Research Article
BibTex RIS Cite

Hiyerarşik Toplayıcı Kümeleme Kullanılarak Duygusal Fizyolojik Durumların Gözetimsiz Keşfi

Year 2025, Volume: 1 Issue: 1 , 39 - 46 , 31.05.2025
https://izlik.org/JA56AA94RZ

Abstract

Fizyolojik sinyallerle otomatik duygu durumu tanıma hayati önem taşımaktadır, ancak Wearable Stress and Affect Detection (WESAD) gibi etiketlenmiş veri setleri üzerindeki denetimli öğrenme, altta yatan nüansları gözden kaçırabilir. Bu çalışma, 15 WESAD denekinden alınan göğüs sensörü verilerindeki (EKG, EDA, RESP, TEMP) içsel yapıları keşfetmek için Ward bağlantılı denetimsiz Hiyerarşik Yığılmalı Kümeleme (HAC) yöntemini kullanmıştır. NeuroKit2 aracılığıyla türetilen ayrıntılı Kalp Hızı Değişkenliği (HRV) metriklerini içeren kapsamlı öznitelikler çıkarılmıştır. Standartlaştırılmış özniteliklere HAC uygulanmış ve EDA, sıcaklık, solunum ve temel HRV indekslerindeki (örn. RMSSD, LF/HF oranı) benzersiz çok değişkenli imzalarla tanımlanan dört ayrı küme elde edilmiştir. Veriye dayalı bu kümeler, orijinal WESAD deneysel etiketleriyle (bazal, stres, eğlence, meditasyon) kısmi uyum göstermiş ancak aynı zamanda önemli farklılıklar sergileyerek önceden tanımlanmış koşullar içindeki fizyolojik heterojenliği ortaya çıkarmıştır. Bulgular, potansiyel olarak farklı otonom sinir sistemi durumlarını (örn. stres, rahatlama/ilgi, uyanık dinlenme) temsil eden fizyolojik olarak yorumlanabilir kümelerin belirlenmesinde HAC'nin etkinliğini göstermektedir. Bu durum, denetimsiz öğrenmenin duyuşsal bilişimde denetimli yöntemleri tamamlamadaki değerini vurgulamakta, deneysel etiketlerin ötesinde fizyolojik durum taksonomilerinin veriye dayalı keşfine olanak tanımakta ve ruh sağlığı takibi ile uyarlanabilir insan-bilgisayar etkileşimi için daha incelikli yapay zeka araçlarının geliştirilmesine yönelik değerli bilgiler sunmaktadır.

References

  • R. Yuvaraj, R. Mittal, A. A. Prince, and J. S. Huang, “Affective Computing for Learning in Education: A Systematic Review and Bibliometric Analysis,” Educ Sci (Basel), vol. 15, no. 1, p. 65, Jan. 2025, doi: 10.3390/educsci15010065.
  • E. Oliver and S. Dakshit, “Cross-Modality Investigation on WESAD Stress Classification,” Feb. 2025.
  • Z. Ahmad and N. Khan, “A Survey on Physiological Signal-Based Emotion Recognition.,” Bioengineering (Basel), vol. 9, no. 11, Nov. 2022, doi: 10.3390/bioengineering9110688.
  • M. A. Al Aleem, R. Mubarak, N. M. Salem, and I. Sadek, “A Deep Learning Approach Using WESAD Data for Multi-Class Classification with Wearable Sensors,” in 2024 6th Novel Intelligent and Leading Emerging Sciences Conference (NILES), IEEE, Oct. 2024, pp. 194–197. doi: 10.1109/NILES63360.2024.10753228.
  • E. Zhou, M. Soleymani, and M. J. Matarić, “Investigating the Generalizability of Physiological Characteristics of Anxiety,” Jan. 2024, doi: 10.1109/BIBM58861.2023.10385292.
  • T. Iqbal, A. Elahi, W. Wijns, and A. Shahzad, “Exploring Unsupervised Machine Learning Classification Methods for Physiological Stress Detection,” Front Med Technol, vol. 4, Mar. 2022, doi: 10.3389/fmedt.2022.782756.
  • D. Makowski et al., “NeuroKit2: A Python toolbox for neurophysiological signal processing,” Behav Res Methods, vol. 53, no. 4, pp. 1689–1696, Aug. 2021, doi: 10.3758/s13428-020-01516-y.
  • D. Bajpai and L. He, “Evaluating KNN Performance on WESAD Dataset,” in 2020 12th International Conference on Computational Intelligence and Communication Networks (CICN), IEEE, Sep. 2020, pp. 60–62. doi: 10.1109/CICN49253.2020.9242568.
  • Philip Schmidt, Attila Reiss, Robert Duerichen, Claus Marberger, and Kristof Van Laerhoven, “Introducing WESAD, a Multimodal Dataset for Wearable Stress and Affect Detection,” in Proceedings of the 20th ACM International Conference on Multimodal Interaction, New York, NY, USA: ACM, 2018, pp. 400–408.
  • A. Almadhor, G. A. Sampedro, M. Abisado, and S. Abbas, “Efficient Feature-Selection-Based Stacking Model for Stress Detection Based on Chest Electrodermal Activity.,” Sensors (Basel), vol. 23, no. 15, Jul. 2023, doi: 10.3390/s23156664.
  • R. Lima, D. Osório, and H. Gamboa, “Heart Rate Variability and Electrodermal Activity in Mental Stress Aloud: Predicting the Outcome,” in Proceedings of the 12th International Joint Conference on Biomedical Engineering Systems and Technologies, SCITEPRESS - Science and Technology Publications, 2019, pp. 42–51. doi: 10.5220/0007355200420051.

