Classification of EEGs via Multi-layer Wavelet Transform and Feature Fusion to Enhance Depression Screening

It is worth noting that depression is a serious mental illness, and early diagnosis and treatment are crucial. This study aims to achieve automatic screening for depression by analyzing electroencephalogram (EEG) signals. However, EEG signal processing faces various challenges with noise. To solve this problem, we propose an adaptive threshold-based multi-layer wavelet transform (MWT) denoising method. This method combines the multi-scale decomposition properties and adaptive threshold processing technology. It retains the high frequency information of the signal while reducing noise interference. In order to compensate for the information loss after signal denoising, this experiment uses fusion features to integrate multiple information to help restore and retain the characteristics of useful signals, thereby enhancing the performance of the classification model. For the purpose of the experiment, EEGs of four subjects with depression and five healthy subjects were first obtained when they performed five activities. The classification results verified the effectiveness of the method in improving the classification accuracy, classification sensitivity, and classification specificity of depression. The classification accuracy of this method is as high as 87.68<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>• In addition, the non-invasive EEG acquisition headband used in this study is simple and easy to use, suitable for a variety of occasions, and provides a feasible solution for large-scale screening and early intervention of depression.