EEG-based stress recognition using deep learning techniques

Successful implementation of deep learning technique for electroencephalography-based stress recognition has not been accomplished yet regardless of the efficacious application of deep learning in the brain-computer interface systems. This paper proposes utilising a Convolutional Neural Network to detect stress levels. To provide a baseline so as to compare the performance between the classifiers, a Support Vector Machine, a commonly used supervised machine learning technique to detect stress has additionally been adopted in this paper. The ramifications of two different input electroencephalography representations was also further explored. The highest classification accuracies attained using the Support Vector Machine are 83.7% and 82.7% for the separate input representations, detecting two classes of stress. This is a great improvement in outcomes as compared to other similar studies. However, for the Convolutional Neural Network, an average accuracy of 50% was achieved in detecting the two classes of stress. Regardless, this project may shed light in additional methods that may be adopted to detect stress successfully using a Convolutional Neural Network.Bachelor of Engineering (Electrical and Electronic Engineering