Search

Functional Near-Infrared Spectroscopy (fNIRS) has shown promise for being potentially more suitable (than e.g. EEG) for brain-based Human Computer Interaction (HCI). While some machine learning approaches have been used in prior HCI work, this paper explores different approaches and configurations for classifying Mental Workload (MWL) from a continuous HCI task, to identify and understand potential limitations and data processing decisions. In particular, we investigate three overall approaches: a logistic regression method, a supervised shallow method (SVM), and a supervised deep learning method (CNN). We examine personalised and generalised models, as well as consider different features and ways of labelling the data. Our initial explorations show that generalised models can perform as well as personalised ones and that deep learning can be a suitable approach for medium size datasets. To provide additional practical advice for future brain-computer interaction systems, we conclude by discussing the limitations and data-preparation needs of different machine learning approaches. We also make recommendations for avenues of future work that are most promising for the machine learning of fNIRS data.

In the overlap between Human-Computer Interaction (HCI) and Cinematics, sits an interest in physiological responses to experiences. Focusing particularly on brain data, Neurocinematics has emerged as a research field using Brain-Computer Interface (BCI) sensors. Where previous work found inter subject correlations (ISC) between brain measurements of people watching movies in constrained conditions using functional magnetic resonance imaging (fMRI), we seek to examine similar responses in more naturalistic settings using functional Near Infrared Spectroscopy (fNIRS). fNIRS has been shown to be highly suitable for HCI studies, being more portable than fMRI and more tolerant of many natural movements than Electroencephalography (EEG). Early results found significant ISC, which gives a lot of hope and potential for using fNIRS in Neurocinematics.

We present the first stage of our on-going artist-driven BCI collaboration, where we equipped an artist with the brain scanning technique functional Near Infrared Spectroscopy (fNIRS) in order to record mental workload levels during her creative practice. The artists are interested in exposing the hidden cognitive processes involved in their creative practice, in order to reuse or integrate the data into their performances. The researchers are interested in collecting unstructured ‘in the wild’ fNIRS data, and to see how the artists interpret the data retrospectively. We highlight some interesting early examples from the data and describe our on-going plans. We will have completed a second data collection before the workshop.

Recent work has demonstrated that functional Near-Infrared Spectroscopy has the potential to measure changes in Mental Workload with increasing ecological validity. It is not clear, however, whether these measurements are affected by anxiety and stress of the workload, where our informal observations see some participants enjoying the workload and succeeding in tasks, while others worry and struggle with the tasks. This research evaluated the effects of stress on fNIRS measurements and performance, using the Montreal Imaging Stress Task to manipulate the experience of stress. While our results largely support this hypothesis, our conclusions were undermined by data from the Rest condition, which indicated that Mental Workload and Stress were often higher than during tasks. We hypothesize that participants were experiencing anxiety in anticipation of subsequent stress tasks. We discuss this hypothesis and present a revised study designed to better control for this result.

Recent efforts have shown that functional near-infrared spectroscopy (fNIRS) has potential value for brain sensing in HCI user studies. Research has shown that, although large head movement significantly affects fNIRS data, typical keyboard use, mouse movement, and non-task-related verbalisations do not affect measurements during Verbal tasks. This work aims to examine the Reliability of fNIRS, by 1) confirming these prior findings, and 2) significantly extending our understanding of how artefacts affect recordings during Spatial tasks, since much of user interfaces and interaction is inherently spatial. Our results show that artefacts have a significantly different impact during Verbal and Spatial tasks. We contribute clearer insights into using fNIRS as a tool within HCI user studies.

The Think Aloud Protocol (TAP) is a verbalisation technique widely employed in HCI user studies to give insight into user experience, yet little work has explored the impact that TAPs have on participants during user studies. This paper utilises a brain sensing technique, fNIRS, to observe the effect that TAPs have on participants. Functional Near-Infrared Spectroscopy (fNIRS) is a brain sensing technology that offers the potential to provide continuous, detailed insight into brain activity, enabling an objective view of cognitive processes during complex tasks. Participants were asked to perform a mathematical task under 4 conditions: nonsense verbalisations, passive concurrent think aloud protocol, invasive concurrent think aloud protocol, and a baseline of silence. Subjective ratings and performance measures were collected during the study. Our results provide a novel view into the effect that different forms of verbalisation have on workload during tasks. Further, the results provide a means for estimating the effect of spoken artefacts when measuring workload, which is another step towards our goal of proactively involving fNIRS analysis in ecologically valid user studies.