EEG Based Depression Detection Using Singular Value Decomposition Technique with Machine Learning Classifier

Depression is a prevalent mental health condition that is defined by persistent feelings of sadness, apathy, and hopelessness. In this article, an approach for classifying depression using signal processing techniques is applied to electroencephalogram (EEG) signals. Specifically, the application of Singular Value Decomposition (SVD) and Variational Mode Decomposition (VMD) is used to analyse the EEG signals which are obtained from the depression dataset which consists of eyes opened (EO) and eyes closed condition (EC). By decomposing EEG signals into the underlying components, these techniques offer insights into the temporal dynamics and frequency characteristics associated with depression. The aim is to identify distinctive patterns in EEG signals that correlate with depressive states, enabling the development of diagnostic tools for depression. Furthermore, Signal Processing algorithms are used for classification, and to build classification models capable of discriminating between depressed and non-depressed datasets based on their EEG patterns. The proposed approach helps enhance the accuracy and efficiency of depression diagnosis. The extracted features are fed into the machine learning classifiers. The SVD model demonstrates superior performance with Cubic SVM, Decision Tree, Logistic Regression, and KNN, achieving an accuracy of, 99.9%, sensitivity of 100%, F1-score 0.99, and Precision of 99.9%.