Dataset for the Paper "Comparing Eye-Tracking and SSVEP-BCI Interfaces for Target Selection with a Projected Augmented Reality System"

Description

This is the dataset of the paper titled "Comparing Eye-Tracking and SSVEP-BCI Interfaces for Target Selection with a Projected Augmented Reality System".

The main goal is to assess two different eye-tracking systems and one SSVEP-BCI durin some static and moving tasks in Projected Augemented Reality environments.

Abstract

Problem. Projected Augmented Reality (PAR) systems overlay virtual assets onto real-world horizontal surfaces in an asymmetric manner, making interaction performance highly dependent on user–surface distance. This limitation disproportionately affects people with severe motor disabilities, who cannot close this distance through voluntary movement.
Aim. We investigate and compare three hands‑free input modalities, two eye‑tracking dwell‑based systems (300 ms and 2.5 s dwell times) and a steady‑state visually evoked potential Brain–Computer Interface (SSVEP‑BCI), for selecting both static and moving targets in an asymmetric PAR environment.
Methods. We first derived appropriate metrics by adapting the Shannon Formulation for Fitts’ equation and the Hoffman variation to asymmetric target selection tasks. Then, in a within‑subjects study, 24 participants performed target‑selection tasks under static and moving target conditions using each input modality. Performance metrics included selection accuracy, task completion time, information throughput, and subjective user preference.
Results. The slow eye‑tracking interface (2.5 s) yielded the best balance between error rate and throughput for static target selection, whereas the fast dwell‑based interface (300 ms) excelled at moving‑target selection. The SSVEP‑BCI achieved an average throughput of approximately 1 bit/s, comparable to desktop‑based SSVEP systems, and outperformed the slow dwell interface when selecting moving targets.
Conclusion. Slow and fast dwell‑based eye‑tracking interfaces offer versatile solutions for static and dynamic target selection, respectively, while the SSVEP‑BCI represents a compelling alternative to slow dwell control for moving‑target interactions

Dataset

The dataset is composed of three different .xlsx files:

1. user_data.xlsx
   1. data related to the users' knowledge of PAR systems and their expertise with the NextMind and the HoloLens 2 devices. 
   2. data related to the users' eyesight
2. static_data.xlsx: data collecetd during the static task
   1. time out error
   2. wrong selection error
   3. throughput
   4. ease of use
   5. adoption
   6. fatigue
3. moving_data.xlsx: data collecetd during the moving task
   1. time out error
   2. wrong selection error
   3. throughput
   4. ease of use
   5. adoption
   6. fatigue