计算机应用研究2025，Vol.42Issue(6)：1734-1741,8.DOI:10.19734/j.issn.1001-3695.2024.10.0457
一种面向情绪压力分布外检测的多任务跨模态学习方法
Multi-task cross-modal learning approach for out-of-distribution detection of emotional stress
摘要
Abstract
Recent research indicates that emotional stress detection systems based on PPG signals can be a potential conve-nient solution.However,PPG-based methods usually induce severe OOD issues when detecting stress in previously unseen subjects due to significant variations in PPG signals across individuals.To address this challenge,this paper proposed a cross-modal stress detection model based on multi-task learning(CSMT).By introducing ECG signal reconstruction and multiple cardiovascular feature prediction as auxiliary tasks to enhance the feature extraction capability of PPG signals,the proposed method performed collaborative optimization of PPG-based stress detection in high-dimensional representation space,thereby learning robust stress detection representations across individuals.Experimental results on the WESAD dataset demonstrate that in leave-one-subject-out validation tests,CSMT achieves best accuracy and F1 scores compared to existing methods in both three-class(neutral/stress/amusement)and binary(stress/non-stress)classification tasks,meanwhile effectively mitigating the OOD generalization problem in stress detection.The ablation experiments further validate the effectiveness of CSMT in en-hancing model generalization capability.关键词
多任务学习/光电容积脉搏波/压力检测/分布外问题Key words
multi-task learning/photoplethysmography(PPG)/stress detection/out-of-distribution(OOD)issues分类
信息技术与安全科学引用本文复制引用
万奕晨,邢凯,刘宇,杨慧,徐筠涵,袁艳雪..一种面向情绪压力分布外检测的多任务跨模态学习方法[J].计算机应用研究,2025,42(6):1734-1741,8.基金项目
江苏省重点研发项目(BE2020665) （BE2020665）
