A new EEG-based wireless brain computer interface (BCI) system is presented with which text can be typed on a computer screen (mind-typing). The application is based on detecting P300 event-related potentials (ERPs). The EEG recordings were performed using a prototype of an ultra low-power battery-operated 8-channel wireless EEG system, which consists of two parts: an amplifier coupled with a wireless transmitter and a receiver, which is connected through a USB interface to the PC. We used a braincap with large filling holes and sockets for ring electrodes. We used the same visual stimuli paradigm as in the P300-based speller of Farwell and Donchin [1]: a matrix of 6x6 symbols for which columns and rows of the matrix were intensified in a random manner. The subjects were asked to count the number of intensifications of the attended symbol. Four healthy male subjects (aged 23-36 with average age of 31, three right one lefthanded) participated in the experiments.For the P300 detection, two different machine learning approaches were explored. The first approach uses the Group Method of Data Handling [2] for the optimal selection of a set of simple amplitude-based features. The second approach extracts features by maximizing the mutual information between features and class labels [3, 4]. Finally, a simple linear classifier uses the selected/extracted features in order to detect the P300.The results show that the feature extraction method is more efficient than the feature selection method in terms of speed as well as accuracy. We also compared both methods with the state-of-the-art mind-typer reported in [5]. As it can be seen from the Figure 1, the accuracy of the presented system is comparable. However, our system benefits from its simple design which supports a power-efficient on-chip implementation (on an ASIC).
展开▼
