Unsupervised feature extraction with autoencoders for EEG based multiclass motor imagery BCI

作者：

Highlights：

• Proposed a novel feature extraction method for EEG signal classification.

• The non-separability between the MI classes has been reduced in a new transformation.

• A novel approach for getting higher accuracy in motor imagery classification.

• A new feature fusion for EEG is developed.

• The experimental results outperformed other popular feature extraction techniques.

摘要

•Proposed a novel feature extraction method for EEG signal classification.•The non-separability between the MI classes has been reduced in a new transformation.•A novel approach for getting higher accuracy in motor imagery classification.•A new feature fusion for EEG is developed.•The experimental results outperformed other popular feature extraction techniques.

论文关键词：Autoencoder,BCI,Electroencephalogram,EEG Signal Transformation,Feature Extraction,Motor Imagery,Wavelet Transform

论文评审过程：Received 9 August 2020, Revised 26 August 2022, Accepted 22 September 2022, Available online 26 September 2022, Version of Record 3 October 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.118901