Comparison Between Online and Offline Independent Component Analysis in the Context of Motor Imagery-Based Brain-Computer Interface

Independent component analysis (ICA) is an important tool for recovering brain activity sources from electroencephalogram (EEG) signals. It is a blind technique, hence does not require that reference signals be available. Most ICA algorithms are considered, to a certain extent, offline, as they demand parameter adjustment before the actual filtering step in new data. Online Recursive ICA (ORICA) was proposed as an adaptive algorithm that can estimate filter parameters in a real-time manner. The approach has the advantage of being more capable of tracking changes in brain dynamics and source changes. In this work, we perform a comparative analysis of six offline ICA algorithms and ORICA in the context of EEG signal preprocessing for motor imagery-based brain-computer interfaces (BCIs). The experimental results show that there was no significant difference between ORICA performance and the best offline methods, validating the feasibility of its use in motor imagery BCI.