EEG based analysis and decoding of multiple actions in separate lower limbs for brain computer interfacing

Brain-computer interface (BCI) can be utilised to enhance human abilities or promote neurorehabilitation. One of the frequently-used BCI paradigms is motor imagery (MI)-based BCI, which is usually executed by imagining separate upper limbs or both lower limbs. To date, it is very limited to imagine individual lower limbs. This might be because it is more difficult to distinguish between the left and right lower limbs than between the upper limbs. Meanwhile, a similar situation exists for real movement (RM) and movement observation (MO). To fill the above gap, we designed an EEG-based experiment to investigate separate lower limbs performing MI, RM and MO (a total of six actions: left-MI, right-MI, left-RM, right-RM, left-MO, and right-MO). The EEG analysis results revealed that there is no significant difference in the first few hundred milliseconds after action execution. However, significant differences emerged in the late phase around the brain midline. Differences were also found in the delta and alpha bands, with significant variations in the band of 18-22 Hz across tasks (MI, RM and MO). Furthermore, we attempted to classify six actions using the features extracted based on the above difference findings. We first used simple and classical methods, such as k-Nearest Neighbour and Support Vector Machine, to classify actions, showing these actions are classifiable. We then used deep learning models, including the proposed hybrid models, and found better classification performance. However, there is no significant difference between CNNs and hybrid models except more robust for the hybrid models. We further developed a novel model called ‘NeuroFusionNet’, which achieved an average classification accuracy of 55.21%, outperforming all other compared models. This study revealed the differences among the actions of individual lower limbs and provided evidence on the feasibility of multi-action classification, which is informative for further studies on separate lower limbs.