AU2019101756A4

AU2019101756A4 - Method for analyzing heart rate variability signal based on extremum energy decomposition method

Info

Publication number: AU2019101756A4
Application number: AU2019101756A
Authority: AU
Inventors: Xiaodong Jiang; Hongxing Liu; Xinbao Ning; Binbin Wang; Hua Wang; Peng Zeng; Zuojian ZHOU
Original assignee: Jiangsu Huakang Information Tech Co Ltd
Current assignee: Jiangsu Huakang Information Technology Co Ltd
Priority date: 2019-09-06
Filing date: 2019-11-28
Publication date: 2020-11-12
Anticipated expiration: 2027-11-28
Also published as: CN110464337B; WO2021042591A1; CN110464337A

Abstract

: The present invention discloses a method for analyzing a heart rate variability signal based on an extremum energy decomposition method, including: obtaining an ECG signal in an unknown state at a given time and a given sampling frequency, and performing denoising on the ECG signal, to obtain an RRI signal x(t); using the RRI signal x(t) as an original signal, decomposing the original signal x(t) into n extremum modal function components and one margin, representing components of the original signal in different frequency bands by using the n extremum modal function components obtained by decomposing the original signal x(t), and determining, based on the n extremum modal function components, whether the RRI signal is an abnormal heart rate variability signal. Based on the present invention, the RRI signal is analyzed by using the extremum energy decomposition method, the original signal is decomposed into a plurality of components, that is, an extremum component function, and energy of each component is calculated, to obtain energy distribution of the component.

Description

METHOD FOR ANALYZING HEART RATE VARIABILITY SIGNAL BASED ON EXTREMUM ENERGY DECOMPOSITION METHOD TECHNICAL FIELD

The present invention relates to analysis of electrocardiogram signals, and in particular,

to a method for analyzing a heart rate variability signal based on an extremum energy

decomposition method.

BACKGROUND

Physiological signals are generated through interaction of a plurality of systems in a

living body, and different systems have different action time and action intensity, leading to

complexity of time and space in the physiological signals. Heart rate variability (HRV) refers

to measurement of a quantity of time variations between consecutive cardiac cycles, and to be

precise, should be measurement of a quantity of variations of differences between consecutive

normal P-P intervals. However, since a P wave is not as obvious as an R wave or the top of

the P wave is sometimes wide and blunt, for studies on heart rate variability signals, the heart

rate variability signal is usually replaced by a signal (RRI) whose R-R interval is equal to the

P-P interval. Studies have shown that HRV can be used as a non-invasive detection indicator

of autonomic nervous system activities, and is especially of great significance of judging

prognosis of certain cardiovascular diseases.

In the prior art, a heart rhythm variability signal (HRV signal) is often used as an analysis

object to study a long-term regulation rhythm (<1 Hz) of a heart. A large quantity of studies

have shown that HRV signals of a human body have long-term correlation and nonlinear

dynamics complexity, and age and diseases reduce the dynamics complexity. In the studies of

the heart rate variability (HRV) signals, the RR interval (Interbeat Intervals, RRI) signal is

commonly used, that is, a time interval signal between consecutive RRI signals and the R

wave.

The most common method to study energy changes in HRV signals is using power

spectral density (PSD). Based on the PSD, a power of the HRV signal is converted into a frequency function through Fourier transform, powers of different frequency domain ranges are studied, and generally, an HRV spectrum is divided into several frequency bands, such as a high frequency (HF), a low frequency (LF), and a very low frequency (VLF). An LF/HF ratio has important clinical value. A heart disease can cause changes in the HRV power spectrum. For example, a heart failure and myocardial infarction cause a normalized HF to increase, and the LF and the VLF to decrease. However, the PSD is not a data-driven method, and the frequency domain is relatively roughly segmented, thereby resulting in missing details and that the segmentation is not flexible enough.

Therefore, it is urgent to resolve the foregoing problems.

