Electrocorticogram as a Brain-Computer Interface Signal Source

The use of electrocorticogram (ECoG) as the signal source for brain-computer interfaces (BCIs) has advantages for both the potential BCI user and the BCI researcher. However, research using ECoG can be logistically challenging. Visualization of time- and frequencybased characteristics of movement-related potentials in ECoG illustrates the features available for detection by a BCI and their spatial distribution. A quantitative comparison of the detection possible with EEG and ECoG verifies the signal quality advantages of ECoG and the utility of spatial filtering for improving detection. A quadratic detector based on a twocovariance signal model is presented as the basis for a BCI using ECoG, and the detection achieved by the quadratic detector is compared to BCI methods based on cross-correlation and bandpower. The quadratic detector provides dramatically improved detection and response time over the cross-correlation-based method.