Unsupervised Discovery of Affective Physiological States using Hierarchical Agglomerative Clustering

Year 2025, Volume: 1 Issue: 1 , 39 - 46 , 31.05.2025
https://izlik.org/JA56AA94RZ

Abstract

Automated physiological affect recognition is vital, but supervised learning on labeled datasets like Wearable Stress and Affect Detection (WESAD) may miss underlying nuances. This study used unsupervised Hierarchical Agglomerative Clustering (HAC) with Ward's linkage to explore inherent structures in chest-worn sensor data (ECG, EDA, RESP, TEMP) from 15 WESAD subjects. Comprehensive features, including detailed Heart Rate Variability (HRV) metrics derived via NeuroKit2, were extracted. HAC was applied to standardized features, yielding four distinct clusters defined by unique multivariate signatures in EDA, temperature, respiration, and key HRV indices (e.g., RMSSD, LF/HF ratio). These data-driven clusters showed partial alignment but also significant divergence from original WESAD experimental labels (baseline, stress, amusement, meditation), revealing physiological heterogeneity within predefined conditions. Findings demonstrate HAC's efficacy in identifying physiologically interpretable clusters potentially representing distinct autonomic nervous system states (e.g., stress, relaxation/engagement, alert rest). This underscores the value of unsupervised learning for complementing supervised methods in affective computing, enabling data-driven discovery of physiological state taxonomies beyond experimental labels and offering valuable insights for developing more nuanced AI-driven tools for mental health monitoring and adaptive human-computer interaction.

Ethical Statement

"This article does not require ethics committee approval."

Supporting Institution

"This article has no conflicts of interest with any individual or institution."

References

  • R. Yuvaraj, R. Mittal, A. A. Prince, and J. S. Huang, “Affective Computing for Learning in Education: A Systematic Review and Bibliometric Analysis,” Educ Sci (Basel), vol. 15, no. 1, p. 65, Jan. 2025, doi: 10.3390/educsci15010065.
  • E. Oliver and S. Dakshit, “Cross-Modality Investigation on WESAD Stress Classification,” Feb. 2025.
  • Z. Ahmad and N. Khan, “A Survey on Physiological Signal-Based Emotion Recognition.,” Bioengineering (Basel), vol. 9, no. 11, Nov. 2022, doi: 10.3390/bioengineering9110688.
  • M. A. Al Aleem, R. Mubarak, N. M. Salem, and I. Sadek, “A Deep Learning Approach Using WESAD Data for Multi-Class Classification with Wearable Sensors,” in 2024 6th Novel Intelligent and Leading Emerging Sciences Conference (NILES), IEEE, Oct. 2024, pp. 194–197. doi: 10.1109/NILES63360.2024.10753228.
  • E. Zhou, M. Soleymani, and M. J. Matarić, “Investigating the Generalizability of Physiological Characteristics of Anxiety,” Jan. 2024, doi: 10.1109/BIBM58861.2023.10385292.
  • T. Iqbal, A. Elahi, W. Wijns, and A. Shahzad, “Exploring Unsupervised Machine Learning Classification Methods for Physiological Stress Detection,” Front Med Technol, vol. 4, Mar. 2022, doi: 10.3389/fmedt.2022.782756.
  • D. Makowski et al., “NeuroKit2: A Python toolbox for neurophysiological signal processing,” Behav Res Methods, vol. 53, no. 4, pp. 1689–1696, Aug. 2021, doi: 10.3758/s13428-020-01516-y.
  • D. Bajpai and L. He, “Evaluating KNN Performance on WESAD Dataset,” in 2020 12th International Conference on Computational Intelligence and Communication Networks (CICN), IEEE, Sep. 2020, pp. 60–62. doi: 10.1109/CICN49253.2020.9242568.
  • Philip Schmidt, Attila Reiss, Robert Duerichen, Claus Marberger, and Kristof Van Laerhoven, “Introducing WESAD, a Multimodal Dataset for Wearable Stress and Affect Detection,” in Proceedings of the 20th ACM International Conference on Multimodal Interaction, New York, NY, USA: ACM, 2018, pp. 400–408.
  • A. Almadhor, G. A. Sampedro, M. Abisado, and S. Abbas, “Efficient Feature-Selection-Based Stacking Model for Stress Detection Based on Chest Electrodermal Activity.,” Sensors (Basel), vol. 23, no. 15, Jul. 2023, doi: 10.3390/s23156664.
  • R. Lima, D. Osório, and H. Gamboa, “Heart Rate Variability and Electrodermal Activity in Mental Stress Aloud: Predicting the Outcome,” in Proceedings of the 12th International Joint Conference on Biomedical Engineering Systems and Technologies, SCITEPRESS - Science and Technology Publications, 2019, pp. 42–51. doi: 10.5220/0007355200420051.
There are 11 citations in total.

Details

Primary Language English
Subjects Artificial Life and Complex Adaptive Systems
Journal Section Research Article
Authors

Helen Arya 0009-0003-0009-3091

Muhammad Arya This is me 0009-0008-7057-9296

Submission Date April 19, 2025
Acceptance Date May 21, 2025
Early Pub Date May 30, 2025
Publication Date May 31, 2025
IZ https://izlik.org/JA56AA94RZ
Published in Issue Year 2025 Volume: 1 Issue: 1

Cite

IEEE [1]H. Arya and M. Arya, “Unsupervised Discovery of Affective Physiological States using Hierarchical Agglomerative Clustering”, INNAI, vol. 1, no. 1, pp. 39–46, May 2025, [Online]. Available: https://izlik.org/JA56AA94RZ