Article Preview
TopIntroduction
Human Computer Interface (HCI) is the interaction between people and computers. A Human-Computer Interface which uses brain activity as communication support is referred as Brain Computer Interface (BCI). A Brain Computer Interface is the direct communication pathway between a human brain and an external device. It helps the physically disabled people to communicate their needs to the external environment (Palaniappan et al., 2002). The EEG signals of the physically disabled patients are recorded via the electrodes placed over the scalp according to the specification by International 10-20 system, and further processing is done on these signals (Guger et al., 2003;Curan et al., 2003). This technology do not depend on peripheral nerves and muscles and hence suitable for physically disabled patients. Eleanor et al., (2003) considered different cognitive tasks such as spatial navigation around a familiar environment, auditory imagery of a familiar tune, right and left motor imagery of opening and closing the hand, for use with a brain-computer interface (BCI). Autoregressive (AR) model was used for extracting features from the EEG signals and logistic regression and a nonlinear generative classifier was used. The classification accuracy obtained was around 63-74%.
Lei Qin and BinHe (2005) have applied wavelet based time-frequency analysis approach for classifying motor imagery tasks. Time-frequency distributions (TFDs) were constructed based on wavelet decomposition and event-related (de) synchronization patterns were extracted from symmetric electrode pairs. The weighted energy difference of the electrode pairs was compared to classify the imagery movement. This method gave an averaged classification rate of 78%. Pradeep Shenoy et al., (2006) have made an attempt by using spatial complexity of brain region, field power and frequency of field changes for classifying event related EEG during imagination of left or right hand movement. These three multi- channel linear descriptors were used for analysis of ERD/ERS time course over the left and right brain hemispheres. Fisher discriminant analysis was used to realize the classification of two classes of EEG patterns. The maximum classification accuracy of 90% was achieved.
Guger et al (2003) recorded EEG signals with gold electrodes from two bipolar channels over the right-hand and foot representation areas. For the analysis of EEG patterns, an adaptive autoregressive (AAR) model and band power estimation were applied. AAR parameters were estimated with the recursive-least-square (RLS) algorithm. For band-power estimation, the average power in the alpha and beta band at each electrode position was estimated by digitally band-pass filtering the data in standard frequency ranges of 10-12 Hz (alpha) and 16-20 Hz (beta). Then the samples were squared and the average over several consecutive samples was taken. Linear Discriminant Analysis (LDA) was used for classification which gave the classification accuracy of 97%.