Action Observation with Rhythm Imagery (AORI): A Novel Paradigm to Activate Motor-Related Pattern for High-Performance Motor Decoding

The Motor Imagery (MI) paradigm has been widely used in brain-computer interface (BCI) for device control and motor rehabilitation. However, the MI paradigm faces challenges such as comprehension difficulty and limited decoding accuracy. Therefore, we propose the Action Observation with Rhythm Imagery (AORI) as a natural paradigm to provide distinct features for high-performance decoding. Twenty subjects were recruited in the current study to perform the AORI task. Spectral-spatial, temporal and time-frequency analyses were conducted to investigate the AORI-activated brain pattern. Task-discriminant component analysis (TDCA) was utilized to perform multiclass motor decoding. The results demonstrated distinct lateralized ERD in the alpha and beta bands, and clear lateralized steady-state movement-related rhythm (SSMRR) at the movement frequencies and their first harmonics. The activated brain areas included frontal, sensorimotor, posterior parietal, and occipital regions. Notably, the decoding accuracy reached 92.16% ± 7.61% in the four-class scenario. We proposed the AORI paradigm, revealed the activated motor-related pattern and proved its efficacy for high-performance motor decoding. These findings provide new possibilities for designing a natural and robust BCI for motor control and motor rehabilitation.