Refractive EEG classification joint singular spectrum analysis and tunable Q factor wavelet transform

Abstract

Abstract Compared with the traditional methods, the application of electroencephalogram(EEG) in refractive classification is more portable and more suitable for daily environment. However, there is still a lack of research on refractive classification based on EEG signals. Therefore, this paper proposes a multi-classification of refractive EEG based on single-channel joint singular spectrum analysis and tunable Q factor wavelet transform(SSA-TQWT). In order to improve the classification accuracy of refractive EEG signals and better adapt to the characteristics of nonlinear EEGs, this paper proposes a time-frequency analysis method of SSA-TQWT to denoise EEG signals. Firstly, this method uses SSA algorithm to decompose, and uses component selection algorithm to quickly and automatically screen useful signals, which reduces the labor loss and avoids the loss of useful signals. Then, TQWT algorithm is used to adaptively process the refractive EEG signal through flexible and adjustable Q and r. At the same time, multi-domain features are extracted for fusion. Include nonlinear dynamic features, statistical features and frequency domain features. Finally, machine learning is used to classify them. The experimental results show that the average classification accuracy of this method is 90.61%, which is 10.29% higher than SSA method and 8.05% higher than TQWT method. At the same time, the sensitivity and specificity for low, medium and high refractive states are 93.94%, 81.82% and 80%, 93.75%, 92.11% and 93.48% respectively.