G. Linda Rose, M. Punithavalli



Optimized Convolutional Neural Network Transfer Learning Method for Stress Emotion Classification

pdf PDF


Stress management plays an essential role in predicting stress levels and diagnosing to avoid its effects on certain individual’s socio-economic life. To achieve efficient stress prediction, a hybrid Deep Belief Network and Transfer Learning (DBNTL) method was proposed. Though DBNTL learn domain exact features on the top layers, but it decrease the change among the two domain distributions of different layers. Hence this paper proposes a novel Optimized Convolutional Neural Network and TL (OCNNTL) method that supports OCNN-based classifier on small-scale emotion and stress data domains. This novel model requires two domains that distribute similar OCNN. Two distribution models Marginal Distribution Discrepancy (MDD) at similar layers and Joint Distribution Discrepancy (JDD) of various layers helps OCNN to learn higher quality features at the top layers even the different emotion and stress domains contain similarity elements on feature-level. The OCNNTL layers are trained equally, so it measures both MDD of one layer and JDD of multiple layers. Moreover, a precise trade-off of these two discrepancies can increase transferability between emotion-stress domains. At last, the experimental outcomes exhibit the efficiency of OCNNTL method as compared to the CNN and DBNTL-based stress emotion classification methods.


Mental stress, Emotion analysis, Deep learning, Transfer learning, CNN


Cite this paper

G. Linda Rose, M. Punithavalli. (2022) Optimized Convolutional Neural Network Transfer Learning Method for Stress Emotion Classification. International Journal of Computers, 7, 104-110


Copyright © 2022 Author(s) retain the copyright of this article.
This article is published under the terms of the Creative Commons Attribution License 4.0