Emotion Recognition from Brain Wave using Multitask Machine Learning Leveraging Residual Connections

Abstract: Emotions have a significant influence on both our personal and professional lives. Images, videos, and brain signals can all be used to identify emotion. Electroencephalography (EEG), which is used to determine the state of the human brain, can also be used to recognize emotion. However, emotion recognition from EEG is a complex problem by nature, yet it is more dependable than other emotion detection methods. Despite the efforts of numerous researchers in the past to increase performance to identify emotion more correctly, they were unable to do so. This paper proposes a deep learning strategy to recognize human emotions better using EEG data from the DEAP dataset. We leverage multitask machine learning to improve performance by using the capability of residual connections. Most studies concentrated solely on valence and arousal, which are directly related to emotion but incapable of fully comprehending the emotional state. To detect emotion better, we consider valence, arousal, dominance, liking, and familiarity in this experiment. The results demonstrated that the proposed technique can accurately predict emotions with an accuracy of 96.85%, 97.10%, 97.19%, 97.03%, and 95.24% for 2 class classification and for 3 class classification 95.92%, 96.02%, 96.63%, 96.08%, and 95.39% for valence, arousal, dominance, liking, and familiarity respectively.

Authors

Rumman Ahmed Prodhan^{1[0000-0002-6865-185X]}, Sumya Akter^{1[0000-0001-7114-1845]}, Muhammad Bin Mujib^{1[0000-0003-0283-3439]}, Md. Akhtaruzzaman Adnan^{1[0000-0003-4137-0844]}, Tanmoy Sarkar Pias^{2*[0000-0002-7325-9844]}

1. University of Asia Pacific, Dhaka 1205, Bangladesh
   Email: rumman153@gmail.com, sumyaakter601@gmail.com, binmujib99@gmail.com, adnan.cse@uap-bd.edu
2. Virginia Tech, Blacksburg, VA 24061, United States
   Email: tanmoysarkar@vt.edu

*Corresponding Author

Links

Read Paper
Download Preprint
View Code on GitHub

Recommended citation: Prodhan, R. A., Akter, S., Mujib, M. B., Adnan, M. A., & Pias, T. S. (2023, June). Emotion recognition from brain wave using multitask machine learning leveraging residual connections. In *International Conference on Machine Intelligence and Emerging Technologies* (pp. 121-136). Cham: Springer Nature Switzerland.