基于主动迁移学习的脑力负荷识别研究

doi:10.19678/j.issn.1000-3428.0070039

摘要/Abstract

摘要：

无监督迁移学习在基于生理信号的跨个体脑力负荷识别研究中已得到广泛运用, 但模型性能受限于目标域数据缺乏标注。针对该问题, 提出一种将迁移学习与主动学习相结合的跨个体脑力负荷识别方法。以脑电为研究对象, 首先通过计算源域与目标域间的最大均值差异优选出与目标域分布最接近的源域集合; 其次对优选源域与目标域逐个构建一对一跨个体脑力负荷识别模型, 通过对抗迁移网络拉近两域特征分布后, 采用基于不确定性加权聚类的主动学习挑选并标注少量兼具不确定性和多样性的目标域样本, 这些样本将参与后续模型分类层训练; 最后利用集成学习综合多个单源域模型识别结果。在公开数据集WAUC上的实验结果表明, 源域优选可降低负迁移发生率并节约计算成本, 引入主动学习提升了跨个体迁移学习的性能, 相较于无监督迁移学习, 在不同体力负荷水平下的脑力负荷识别任务中平均识别准确率提高了14.7%;集成学习克服了单源域模型所学知识有限的问题, 进一步提升了模型的识别性能, 最终达到了86.1%的平均识别效果。

关键词: 脑力负荷, 跨个体, 主动迁移学习, 源域优选, 集成学习

Abstract:

Unsupervised transfer learning has been widely used in cross-subject mental workload recognition studies based on physiological signals; however, the model performance is limited by the lack of labeled target domain data. To address this problem, a cross-subject mental workload recognition method that combines transfer learning with active learning is proposed. Using an Electroencephalogram (EEG) as the research object, source domains with distributions similar to the target domain are selected by calculating the maximum mean discrepancy between the source and target domains. Second, a one-to-one cross-subject mental workload recognition model is constructed for each selected source and target domain. The feature distributions of the two domains are brought closer by an adversarial network, and a small number of target domain samples, considering both uncertainty and diversity, are labeled by active learning based on uncertainty-weighted clustering, which participated in the subsequent training of the model classification layers. Finally, ensemble learning is utilized to synthesize the recognition results of multiple single-source domain models. Experiments on the publicly available WAUC dataset reveal that source domain selection reduces the incidence of negative transfers and computational costs. The introduction of active learning effectively improves the performance of cross-subject transfer learning. Compared to unsupervised transfer learning, the average recognition accuracy is improved by 14.7% in the task of recognizing mental workload under different levels of physical workload. Ensemble learning overcomes the shortcomings of the limited knowledge learned by single-source domain models, further improving the recognition performance of the model and achieving an average recognition of 86.1%.

Key words: mental workload, cross subject, active transfer learning, source domain selection, ensemble learning

蒋欣怡, 陈兰岚, 郑时蓬. 基于主动迁移学习的脑力负荷识别研究[J]. 计算机工程, 2026, 52(6): 109-120.

JIANG Xinyi, CHEN Lanlan, ZHENG Shipeng. Research on Mental Workload Recognition Based on Active Transfer Learning[J]. Computer Engineering, 2026, 52(6): 109-120.

https://www.ecice06.com/CN/Y2026/V52/I6/109

图/表 16

图1 基于主动迁移的多源域跨个体脑力负荷识别框架

Fig.1 Multi-source cross-subject mental workload recognition framework based on active transfer learning

图2 主动迁移网络整体结构

Fig.2 Overall structure of active transfer learning network

图3 不确定性加权聚类主动学习

Fig.3 Uncertainty-weighted clustering active learning

图4 实验流程

Fig.4 Experimental procedure

图5 源域数量与分类准确率折线图

Fig.5 Line graph of classification accuracy and number of source domains

图6 不同源域选择顺序下的源域数量与分类准确率折线图

Fig.6 Line graph of classification accuracy and number of source domains under different source domain selection orders

图7 不同主动学习方法的分类准确率折线图

Fig.7 Line graph of classification accuracy for different active learning methods

图8 不同主动学习方法的特征可视化

Fig.8 Visualization of the features of different active learning methods

图9 不同迁移学习方法的分类准确率柱状图

Fig.9 Histogram of classification accuracy for different transfer learning methods

