Hybrid CNN–LSTM Model For Continuous Stress Quantification Via Emotion-Valence Mapping On EEG Signals
DOI:
https://doi.org/10.29284/cppcad56Keywords:
EEG Signals, Mental Stress Estimation, Emotional Valence, Deep Learning, CNN-BiLSTM Hybrid, Affective Computing, Regression Analysis.Abstract
Correctly estimating the level of mental stress is a must for smart affective-computing and cognitive-health systems. Most of the anxiety models based on EEG that are currently available use only binary classification and thus cannot breathe the continuous spectrum of emotional fluctuations. Here, a hybrid CNN-LSTM deep-learning architecture for dual-dataset EEG-based stress analysis is introduced. The proposed framework combines adaptive-noise removal with hierarchical spatiotemporal feature extraction and attains 100 % binary accuracy on the STEW dataset and 97 % on the Neurocom dataset, along with 99 % accuracy for ternary emotional-valence classification on both datasets. The comparative benchmarking of the present framework against the latest models (ConvNeXt-CNN, Attention-BiLSTM, and Temporal-Attention Networks) indicates that the proposed method achieves state-of-the-art performance with excellent cross-dataset generalization capability. The importance of this work is the quantification of stress driven by emotion-valence across the heterogeneous EEG sources, thereby changing stress recognition from discrete detection to fine-grained, interpretable estimation. Such results position the CNN-LSTM hybrid as a stable, extendable, and deployable continuously mental-stress monitoring tool in clinics, workplaces, and human–computer-interaction environments.
References
[1] R. Phutela, S. Goswami, and A. Sikka, “An LSTM model-based EEG signal analysis for mental stress detection,” Biomed. Signal Process. Control, vol. 71, p. 103075, 2022.
[2] M. Khatwani and R. Srivastava, “EEG based human stress detection using hybrid deep learning model,” Neurocomputing, vol. 469, pp. 420–432, 2022.
[3] S. M. Alarcao and M. J. Fonseca, “Emotions recognition using EEG signals: A survey,” IEEE Trans. Affect. Comput., vol. 10, no. 3, pp. 374–393, 2019.
[4] H. S. Lee and K. Kim, “Simultaneous EEG and behavioral measures for mental workload classification,” Front. Hum. Neurosci., vol. 14, 2021.
[5] Z. Gao, X. Cui, and W. Wan, “Multiscale feature fusion with deep CNN for emotion recognition using EEG,” Comput. Mater. Contin., vol. 71, no. 2, pp. 3327–3342, 2022.
[6] U. R. Acharya, S. L. Oh, Y. Hagiwara, J. H. Tan, and H. Adeli, “Deep convolutional neural network for the automated detection and diagnosis of seizure using EEG signals,” Comput. Biol. Med., vol. 100, pp. 270–278, 2018.
[7] Y. Roy, H. Banville, I. Albuquerque, A. Gramfort, T. H. Falk, and J. Faubert, “Deep learning-based electroencephalography analysis: A systematic review,” J. Neural Eng., vol. 16, no. 5, p. 51001, 2019.
[8] P. Bashivan, I. Rish, M. Yeasin, and N. Codella, “Learning representations from EEG with deep recurrent-convolutional neural networks,” Int. Conf. Learn. Represent., 2016.
[9] L. Santamaria, A. de la Torre-Luque, and D. Mier, “Deep learning for mental health classification: A systematic review,” Curr. Psychiatry Rep., vol. 24, no. 3, pp. 83–95, 2022.
[10] J. Heimerl, S. K. Schäfer, and J. Held, “Multimodal stress detection based on physiological signals: A comparison of LSTM and CNN,” Front. Comput. Sci., vol. 3, p. 52, 2021.
[11] A. R. Subhani, S. Wibowo, and N. Kamel, “Evaluation of EEG signal processing and classification techniques for mental stress recognition,” Brain Informatics, vol. 7, no. 1, pp. 1–15, 2020.
[12] X. Chi, X. Zhu, C. Du, and P. Zhang, “Adaptive noise filtering in EEG signals for improved classification performance,” IEEE Access, vol. 10, pp. 52176–52186, 2022.
[13] S. K. Bashar and M. I. H. Bhuiyan, “Automatic sleep scoring from single-channel EEG using multi-scale entropy and ensemble empirical mode decomposition,” Biomed. Signal Process. Control, vol. 24, pp. 1–10, 2016.
[14] J. Galvão, D. Amaral, L. Correia, and P. Ferreira, “Regression models for continuous emotion estimation from EEG signals,” Sensors, vol. 21, no. 10, p. 3414, 2021.
[15] T. Jon, A. Silva, and C. Freitas, “EEG-RegNet: Deep regression network for continuous valence--arousal prediction,” Mathematics, vol. 13, no. 1, p. 87, 2024.
[16] Y. Li, Y. Zheng, and Y. Zhang, “Graph convolutional neural networks for continuous emotion recognition from EEG,” IEEE Trans. Affect. Comput., 2023.
