Motor-imagery-based brain-computer interfaces (BCIs) commonly usethe common spatial pattern filter (CSP) as preprocessing step before featureextraction and classification. The CSP method is a supervised algorithmand therefore needs subject-specific training data for calibration,which is very time consuming to collect. In order to reduce the amountof calibration data that is needed for a new subject, one can apply multitask (from now on called multisubject) machine learning techniques to the preprocessing phase. Here, thegoal of multisubject learning is to learn a spatial filter for a new subjectbased on its own data and that of other subjects. This paper outlinesthe details of the multitask CSP algorithm and shows results on two datasets. In certain subjects a clear improvement can be seen, especially whenthe number of training trials is relatively low.
展开▼