参考文献 33

1	明东, 柯余峰, 何峰, 等. 基于生理信号的脑力负荷检测及自适应自动化系统研究: 40年回顾与最新进展. 电子测量与仪器学报, 2015, 29 (1): 1- 13.
	MING D , KE Y F , HE F . Psychophysiological measures based studies on mental workload assessment and adaptive automation: review of the last 40 years and the latest developments. Journal of Electronic Measurement and Instrumentation, 2015, 29 (1): 1- 13.
2	WICKENS C D . Multiple resources and mental workload. Human Factors, 2008, 50 (3): 449- 455. doi: 10.1518/001872008X288394
3	王韬. 融合多生理信号的跨时间脑力负荷检测研究[D]. 天津: 天津大学, 2019.
	WANG T. Research on cross-time mental workload detection with multiple physiological signals[D]. Tianjin: Tianjin University, 2019. (in Chinese)
4	张效艇, 陈兰岚, 陈长德. 基于图卷积神经网络的脑力负荷识别. 华东理工大学学报(自然科学版), 2023, 49 (6): 882- 889.
	ZHANG X T , CHEN L L , CHEN C D . Recognition of mental workload based on graph convolutional neural network. Journal of East China University of Science and Technology, 2023, 49 (6): 882- 889.
5	LONG M, CAO Y, WANG J, et al. Learning transferable features with deep adaptation networks[C]//Proceedings of the 32nd International Conference on Machine Learning. Washington D. C., USA: IEEE Press, 2015: 97-105.
6	YU C H, WANG J D, CHEN Y Q, et al. Transfer learning with dynamic adversarial adaptation network[C]//Proceedings of the IEEE International Conference on Data Mining. Washington D. C., USA: IEEE Press, 2019: 778-786.
7	陈长德, 陈兰岚, 张效艇. 基于多源域深度域自适应的脑力负荷识别. 华东理工大学学报(自然科学版), 2023, 49 (5): 744- 753.
	CHEN C D , CHEN L L , ZHANG X T . Mental workload recognition based on multi-source deep domain adaptation. Journal of East China University of Science and Technology, 2023, 49 (5): 744- 753.
8	CHUNG Y A, LEE H Y, GLASS J. Supervised and unsupervised transfer learning for question answering[C]//Proceedings of NAACL-HLT'18. New Orleans, USA: Association for Computational Linguistics, 2018: 1585-1594.
9	MOTIIAN S, PICCIRILLI M, ADJEROH D A, et al. Unified deep supervised domain adaptation and generalization[C]//Proceedings of the IEEE International Conference on Computer Vision. Washington D. C., USA: IEEE Press, 2017: 5715-5725.
10	ZHOU F , SHUI C J , YANG S C , et al. Discriminative active learning for domain adaptation. Knowledge-Based Systems, 2021, 222, 106986. doi: 10.1016/j.knosys.2021.106986
11	刘大鹏, 曹永锋, 张伦. 深度迁移主动学习研究综述. 现代计算机, 2021 (10): 88- 93.
	LIU D P , CAO Y F , ZHANG L . Review of research on deep transfer active learning. Modern Computer, 2021 (10): 88- 93.
12	MA X H, GAO J Y, XU C S. Active universal domain adaptation[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Washington D. C., USA: IEEE Press, 2021: 8968-8977.
13	江彪, 龙坤, 谢佳鑫, 等. 基于主动迁移学习的图像目标自动标注. 指挥信息系统与技术, 2021, 12 (5): 61- 69.
	JIANG B , LONG K , XIE J X , et al. Automatic annotation for image objects based on active transfer learning. Command Information Systems and Technology, 2021, 12 (5): 61- 69.
14	DENG C , LIU X L , LI C , et al. Active multi-kernel domain adaptation for hyperspectral image classification. Pattern Recognition, 2018, 77, 306- 315. doi: 10.1016/j.patcog.2017.10.007
15	REN P Z , XIAO Y , CHANG X J , et al. A survey of deep active learning. ACM Computing Surveys, 2022, 54 (9): 1- 40.
16	FU B, CAO Z J, WANG J M, et al. Transferable query selection for active domain adaptation[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville, USA: IEEE Press, 2021: 7272-7281.
17	HUANG S J, ZHAO J W, LIU Z Y. Cost-effective training of deep CNNs with active model adaptation[C]//Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. New York, USA: ACM Press, 2018: 1580-1588.
18	SU J C, TSAI Y H, SOHN K, et al. Active adversarial domain adaptation[C]//Proceedings of the IEEE Winter Conference on Applications of Computer Vision. Washington D. C., USA: IEEE Press, 2020: 739-748.
19	ZHAO Y , DAI G J , BORGHINI G , et al. Label-based alignment multi-source domain adaptation for cross-subject EEG fatigue mental state evaluation. Frontiers in Human Neuroscience, 2021, 15, 706270. doi: 10.3389/fnhum.2021.706270
20	SHE Q S , SHI X S , FANG F , et al. Cross-subject EEG emotion recognition using multi-source domain manifold feature selection. Computers in Biology and Medicine, 2023, 159, 106860.
21	GANIN Y, LEMPITSKY V. Unsupervised domain adaptation by backpropagation[EB/OL]. [2024-05-10]. http://de.arxiv.org/pdf/1409.7495.
22	PRABHU V, CHANDRASEKARAN A, SAENKO K, et al. Active domain adaptation via clustering uncertainty-weighted embeddings[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Washington D. C., USA: IEEE Press, 2021: 8505-8514.
23	KUMAR P , GUPTA A . Active learning query strategies for classification, regression, and clustering: a survey. Journal of Computer Science and Technology, 2020, 35 (4): 913- 945.
24	ALBUQUERQUE I , TIWARI A , PARENT M , et al. WAUC: a multi-modal database for mental workload assessment under physical activity. Front Neurosci, 2020, 14, 549524. doi: 10.3389/fnins.2020.549524
25	CASTELLANOS N P , MAKAROV V A . Recovering EEG brain signals: artifact suppression with wavelet enhanced independent component analysis. Journal of Neuroscience Methods, 2006, 158 (2): 300- 312. doi: 10.1016/j.jneumeth.2006.05.033
26	XIE B H , YUAN L H , LI S , et al. Active learning for domain adaptation: an energy-based approach. Artificial Intelligence, 2022, 36 (8): 8708- 8716.
27	HAN K , KIM Y , HAN D , et al. TL-ADA: transferable loss-based active domain adaptation. Neural Networks, 2023, 161, 670- 681.
28	PAN S J , TSANG I W , KWOK J T , et al. Domain adaptation via transfer component analysis. IEEE Transactions on Neural Networks, 2010, 22 (2): 199- 210.
29	LONG M S, WANG J M, DING G G, et al. Transfer feature learning with joint distribution adaptation[C]//Proceedings of the IEEE International Conference on Computer Vision. Washington D. C., USA: IEEE Press, 2013: 2200-2207.
30	TZENG E, HOFFMAN J, ZHANG N, et al. Deep domain confusion: Maximizing for domain invariance[C]//Proceedings of the IEEE Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2014: 358-366.
31	ASH J T, ZHANG C, KRISHNAMURTHY A, et al. Deep batch active learning by diverse, uncertain gradient lower bounds[C]//Proceedings of the International Conference on Learning Representations. Washington D. C., USA: IEEE Press, 2019: 2231-2245.
32	ALBUQUERQUE I , MONTEIRO J , ROSANNE O , et al. Estimating distribution shifts for predicting cross-subject generalization in electroencephalography-based mental workload assessment. Frontiers in Artificial Intelligence, 2022, 5, 992732.
33	ZHOU Y , WANG P , GONG P , et al. Cross-subject cognitive workload recognition based on EEG and deep domain adaptation. IEEE Transactions on Instrumentation and Measurement, 2023, 72, 3276515.

