Facial Expression Recognition Via Enhanced Stress Convolution Neural Network for Stress Detection

Citation

Chew, Wan Ting and Chong, Siew Chin and Ong, Thian Song and Chong, Lee Ying (2022) Facial Expression Recognition Via Enhanced Stress Convolution Neural Network for Stress Detection. IAENG International Journal of Computer Science, 49 (3). pp. 1-10. ISSN 1819-9224

[img] Text
IJCS_49_3_20.pdf - Published Version
Restricted to Repository staff only

Download (1MB)

Abstract

The analysis of facial expressions for human stress detection has recently received wide attention. A facial expression is commonly used to explain a person’s emotional state to an observer. Based on a deep learning approach, we propose a lightweight and reliable method to detect stress by using facial expression recognition. This method is called Enhanced Stress Convolutional Neural Network (ESCNN). The ESCNN identifies a person’s facial expression and categorizes them into stress or non-stress categories based on their emotional state. Anger, sadness, disgust, and fear are categorized as stress outputs, whereas facial expressions of happiness, neural activity, and surprise are categorized as non-stress outputs. With this approach, Transfer Learning with MobileNet V2 and Tensor Flow is used to run the pre-trained FER2013 dataset and determine stress or no-stress. The MobileNet V2 model will make the proposed method lighter and faster than other algorithms. To get the highest accuracy, different epochs and parameters were tested. After several experiments, ESCNN, the combination of Transfer Learning and Haar Cascade face detection, produced the most effective performance for stress recognition. The experimental analysis vindicated the superiority of ESCNN over state-of-the-art methods.

Item Type: Article
Uncontrolled Keywords: Convolution neural network, Facial expression, Stress detection, Transfer learning
Subjects: Q Science > QP Physiology > QP351 Neurophysiology and Neuropsychology
Divisions: Faculty of Information Science and Technology (FIST)
Depositing User: Ms Nurul Iqtiani Ahmad
Date Deposited: 11 Oct 2022 06:00
Last Modified: 11 Oct 2022 06:00
URII: http://shdl.mmu.edu.my/id/eprint/10534

Downloads

Downloads per month over past year

View ItemEdit (login required)