Pre-movement pattern decoding from motor evoked potentials for reducing interaction delay
Rongrong Fu,Fangyao Xu,Haifeng Liang,Yang Liu,Shiwei Wang,Yaodong Wang,Junxiang Chen
DOI: https://doi.org/10.1016/j.bspc.2024.107275
IF: 5.1
2024-11-30
Biomedical Signal Processing and Control
Abstract:Brain-computer interfaces (BCIs) represent one of the most successful integrations of neuroscience and artificial intelligence, and have been applied in intelligent healthcare. However, the high complexity and low signal-to-noise ratio of electroencephalogram (EEG) signals make them susceptible to external interference and noise, leading to low decoding rates and high latency. To effectively enhance the performance of pre-movement pattern decoding from motor evoked potentials, this study proposed a robust method. We employed the filter bank common spatial patterns (FBCSP) to extract and identify event-related desynchronization (ERD) features from motor evoked potentials, and used FBCSP to investigate the ERD bands associated with actual movements. Feature vectors were classified using Cubic-SVM. The results indicated that extracting ERD features in sub-frequency bands yields more discriminative information. Approximately 530 ms before actual movement, distinct ERD patterns were observed for two movement tasks at electrode sites C3 and C4. Furthermore, frequency band selection revealed that ERD features in motor evoked potentials predominantly exist in the 17–29 Hz range. The robustness and effectiveness of the method were validated, achieving an average accuracy of 93.4 %, an average AUC of 0.954, and an average MSE of 0.195 on the self-collected hand fisting experiment dataset, and an average accuracy of 95.8 % on the BCI Competition II dataset, representing a 4.7 % improvement over the current highest classification accuracy. Overall, this study offered a promising approach for decoding pre-movement patterns from motor evoked potentials, which was crucial for enhancing the operational efficiency of brain-computer interfaces.
engineering, biomedical