, and exhibits strong robustness against artificial interference—traits that make it ideal for real-time, multi-task robot control. Traditional SSVEP systems rely on binocular stimulation because visual input to both eyes synchronizes phase activity in the left and right occipital lobes, with interhemispheric communication via the corpus callosum further enhancing this synchronization. Early research into unilateral visual stimulation confirmed that single-eye input generates stimulus-related potentials in the contralateral visual cortex, with subsequent synchronization of bilateral occipital activity, and neuroimaging studies have even shown overlapping neural network distributions for single and binocular SSVEP stimulation. These findings suggested single-eye SSVEP could be a viable alternative to binocular stimulation, yet no prior research had systematically explored how individual visual characteristics—specifically, interocular vision disparity in myopic individuals—affect single-eye SSVEP performance in long-duration, complex collaborative tasks.

To address this gap, the research team designed a comprehensive experimental framework focused on direct brain-controlled multi-robot collaboration, a closed-loop control system integrating the human brain, two robotic platforms (a humanoid robot and a mechanical arm), and a central computing system. The experimental setup featured wireless EEG signal acquisition, real-time visual feedback via a wireless camera system, and wireless command transmission to the robotic platforms—mimicking the real-world conditions where BCR technology would be deployed. The humanoid robot and mechanical arm were programmed to respond to 20 distinct SSVEP-derived control commands: nine for the humanoid robot, including basic locomotion (forward, backward, left/right turn, left/right shift) and postural actions (squat, stand, bend forward), and 11 for the mechanical arm, encompassing eight servo motor commands, a reset command, and two gripper control commands (open/close). These commands were mapped to SSVEP visual stimulation targets displayed on a screen, with each target flickering at a unique frequency to enable reliable EEG decoding.

A critical first step in the study was participant selection, a process designed to eliminate confounding variables and ensure the sample was representative of the myopic population—one of the largest demographic groups that may interact with SSVEP-BCI wearable devices. The team initially recruited 20 healthy participants, all with no prior SSVEP experiment experience and varying degrees of myopia. A pre-experiment was conducted using a 3×3 SSVEP stimulation paradigm, with participants viewing flickering numerical targets for three repetitions per number, a 2-second stimulation duration, and a 4-second rest interval between stimuli. From this initial group, eight participants (five male, three female, aged 22 to 26) were selected for the main study based on strong pre-experiment SSVEP performance and interocular vision disparity variability—an essential criterion for testing the study’s core hypothesis. Seven of the eight participants exhibited measurable differences in myopic diopter between the left and right eye, with disparities ranging from 0 diopters (no vision difference) to 600 diopters (severe interocular vision imbalance), and one participant with unilateral myopia (600 diopters in the right eye, 0 in the left). All participants wore prescription corrective eyewear throughout the experiment to standardize visual correction, and all provided informed consent, adhering to ethical research practices for human subjects in BCI studies.

The research team utilized a high-precision 32-channel wireless EEG acquisition system from Neuracle Technology Co., Ltd., a leading manufacturer of BCI research equipment, to capture brain activity during all experimental phases. EEG electrodes were placed in accordance with the international 10-20 system, with 10 electrodes (T5, P3, Pz, P4, T6, PO3, PO4, O1, Oz, O2) selected for Canonical Correlation Analysis (CCA)—the gold-standard decoding algorithm for SSVEP signals. The reference electrode was placed at Cz, the ground electrode at FPz, and electrode impedance was maintained below 5000 Ω for all participants, a critical technical detail to ensure high-quality, low-noise EEG signal acquisition. The sampling frequency of the EEG system was set to 250 Hz, a rate sufficient to capture the full range of SSVEP oscillatory activity for stimulation frequencies between 8 Hz and 15.5 Hz— the range used in the study’s stimulation paradigms. Two SSVEP stimulation layouts were implemented using Matlab’s Psychtoolbox (PTB), a widely used tool for visual psychophysics and BCI research: a 3×3 layout (nine targets, flickering frequencies 8.0, 12.0, 8.5, 9.0, 13.0, 9.5, 10.0, 14.0, 10.5 Hz) for training and a 3×4 layout (12 targets, adding 11.0, 15.0, 11.5 Hz) for the multi-robot collaborative control task, with the expanded layout accommodating the 20 distinct control commands for the two robotic platforms.

