Abstract:

Although existing EEG compression algorithms can achieve good compression rates, they lack attention to task-related data and are also unable to meet the real-time requirements of brain-computer interface (BCI) applications, which will significantly reduce the performance of BCI systems. Based on the subspace method, when the task-related information of the BCI system is known, the task related EEG signals can be preserved as much as possible during the compression process, which can significantly reduce the amount of data that needs to be transmitted without affecting the performance of the BCI system. By using a finite impulse response filter bank to approximate the signal subspace of task related components, the EEG signal can be segmented into the smallest possible compression ratio and processed in real-time. On the premise that there is no significant difference in the performance of some classification algorithms compared to the original data, an algorithm that only transmits 8% of the data can be proposed. This algorithm can not only transmit less data while minimizing the impact on the performance of the BCI system, but also achieve real-time compression, whitch has important application value.

Key words: brain-computer interface, electroencephalography(EEG), signal compression, task dependence