[17] S. Tripathi, S. Acharya, R. D. Sharma, S. Mittal, and S. Bhattacharya, “Using deep and convolutional neural networks for accurate emotion classification on DEAP dataset,” Twenty-Ninth AAAI Conf. Artif. Intell., 2017.
[18] Z. Huang, Y. Ma, R. Wang, W. Li, and Y. Dai, “A Model for EEG-Based Emotion Recognition: CNN-Bi-LSTM with Attention Mechanism,” Electron. 2023, Vol. 12, Page 3188, vol. 12, no. 14, p. 3188, Jul. 2023, doi: 10.3390/ELECTRONICS12143188.
[19] J. Posner, J. A. Russell, and B. S. Peterson, “The circumplex model of affect: An integrative approach to affective neuroscience, cognitive development, and psychopathology,” Dev. Psychopathol., vol. 17, no. 3, p. 715, Jul. 2005, doi: 10.1017/S0954579405050340.
[20] X. Wang, L. Li, and H. Zhang, “EEG classification using sparse deep convolutional neural networks,” Med. Biol. Eng. Comput., vol. 58, no. 6, pp. 1175–1187, 2020.
[21] R. Chakravarthi and N. Rajagopalan, “Hybrid CNN--LSTM architecture for emotion classification based on EEG,” Expert Syst. Appl., vol. 210, p. 118513, 2023.
[22] Z. Cheng, X. Bu, Q. Wang, T. Yang, and J. Tu, “EEG-based emotion recognition using multi-scale dynamic CNN and gated transformer,” Sci. Rep., vol. 14, Dec. 2024, doi: 10.1038/s41598-024-82705-z.
[23] Y. Xu et al., “AMDET: Attention Based Multiple Dimensions EEG Transformer for Emotion Recognition,” IEEE Trans. Affect. Comput., vol. 15, no. 3, pp. 1067–1077, 2024, doi: 10.1109/TAFFC.2023.3318321.
[24] S. Ding et al., “DGAT: a dynamic graph attention neural network framework for EEG emotion recognition,” Front. Psychiatry, vol. 16, p. 1633860, Jul. 2025, doi: 10.3389/FPSYT.2025.1633860/BIBTEX.
[25] Y. Pan et al., “DuA: Dual Attentive Transformer in Long-Term Continuous EEG Emotion Analysis,” Jul. 2024, Accessed: Oct. 28, 2025. [Online]. Available: https://arxiv.org/pdf/2407.20519
[26] A. Shahzadi, A. Salman, F. Ahmed, and M. Lee, “EEG-based human stress detection using hybrid deep learning model,” Biomed. Signal Process. Control, vol. 68, p. 102741, 2021.
[27] F. Nirabi, S. K. Setarehdan, and A. Mahnam, “A framework for mental workload estimation using preprocessing and deep learning of EEG signals,” Comput. Biol. Med., vol. 137, p. 104808, 2021.
[28] J. S. Kang, S. Kavuri, and M. Lee, “Adaptive EEG noise filtering for coherence analysis,” 2014 Int. Winter Work. Brain-Computer Interface, BCI 2014, 2014, doi: 10.1109/IWW-BCI.2014.6782569.
[29] A. Patel, D. Nariani, and A. Rai, “Mental Stress Detection using EEG and Recurrent Deep Learning,” APSCON 2023 - IEEE Appl. Sens. Conf. Symp. Proc., 2023, doi: 10.1109/APSCON56343.2023.10100977.
[30] A. M. Alghamdi, M. U. Ashraf, A. A. Bahaddad, K. A. Almarhabi, W. A. Al Shehri, and A. Daraz, “Cross-subject EEG signals-based emotion recognition using contrastive learning,” Sci. Rep., vol. 15, no. 1, pp. 1–17, Dec. 2025, doi: 10.1038/S41598-025-13289-5;SUBJMETA.
[31] F. Wang, Y. C. Tian, and X. Zhou, “Cross-dataset EEG emotion recognition based on pre-trained Vision Transformer considering emotional sensitivity diversity,” Expert Syst. Appl., vol. 279, p. 127348, Jun. 2025, doi: 10.1016/J.ESWA.2025.127348.
[32] Y. Zhao, L. Cao, Y. Ji, B. Wang, and W. Wu, “Interpretable EEG Emotion Classification via CNN Model and Gradient-Weighted Class Activation Mapping,” Brain Sci., vol. 15, no. 8, p. 886, Aug. 2025, doi: 10.3390/BRAINSCI15080886.
[33] J. Yedukondalu, K. Sunkara, V. Radhika, S. Kondaveeti, M. Anumothu, and Y. Murali Krishna, “Cognitive load detection through EEG lead wise feature optimization and ensemble classification,” Sci. Rep., vol. 15, no. 1, pp. 1–18, Dec. 2025, doi: 10.1038/S41598-024-84429-6;SUBJMETA.
