Research on Sensing Cardiac Cycles Using a Microbend Fiber Sensor
Objective The detection of the cardiac cycle is essential for the diagnosis and analysis of cardiovascular diseases.In addition to bioelectric signals,a ballistocardiogram(BCG)is another physiological signal that can be used for cardiac cycle detection.Unlike electrocardiograms(ECGs),the collection of BCG signals does not require direct contact with the skin and is safe and noninvasive.The use of microbend fiber optic sensors for BCG sensing has the advantages such as simple structure,low cost,resistance to electromagnetic interference,and high sensitivity.When collecting BCG using a microbend fiber optic sensor,environmental noise,circuit noise,optical path noise,respiratory signals,and motion artifacts affect the performance of the sensor.These noise sources jointly destroy the BCG waveform,and the BCG exhibits nonlinear,nonstationary characteristics and significant individual differences in both the time and frequency domains,thereby affecting the accurate identification of physiological parameters such as the cardiac cycle from the BCG.Therefore,the corresponding signal-processing algorithms should be studied to achieve BCG waveform extraction and feature recognition.Methods The paper investigates the use of a microbend fiber optic sensor to obtain BCGs for cardiac cycle detection,which primarily includes three processes:BCG optical fiber sensing,BCG waveform extraction,and BCG feature recognition.For BCG optical fiber sensing,the microbend fiber optic sensor consists of a multimode optical fiber,grid structure,light source,photodetector,and signal processing circuit,which are embedded in the seat cushion to acquire the heart and lung vibration signals.An algorithm based on the smoothness prior approach(SPA)combined with improved variational mode decomposition(IVMD)is proposed for BCG waveform extraction.Based on the principle of regularized least squares,the SPA is first used to suppress the low-frequency trend term of the acquired signal.Subsequently,IVMD combined with central frequency and correlation analysis is used to suppress the high-frequency noise of the acquired signal.The center frequency of the intrinsic mode function(IMF)is used to determine the optimal number of layers for VMD decomposition,and the IMF correlation coefficient is used to select the IMF for signal reconstruction.For BCG feature recognition,the proposed algorithm uses prior information such as amplitude features and peak time intervals to locate feature peaks and then extracts parameters such as the heart rate and cardiac cycle.Results and Discussions The weak heart and lung vibration signals acquired by the microbend fiber optic sensor contain an apparent periodic signal with a frequency of 5 Hz[Fig.5(b)],that is,the BCG,which provides a basis for the cardiac cycle segmentation of the signal.The SPA can effectively suppress the low-frequency trend term of the acquired signal(blue dotted line in Fig.7),and IVMD can effectively suppress the high-frequency noise of the acquired signal(red solid line in Fig.7).The J-peak of the BCG signal can be effectively located using the amplitude features and peak time intervals(Fig.8).The BCG and ECG signals of five participants are acquired simultaneously.With the ECG cardiac cycle as the reference standard,the algorithm obtains cardiac cycles from the BCG with a maximum standard deviation of 0.0287 s(Fig.10).The average heart rate is calculated using peak localization/wave group segmentation methods such as short-term energy(STE),template matching(TM),clustering approach(CA),and the proposed algorithm.The error of the average heart rate using the proposed algorithm is 0.69％,which is better than the three feature localization algorithms STE,TM,and CA(Table 3).Conclusions The proposed BCG waveform extraction algorithm combined with SPA and IVMD effectively suppresses the low-frequency trend term and high-frequency noise of an acquired signal.The proposed BCG feature recognition algorithm for locating feature peaks utilizing prior information such as amplitude features and peak time intervals can accurately obtain the cardiac cycle.The BCG-ECG signals of five participants are acquired.With the ECG cardiac cycle as the reference standard,the algorithm obtains cardiac cycles from the BCG with a maximum standard deviation of 0.0287 s,and the error in the average heart rate is 0.69％,which is better than that of the three feature location algorithms STE,TM,and CA.The BCG waveform extraction and feature recognition algorithm can effectively extract the cardiac cycle from the weak heart and lung vibration signals obtained by the microbend fiber optic sensor.
万方数据
维普