SUMMARY

An Objective of the present invention: The objective of the present invention is to

provide a method for analyzing a heart rate variability signal based on an extremum energy

decomposition method, which can intuitively reflect a true rule of electrocardiogram energy

distribution and signal fluctuations with less data.

Technical solution: To implement the foregoing objective, the present invention discloses

a method for analyzing a heart rate variability signal based on an extremum energy

decomposition method, including the following steps:

(1) obtaining an ECG signal in an unknown state at a given time and a given sampling

frequency, then performing denoising preprocessing on the ECG signal, and extracting an RRI

signal from the ECG signal, to obtain an RRI signal x(t) in the unknown state;

(2) using the RRI signal x(t) as an original signal, to obtain all local extremum points of

the original signal, and then connecting all maximum points and all minimum points of the

original signal by using spline curves, to respectively form an upper envelope emax and a

lower envelope emi, and obtain an envelope mean signal m(t) = (emax+ emin)/ 2 of the upper

and lower envelopes;

(3) subtracting the envelope mean signal m(t) from the original signal x(t), to obtain h(t)

= x(t) - m(t); and then determining whether h(t) meets a determining condition of an

extremum modal function; and if h(t) does not meet the determining condition of the extremum modal function, using h(t) as the original signal and returning to step (3), until h(t) meets the determining condition of the extremum modal function, where ci(t) = hk(t) is recorded as a first extremum modal function component;

(4) subtracting the first extremum modal function component ci(t) from the original

signal x(t), to obtain a margin ri(t) = x(t) - ci(t); then determining whether hk(t) meets a

stopping criterion; if h(t) does not meet the stopping criterion, using ri(t) as a new original

sequence x(t), and returning to steps (2) and (3); if h(t) meets the stopping criterion and n<8,

returning to step (1) to re-obtain an original signal; if h(t) meets the stopping criterion and

n>8, obtaining second, third, ... , and nth extremum modal function components and a margin rn(t); and then decomposing the original signal x(t) into n extremum modal function

components and one margin, that is n x(t)= ci(t)+rn(t) i=1

(5) performing spectral analysis on an extremum modal function component ci(t),

where i = 1, 2, ... , and n, to obtain a center frequency of each extremum modal function

component;

(6) representing components of the original signal in different frequency bands by using

the n extremum modal function components obtained by decomposing the original signal x(t),

and then calculating energy of each component, where

Ej = f ci(t)|2 dt, where i=1, 2, . . , and n; and

normalizing each energy value, to obtain a normalized energy distribution vector, where

pi= Ej/E, where i=1, 2, ... , and n; and

E = l1 Ej, a first component pi indicates energy in a highest frequency band, and

represents a ratio of energy distribution of a signal within a highest frequency band range, and

a last component pn indicates a ratio of energy distribution of the signal within a lowest

frequency band range; a normalized energy distribution diagram is drawn based on the

normalized energy distribution vector, where a horizontal coordinate indicates a component

level, a vertical coordinate indicates a normalized energy distribution vector value, a curve

indicates a mean, and an error bar indicates a standard deviation;

(7) selecting a second component P2 to a seventh component P7, calculating an energy

difference value EDV, where EDV = (P2 + P3 + p4) - (p5 + P6 + p7), and when the EDV a first standard value, determining that the RRI signal is a normal heart rate variability signal;

when the first standard value < the EDV < a second standard value, determining that the RRI

signal is a suspected abnormal heart rate variability signal; and when the EDV > the second

standard value, determining that the RRI signal is an abnormal heart rate variability signal.

A minimum data volume required by the original signal x(t) is N = 2 "+r, where n is a

quantity of extremum modal function components obtained through decomposition.

Preferably, a specific denoising preprocessing method in step (1) is: filtering the ECG

signal by using a 40 Hz zero-phase FIR low-pass filter to remove high-frequency noise, and

then removing baseline drift by using a median filter.

Then, the determining condition of the extremum modal function in step (3) is as follows:

(a) in an entire data sequence, a quantity of extremum points and a quantity of zero crossing

points are equal or differ from each other by one; (b) at any moment, the upper and lower

envelopes are symmetric for a time axis.

