基于光电容积脉搏波的精神疲劳评估方法

Mental fatigue assessment method based on photoplethysmography

目的针对精神疲劳难于定量评估的问题,本文探索一种非侵入式可穿戴检测方法获取人体生理参数,从而实现对人体精神疲劳的定量评估。方法搭建光电容积脉搏波(photoplethysmography,PPG)采集平台,采集20名健康在校生的PPG信号,对PPG信号进行预处理和特征提取,获取时域、频域共143维特征。使用机器学习算法建立分类模型,对于Pearson相关系数法、F检验和relief-F得到的特征权值,选择最优的特征子集,使用降维后的特征子集训练模型,减少复杂度和过拟合概率。结果与实际状态对比,基于该方法的单个体疲劳检测平均准确率为92.48%,多个体疲劳检测准确率最大值为92.2%,可以有效地识别精神疲劳。结论光电容积脉搏波信号经过时域和频域分析构建的特征能够使用机器学习算法进行准确的精神疲劳状态分类评估。

ObjectiveIn view of the difficulty in quantitative assessment of mental fatigue, this paper explores a non-invasive wearable detection method to obtain human physiological parameters, so as to achieve quantitative assessment of human mental fatigue.MethodsPhotoplethysmography(PPG) signals of 20 healthy school students were collected by PPG acquisition platform. PPG signals were preprocessed and features were extracted. A total of 143 dimensional features were obtained in time and frequency domains. The machine learning algorithm was used to establish the classification model, and the feature weights were obtained by comparing Pearson correlation coefficient method,Ftest and relief-F. The optimal feature subset was selected,and the feature subset training model after dimension-reduction was used to reduce the complexity and overfitting probability.ResultsCompared with the actual state, the average accuracy of individual fatigue detection based on this method was 92.48%,the maximum accuracy of multi-individual fatigue detection was92.2%, which could effectively identify mental fatigue.ConclusionsThe features constructed by timefrequency domain analysis of photoplethysmography signals can be used for accurate mental fatigue state classification assessment using machine learning algorithms.