基于HRV特征和机器学习融合模型的精神压力评估方法研究  

Research of Mental Stress Assessment Based on HRV Features and Machine Learning Fusion Model

作　　者：俞锴婷 俞政涛 王俊青 叶家浩 赵兴群[3] YU Kaiting;YU Zhengtao;WANG Junqing;YE Jiahao;ZHAO Xingqun(College of Software,Southeast University,Suzhou Jiangsu 215000,China;Nanjing Osteotech Biotechnology Co.,Ltd.,Nanjing Jiangsu,210000,China;College of Biological Science and Medical Engineering,Southeast University,Nanjing Jiangsu 210096,China)

机构地区：[1]东南大学软件学院,江苏苏州215000 [2]南京澳思泰生物科技有限公司,江苏南京210000 [3]东南大学生物科学与医学工程学院,江苏南京210096

出　　处：《电子器件》2025年第1期91-97,共7页Chinese Journal of Electron Devices

摘　　要：在当代生活方式中,与压力相关的健康状况日益普遍。近年来采用机器学习识别精神压力已成为研究的热点,然而对压力指数的量化计算依旧是难题,且单一模型的预测准确性难以提升,因此提出利用机器学习融合模型来提供一种评估压力指数的方法。从光电容积脉搏波信号中提取HRV特征,结合DASS-21问卷调查结果,在考虑性别年龄的基础上,测试不同的机器学习方法,并进行模型融合以提高预测的准确性,最终得到范围在0至42的精神压力指数。以均方根误差作为预测准确性的评价指标,结果表明融合模型的预测准确率高于单一模型,其中GB与MLP的组合是最理想的压力预测器,均方根误差不超过1.35。所提出的方法将对无创检测,量化压力水平以及临床实践具有潜在的价值。There is a rising prevalence of stress-related health conditions associated with the stressful contemporary lifestyle.Assessing mental stress by using machine learning method has become a research hotspot.However,it is still a difficult subject to quantify the stress index and the accuracy of a single model is difficult to improve.Hence a method to quantize mental stress by using machine learn-ing model fusion is proposed.Heart rate variability(HRV)features extracted from photoplethysmogram(PPG)and the results of DASS-21 questionnaire are combined.Different machine learning algorithms are tested to quantify stress indices ranging from 0 to 42 in considera-tion of gender and age.Moreover,fusion models are trained to obtain higher accuracy.Root mean square error is used as the evaluation index of prediction accuracy.The result suggests that the prediction accuracy of the fused model is higher than that of each single model.GB fused with MLP is the most ideal predictor of stress indices with the root mean square error less than 1.35.The method proposed will have potential value for noninvasive detection,quantification of stress levels and clinical practice.

关 键 词：精神压力 HRV 机器学习 模型融合 

分 类 号：TN911.7[电子电信—通信与信息系统] R318[电子电信—信息与通信工程]