Chou Yongxin, Zhang Aihua, Liu Jicheng, Lin Jiajun, Huang Xufeng
School of Electrical and Automatic Engineering, Changshu Institute of Technology, Suzhou, Jiangsu 215500, P.R.China;The East China Science and Technology Research Institute of Changshu Co., Ltd, Suzhou, Jiangsu 215500, P.R.China.
College of Electrical and Information Engineering, Lanzhou University of Technology, Lanzhou 730050, P.R.China.
Sheng Wu Yi Xue Gong Cheng Xue Za Zhi. 2020 Feb 25;37(1):61-70. doi: 10.7507/1001-5515.201904024.
In order to quantitatively analyze the morphology and period of pulse signals, a time-space analytical modeling and quantitative analysis method for pulse signals were proposed. Firstly, according to the production mechanism of the pulse signal, the pulse space-time analytical model was built after integrating the period and baseline of pulse signal into the analytical model, and the model mathematical expression and its 12 parameters were obtained for pulse wave quantification. Then, the model parameters estimation process based on the actual pulse signal was presented, and the optimization method, constraints and boundary conditions in parameter estimation were given. The spatial-temporal analytical modeling method was applied to the pulse waves of healthy subjects from the international standard physiological signal sub-database Fantasia of the PhysioNet in open-source, and we derived some changes in heartbeat rhythm and hemodynamic generated by aging and gender difference from the analytical models. The model parameters were employed as the input of some machine learning methods, e.g. random forest and probabilistic neural network, to classify the pulse waves by age and gender, and the results showed that random forest has the best classification performance with Kappa coefficients over 98%. Therefore, the space-time analytical modeling method proposed in this study can effectively quantify and analyze the pulse signal, which provides a theoretical basis and technical framework for some related applications based on pulse signals.
为了定量分析脉搏信号的形态和周期，提出了一种脉搏信号的时空分析建模与定量分析方法。首先，根据脉搏信号的产生机制，将脉搏信号的周期和基线纳入分析模型，构建了脉搏时空分析模型，得到了用于脉搏波量化的模型数学表达式及其12个参数。然后，给出了基于实际脉搏信号的模型参数估计过程，并给出了参数估计中的优化方法、约束条件和边界条件。将时空分析建模方法应用于开源的PhysioNet国际标准生理信号子数据库Fantasia中健康受试者的脉搏波，从分析模型中得出了衰老和性别差异所产生的心跳节律和血流动力学的一些变化。将模型参数作为随机森林和概率神经网络等一些机器学习方法的输入，按年龄和性别对脉搏波进行分类，结果表明随机森林具有最佳的分类性能，卡帕系数超过98%。因此，本研究提出的时空分析建模方法能够有效地对脉搏信号进行量化和分析，为基于脉搏信号的一些相关应用提供了理论基础和技术框架。
