Estimating cognitive workload in an interactive virtual reality environment using EEG

Motivation

BCI has seen various applications in a wide range use-cases, such as:

• enhancing experiences, i.e. influence the experience via the BCI
• analysis tool
• complementing existing features
• ...

The recent surge of affordable, high-performance virtual reality(VR) headsets has lead to a rise of a new platform, which distinctly seperates it from the other traditional media and its usage seems to grow with time. Different hardware options at different price points, empowering customers to choose the one that suits them best, reinforces that point, e.g.:

• VIVE
• Oculus
• Valve Index
• Sony VR

VR headsets represent an enclosed hardware system, which conveniently plays nicely with BCI as it provides an unobtrusive way of mounting EEG sensors. Interestingly enough, this shows that BCI could enhance the VR platform and benefit from the extened or complemented applications, especially in fields where VR is already being used:

• education
• training
• entertainment
• fitness
• ...

Use-case and Goal

The specific use-case this study is focusing on, is the enhancement of VR sessions and to influence the virtual environment based on the cognitive state of the user via EEG. Since the use of VR headsets already require users to mount hardware onto your head, simply adding additional sensors usually does not seem to be problematic and thus provides an easy gateway for BCI, in this case EEG. Although at first glance, this seems like it would simply provide developers another degree of freedom of control, in actuality its adoption, implementation and support is unclear. Due to fact that when keeping in mind that research in BCI and VR are still ongoing and far from being well-known, the question of scope, precision and properties of the new input quickly arises.

The goal of this study is therefore evaluate the feasibility of passively monitoring cognitive workload via EEG, with the intention of hopefully gaining the knowledge along the way. The evaluation will be done in special setting that has intentionally been designed to be more oriented towards practicality, rather than sterile and academic lab settings. Based on the n-back task, 15 participants are asked to perform a certain task in rounds or blocks. The authors used spatio-spectral features, along with some pre- and post-processing steps, to analyze and evaluating the results and could show that it is possible to discriminate between three workload levels.

Table of Contents

1. Background
2. Methodology
3. Results

The source paper can be found here.