面向个性化人体行为识别的双重动态加权集成方法

doi:10.6040/j.issn.1672-3961.0.2025.131

摘要/Abstract

摘要： 针对传统静态模型在人体行为识别中难以有效应对用户行为个性化及动态变化的问题,提出一种面向个性化人体行为识别的双重动态加权集成(dual dynamic weighted ensemble, DDWE)方法。通过融合多种概念漂移检测技术与基分类器自适应机制,实现对动态数据流中用户个性化行为的实时检测与自适应学习;设计基于类别概率与KappaM的权重矩阵更新策略,从类别置信度和模型识别能力两方面动态调整基分类器权重,提升集成分类器在特定用户行为识别任务中的适应能力与精度。在多个人体行为识别公开数据集上的试验结果表明,DDWE方法优于现有方法,在特定用户个性化人体行为识别场景下展现出显著优势,为医疗监护、运动康复和智慧养老等相关领域提供借鉴与参考。

关键词: 人体行为识别, 个性化, 数据流, 集成学习, 动态

Abstract: Aiming at the problem that traditional static models could not effectively address the personalization and dynamic changes in human activity recognition, a dual dynamic weighted ensemble(DDWE)method for personalized human activity recognition was proposed. By integrating multiple concept drift detection techniques with adaptive mechanisms of base classifiers, real-time detection and adaptive learning of user personalized activities in dynamic data streams were enabled. A weight matrix updating strategy based on class probability and KappaM was designed to dynamically adjust the weights of base classifiers from the perspectives of class confidence and recognition ability, improving the adaptability and accuracy of the ensemble classifier in user-specific recognition tasks. Experiments conducted on several public human activity recognition datasets showed that DDWE method outperformed existing approaches and demonstrated clear advantages in personalized human activity recognition, providing guidance and reference for related fields such as medical monitoring, sports rehabilitation, and intelligent elderly care.

Key words: human activity recognition, personalized, data stream, ensemble learning, dynamic

中图分类号:

TP181

忽丽莎,高海辰,王素贞. 面向个性化人体行为识别的双重动态加权集成方法[J]. 山东大学学报 (工学版), 2026, 56(4): 38-51.

Hu Lisha, Gao Haichen, Wang Suzhen. A dual dynamic weighted ensemble method for personalized human activity recognition[J]. Journal of Shandong University(Engineering Science), 2026, 56(4): 38-51.