[1]	游雅倩, 闫辉, 苏耀峰, 王晓双, 鄢睿丞. 基于EnsembleBRB-SHAP的航空发动机健康状态可解释预测方法[J]. 计算机工程, 2026, 52(4): 386-397.
[2]	董方和, 石琼, 师智斌. 基于集成学习与异常检测的对抗流量检测方法[J]. 计算机工程, 2026, 52(2): 275-286.
[3]	谢久超, 苌道方. 基于AGSCOA-Stacking特征加权的船用钢板焊接余量预测[J]. 计算机工程, 2026, 52(1): 414-426.
[4]	王光明, 李冬青, 蒋从锋. 不平衡数据集下的数据中心网络流量异常检测[J]. 计算机工程, 2025, 51(8): 227-237.
[5]	单永航, 张希, 胡川, 丁涛军, 姚远. 基于集成学习的交通事故严重程度预测研究与应用[J]. 计算机工程, 2024, 50(2): 33-42.
[6]	曲坤, 王震龙, 刘志锋. 基于多表征学习的交叉熵集成图像分类方法[J]. 计算机工程, 2024, 50(10): 322-333.
[7]	包善书, 车波, 邓林红. 基于双源域迁移学习的肺音信号识别[J]. 计算机工程, 2023, 49(9): 295-302, 312.
[8]	陈仲磊, 伊鹏, 陈祥, 胡涛. 基于集成学习的系统调用实时异常检测框架[J]. 计算机工程, 2023, 49(6): 162-169,179.
[9]	艾成豪, 高建华, 黄子杰. 混合特征选择和集成学习驱动的代码异味检测[J]. 计算机工程, 2022, 48(7): 168-176,198.
[10]	王萌铎, 续欣莹, 阎高伟, 史丽娟, 郭磊. 基于AdaBoost集成加权宽度学习系统的不平衡数据分类[J]. 计算机工程, 2022, 48(4): 99-105,112.
[11]	宝音图, 刘伟, 牛朝阳, 李润生, 张浩波. 联合集成学习与EfficientNet的光学遥感图像场景分类[J]. 计算机工程, 2021, 47(10): 226-235.
[12]	王旭, 陈永乐, 王庆生, 陈俊杰. 结合特征选择与集成学习的密码体制识别方案[J]. 计算机工程, 2021, 47(1): 139-145,153.
[13]	李培媛, 黄迟. 基于CSPPNet与集成学习的人类蛋白质图像分类方法[J]. 计算机工程, 2020, 46(8): 235-242.
[14]	仇瑜, 程力. 面向财税领域的实体识别与标注研究[J]. 计算机工程, 2020, 46(5): 312-320.
[15]	周胜, 刘三民. 基于动态策略的多源迁移学习数据流分类研究[J]. 计算机工程, 2020, 46(5): 139-143,149.

选择文件类型/文献管理软件名称

选择包含的内容