Leveraging Transfer Superposition Theory for Stable-State Visual Evoked Potential Cross-Subject Frequency Recognition

• Published in: IEEE transactions on biomedical engineering Vol. 71; no. 11; pp. 3071 - 3084
• Main Authors: He, Xinjie, Allison, Brendan Z., Qin, Ke, Liang, Wei, Wang, Xingyu, Cichocki, Andrzej, Jin, Jing
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.11.2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Adult; Algorithms; Benchmark testing; Brain computer interface (BCI); Brain-Computer Interfaces; Calibration; Correlation; Decoding; Electroencephalography; electroencephalography (EEG); Electroencephalography - methods; Evoked Potentials, Visual - physiology; Fatigue; Female; Filters; Frequency dependence; Human-computer interface; Humans; Identification methods; Iterative methods; Knowledge management; Male; Recognition; Signal Processing, Computer-Assisted; Spatial filtering; Spatial filters; steady-state visual evoked potential (SSVEP); superposition theory; Transfer learning; transfer mode; Visual evoked potentials; Visualization; Young Adult
• ISSN: 0018-9294; 1558-2531; 1558-2531
• DOI: 10.1109/TBME.2024.3406603

In steady-state visual evoked potential (SSVEP)-based brain-computer interfaces (BCIs), various spatial filtering methods based on individual calibration data have been proposed to alleviate the interference of spontaneous activities in SSVEP signals for enhancing the SSVEP detection performance. However, the necessary calibration procedures take time, cause visual fatigue and reduce usability. For the calibration-free scenario, we propose a cross-subject frequency identification method based on transfer superimposed theory for SSVEP frequency decoding. First, a multi-channel signal decomposition model was constructed. Next, we used the cross least squares iterative method to create individual specific transfer spatial filters as well as source subject transfer superposition templates in the source subject. Then, we identified common knowledge among source subjects using a prototype spatial filter to make common transfer spatial filters and common impulse responses. Following, we reconstructed a global transfer superimposition template with SSVEP frequency characteristics. Finally, an ensemble cross-subject transfer learning method was proposed for SSVEP frequency recognition by combining the source-subject transfer mode, the global transfer mode, and the sine-cosine reference template. Offline tests on two public datasets show that the proposed method significantly outperforms the FBCCA, TTCCA, and CSSFT methods. More importantly, the proposed method can be directly used in online SSVEP recognition without calibration. The proposed algorithm was robust, which is important for a practical BCI.