Enhancing Human-Computer Interaction with Input from Active and Passive Brain-Computer Interfaces

Abstract

This chapter introduces a formal categorization of BCIs, according to their key characteristics within HCI scenarios. This comprises classical approaches, which we group into active and reactive BCIs, and the new group of passive BCIs. Passive BCIs provide easily applicable and yet efficient interaction channels carrying information on covert aspects of user state, while adding little further usage cost. All of these systems can also be set up as hybrid BCIs, by incorporating information from outside the brain to make predictions, allowing for enhanced robustness over conventional approaches. With these properties, passive and hybrid BCIs are particularly useful in HCI. When any BCI is transferred from the laboratory to real-world situations, one faces new types of problems resulting from uncontrolled environmental factors—mostly leading to artifacts contaminating data and results. The handling of these situations is treated in a brief review of training and calibration strategies. The presented theory is then underpinned by two concrete examples. First, a combination of Event Related Desynchronization (ERD)-based active BCI with gaze control, defining a hybrid BCI as solution for the midas touch problem. And second, a passive BCI based on human error processing, leading to new forms of automated adaptation in HCI. This is in line with the results from other recent studies of passive BCI technology and shows the broad potential of this approach.