Further, a formula of hk(t) in step (4) meeting the stopping criterion is as follows:

SD = t=o[hk- (t)hkt E, where , indicates a screening threshold, and ranges

between 0.2 and 0.3. Preferably, in step (7), the first standard value is -0.15 and the second standard value is

0.08. Beneficial effects: Compared with the prior art, the present invention has the following

significant advantages:

Based on the present invention, the RRI signal is analyzed by using the extremum energy

decomposition (EED) method, the original signal is decomposed into a plurality of

components, that is, an extremum component function, and energy of each component is

calculated, to obtain energy distribution of the component. Based on the present invention,

based on a fluctuation rule of the RRI signal, the signal can be decomposed into a

high-frequency signal to a low-frequency signal at different time levels, and a frequency band

is relatively meticulously segmented. Data lengths obtained by decomposing an extremum at all levels are the same. Therefore, the data length is not reduced, so that the EED can be used for short-term data analysis, that is, an accurate result can be obtained through analysis by using a very small data volume. For analysis of energy of components at different levels, the

EED is not easily affected by noise.

BRIEF DESCRIPTION OF DRAWINGS

Fig. 1 is a schematic diagram of an original signal according to the present invention;

Fig. 2 is a schematic diagram of obtaining an envelope of an original signal according to

the present invention;

Fig. 3 is a schematic diagram of subtracting an envelope mean signal from an original

signal according to the present invention;

Fig. 4 is a schematic diagram of obtaining a first extremum modal function component

according to the present invention;

Fig. 5 is a schematic flowchart of an extremum energy decomposition method according

to the present invention;

Fig. 6 is a schematic exploded view of EED for an RRI signal according to Embodiment

1 of the present invention; and

Fig. 7 is a schematic diagram of distribution of normalized energy in an RRI signal

according to Embodiment 1 of the present invention.

DETAILED DESCRIPTION

Technical solutions of the present invention are further described below with reference to

accompanying drawings.

As shown in Fig. 1, Fig. 2, Fig. 3, Fig. 4, and Fig. 5, a method for analyzing a heart rate

variability signal based on an extremum energy decomposition method in the present

invention includes the following steps:

(1) Obtain an ECG signal in an unknown state at a given time and a given sampling

frequency, then perform denoising preprocessing on the ECG signal, and extract an RRI

signal from the ECG signal, to obtain an RRI signal x(t) in the unknown state, where a specific denoising preprocessing method is: because ECG energy is mainly concentrated within 0 Hz to 40 Hz, filtering the ECG signal by using a 40 Hz zero-phase FIR low-pass filter to remove high-frequency noise, and then removing baseline drift by using a median filter.

(2) Use the RRI signal x(t) as an original signal, to obtain all local extremum points of

the original signal, where a minimum data volume required by the original signal x(t) is N =

2"1, and n is a quantity of extremum modal function components obtained through

decomposition; and then connecting all maximum points and all minimum points of the

lower envelope emin, and obtain an envelope mean signal m(t) = (emax+ emi)/ 2 of the upper

and lower envelopes.

(3) Subtract the envelope mean signal m(t) from the original signal x(t), to obtain h(t)=

x(t) - m(t); and then determine whether h(t) meets a determining condition of an extremum

modal function; and if h(t) does not meet the determining condition of the extremum modal

function, use h(t) as the original signal and return to step (2), until h(t) meets the determining

condition of the extremum modal function, where ci(t) = hk(t) is recorded as a first extremum

modal function component; the determining condition of the extremum modal function is as

follows: (a) in an entire data sequence, a quantity of extremum points and a quantity of zero

crossing points are equal or differ from each other by one; (b) at any moment, the upper and

lower envelopes are symmetric for a time axis.

