No abstract

Brain-Computer Interfaces (BCIs) ensure non-muscular communication between a user and external device by using brain activity. Currently, BCIs were applied in the medical field to increase quality of life of patients suffering from neuromuscular disorders. Most BCI systems use scalp recorded electroencephalographic activity, while Electrocorticography (ECoG) is a minimally-invasive alternative to Electroencephalogram (EEG), which ensures higher and superior signal characteristics enabling rapid user training and quicker communication. This paper presents a BCI system; ECoG signals are pre-processed and features are extracted from using Wavelet Packet Tree and Common Spatial Pattern. The extracted features are fused using Median Absolute Deviation (MAD) to improve the discrimination power of the feature vector. BCI Competition III, Data Set I having ECoG recordings motor imagery is used in investigation to evaluate the presented methodology.