Universität Bielefeld › Technische Fakultät › NI
Recently Brain-Computer Interfaces have mainly been used as a tool for paralyzed people to communicate with the outside world by spelling letters using only their measurable brain potentials. Naturally this relies on a set of computer generated visual, auditory or tactile stimuli which serve as items to communicate the subjects intentions.
A prime example of this kind is the well known P300-Speller paradigm which lets the subject spell letters by focusing attention on the target letters. Using this paradigm as a basis, we aim to develop a system that treats items from real-world scenes as symbols that carry a specific semantic. These real-world items could consist of a telephone, a bottle or just an abstract entity like a specific location in a room. A system that is able to associate semantics with certain symbols in a (semi-)automatic way will be able to offer a rich set of associated functions which are offered as selectable items to the user. Augmented-Reality (AR) techniques are used to augment the real-world items seen through the camera with specific visual stimuli and transform the task into a P300-Speller paradigm. This technique provides the link between physical obejects and computer generated stimuli which is key to the successful application of Brain-Computer Interfaces as assitive device for completely paralyzed people.
To increase the applicability of BCI in every day usage, we focus on the development and improvement of specific aspects:
For spelling tasks it is common to define the number of stimulus presentation which results in a pre-defined time span until a classification is possible. We developed an advanced classification method which adapts the amount of data needed to infere the users intention on-line; resulting in faster response times while maintaining high accuracies.