An Efficient P300-based Brain-Computer Interface with Minimal Calibration Time
published: Jan. 19, 2010, recorded: December 2009, views: 5320
Report a problem or upload filesIf you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.
Brain-Computer Interfaces (BCI) are communication systems that enable subjects to send commands to computers by using only their brain activity . Most existing BCI are based on ElectroEncephaloGraphy (EEG) as the measure of brain activity . So far, BCI have been proven to be very promising communication and control tools for disabled people . A promising brain signals used in the design of assistive BCI for disabled people is the P300, a positive waveform occuring roughly 300 ms after a rare and relevant stimulus [1, 2]. In order to use a P300-based BCI, subjects have to focus their attention on a given stimulus randomly appearing among many others, each stimulus corresponding to a given command. The appearance of the desired stimulus being rare and relevant, it is expected to trigger a P300 in the subject’s brain activity. As such, detecting the P300 enables the system to identify the desired stimulus and hence the desired command. Interestingly enough, P300-based BCI have been successfully used to control a wheelchair (see, e.g., ) or to enable severely disabled users to spell words [2, 4].
However, current P300-based BCI as well as other BCI systems still suffer from several limitations which prevent them from being widely used . One of these limitations is that to use a BCI, many examples of the subject’s EEG signals must be recorded in order to calibrate the BCI, which is unconvenient and time consuming. Moreover, this calibration process generally has to be repeated at regular intervals (e.g., from one day to the other) in order to accomodate sources of variations such as changes in electrode positions or changes in the subject’s mental state and fatigue level. Therefore, the calibration time should be maintained as brief as possible. Until now, reducing the calibration time of P300-based BCI has been scarcely addressed by the literature. Exceptions are the works of Li et al  and Lu et al . Li et al suggested to use initially a BCI calibrated with few training samples, and then to incrementally adapt this BCI online, thanks to semi-supervised learning . Lu et al proposed to use a subject-independent BCI, previously learnt from the data of many other subjects, also followed by online adaptation . However, the main limitiation of these two approaches is that such BCI would have initially poor detection performances, becoming efficient only after adaptation. An ideal P300-based BCI would have initially high performances, even if trained with very few examples. In this paper, we propose a new P300-based BCI design which can be trained using much fewer examples than current BCI designs, without sacrifying the detection performances.
Link this pageWould you like to put a link to this lecture on your homepage?
Go ahead! Copy the HTML snippet !