| A large amount of information about heart function and its nervous activity is containedin ECG and HRV signal. Therefore, it is greatly significant to extract and process theinformation for the prevention, diagnosis and therapy etc. of heart disease. The thesis mainlydiscusses the ECG preprocessing, feature analysis and heart rate variability, which is based onthe predecessors’ research results and literature material.The physiologic principle and mechanism, typical feature and clinical diagnostic value ofECG and HRV signal are briefly illustrated in this thesis; the present developing status ofpreprocess methods of ECG signal and the methods of feature waveform recognition of ECGsignal and HRV analysis methods are summarized. According to the existing deficiencies ofthe typical analysis algorithm, some studies have been tried to improve and innovate theanalysis methods of ECG and HRV signal based on the new signal analysis methods, andsome research outcomes have been obtained, which turns out to be significant to develop newECG automatic analysis system. It mainly includes the following research works:1) As to the de-noising method of wavelet threshold, Gibbs oscillatory phenomena easilyoccurs in Q,S waveform of ECG signal and the serious distortion for T waveform and STsegment waveform of ECG signal can be easily caused when baseline drift is eliminated. Tosolve this problem, the one-off denoising method of ECG signal combining stationary wavelettransform and adaptive filter is proposed. The power-line interference and EMG noise iseliminated based on the de-noising method of wavelet threshold in low scale decompositioncomponent, and the adaptive filter is introduced so as to eliminate baseline drift in large scaledecomposition component,the ECG waveform form remain well after the reconstruction, theone-off denoising is achieved. Simulation and comparison test result shows that thissynthetical filter can avoid Gibbs oscillatory phenomena effectively and reduce the distortionfor T waveform and ST segment waveform of ECG signal, which laid foundation for accuratefeature information extraction of ECG signal.2) The combination algorithm of wavelet transform and nonlinear energy operator isproposed for R waveform detection of ECG signal, which uses Marr wavelet as wavelet basis.Through demarcating the location of R waveform by the extreme value point on3scalecomponent of wavelet decomposition, the effect of noise is effectively restrains,and thedisadvantage that the location of zero-crossings are needed for R waveform detection byspline wavelet transform is overcome. Smooth nonlinear energy operator is operated on3scale component so as to protrude R waveform peak and suppress high T waveform and Pwaveform. In comparison with the R waveform detection method of traditional wavelet transform, the method shows high detection rate, better ability to anti-interference, much lesscomputation and good real-time performance. Simulation test shows that the method iseffective for actual ECG data in the MIT/BIH arrhythmia data.3) A time-frequency analysis method is proposed for HRV signal based on Hilberttime-frequency spectrum. The method analyzes the difference of Hilbert time-frequencyspectrum for HRV signal under different physiological or pathological conditions, and theHilbert energy bar chart of HRV signal is plotted in the different physiological frequency bandby the characteristic index of short-time HRV signal in the linear frequency domain,theenergy features are extracted in the different physiological frequency band as thetime-frequency feature of quantitative evaluating for heart rate variability, the correlationbetween the heart rate change and the time-frequency features is studied in this paper, andpoints out that the change of heart rate is a important factor considered in HRV signalanalysis.Simulation tests to HRV signal of the young, old and heart failure patients show thatthe new time-frequency features have good differentiating performance, and can accuratelyreflect the regulatory function and damage of sympathetic and vagal nervous system of heart.4) The ventricular tachycardia (VT) and ventricular fibrillation (VF) event is predictedbased on improved HHT method. Firstly, the RR interval series preceding the onset of VT/VFevents is transformed instantaneous heart rate (IHR) series, and it is pre-decomposed into aseries of narrow-bands signal by wavelet packet transform, then EMD decomposition iscarried out for each narrow-band component, the undesirable components are removed by thecorrelation threshold identification, the new intrinsic mode function (IMF) components andHilbert marginal spectrum of IHR series are obtained, the amplitude features of differentphysiological frequency band are extracted to be used for prediction features of VT/VF events.Simulation test shows that high frequency amplitude, very high frequency amplitude and totalamplitude of IHR series preceding the onset of VT/VF are significantly higher than that ofnormal sinus rhythm, and low-to-high frequency amplitude ratio is significantly lower thanthat of normal sinus rhythm. It suggests that the activity of sympathetic and vagal nerveappears to increase preceding the onset of VT/VF event, and the activity of vagal nerve ismore marked, which causes imbalance of the sympathetic and vagal nerve activity.Simulation test is carried out for the improved HHT method and the traditional HHT, thestatistical analysis show that the differentiated performance of the improved HHT is morebetter than that of traditional HHT.5) The conception of HHT marginal spectrum entropy and energy spectrum entropy andthe analysis method of HRV signal are proposed. The complexity analysis is processed for theconventional signal with some degree of white noise and chaotic time series. The results show that the method is superior to the analysis method of the traditional complexity and entropy indepicting signal complexity and anti-pulse interference. Applying the new approach to HRVsignal of the young, older and atrial fibrillation patients, the results show that the two entropyof frequency domain can sensitively detect the physiological and pathological changes fromHRV signal, and its analytical performace is better than that of the traditional power spectrumentropy method.6) The new method of HRV signal analysis is proposed based on the Hilbert spectrumentropy dividing frequency range. the Hilbert spectrum entropy of HRV signal in differentfrequency range is calculated, which is used as the characteristic evaluation index of heart ratevariability. The Hilbert spectrum entropy can reflect the uncertainty of HRV signal energy ontime-frequency distributions, the Hilbert spectrum entropy features of HRV signal based onappropriate separation for various physiological factors by dividing frequency range are moreclosely linked to different physiological and pathological functions and more conducive tomanifest the regulation law of autonomic nervous system of heart. The results of thesimulation test to HRV signal of the sample group for the young, elder and patients with atrialfibrillation, and the sample group for health person and congestive heart failure(CHF) patientsshow that this method can effectively differentiate different sample group, and can accuratelyreflect the regulation law of sympathetic and vagal nervous system of heart in differentphysiological and pathological functions.At last, the dissertation sums up the research of this project and makes the expectationabout the development direction of the assignment. |
PDF Full Text Request