Study Of Heart Rate Variability Analysis Methods And Application

Heart rate variability(HRV) refers to the heart rate or heart rate changes that may varywith time nuances, which can reflect whether the human autonomic nerve, sympathetic nerveand vagus nerve regulation ability is strong or not on the cardiovascular system, at the sametime it can be used to analyze the strength of the heart to do work. Because of HRV caneffectively from the heart activity to reflect the state of human body, and HRV in research andautonomic nervous activity related diseases such as diabetes, high blood pressure, coronaryheart disease, heart transplantation and hyperthyroidism also has important significance andvalue, so the HRV signal is a hot spot of medical research in recent years. If it can effectivelydetect the characteristic value of HRV signal such as RR interphase, power spectral density, itwill be able to use these information to judge and predict heart related disease and put forwardeffective cure methods.This paper is based on the wavelet transform to conduct the contrast analysis of the effectof several different kinds of wavelet signal decomposition, the results found that in ecg HRVsignal, high SNR gain Db6 wavelet base, decomposed the effect is good. Secondly, this paperchooses the Db6 wavelet on the basis to conduct the comparative analysis of four differentbased on wavelet transform denoising method including soft threshold denoising, hardthreshold denoising, Garrote threshold denoising and the improved Garrote thresholddenoising method. Through the analysis, it found that based on the improved Garrotethreshold denoising method in the treatment of the ecg signal method has obvious advantages.Again, this paper uses Mallat algorithm to realize the QRS complex of ecg signal HRVcharacteristic information, RR interphase feature information extraction, and drawn from theMIT/BIH ecg database 15 set of ecg signals validate HRV QRS complex detection results.Finally, it uses AR model to conduct analysis of HRV signal power spectrum, and extracts theSDNN, LF/HF, HR, SDANN, TP, PNN50 and other features of ecg signal. Through 5 groupsthat can be drawn from MIT- BIH normal ecg signals and arrhythmia ecg signals of 200 groups of data analysis, it finds that the SDNN, LF/HF, HR and PNN50 correlation is small,can be used as the characteristic variable which could be used to separate signal classification.So this paper chooses the SDNN, LF/HF, HR and PNN50 characteristic values to conduct theclassification of ecg signal.This papers uses the HRV eigenvalue extraction method including signal denoising, QRScomplex detection, power spectrum analysis and other methods to design the interface of ecgsignals classification and recognition based on the characteristics of HRV values. It draws fivekinds groups of ecg data from the MIT/BIH database for training and testing. This paperdesigns three kinds of classification system: the classification system based on the decisiontree the classification system based on BP neural network, the classification system basedon support vector machine(SVM). At the same time, this paper compares two methods ofdimensionality reduction: traditional dimensionality of PCV, and nonnegative matrixdecomposition NMF dimension reduction. Based on the two dimension reduction method andthe combination of three kinds of classification algorithms, it conducts ecg signalclassification experiment, the results found in this paper that the classification system basedon the decomposition of nonnegative matrices and the decision tree methods has highaccuracy to classify ecg signals, and it has obvious advantage diagnosis. At the same time,through the choice of different electrical signal eigenvalue analysis, this paper extracts thefour features based on HRV signal including SDNN, HR, LF/HF, PNN50, it finds that the fourvalues can effectively improve the classification and diagnosis of ecg signal.