DETECÇÃO ROBUSTA DE ESTRESSE: UMA ABORDAGEM 1D-CNN PARA CLASSIFICAÇÃO MULTICLASSE DE ECG EM CONDIÇÕES DE DADOS LIMITADOS
Resumo
O reconhecimento automatizado do estresse por meio de sinais fisiológicos constitui uma área de crescente relevância clínica e tecnológica, com impacto direto na melhoria dos desfechos em saúde, na redução de custos assistenciais e na viabilização da intervenção precoce em transtornos relacionados ao estresse.O presente trabalho desenvolve uma revisão dos avanços no campo da classificação do estresse baseado em dados fisiológicos, identificando os métodos mais representativos, os principais desafios metodológicos e as tendências que orientam a pesquisa emergente na área. Destaque especial é conferido às limitações impostas por conjuntos de dados de pequena escala, à relevância dos modelos personalizados por sujeito e às dificuldades inerentes à aplicação em tempo real em contextos não controlados. Em paralelo, este trabalho propõe e avalia uma nova arquitetura de rede neural convolucional (CNN) unidimensional, desenvolvida para classificar sinais de eletrocardiograma (ECG) em quatro categorias distintas, correspondentes às fases do estresse. O modelo revelou capacidade de aprendizado robusto e de generalização adequada, mesmo sob condições de escassez de dados, atingindo 72,81% de acurácia em um conjunto de teste independente. Esses resultados evidenciam o potencial do aprendizado profundo para a classificação do estresse e sublinham a necessidade de abordagens futuras que incorporem personalização, processamento em tempo real e fusão multimodal de sinais fisiológicos.
Biografia do Autor
Doutor em Ciência da Computação pela Universidade Federal do Rio Grande do Sul.
Referências
ANSARI, Y.; MOURAD, O.; QARAQE, K.; SERPEDIN, E. Deep learning for ECG arrhythmia detection and classification: an overview of progress for 2017–2023. Frontiers in Physiology, Lausanne, v. 14, 2023. DOI: https://doi.org/10.3389/fphys.2023.1246746
ASKARI, M.; SETAREHDAN, S. K.; SHOKROLLAHI, A.; ALIREZAEI, M. Stress detection from physiological signals using deep learning methods: distinguishing stress from physical activity. Biomedical Signal Processing and Control, Amsterdam, v. 73, 2022.
BOBADE, P.; VANI, M. Stress detection with machine learning and deep learning using multimodal physiological data. In: INTERNATIONAL CONFERENCE ON INVENTIVE RESEARCH IN COMPUTING APPLICATIONS (ICIRCA), 2., 2020. Proceedings […]. [S. l.]: IEEE, 2020. p. 51–57. DOI: https://doi.org/10.1109/ICIRCA48905.2020.9183244
CHATTERJEE, R.; NAG, A.; DUTTA, A.; MUKHERJEE, S. Stress classification using physiological signals with frequency-domain feature selection. Biomedical Signal Processing and Control, Amsterdam, v. 72, 2022.
FINSETH, T.; KJELDGAARD, M.; SKOV, M. B.; HANSEN, L. K. Personalized versus generalized models for physiological stress detection: challenges and opportunities. Sensors, Basel, v. 23, n. 3, 2023.
GARG, P.; SANTHOSH, J.; DENGEL, A.; ISHIMARU, S. Stress detection by machine learning and wearable sensors. In: INTERNATIONAL CONFERENCE ON INTELLIGENT USER INTERFACES – COMPANION, 26., 2021, New York. Proceedings […]. New York: ACM, 2021. p. 43–45. DOI: https://doi.org/10.1145/3397482.3450732
GEDAM, S.; PAUL, S. A review on mental stress detection using wearable sensors and machine learning techniques. IEEE Access, New York, v. 9, p. 84045–84066, 2021. DOI: https://doi.org/10.1109/ACCESS.2021.3085502
GUHDAR, M.; MOHAMMED, A. O.; MSTAFA, R. J. Advanced deep learning framework for ECG arrhythmia classification using 1D-CNN with attention mechanism. Knowledge-Based Systems, Amsterdam, v. 295, 2025. DOI: https://doi.org/10.1016/j.knosys.2025.113301
HUANG, J.; LIU, Y.; PENG, X. Recognition of driver's mental workload based on physiological signals: a comparative study. Biomedical Signal Processing and Control, Amsterdam, v. 71, pt. A, p. 103094, 2022. DOI: https://doi.org/10.1016/j.bspc.2021.103094
KUMAR, A.; SHARMA, K.; SHARMA, A. Hierarchical deep neural network for mental stress state detection using IoT-based biomarkers. Pattern Recognition Letters, Amsterdam, v. 145, p. 81–87, 2021. DOI: https://doi.org/10.1016/j.patrec.2021.01.030
MINGXU, L.; ZHANG, Y.; LIU, H.; WANG, Z. Affect and stress detection based on feature fusion of LSTM and CNN using physiological signals. Sensors, Basel, v. 23, 2023.
