![]() | Only 14 pages are availabe for public view |
Abstract In this thesis we study a new technique for BCI data that requires no earlier training. The new approach is applied to experimental data for motor imagery and P300-based BCI for both healthy and disabled subjects and compared to the classification output results of the same data utilizing the conventional processing techniques requiring earlier training. Regarding P300 based-BCI, The fundamental rule of this new class of unsupervised methodologies is that the trial with true activation signal inside every block must be distinctive from whatever remains of the trials inside that block. Consequently, a measure that is delicate to this difference can be utilized to settle on a choice taking into account a single block with no earlier training. As well, we tend to study different algorithms of aggregating info from many trials to extend communication speed and bit rate. Such aggregation strategies include simple average, PCA and PPCA. The results by averaging, from the sample individual cases show that the proposed technique supported SVD provided the most effective performance reaches 98.61%. Regarding the motor imagery part, we tend to used different classification methodologies as in time and frequency domain. And then we found that wavelet transform get best performance reaches 82.14%. So, these promising results recommend that this approach can reach accuracies not extremely far from those got with training while keeping up robust performance in practice |