[34] R. Sharma and K. Chopra, “EEG signal analysis and detection of stress using classification techniques,” J. Inf. Optim. Sci., vol. 41, no. 1, pp. 229–238, Jan. 2020, doi: 10.1080/02522667.2020.1714187.
[35] D. Steyrl, G. Krausz, and K. Koschutnig, “Early Stress Detection and Analysis using EEG signals in Machine Learning Framework,” IOP Conf. Ser. Mater. Sci. Eng., vol. 1116, no. 1, p. 012134, Apr. 2021, doi: 10.1088/1757-899X/1116/1/012134.
[36] R. Suryawanshi and S. Vanjale, “International Journal of INTELLIGENT SYSTEMS AND APPLICATIONS IN ENGINEERING Brain Activity Monitoring for Stress Analysis through EEG Dataset using Machine Learning.” [Online]. Available: www.ijisae.org
[37] S. M. Atoar Rahman et al., “Advancement in Graph Neural Networks for EEG Signal Analysis and Application: A Review,” IEEE Access, vol. 13, no. February, pp. 50167–50187, 2025, doi: 10.1109/ACCESS.2025.3549120.
[38] Y. Badr, U. Tariq, F. Al-Shargie, F. Babiloni, F. Al Mughairbi, and H. Al-Nashash, “A review on evaluating mental stress by deep learning using EEG signals,” Neural Comput. Appl., vol. 36, no. 21, pp. 12629–12654, Jul. 2024, doi: 10.1007/S00521-024-09809-5/FIGURES/27.
[39] W. Wang, X. Zhou, and H. Lin, “DuA Transformer for long-term continuous EEG emotion estimation,” Neurocomputing, vol. 601, pp. 127–141, 2024.
[40] M. Safari, R. Shalbaf, S. Bagherzadeh, and A. Shalbaf, “Classification of mental workload with EEG analysis by using effective connectivity and a hybrid model of CNN and LSTM,” Comput. Methods Biomech. Biomed. Engin., pp. 1–15, Jul. 2024, doi: 10.1080/10255842.2024.2386325.
[41] A. Nirabi, F. A. Rahman, M. H. Habaebi, K. A. Sidek, and S. Yusoff, “Machine Learning-Based Stress Level Detection from EEG Signals,” in 2021 IEEE 7th International Conference on Smart Instrumentation, Measurement and Applications (ICSIMA), 2021, pp. 53–58. doi: 10.1109/ICSIMA50015.2021.9526333.
[42] M. Tahira and P. Vyas, “EEG based Mental Stress Detection using Deep Learning Techniques,” 2nd IEEE Int. Conf. Distrib. Comput. Electr. Circuits Electron. ICDCECE 2023, 2023, doi: 10.1109/ICDCECE57866.2023.10150574.
[43] A. Joshi, P. S. Matharu, L. Malviya, M. Kumar, and A. Jadhav, “Advancing EEG based stress detection using spiking neural networks and convolutional spiking neural networks,” Sci. Rep., vol. 15, no. 1, pp. 1–14, Dec. 2025, doi: 10.1038/S41598-025-10270-0;SUBJMETA.
[44] L. Fiorini, F. Bossi, and F. Di Gruttola, “EEG-based emotional valence and emotion regulation classification: a data-centric and explainable approach,” Sci. Rep., vol. 14, no. 1, pp. 1–18, Dec. 2024, doi: 10.1038/S41598-024-75263-X;SUBJMETA.
[45] F. Al-Shargie, T. B. Tang, N. Badruddin, and M. Kiguchi, “Mental stress assessment using simultaneous measurement of EEG and fNIRS,” Biomed. Opt. Express, vol. 11, no. 7, pp. 3563–3578, 2020.
[46] M. Murugappan, N. Ramachandran, and Y. Sazali, “Classification of human emotion from EEG using discrete wavelet transform,” J. Biomed. Sci. Eng., vol. 3, no. 4, pp. 390–396, 2010.
[47] J. Z. Xiang, Q. Y. Wang, Z. Bin Fang, J. A. Esquivel, and Z. X. Su, “A multi-modal deep learning approach for stress detection using physiological signals: integrating time and frequency domain features,” Front. Physiol., vol. 16, p. 1584299, 2025, doi: 10.3389/FPHYS.2025.1584299.
[48] A. Khosrowabadi, N. Sulaiman, and M. Ariffin, “EEG-based emotion recognition using hybrid CNN and SVM,” IEEE Trans. Affect. Comput., vol. 11, no. 4, pp. 585–597, 2020.
[49] M. Maji, P. Kumar, and R. Mukherjee, “EEG-based stress detection using CNN--LSTM hybrid model on STEW dataset,” IEEE Access, vol. 9, pp. 151230–151242, 2021.
[50] M. Alotaibi, H. Abdullah, and S. Alrashid, “Deep hybrid CNN--BiLSTM--GRU network for EEG-based emotion classification,” Sensors, vol. 24, no. 7, p. 3512, 2024.
Downloads
Published
Issue
Section
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
This work is licensed under a Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
