Stacked Sparse Autoencoder and Machine Learning Based Anxiety Classification Using EEG Signals

Anxiety is an emotion characterized by trepidation, stress, or uneasiness that involves extreme worry or fear over future unwanted events or an actual situation. Careful analysis for anxiety is critical since approximately 2 to 4% of the general population have experienced adequate symptoms indicating an anxiety disorder. This paper aims to classify anxiety levels based on machine learning and deep learning algorithms with improved performance. This work uses the publically available DASPS Database (Database for Anxious States based on a Psychological Stimulation). The dataset consists of EEG recordings from 23 participants during anxiety elicitation through face-to-face psychological stimuli. This work uses RFECV with the classifiers to reduce redundancy between features and improve results. We achieve the highest classification accuracy of 83.93% and 70.25% using Stacked Sparse Autoencoder and Decision Tree for two-class anxiety classification.