(4) Subtract the first extremum modal function component ci(t) from the original signal

x(t), to obtain a margin ri(t) = x(t) - ci(t); then determine whether hk(t) meets a stopping

criterion; if hk(t) does not meet the stopping criterion, use ri(t) as a new original sequence x(t),

and returning to steps (2) and (3); if h(t) meets the stopping criterion and n<8, return to step

(1) to re-obtain an original signal; if hk(t) meets the stopping criterion and n>8, obtain second,

third, ... , and nth extremum modal function components and a margin r(t); and then

decompose the original signal x(t) into n extremum modal function components and one

margin, that is n x(t) = (t)+ rn(t) i=1 a formula of hk(t) meeting the stopping criterion is as follows:

SD = t=o[hk- (t)-hk E, where , indicates a screening threshold, and ranges

between 0.2 and 0.3; extremum modal decomposition meeting the stopping criterion meets the following two conditions: (a) a finally obtained extremum modal function component ce(t) or a margin rn(t) is less than a preset threshold; (b) a residual signal rn(t) becomes a monotonic signal, and an extremum modal function signal can no longer be extracted from the residual signal.

(5) Perform spectral analysis on an extremum modal function component ci(t), where i

= 1, 2, ... , and n, to obtain a center frequency of each extremum modal function component.

(6) Represent components of the original signal in different frequency bands by using the

n extremum modal function components obtained by decomposing the original signal x(t),

and then calculating energy of each component, where

Ej = f ci(t)I2 dt, where i=1, 2, . . , and n; and

normalize each energy value, to obtain a normalized energy distribution vector, where

p i =Ej/E, where i=1, 2, ... , and n; and

E =Z7 MEi, a first component pi indicates energy in a highest frequency band, and

indicates a mean, and an error bar indicates a standard deviation.

(7) Select a second component P2 to a seventh component P7, calculate an energy

difference value EDV, where EDV = (p2 + P3 + p4) - (p5 + P6 + p7), and when the EDV a

first standard value, determine that the RRI signal is a normal heart rate variability signal;

when the first standard value < the EDV < a second standard value, determine that the RRI

standard value, determine that the RRI signal is an abnormal heart rate variability signal, where the first standard value is -0.15 and the second standard value is 0.08. The extremum energy decomposition (EED) method used in the present invention is a method based on a concept of the extremum modal function. The extremum modal function is a type of signal that satisfies the following two conditions at the same time and that is at a single frequency. The two conditions are as follows: (a) In an entire data sequence, a quantity of extremum points (including maximums and minimums) and a quantity of zero crossing points must be equal or differ from each other by one at most. (b) At any moment, a mean of an upper envelope formed by local maximum points and a lower envelope formed by local minimum points is zero, that is, the local upper and lower envelopes are locally symmetric for a time axis. For the above two conditions, the condition (a) is similar to a requirement of a Gaussian normal stationary process for a traditional narrowband, and the condition (b) ensures that an instantaneous frequency calculated by using the extremum modal function has a physical meaning. A standard of decomposition termination of the extremum modal function in the present invention needs to be appropriately selected. If the condition is too strict, last few extremum modal function components are caused to lose their meanings. If the condition is too loose, a useful component is caused to be lost. In practical applications, a quantity of levels of extremum modal function components to be obtained through decomposition may be set according to needs. Calculation is terminated when the quantity of levels obtained through decomposition is met. Embodiment 1 The EED analysis method is used to analyze energy distribution of ECGs of healthy persons and patients with CHF at different levels. A method for analyzing a heart rate variability signal of a healthy person based on an extremum energy decomposition method includes the following steps: (1) Obtain ECG signals of healthy people from an R-R interval database NSR2DB of PhysioNet, where data includes 54 healthy persons (at an age of 28 to 76 with a mean of 61), then perform denoising preprocessing on the ECG signal, and extract an RRI signal from the

ECG signal, to obtain an RRI signal x(t), where a specific denoising preprocessing method is:

because ECG energy is mainly concentrated within 0 Hz to 40 Hz, filtering the ECG signal by

using a 40 Hz zero-phase FIR low-pass filter to remove high-frequency noise, and then

removing baseline drift by using a median filter.

