Enhancing HAR Novelty Detection with Activity Confusion Analysis and Clustering

• Thaís B. de M. B. de Sousa UFC
• Lívia A. Cruz UFC
• Criston P. de Souza UFC
• Regis P. Magalhães UFC
• José A. F. de Macêdo UFC

DOI: https://doi.org/10.5753/sbcas.2025.6851

Resumo

The field of medicine and healthcare, especially Human Activity Recognition (HAR), has been experiencing the benefits of technology for tracking people’s habits. In this context, novelty detection aims to detect if an activity performed by a subject hasn’t been seen before, in the past training data. However, the novelty may not be properly recognized if there is a similar action in the train set. Thus, this work proposes an analysis of activity pairs to determine which pairs are most confused when using three traditional models for novelty detection: Local Outlier Factor, One-class SVM, and Isolation Forest. After that, we employ agglomerative clustering to gather similar activities, and the clustered activities are used as input to a leave-one-activity-out novelty detection approach. We conclude that the Isolation Forest model achieves the best results in the activity pairs analysis, yielding an F1 score of 88.4%. Finally, the leave-one-activity-out methodology is evaluated with a sliding window technique, and the F1 score on the Local Outlier Factor model increases from 66.3% (without grouping similar activities) to 74.6% (grouping similar activities).

Referências

Boyer, P., Burns, D., and Whyne, C. (2021). Out-of-distribution detection of human activity recognition with smartwatch inertial sensors. Sensors, 21(5):1669.

Breunig, M. M., Kriegel, H.-P., Ng, R. T., and Sander, J. (2000). LOF: identifying density-based local outliers. In Proceedings of the 2000 ACM SIGMOD international conference on Management of data, pages 93–104.

Das Antar, A., Ahmed, M., and Ahad, M. A. R. (2019). Challenges in sensor-based human activity recognition and a comparative analysis of benchmark datasets: A review. In 2019 Joint 8th International Conference on Informatics, Electronics & Vision (ICIEV) and 2019 3rd International Conference on Imaging, Vision & Pattern Recognition (icIVPR), pages 134–139.

Jin, L., Wang, X., Chu, J., and He, M. (2022). Human activity recognition machine with an anchor-based loss function. IEEE Sensors Journal, 22(1):741–756.

Kim, H. and Lee, D. (2024). Clan: A contrastive learning based novelty detection framework for human activity recognition.

Liu, F. T., Ting, K. M., and Zhou, Z.-H. (2008). Isolation forest. In IEEE International Conference on Data Mining, pages 413–422. IEEE.

Lubba, C. H., Sethi, S. S., Knaute, P., Schultz, S. R., Fulcher, B. D., and Jones, N. S. (2019). catch22: Canonical time-series characteristics: Selected through highly comparative time-series analysis. Data Mining and Knowledge Discovery, 33(6):1821–1852.

Lukasová, A. (1979). Hierarchical agglomerative clustering procedure. Pattern Recognition, 11(5):365–381.

Munoz-Organero, M. (2019). Outlier detection in wearable sensor data for human activity recognition (har) based on drnns. IEEE Access, 7:74422–74436.

Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., and Duchesnay, E. (2011). Scikit-learn: Machine learning in Python. Journal of Machine Learning Research, 12:2825–2830.

Pimentel, M. A., Clifton, D. A., Clifton, L., and Tarassenko, L. (2014). A review of novelty detection. Signal Processing, 99:215–249.

Reiss, A. and Stricker, D. (2012). Introducing a new benchmarked dataset for activity monitoring. 2012 16th International Symposium on Wearable Computers, pages 108–109.

Schölkopf, B., Platt, J. C., Shawe-Taylor, J., Smola, A. J., and Williamson, R. C. (2001). Estimating the support of a high-dimensional distribution. Neural computation, 13(7):1443–1471.

Staab, S., Krissel, S., Luderschmidt, J., and Martin, L. (2022). Recognition models for distribution and out-of-distribution of human activities. In 2022 18th International Conference on Wireless and Mobile Computing, Networking and Communications (WiMob), pages 6–12. IEEE.

Weiss, G. M. (2019). WISDM smartphone and smartwatch activity and biometrics dataset. UCI Machine Learning Repository: WISDM Smartphone and Smartwatch Activity and Biometrics Dataset Data Set, 7:133190–133202.

Xu, H., Pan, Y., Li, J., Nie, L., and Xu, X. (2019). Activity recognition method for home-based elderly care service based on random forest and activity similarity. IEEE Access, 7:16217–16225.

Yang, J., Zhou, K., Li, Y., and Liu, Z. (2024). Generalized out-of-distribution detection: A survey.