参考文献

[1] Karim M, Khalid S, Aleryani A, et al. Human action recognition systems: a review of the trends and state-of-the-art[J]. IEEE Access, 2024, 12: 36372-36390.
[2] Liang F, Su Z, Sheng W. Multimodal monitoring of activities of daily living for elderly care[J]. IEEE Sensors Journal, 2024, 24(7): 11459-11471.
[3] Yang G, Wu X, Zhang J. A dynamic balanced quadtree for real-time streaming data[J]. Knowledge-Based Systems, 2023, 263: 110291.
[4] 文益民, 刘帅, 缪裕青, 等. 概念漂移数据流半监督分类综述[J]. 软件学报, 2022, 33(4): 1287-1314. Wen Yimin, Liu Shuai, Miao Yuqing, et al. Survey on semi-supervised classification of data streams with concept drifts[J]. Journal of Software, 2022, 33(4): 1287-1314.
[5] Rutkowski L, Jaworski M, Duda P. Stream data mining: algorithms and their probabilistic properties[M]. Cham, Switzerland: Springer, 2019: 13-33.
[6] 翟婷婷, 高阳, 朱俊武. 面向流数据分类的在线学习综述[J]. 软件学报, 2020, 31(4): 912-931. Zhai Tingting, Gao Yang, Zhu Junwu. Survey of online learning algorithms for streaming data classification[J]. Journal of Software, 2020, 31(4): 912-931.
[7] 梁斌, 李光辉, 代成龙. 面向概念漂移且不平衡数据流的G-mean加权分类方法[J]. 计算机研究与发展, 2022, 59(12): 2844-2857. Liang Bin, Li Guanghui, Dai Chenglong. G-mean weighted classification method for imbalanced data stream with concept drift[J]. Journal of Computer Research and Development, 2022, 59(12): 2844-2857.
[8] Bifet A, Gavaldà R. Learning from time-changing data with adaptive windowing[C] //Proceedings of the 2007 SIAM International Conference on Data Mining. Minneapolis, USA: SIAM, 2007: 443-448.
[9] Gama J, Medas P, Castillo G, et al. Learning with drift detection[C] //Advances in Artificial Intelligence-SBIA 2004. Sao Luis, Brazil: Springer, 2004: 286-295.
[10] Page E S. Continuous inspection schemes[J]. Biome-trika, 1954, 41(1/2): 100-115.
[11] Gomes H M, Bifet A, Read J, et al. Adaptive random forests for evolving data stream classification[J]. Machine Learning, 2017, 106(9): 1469-1495.
[12] Gomes H M, Read J, Bifet A. Streaming random patches for evolving data stream classification[C] //2019 IEEE International Conference on Data Mining(ICDM). Beijing, China: IEEE, 2020: 240-249.
[13] Bifet A, De Francisci Morales G, Read J, et al. Efficient online evaluation of big data stream classifiers[C] //Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Sydney, Australia: ACM, 2015: 59-68.
[14] Ferrari A, Micucci D, Mobilio M, et al. On the personalization of classification models for human activity recognition[J]. IEEE Access, 2020, 8: 32066-32079.
[15] Tasnim U, Islam R, Desai K, et al. Investigating personalization techniques for improved cybersickness prediction in virtual reality environments[J]. IEEE Transactions on Visualization and Computer Graphics, 2024, 30(5): 2368-2378.
[16] Gan G, Ma C, Wu J. Data clustering: theory, algorithms, and applications[M]. Philadelphia, USA: SIAM, 2020: 243-286.
[17] Chai Y D, Liu H X, Zhu H Y, et al. A profile similarity-based personalized federated learning method for wearable sensor-based human activity recognition[J]. Information & Management, 2024, 61(7): 103922.
[18] Amrani H, Micucci D, Mobilio M, et al. Leveraging dataset integration and continual learning for human activity recognition[J]. International Journal of Machine Learning and Cybernetics, 2025, 16(7): 5213-5234.
[19] Fereidoonian F, Ali Ahmadi-Pajouh M. A personalized deep neural network to recognize human activities in healthy subjects[C] //2022 29th National and 7th International Iranian Conference on Biomedical Engineering(ICBME). Tehran, Islamic Republic of Iran: IEEE, 2023: 325-332.
[20] Ferrari A, Micucci D, Mobilio M, et al. Deep learning and model personalization in sensor-based human activity recognition[J]. Journal of Reliable Intelligent Environ-ments, 2023, 9(1): 27-39.
[21] Dhekane S G, Ploetz T. Transfer learning in sensor-based human activity recognition: a survey[J]. ACM Computing Surveys, 2025, 57(8): 1-39.
[22] Albogamy F R. Federated learning for IoMT-enhanced human activity recognition with hybrid LSTM-GRU networks[J]. Sensors, 2025, 25(3): 907.
[23] Li J X, Kang P Q, Bradley Shull P. Multitarget unsupervised transfer learning improves wearable-sensor cross-body-position activity recognition via inter-target shared representation[J]. IEEE Sensors Journal, 2025, 25(5): 9101-9112.
[24] Singh G, Chowdhary M, Kumar A, et al. A personalized classifier for human motion activities with semi-supervised learning[J]. IEEE Transactions on Consumer Electronics, 2020, 66(4): 346-355.
[25] Presotto R, Civitarese G, Bettini C. Semi-supervised and personalized federated activity recognition based on active learning and label propagation[J]. Personal and Ubiquitous Computing, 2022, 26(5): 1281-1298.
[26] Soviany P, Ionescu R T, Rota P, et al. Curriculum learning: a survey[J]. International Journal of Computer Vision, 2022, 130(6): 1526-1565.
[27] Mannini A, Intille S S. Classifier personalization for activity recognition using wrist accelerometers[J]. IEEE Journal of Biomedical and Health Informatics, 2019, 23(4): 1585-1594.
[28] Saha U, Saha S, Kabir M T, et al. Decoding human activities: analyzing wearable accelerometer and gyroscope data for activity recognition[J]. IEEE Sensors Letters, 2024, 8(8): 7003904.
[29] Siirtola P, Röning J. Context-aware incremental learning-based method for personalized human activity recognition[J]. Journal of Ambient Intelligence and Humanized Computing, 2021, 12(12): 10499-10513.
[30] Cruciani F, Nugent C D, Quero J M, et al. Personalizing activity recognition with a clustering based semi-population approach[J]. IEEE Access, 2020, 8: 207794-207804.
[31] Hu R, Chen L, Miao S H, et al. SWL-Adapt: an unsupervised domain adaptation model with sample weight learning for cross-user wearable human activity recognition[C] //Proceedings of the AAAI Conference on Artificial Intelligence. Washington, DC, USA: AAAI, 2023: 6012-6020.
[32] Wu T, Liu Y B, Yongchareon S. Class-aware sample weight learning for cross-modal unsupervised domain adaptation in cross-user wearable human activity recognition[C] //Proceedings of the 27th European Conference on Artificial Intelligence. Santiago de Compostela, Spain: IOS, 2024: 569-576.
[33] Sucerquia A, López J D, Vargas-Bonilla J F. SisFall: a fall and movement dataset[J]. Sensors, 2017, 17(1): 198.
[34] Casilari E, Santoyo-Ramón J A, Cano-García J M. Analysis of a smartphone-based architecture with multiple mobility sensors for fall detection[J]. PLoS One, 2016, 11(12): e0168069.
[35] Roggen D, Calatroni A, Rossi M, et al. Collecting complex activity datasets in highly rich networked sensor environments[C] //2010 Seventh International Confe-rence on Networked Sensing Systems(INSS). Kassel, Germany: IEEE, 2010: 233-240.
[36] Reiss A, Stricker D. Introducing a new benchmarked dataset for activity monitoring[C] //2012 16th Interna-tional Symposium on Wearable Computers. Newcastle, UK: IEEE, 2012: 108-109.
[37] Altun K, Barshan B, Tunçel O. Comparative study on classifying human activities with miniature inertial and magnetic sensors[J]. Pattern Recognition, 2010, 43(10): 3605-3620.
[38] Anguita D, Ghio A, Oneto L, et al. A public domain dataset for human activity recognition using smartphones [C] //Proceedings of the 21st European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning. Bruges, Belgium: [s.n.] : 437-442.
[39] Sztyler T, Stuckenschmidt H. Online personalization of cross-subjects based activity recognition models on wearable devices[C] //2017 IEEE International Confe-rence on Pervasive Computing and Communications(PerCom). Kona, USA: IEEE, 2017: 180-189.
[40] Yang L, Manias D M, Shami A. PWPAE: an ensemble framework for concept drift adaptation in IoT data streams[C] //2021 IEEE Global Communications Conference(GLOBECOM). Madrid, Spain: IEEE, 2022: 9685338.
[41] Wu Y F, Liu L, Yu Y J, et al. Online ensemble learning-based anomaly detection for IoT systems[J]. Applied Soft Computing, 2025, 173: 112931.

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed