Zobrazit v EDS

Deep transfer learning-based SSVEP frequency domain decoding method.

Uloženo v:
Podrobná bibliografie
Název: Deep transfer learning-based SSVEP frequency domain decoding method.
Autoři: Xiong, Hui1,2 (AUTHOR) xionghui@tiangong.edu.cn, Song, Jinlong1,2 (AUTHOR), Liu, Jinzhen1,2 (AUTHOR), Han, Yuqing3 (AUTHOR)
Zdroj: Biomedical Signal Processing & Control. Mar2024, Vol. 89, pN.PAG-N.PAG. 1p.
Témata: Convolutional neural networks, Deep learning, Fast Fourier transforms, Feature extraction, Brain-computer interfaces, Filter banks
Abstrakt: • A feature extraction method combining the filter bank technique and zero-padding fast Fourier transform was designed, which effectively obtain the rich spatial and frequency domain features in EEG. • A efficient deep learning model was constructed, which is able to learn the potential semantic features of input well and realize high information transfer rate and decoding accuracy. • The proposed transfer learning training strategy effectively alleviated the problem of insufficient model training data, reduced the inter-subject variability, and improved the generalization performance of the model. Improving the decoding accuracy and information transfer rate (ITR) of a steady-state visual evoked potential-based brain-computer interface (SSVEP-BCI) system and narrowing the inter-subject variance are key to the application of the SSVEP-BCI system. To this end, we proposed a deep transfer learning-based SSVEP frequency domain decoding method to improve the decoding performance. Input data representations with rich spatial and frequency domain features were extracted using filter bank and zero-padding-based fast Fourier transform techniques. A concise and efficient 3-dimensional convolutional neural network (3DCNN) model was designed for feature extraction and decoding of the input data. A transfer learning strategy was proposed to further improve the decoding accuracy and narrow the inter-subject variance. Our proposed 3DCNN achieves 89.35 % average classification accuracy and 173.02 bits/min ITR on the benchmark dataset with 1 s signal length. On our laboratory dataset, the average classification accuracy and ITR of 3DCNN reach 88.75 % and 120.33 bits/min, respectively, when the signal length is 0.6 s. In this study, we experimentally confirmed the effectiveness and superiority of our proposed method for SSVEP decoding, which provides a promising decoding method for the application of SSVEP-BCI. [ABSTRACT FROM AUTHOR]
Databáze: Supplemental Index
Popis
Abstrakt:• A feature extraction method combining the filter bank technique and zero-padding fast Fourier transform was designed, which effectively obtain the rich spatial and frequency domain features in EEG. • A efficient deep learning model was constructed, which is able to learn the potential semantic features of input well and realize high information transfer rate and decoding accuracy. • The proposed transfer learning training strategy effectively alleviated the problem of insufficient model training data, reduced the inter-subject variability, and improved the generalization performance of the model. Improving the decoding accuracy and information transfer rate (ITR) of a steady-state visual evoked potential-based brain-computer interface (SSVEP-BCI) system and narrowing the inter-subject variance are key to the application of the SSVEP-BCI system. To this end, we proposed a deep transfer learning-based SSVEP frequency domain decoding method to improve the decoding performance. Input data representations with rich spatial and frequency domain features were extracted using filter bank and zero-padding-based fast Fourier transform techniques. A concise and efficient 3-dimensional convolutional neural network (3DCNN) model was designed for feature extraction and decoding of the input data. A transfer learning strategy was proposed to further improve the decoding accuracy and narrow the inter-subject variance. Our proposed 3DCNN achieves 89.35 % average classification accuracy and 173.02 bits/min ITR on the benchmark dataset with 1 s signal length. On our laboratory dataset, the average classification accuracy and ITR of 3DCNN reach 88.75 % and 120.33 bits/min, respectively, when the signal length is 0.6 s. In this study, we experimentally confirmed the effectiveness and superiority of our proposed method for SSVEP decoding, which provides a promising decoding method for the application of SSVEP-BCI. [ABSTRACT FROM AUTHOR]
ISSN:17468094
DOI:10.1016/j.bspc.2023.105931