Designing a passive brain computer interface using real time classification of functional near-infrared spectroscopy

Abstract

Passive brain-computer interfaces consider brain activity as an additional source of information, to augment and adapt the interface instead of controlling it. We have developed a software system that allows for real time brain signal analysis and machine learning classification of affective and workload states measured with functional near-infrared spectroscopy (fNIRS) called the online fNIRS analysis and classification (OFAC). Our system reproduces successful offline procedures, adapting them for real-time input to a user interface. Our first evaluation compares a previous offline analysis with our online analysis. While results show an accuracy decrease, they are outweighed by the new ability of interface adaptation. The second study demonstrates OFAC's online features through real-time classification of two tasks, and interface adaptation according to the predicted task. Accuracy averaged over 85%. We have created the first working real time passive BCI using fNIRS, opening the door to build adaptive user interfaces.

Classifying driver workload using physiological and driving performance data

“…Most previous work was either done in a simulator or with a small number of participants. In addition, the methods we described would apply to classifying workload in other contexts, such as game user experience evaluation or passive, adaptive user interfaces (similar to passive BCI work [1,9,35,44]), and our sample size is larger than most papers in those areas. This has implications for broader applications for real-time cognitive load assessment and evaluating user interface technology in the wild, beyond driver user interfaces.…”

A feature-based trust sequence classification algorithm

“…It is one of the essential techniques in text classification [3,41,52,71] and speech recognition [65], and has been employed in recent years in brain-computer interfaces [29,53] and other artificial intelligence applications. Data observed and collected from many environments are sequential in nature which explains the importance of sequence classification and the vast spread of its applications.…”

Brain-based target expansion

“…Girouard et. al built an application which altered music based on task predictions from an fNIRS processing algorithm [17]. Solovey et al's Brainput system adapted in real-time to a scenario where an interactive human-robot system changed its state of autonomy based on whether it detected a particular state of multitasking [46] and Peck et al demonstrated a passive adaptive movie recommendation system [37].…”