Cognitive Stress Recognition During Mathematical Task and EEG Changes Following Audio-Visual Stimuli for Relaxation

Solving mathematical tasks or puzzles is one of the most common learning processes in many learning activities. This task can be either simple or complex depending on the level of study. Many individuals feel the fear of math tasks which will hamper the learning process due to stress. Hence, recognition of stress plays a crucial role in many learning activities. In this work, machine learning-based automatic recognition of mental stress during math puzzle solving is developed using EEG signals. EEG signals are recorded using Enobio 8 device in two categories, one during math puzzle solving and another in happy state with audio-visual stimuli with funny video clips. The raw signals are preprocessed to remove the artifacts using 2-40Hz band pass filter. Filtered data is used to extract EEG bands. EEG relative band power is analyzed to observe the changes during math task and relaxation. The two mental states are classified using Long Short Term Memory (LSTM) deep learning model. The classification accuracy of 86% is achieved in beta band.