2"+' = 29, n is a quantity of extremum modal function components obtained through

decomposition, and n = 8; and then connect all maximum points and all minimum points of

the original signal by using spline curves, to respectively form an upper envelope emax and a

and lower envelopes.

lower envelopes are symmetric for a time axis.

(5) Subtract the first extremum modal function component ci(t) from the original signal

x(t), to obtain a margin ri(t)= x(t) - ci(t); use ri(t) as a new original sequence x(t), and return

to steps (2) and (3), to obtain second, third, ... , and eighth extremum modal function

components and a margin r8 (t); and then decompose the original signal x(t) into eight

extremum modal function components and one margin, that is 8

x(t) ci (t) + r8 (t)

An extremum modal decomposition diagram of a healthy person shown in Fig. 6 is obtained. It can be seen that a component 1 has a highest frequency, and the signal fluctuates on a shortest time scale. As a component number increases, the frequency gradually decreases. (6) Perform spectral analysis on an extremum modal function component c1 (t), where i = 1, 2, ... , and 8, to obtain a center frequency of each extremum modal function component, and a frequency domain analysis result diagram. (7) Represent components of the original signal in different frequency bands by using the 8 extremum modal function components obtained by decomposing the original signal x(t), and then calculate energy of each component, where Ei = fIci(t)|2 dt, where i=1, 2, ... , and 8; and normalize each energy value, to obtain a normalized energy distribution vector, where pi= Ej/E, where i=1, 2, ... , and 8; and

E =nl 1 Ej, a first component pi indicates energy in a highest frequency band, and

represents a ratio of energy distribution of a signal within a highest frequency band range, and a last component pn indicates a ratio of energy distribution of a signal within a lowest frequency band range. (8) Select a second component P2 to a seventh component py, calculate an energy difference value EDV, where EDV = (p2+p3 +p4) - (p5 +p6 +p7)= -0.2092 ±0.2940.

A method for analyzing a heart rate variability signal of a patient with CHF based on an extremum energy decomposition method includes the following steps: (1) Obtain ECG signals from an R-R interval database CHFDB of PhysioNet, where the CHFDB database includes 29 patients with the congestive heart failure (CHF) (at an age of 34 to 79 with a mean of 55), then perform denoising preprocessing on the ECG signal, and extract an RRI signal from the ECG signal, to obtain an RRI signal x(t), where a specific denoising preprocessing method is: because ECG energy is mainly concentrated within 0 Hz to 40 Hz, filtering the ECG signal by using a 40 Hz zero-phase FIR low-pass filter to remove high-frequency noise, and then removing baseline drift by using a median filter. (2) Use the RRI signal x(t) as an original signal, to obtain all local extremum points of the original signal, where a minimum data volume required by the original signal x(t) is N=

2 "+' = 29, n is a quantity of extremum modal function components obtained through

and lower envelopes.

function, use h(t) as the original signal and return to step (2), until hk(t) meets the determining

lower envelopes are symmetric for a time axis.

extremum modal function components and one margin, that is 8

x(t) ci (t) + r8 (t) i=1

= 1, 2, ... , and 8, to obtain a center frequency of each extremum modal function component.

8 extremum modal function components obtained by decomposing the original signal x(t),

and then calculate energy of each component, where

EI = fIc (t)12 dt, where i=1, 2, ..., and 8; and normalize each energy value, to obtain a normalized energy distribution vector, where

pi = Ej/E, where i=1, 2, ... , and 8; and

normalized energy distribution vectors of healthy people and patients with CHF, where a

horizontal coordinate indicates a component level, a vertical coordinate indicates a

normalized energy distribution vector value, a curve indicates a mean, and an error bar

indicates a standard deviation.

(7) Select a second component P2 to a seventh component P7, calculate an energy

difference value EDV, where EDV = (p2+ p3 +p4) - (p5 +P6 + p7) = 0.2642 ±0.4070. In the present invention, an average center frequency of components of HRV signals of

the healthy people and the patients with CHF at different levels is further analyzed, to obtain

results in Table 1.

