运动表象
模式识别(心理学)
计算机科学
自回归模型
脑-机接口
人工智能
脑电图
主成分分析
支持向量机
分类器(UML)
特征提取
数学
统计
心理学
精神科
作者
Zhenfei Liu,Lina Wang,Shijie Xu,Kunfeng Lu
标识
DOI:10.1016/j.compbiomed.2022.106196
摘要
Brain-computer Interface (BCI) system based on motor imagery (MI) heavily relies on electroencephalography (EEG) recognition with high accuracy. However, modeling and classification of MI EEG signals remains a challenging task due to the non-linear and non-stationary characteristics of the signals. In this paper, a new time-varying modeling framework combining multiwavelet basis functions and regularized orthogonal forward regression (ROFR) algorithm is proposed for the characterization and classification of MI EEG signals. Firstly, the time-varying coefficients of the time-varying autoregressive (TVAR) model are precisely approximated with the multiwavelet basis functions. Then a powerful ROFR algorithm is employed to dramatically alleviate the redundant model structure and accurately recover the relevant time-varying model parameters to obtain high resolution power spectral density (PSD) features. Finally, the features are sent to different classifiers for the classification task. To effectively improve the accuracy of classification, a principal component analysis (PCA) algorithm is utilized to determine the best feature subset and Bayesian optimization algorithm is performed to obtain the optimal parameters of the classifier. The proposed method achieves satisfactory classification accuracy on the public BCI Competition II Dataset III, which proves that this method potentially improves the recognition accuracy of MI EEG signals, and has great significance for the construction of BCI system based on MI.
科研通智能强力驱动
Strongly Powered by AbleSci AI