Nonlinear Characteristic Analysis Of BCG Signals Based On Chaotic Theory

With the change of people’s lifestyle,the number of cardiovascular disease patients in China is increasing continuously,and a large number of heart disease outbreaks occur outside of the hospital.Thus,it is necessary to find a non-invasive family cardiac monitoring method to monitor the state of the heart outside of the hospital continuously,and to warn the cardiac abnormalities in advance,so as to reduce the risk and loss caused by sudden heart disease.Ballistocardiogram(BCG)is a kind of cardiac physiological signal which can be acquired by non-contact method.It is a tiny vibration of the body caused by the impact of heart contraction and blood ejection on blood vessels.In this thesis,a sitting BCG signal acquisition system based on piezoelectric film sensor is designed to realize the non-contact acquisition of cardiac physiological signals.Then,the characteristics of the signals are extracted and analyzed.The main work is as follows:(I)According to the principle of BCG signal generation,the piezoelectric film sensor is placed in the cushion.Composed of charge amplifier circuit,filter circuit,post-amplifier circuit and data acquisition card,the signal acquisition system is first designed,then manufactured.ECG and pulse signal are collected synchronously.Being able to collect and record three kinds of cardiac physiological signals effectively,the acquisition system has good stability.(2)The denoising method of BCG signal based on wavelet transform is studied.According to the characteristics of BCG signal and noise sources,Chebyshev Ⅱ ⅡR digital filter is selected to remove the respiratory trend and noise introduced by analog circuit and environment in frequency domain.Then the influence of wavelet base,decomposition scale and threshold function on the denoising effect of wavelet is discussed.The improved threshold method based on Sym8 wavelet base is used to denoise BCG signal.This method can effectively remove the noise introduced in the process of signal acquisition,and obtain clear waveforms that conform to the standard BCG signal characteristics.(3)The method of extracting and analyzing characteristic points of respiratory rate and BCG signal was studied.Based on the respiratory trend and normal respiratory frequency of the original signal,the basic waveform of the respiratory signal is extracted by low-pass filtering,and then the respiratory rate is obtained by marking the feature points.The method of threshold selection in feature point marking is discussed.The method of combining amplitude and interval thresholds is used to accurately mark J wave of BCG signal.Then K,I and L wave are sequentially marked by extremum recursion.And QRS wave group of ECG signal is also marked in the same way.Finally,the characteristic point interval of BCG signal and J-R interval between two kinds of signals are calculated,drawing and comparing the R-R interval scatter plots of ECG signal and J-J interval scatter plots of BCG signal respectively.Statistical analysis showed that each band of BCG signal was stable and synchronized with ECG signal in a single cycle,which could accurately reflect the cardiac cycle.(4)The method of non-linear characteristic analysis of BCG signal based on chaos theory is introduced,calculating the characteristic parameters.The power spectrum analysis of the signal confirms that the BCG signal is a chaotic signal with deterministic law;the C-C algorithm is used to calculate the delay time and embedding dimension,then the phase space of the BCG signals are reconstructed.It is found that the phase space reconstructed map of healthy subjects is more regular than that of abnormal subjects;the correlation dimension and the maximum Lyapunov exponent of the signals are calculated by G-P algorithm and Wolf method.The results also show that the BCG signal of healthy people is more chaotic.(5)Support vector machine(SVM)is used to classify signals and recognize anomalies The anomaly recognition rate is compared when different eigenvectors are input.The results show that the recognition rate of signal anomaly is obviously improved by adding the non-linear chaotic features.