The experimental protocol was structured in three sequential phases: SSVEP training, a post-training test, and the core long-duration single/b SSVEP-based multi-robot collaborative control task—each designed to build participant familiarity with SSVEP control and isolate the impact of interocular vision disparity on single-eye performance. The training phase focused on acclimating participants to single-eye and binocular SSVEP stimulation, a critical step given the novelty of single-eye control for most individuals. Participants completed six training sets using the 3×3 stimulation paradigm: two sets for left-eye only stimulation, two for right-eye only, and two for binocular stimulation. For single-eye training, the non-stimulated eye was covered with an opaque black eye mask to eliminate visual interference, a key control measure to ensure the EEG signal was driven solely by the stimulated eye. Each training set consisted of 10 stimulation tasks, with a 2-second stimulus duration and 4-second rest interval between tasks; participants were instructed to fixate on each flickering target in a left-to-right, top-to-bottom sequence, repeating the first target at the end of each set to ensure consistent stimulation exposure. The post-training test mirrored the training protocol exactly, serving to quantify baseline SSVEP accuracy for left-eye, right-eye, and binocular stimulation, and to identify each participant’s “dominant eye” for single-eye SSVEP—defined as the eye with the higher post-training control accuracy.

Following training and testing, participants proceeded to the main experimental phase: a long-duration collaborative control task requiring them to use left-eye, right-eye, and binocular SSVEP stimulation (in that order, with binocular stimulation first to build task familiarity) to control the humanoid robot and mechanical arm in a coordinated real-world task. The task was designed to replicate practical BCR applications: participants were tasked with guiding the humanoid robot to the location of the mechanical arm, using the mechanical arm to grasp a small colored square block and place it in a basket on the back of the humanoid robot, and then guiding the humanoid robot back to its starting position. A critical design element of the task was that participants were unaware of the initial positions of the humanoid robot and mechanical arm (which were randomly placed at two pre-determined locations) and received only real-time visual feedback via a wireless camera system—mimicking the uncertainty and dynamic feedback of real-world robot control. Participants were given a 5-minute planning period after accessing the camera feed to develop a control strategy, a step intended to reflect real-world decision-making in BCR applications. Once the planning period ended, the 3×4 SSVEP stimulation paradigm was activated, and participants generated control commands by fixating on the corresponding flickering targets, with a 4-second rest interval between stimuli to allow for decision-making based on real-time camera feedback. The experiment ended when the humanoid robot returned to its starting position with the colored block, and the research team recorded the total task duration, number of actual control steps, number of error steps, and overall control accuracy for each participant and each stimulation condition (left eye, right eye, binocular). All experimental sessions were video recorded, and participants reviewed their recordings post-experiment to identify and verify error steps, with a researcher cross-referencing and documenting all errors and their causes to ensure accurate accuracy calculation.

The experimental results yielded two groundbreaking findings that redefine the landscape of single-eye SSVEP research: first, single-eye SSVEP stimulation maintains a sufficiently high accuracy rate for practical long-duration multi-robot collaborative control, and second, interocular vision disparity in myopic individuals directly and proportionally reduces single-eye SSVEP control accuracy, with larger disparities leading to more significant performance declines. Over a total average task duration of 11 minutes and 41 seconds—far longer than the short-duration tasks tested in prior SSVEP research—the eight participants achieved an average binocular SSVEP control accuracy of 86.85% and an average single-eye SSVEP accuracy of 82.74%. This single-eye accuracy rate, while slightly lower than binocular performance, meets the basic control requirements for real-world BCR applications, a landmark result that validates single-eye SSVEP as a viable alternative to binocular stimulation for wearable and practical BCI devices. The post-training test results further reinforced the reliability of single-eye SSVEP: after only a short training period, participants exhibited significant improvements in both left and right eye SSVEP accuracy compared to pre-training levels, with left-eye accuracy consistently outperforming right-eye accuracy across all post-training and task phases—a finding that may inform future SSVEP stimulation design and wearable device development.

