Exploiting dimensionality reduction and neural network techniques for the development of expert brain–computer interfaces
作者：
Highlights：
• Two-step filtering technique was adopted for cognitive and external noise removal.
• Automated correlation-based criteria was proposed to select relevant components and coefficients for PCA, ICA and LDA respectively.
• The regularization parameters for NCA were tuned to reduce the classification loss.
• Extensive experiments with PCA, ICA, LDA, NCA techniques with several channel selection, neural networks and statistical measures were conducted in EWT domain.
• The proposed framework provide 100% and 92.9% classification accuracy for subject dependent and independent experiments.
摘要
•Two-step filtering technique was adopted for cognitive and external noise removal.•Automated correlation-based criteria was proposed to select relevant components and coefficients for PCA, ICA and LDA respectively.•The regularization parameters for NCA were tuned to reduce the classification loss.•Extensive experiments with PCA, ICA, LDA, NCA techniques with several channel selection, neural networks and statistical measures were conducted in EWT domain.•The proposed framework provide 100% and 92.9% classification accuracy for subject dependent and independent experiments.
论文关键词：Electroencephalography,Brain–computer interface,Empirical wavelet transform,Motor imagery,Neighborhood component analysis,Neural networks
论文评审过程：Received 10 February 2020, Revised 23 August 2020, Accepted 14 September 2020, Available online 17 September 2020, Version of Record 28 September 2020.
论文官网地址：https://doi.org/10.1016/j.eswa.2020.114031
