A deep learning approach for decoding visually imagined digits and letters using time–frequency–spatial representation of EEG signals
作者：
Highlights：
• An EEG-based two-phase approach is proposed for decoding visually imagined objects.
• The approach focuses on decoding visually imagined digits and letters.
• The first phase builds a new time–frequency–spatial representation of EEG signals.
• The second phase presents a new deep learning framework to decode imagined objects.
• The results show that the approach enables accurate decoding of imagined objects.
摘要
•An EEG-based two-phase approach is proposed for decoding visually imagined objects.•The approach focuses on decoding visually imagined digits and letters.•The first phase builds a new time–frequency–spatial representation of EEG signals.•The second phase presents a new deep learning framework to decode imagined objects.•The results show that the approach enables accurate decoding of imagined objects.
论文关键词：Electroencephalography,Time–frequency–spatial representation,Visual imagery,Deep learning,Convolutional neural networks,Brain–computer interface
论文评审过程：Received 20 December 2021, Revised 25 March 2022, Accepted 25 April 2022, Available online 9 May 2022, Version of Record 13 May 2022.
论文官网地址：https://doi.org/10.1016/j.eswa.2022.117417