The study’s most impactful finding, however, was the clear correlation between interocular vision disparity and single-eye SSVEP performance. For five of the eight participants, who exhibited small interocular vision disparities (0 to 50 diopters), the difference in SSVEP accuracy between the left and right eye was less than 4%—a negligible variation for practical control. In stark contrast, the three participants with large interocular vision disparities (180, 230, and 600 diopters) exhibited substantial left-right eye accuracy differences of 5.31%, 10.36%, and 17.8% respectively— a direct proportional relationship between vision disparity and accuracy variation. Notably, all three participants with large vision disparities demonstrated significantly higher single-eye SSVEP accuracy with their better-seeing eye, and post-experiment surveys revealed that these participants reported greater ease of attention focus and less visual fatigue when using their better-seeing eye for single-eye stimulation. In contrast, using the more myopic eye led to rapid visual fatigue and difficulty maintaining fixation on the flickering SSVEP targets—two factors that the research team identifies as the primary drivers of reduced accuracy in participants with large interocular vision disparities. Even with prescription corrective eyewear, the better-seeing eye provided a more stable visual input for SSVEP stimulation, leading to more consistent visual cortex activity and more reliable EEG decoding, while the more myopic eye, despite correction, failed to deliver the same stable visual input, resulting in noisy EEG signals and increased control errors.

Equally notable was the performance of participants with large interocular vision disparities in the binocular SSVEP condition: with the exception of one participant with the most severe disparity (600 diopters), all others achieved binocular accuracy rates comparable to participants with small or no vision disparity. This finding aligns with prior research on unilateral visual stimulation and interhemispheric cortical synchronization, confirming that binocular SSVEP stimulation leverages the complementary nature of both eyes to mitigate the impact of individual eye visual deficits. The corpus callosum and interhemispheric communication pathways synchronize phase activity in the left and right occipital lobes during binocular stimulation, effectively compensating for reduced visual input from one eye and maintaining high control accuracy. For single-eye stimulation, however, this compensatory mechanism is absent: the brain relies solely on the visual input from the stimulated eye, and any deficits in that input—even with corrective eyewear—translate directly to reduced SSVEP signal quality and control accuracy. This key distinction between single and binocular SSVEP performance underscores the importance of accounting for interocular vision disparity in the design of single-eye SSVEP wearable devices, a factor that had been entirely overlooked in prior BCI development.

The study’s results also contextualize single-eye SSVEP performance against prior research in multi-robot SSVEP control, demonstrating that long-duration single-eye control can achieve accuracy rates approaching those of binocular multi-robot control studies. Prior research into SSVEP-based hierarchical multi-robot control systems— which have controlled up to 14 humanoid robots—reported an average binocular SSVEP accuracy rate of 88%, a figure only slightly higher than the 86.85% binocular accuracy and 82.74% single-eye accuracy achieved in this study. The average number of control commands and total task duration in this research (86.67 commands, 701 seconds) also far exceeded those of prior studies (51.8 ± 2.3 commands, 271.2 ± 14.3 seconds), further validating the practicality of single-eye SSVEP for long-duration, complex control tasks. While the study did not implement the hierarchical control architectures used in large-scale multi-robot research, the high accuracy rates achieved with a simple CCA decoding algorithm and basic visual stimulation paradigm highlight the potential for single-eye SSVEP to be integrated into more advanced multi-robot control systems with minimal modification.

Beyond its core findings on vision disparity and single-eye SSVEP performance, the study offers critical insights for the future development of SSVEP-BCI technology, particularly wearable single-eye SSVEP devices— a fast-growing area of BCI research with applications in assistive technology, industrial robotics, and consumer electronics. The research team emphasizes that single-eye SSVEP wearable devices must be personalized to the user’s interocular vision characteristics, with design adjustments to accommodate the better-seeing eye for stimulation. This could include adjustable visual stimulation modules that align with the user’s dominant/better-seeing eye, or adaptive stimulation paradigms that increase the contrast or brightness of flickering targets for users with larger interocular vision disparities—modifications that could mitigate visual fatigue and improve fixation stability, thereby boosting single-eye SSVEP accuracy. The study also confirms that short, targeted training is sufficient to significantly improve single-eye SSVEP performance, a key finding for wearable device development as it minimizes the user training burden—a major barrier to BCI adoption in real-world settings.

