An open Steady-State Visually Evoked Potentials Dataset for Augmented Reality-based Brain-Computer Interfaces

This paper presents an open benchmark dataset related to Steady-State Visually Evoked Potential (SSVEP) signals elicited by means of an augmented reality (AR) device. In addition to being the first open dataset that combines SSVEP elicited through AR, another important novelty is the fact this dataset was designed with a metrology-driven approach, monitoring and/or controlling influence factors that can impair target recognition. The dataset consists of 8-channel electroencephalogram (EEG) data acquired from 30 healthy subjects (15 experienced) while they performed a cue-guided target selection task. The flickering frequencies ranged from 8 Hz to 15 Hz with an interval of 1 Hz, while the phase difference between two adjacent flickers was set to π/2. The dataset comprises five acquisition cycles for each subject, with each cycle corresponding to signals recorded from all eight flickers, presented by the visual cues in a randomized order. The stimulation duration in each selection task was five seconds. To ensure the dataset’s trust-worthiness, two crucial parameters for effective stimulus recognition were monitored and/or controlled, namely the illuminance of the environment and the frames per second variation of the AR device, which are primary factors influencing the performance of AR-based SSVEP Brain-Computer Interfaces (BCIs). Additionally, acceleration and angular velocity data related to the subjects’ head movements were acquired. In this way, the proposed dataset addresses the gap in the availability of public datasets for AR-based SSVEP BCIs, serving as a benchmark for comparing different algorithms in i) target identification and ii) motion artifact detection. This contribution is crucial for advancing the development of portable BCI systems aligned with the human-centric principles of the Industry 5.0 paradigm.