Research On Entropy Based On Detrending Term RR Interval Series

Heart is one of the important organs of the human body to maintain normal life activities,with very complex structure and physiological function.The heart system is a typical chaotic system in nature,and its output signal often exhibits complex variability,which is called heart rate variability(HRV)signal,also known as RR interval sequence.This signal contains a wealth of information about cardiovascular regulation,and has been used as an important subject to evaluate the level of autonomic nervous activity and clinical detection of cardiac health.The output waveform of the RR interval sequence appears to have regular and actual sight differences,which is a complex chaotic signal.The research on RR interval sequence must be characterized by nonlinear chaotic index.Among many nonlinear chaotic indicators,entropy analysis is a method to measure the irregularity and complexity of time series,and can be used as an index to identify different physiological states of RR interval sequence.Entropy analysis is essentially a method in which the incidence of new temporal models changes with the embedding dimensions,reflecting the complexity of the data structure.However,when the entropy analysis is used as an indicator to measure nonlinear sequence,it is susceptible to various trends or noises of the signal,resulting in the instability and uncertainty of the calculation results.In the actual ECG signal acquisition process,due to the influence of sensors,ECG signal is usually exposed to various low-frequency noises,such as human breathing,sweating,slight electrode movement,or even slow muscle movement.These low-frequency noise interference or trend term infiltration usually affects the entropy value,making the calculation results inaccurate,thus affecting the accuracy of the experimental results.Therefore,reducing and eliminating the trend term in test signal is an important step in data processing.However,there are still limitations in the elimination of nonlinear trend items.Based on the previous research,this paper makes the following innovations:(1)using three nonlinear methods:wavelet analysis,ensemble empirical mode decomposition and smoothness prior approach,the low-frequency slowly-changing components(the trend term part)are detrended of the chaotic classical model Logistic mapping sequence,and then combining Welch-PSD spectrogram to analysis of the Logistic sequences before and after detrending processing.The results show that the wavelet analysis method eliminates the effective components whose frequencies are less than 100 Hz,which leads to a missing signal of the component of interest and makes a great error in the experimental results;EEMD has some effect on eliminating baseline drift,but it has not been eliminated obviously when the frequencies are less than 12 Hz,which is not obvious for eliminating signal trend item and has no advantage;The SPA algorithm obviously eliminates this part of the trend item,and the frequency response shows that the SPA algorithm eliminates the trend items obviously.Meantime,the frequency response shows that the SPA algorithm to eliminate the trend items whose frequencies are below the cut-off frequency of 0.1668 Hz,and finally obtains the optimality of SPA to remove the nonlinear trend.(2)The approximate entropy and the fuzzy entropy of HRV signal superimposed on healthy people under circadian rhythm are calculated,using a t test,P(day,night)>0.05,the results show that the two algorithms are very sensitive,and are susceptible to the trend among the signals,making it difficult to clearly distinguish RR interval sequence between day and night.(3)In order to improve the recognition of the two entropy algorithms under strong interference of noise,using the SPA to remove the nonlinear trend items when the RR interval sequence has different trends or noise effects,and calculate the approximate entropy and fuzzy entropy again.It is found that the two entropy values are stable and using a t Test,P(day,night)<0.05,which shows that the smoothness prior approach effectively improves the stability and recognition of the two entropy analysis methods.In the future research,there is no longer any fear of the influence of trend or noise on the experimental results.It provides a feasible implementation method for nonlinear signal processing and eigenvalue selection.