中文 | English
Hefei University of Technology Logo J. Hefei Univ. Tech. (Nat. Sci.)

Navigation

Method of ballistocardiogram signals modeling and feature extraction

Download Full Text

心冲击信号建模与特征提取方法

Journal Information

J. Hefei Univ. Tech. (Nat. Sci.), May.2025, Vol.48 No.5: 665-669, 676

DOI: 10.3969/j.issn.1003-5060.2025.05.013

Authors

REN Xueqian, CHEN Enwei, DING Jinlei, PENG Weibin

(Institute of Sound and Vibration Research, Hefei University of Technology, Hefei 230009, China)

Abstract and Keywords

Abstract: In this paper, a method of normal distribution based on Gaussian process regression (GPR) was used to derive the posterior distribution, which was obtained by adding the single-cycle cardiac phase parameter prediction on the basis of the prior distribution. The morphological model of human ballistocardiogram (BCG) signals in the quiet state and in the post-exercise state was developed. The BCG signals were extracted from experimentally acquired raw vibration signals to provide reference for model features in the time and frequency domains. The body micro-vibration signal collected by the sensor in the experiment was decomposed into multiple intrinsic mode component signals by the variational mode decomposition (VMD) method. The energy ratio of the energy in the eigenfrequency domain of each intrinsic mode component to the energy of each intrinsic mode component was calculated separately, and then the intrinsic mode components whose energy ratios in the eigenfrequency domain were far greater than those of other components were selected to perform adaptive reconstruction of BCG signals. The agreement verification between model and experimentally reconstructed peaks was achieved by the Bland-Altman method. The adaptively reconstructed BCG signals were processed by the Shannon energy envelope method, and after that the characteristic peaks were detected by the sliding window method, and the characteristic peak time intervals were calculated to obtain the cardiac cycle and heart rate. This non-contact heart rate detection method can warn cardiovascular diseases.

Keywords: ballistocardiogram(BCG) signals; Gaussian process regression(GPR); heart rate; noninvasive monitoring

作者信息

任雪倩，陈恩伟，丁金磊，彭伟滨

（合肥工业大学噪声振动工程研究所，安徽合肥230009）

摘要和关键词

摘要: 文章采用高斯过程回归的正态分布方法, 在先验分布的基础上由单周期心相参数推导后验分布, 建立人体在安静状态和运动后状态的心冲击(ballistocardiogram, BCG)信号形态学模型, 从实验采集原始振动信号中提取心冲击信号为时域和频域的模型特征提供参考。对实验中由传感器采集的身体微振动信号, 通过变分模态分解方法得到多个固有模态分量信号, 结合建模信号的主要频域特征分布范围, 通过计算各固有模态信号中特征频域范围内的能量占各固有模态信号分量能量比, 筛选特征频域能量占比远大于其他分量的固有模态分量, 进行心冲击信号的自适应性重构; 使用Bland-Altman方法验证模型特征峰和实验重构特征峰的一致性; 对重构的心冲击信号采用香农能量包络方法, 通过滑动窗口寻峰方法检测特征峰以计算心率, 该非接触式心率检测方法可预警心血管疾病。

关键词: 心冲击（BCG）信号；高斯过程回归；心率；非侵入式监测

Funding

安徽省自然科学基金资助项目（2208085ME130）

Previous Next
Back to Contents

Account

Submission Login Reviewer Login Editor Login