Table 1 Average center frequency at different component levels of HRV signals of the

healthy people and the patients with CHF

Level 1 2 3 4 5 6 7 8

Frequency Healthy 0.4677 0.1664 0.0878 0.0437 0.0216 0.0087 0.0034 0.0010 /Hz CHF 0.4926 0.2014 0.1067 0.0550 0.0275 0.0136 0.0063 0.0019

It may be learned that, as component levels increase, the center frequency of the HRV

signal gradually decreases.

In a traditional power spectral density (PSD), a typical manner of segmenting a

frequency domain is: an HF (0.15 Hz to 0.4 Hz), an LF (0.04 Hz to 0.15 Hz), and a VLF

(0.0033 Hz to 0.04 Hz). In the EED method in the present invention, for a component level 1,

a frequency is higher than the HF; a level 2 falls within the frequency range of the HF; levels

3 and 4 fall within the range of the LF; levels 5 to 7 fall within the frequency range of the

VLF; and a level 8 is lower than the VLF. In addition, it can be seen that a frequency of

patients with CHF at a same level is slightly higher than that of healthy people, reflecting an

influence of a heart disease on an HRV fluctuation rhythm. At a same level, patients with CHF

have a faster HRV fluctuation.

In the present invention, HRV signals of two groups are decomposed to obtain extremum

modal function components Ci(t), a normalized energy distribution vector is obtained through

calculation, and an EED curve graph is drawn, as shown in Fig. 7. Fig. 7 is a schematic

diagram of EED analysis results of RR interval signals of healthy people and patients with

CHF. A data length is 10000 points, a curve represents a mean, and an error bar represents a

standard deviation. A symbol * above the curve indicates that p < 0.01 of a T test of energy of

the two groups of people. During selection of levels, the level 1 and a level higher than 7,

including a margin, are removed. The level 1 is easily affected by noise, thereby leading to a

relatively large energy fluctuation, causing an excessively large standard deviation of the

result. A frequency of the level 1 can be as high as several kHz. Therefore, the level 1 has no

clear physiological significance. The level higher than 7 reflects a long-term signal rhythm,

and is very easily affected by an external environment. In addition, a frequency of the level

higher than 7 is very low. Therefore, the level higher than 7 has unknown physiological

significance. In Fig. 7, at the low component levels (the levels 2 and 3), a normalized energy

value of the patients with CHF is significantly higher than that of the healthy people. At the

high component level (the level > 5), an opposite change occurs, and the healthy people have

higher energy than that of the patients with CHF. The energy of the healthy people is

relatively stable at the levels 2 to 5. When the level is higher than 5, energy rises slowly, while

energy of the patients with CHF declines rapidly at the levels 2 to 4, and then stabilizes. In the

4 levels (2, 3, 6, and 7), there is a significant difference in energy between the healthy people

and the patients with CHF (p < 0.01). For comparison, in the present invention, an EED

analysis result of surrogate data (healthy surrogate, CHF surrogate) is added. As shown in Fig.

7, the surrogate data is generated by randomizing original data. Energy distribution of the

surrogate data monotonically decreases as a scale increases. Compared with the patients with

CHF, energy is higher on a small time scale, and is lower on a long time scale.

To evaluate an energy distribution difference between low-level decomposition and

high-level decomposition of the EED curve, an energy difference value EDV of people groups

is calculated. A high EDV value indicates a higher component low-level energy distribution

and a lower component high-level energy distribution of the RRI signal. The EDV values of the healthy people, the patients with CHF, and the surrogate data thereof are calculated, as shown in Table 2.

Table 2 EDV values of the healthy people and the patients with CHF

Groups healthy CHF healthy surrogate CHF surrogate

EDV (mean -0.2092 0.2642 0.7306 0.0226** 0.7230±

std) 0.2940 0.4070* 0.0328**

* represents a result p < 0.0001 of a T test of the healthy people and the patients with

CHF.

** represents a result p < 0.0001 of a T test of the surrogate data and the original data of

the surrogate data.

