Independent component analysis combined with compressed sensing for EEG compression in BCI

Considering the limitation of hardware requirement and power dissipation in wearable brain-computer interface (BCI), the electroencephalogram (EEG) data compression implemented by independent component analysis (ICA) combined with compressed sensing (CS) is proposed in this paper. A simple and effective ICA spatial filtering method is used to obtain motor related independent components (MRICs). Furthermore, CS algorithm is introduced to compress MRICs, which have advantage of frequency sparse. So the proposed scheme can make few MRICs compressed transmission instead of the multi-channel EEG data transmission. Based on the measured motor imagery EEG data, the proposed EEG compression scheme is compared with the traditional CS compression scheme. The experimental results show that, the two system schemes have the similar classification accuracy. However, in the proposed compression scheme, the amount of transmission data can be reduced by 75%.