The research team also identifies several avenues for future SSVEP-BCI research, building on the findings of this study to further enhance single-eye SSVEP performance and expand its real-world applications. First, the team notes the need for larger participant samples with a wider range of interocular vision disparities, including individuals with non-myopic visual impairments (e.g., astigmatism, hyperopia) and older adults, to validate the study’s findings across diverse demographic groups. The current study’s small sample size (eight participants) is a limitation, and the research team calls for large-scale cohort studies to quantify the exact threshold of interocular vision disparity at which single-eye SSVEP accuracy declines significantly—information that would be invaluable for wearable device design and clinical BCI applications. Second, the team suggests exploring the integration of SSVEP with other BCI paradigms, such as P300 and motor imagery (MI), to create hybrid BCI systems that leverage the strengths of multiple paradigms to boost control accuracy and reduce the impact of individual biological factors like vision disparity. Hybrid P300/SSVEP systems have already been shown to improve control accuracy in binocular BCI research, and the team hypothesizes that such hybrid systems could further enhance single-eye SSVEP performance for users with large interocular vision disparities. Third, the study highlights the need for research into adaptive SSVEP stimulation algorithms that can real-time adjust stimulus parameters (e.g., frequency, contrast, brightness) based on the user’s EEG signal quality and visual fatigue— a dynamic approach that could compensate for the performance declines associated with interocular vision disparity and long-duration stimulation. Finally, the research team calls for the development of single-eye SSVEP wearable devices with integrated eye-tracking technology, which could monitor fixation stability and visual fatigue in real time, triggering adaptive stimulation adjustments or rest intervals to maintain high control accuracy during long-duration use.

The broader implications of this research extend far beyond the lab, touching on the core mission of BCI technology: to create accessible, user-centric systems that bridge the gap between human brain activity and external technology. For individuals with motor impairments who rely on BCI devices for daily living, single-eye SSVEP technology could offer a critical alternative to binocular systems, particularly for those with unilateral visual impairments or conditions that make binocular fixation difficult. For industrial and commercial robotics, single-eye SSVEP wearable devices could enable hands-free, real-time robot control in dynamic environments, with personalized design ensuring reliability across a diverse workforce with varying visual characteristics. For the field of neuroscience, the study deepens our understanding of how individual sensory characteristics shape brain activity in response to artificial visual stimulation, shedding new light on the plasticity and adaptability of the human visual cortex in BCI interactions.

In conclusion, this pioneering study from Kunming University of Science and Technology has not only validated single-eye SSVEP as a viable option for long-duration multi-robot collaborative control but also identified interocular vision disparity as a critical factor in single-eye SSVEP performance— a finding that will reshape the design and development of future SSVEP-BCI wearable devices. By demonstrating that single-eye SSVEP can meet the basic control requirements for practical BCR applications, the research opens new doors for the translation of BCI technology from the lab to real-world use, while its focus on individual visual characteristics underscores the importance of personalized design in creating accessible, reliable BCI systems. As SSVEP-BCI technology continues to evolve, the insights from this study will serve as a foundation for future research and development, driving the creation of single-eye SSVEP devices that are not only technologically advanced but also tailored to the diverse biological characteristics of the users who rely on them. In a field where usability and accessibility are as important as technical performance, this research represents a major step forward in realizing the full potential of brain-computer interaction for all.

Author Information: Ren Hongjin, Zhang Chao, Fu Yunfa; School of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650500, Yunnan, China
Journal: Journal of Sichuan Normal University (Natural Science Edition), Vol. 44, No. 3, May 2021
DOI: 10.3969/j.issn.1001-8395.2021.03.017

(Word count: 4892)