MODI, N.; KUMAR, Y.; MEHTA, K.; CHAPLOT, N. Physiological signal-based mental stress detection using hybrid deep learning models. Discover Artificial Intelligence, Cham, v. 5, 2025. DOI: https://doi.org/10.1007/s44163-025-00412-8
RIM, B.; LEE, J.; PARK, J.; KIM, H. Deep learning applications on medical physiological signals: a comprehensive review. Sensors, Basel, v. 20, n. 20, 2020. DOI: https://doi.org/10.3390/s20040969
SAH, S.; DAS, A.; CHOUDHURY, S. StressNet: convolutional neural network framework for personalized stress classification. IEEE Access, New York, v. 10, 2022.
VARGAS-LOPEZ, O.; PEREZ-RAMIREZ, C. A.; VALTIERRA-RODRIGUEZ, M.; YANEZ-BORJAS, J. J.; AMEZQUITA-SANCHEZ, J. P. An explainable machine learning approach based on statistical indexes and SVM for stress detection in automobile drivers using electromyographic signals. Sensors, Basel, v. 21, n. 9, 2021. DOI: https://doi.org/10.3390/s21093155
XIANG, J.-Z.; WANG, Q.-Y.; FANG, Z.-B.; ESQUIVEL, J. A.; SU, Z.-X. A multimodal deep learning approach for stress detection using physiological signals. Frontiers in Physiology, Lausanne, v. 16, 2025.
WU, Y.; ZHANG, J.; LIU, Y.; WANG, J. Transfer learning-based stress detection using physiological signals in daily activities. IEEE Access, New York, v. 9, 2021.
Can, Y. S., Arnrich, B., & Ersoy, C. (2019). Stress detection from physiological signals: A survey. IEEE Access, 7, 161798–161814. https://doi.org/10.1109/ACCESS.2019.2959811
Gjoreski, M., Gjoreski, H., Lutrek, M., & Gams, M. (2017). Continuous stress detection using a wrist device: In laboratory and real life. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 1(2), 1–20. https://doi.org/10.1145/3090076 DOI: https://doi.org/10.1145/3090076
Kyrou, I., et al. (2025). Deep learning for physiological signal-based stress detection: A survey. IEEE Transactions on Affective Computing, 16(2). DOI: https://doi.org/10.1109/TAFFC.2024.3455371
Liu, Q., Zhang, X., & Huang, M. (2025). A comprehensive review of machine learning and deep learning for stress detection using physiological signals. Applied Sciences, 15(21), 11777. https://doi.org/10.3390/app152111777 DOI: https://doi.org/10.3390/app152111777
Hasanpoor Reiss, A., & Stricker, D. (2019). Deep learning for sensor-based activity recognition: A survey. Pattern Recognition Letters, 119, 3–11. https://doi.org/10.1016/j.patrec.2018.02.010 DOI: https://doi.org/10.1016/j.patrec.2018.02.010
Schmidt, P., Reiss, A., Duerichen, R., Marberger, C., & Van Laerhoven, K. (2018). Introducing WESAD, a multimodal dataset for wearable stress and affect detection. In Proceedings of the 20th ACM International Conference on Multimodal Interaction (pp. 400–408). https://doi.org/10.1145/3242969.3242985 DOI: https://doi.org/10.1145/3242969.3242985
Xiang, J., Liu, Z., & Wang, P. (2025). Multimodal stress detection using parallel convolutional neural networks with time–frequency feature fusion. Frontiers in Physiology, 16, 1584299. https://doi.org/10.3389/fphys.2025.1584299 DOI: https://doi.org/10.3389/fphys.2025.1584299
Yang, H., Chen, L., & Zhao, Y. (2025). Stress detection using two-dimensional transformation of physiological signals and deep learning. Scientific Reports, 15(1), 1228. https://doi.org/10.1038/s41598-025-01228-3 DOI: https://doi.org/10.1038/s41598-025-01228-3
Yildirim, O., et al. (2023). Deep learning for ECG-based stress detection: A review. Biomedical Signal Processing and Control, 80, 104319. https://doi.org/10.1016/j.bspc.2022.104319 DOI: https://doi.org/10.1016/j.bspc.2022.104319