It can be seen from Table 2 that the EDV value of the healthy people and the EDV value

of the patients with CHF are significantly different, and the EDV value of the patients with

CHF is much higher than that of the healthy people. The EDV value of the healthy people is

less than 0, indicating that there is higher energy in a high component level, which indicates

that the high-level component has higher adjustment intensity. The EDV values of the two

people groups are significantly different from the EDV values of the surrogate data thereof,

and an EDV of a human HRV is significantly smaller than a random sequence.

Heart rate variability (HRV) refers to measurement of a quantity of time variations

between consecutive cardiac cycles, and to be precise, should be measurement of a quantity of

variations of differences between consecutive normal P-P intervals. However, since a P wave

is not as obvious as an R wave or the top of the P wave is sometimes wide and blunt, for

studies on heart rate variability signals, the heart rate variability signal is usually replaced by a

signal (RRI) whose R-R interval is equal to the P-P interval. Studies have shown that HRV

can be used as a non-invasive detection indicator of autonomic nervous system activities, and

is especially of great significance of judging prognosis of certain cardiovascular diseases.

In clinical and practical applications, it is proposed in the present invention that 512 (n=8)

consecutive RR intervals of short-term heartbeat signals can be used for the above-mentioned

EED analysis, which is effective, and there is a certain quantity of data margins. The above studies show that the EED decomposition reaches 7 component levels (n=7) and there are good results. Therefore, a minimum data volume required can be N = 2 "+' = 27+1 = 256.

CLAIMES:

1. A method for analyzing a heart rate variability signal based on an extremum energy

decomposition method, comprising the following steps:

signal from the ECG signal, to obtain an RRI signal x(t) in the unknown state;

and lower envelopes;

extremum modal function; and if h(t) does not meet the determining condition of the

extremum modal function, using h(t) as the original signal and returning to step (3), until h(t)

meets the determining condition of the extremum modal function, wherein ci(t) = hk(t) is recorded as a first extremum modal function component;

n>8, obtaining second, third, ... , and nth extremum modal function components and a margin

rn(t); and then decomposing the original signal x(t) into n extremum modal function

components and one margin, that is n

x(t)= ci(t)+rn(t)

(5) performing spectral analysis on an extremum modal function component ci(t),

Claims

wherein i = 1, 2, ... , and n, to obtain a center frequency of each extremum modal function

component;

(6) representing components of the original signal in different frequency bands by using

the n extremum modal function components obtained by decomposing the original signal x(t),

and then calculating energy of each component, wherein

Ej = fIct (t)|2 dt, wherein i=1, 2, . . , and n; and

normalizing each energy value, to obtain a normalized energy distribution vector,

wherein

pi= Ej/E, wherein i=1, 2, ... , and n; and

E =nl 1 Ej, a first component pi indicates energy in a highest frequency band, and

represents a ratio of energy distribution of a signal within a highest frequency band range, and

a last component pn indicates a ratio of energy distribution of the signal within a lowest

frequency band range; a normalized energy distribution diagram is drawn based on the

normalized energy distribution vector, wherein a horizontal coordinate indicates a component

level, a vertical coordinate indicates a normalized energy distribution vector value, a curve

indicates a mean, and an error bar indicates a standard deviation; and

(7) selecting a second component P2 to a seventh component py, calculating an energy

difference value EDV, wherein EDV = (P2 + P3 + p4) - (p5 + P6 + p7), and when the EDV ! a first standard value, determining that the RRI signal is a normal heart rate variability signal;

when the first standard value < the EDV < a second standard value, determining that the RRI

signal is a suspected abnormal heart rate variability signal; and when the EDV > the second

standard value, determining that the RRI signal is an abnormal heart rate variability signal.
2. The method for analyzing the heart rate variability signal based on the extremum

energy decomposition method according to claim 1, wherein a minimum data volume

required by the original signal x(t) is N = 2 "-, wherein n is a quantity of extremum modal

function components obtained through decomposition.
3. The method for analyzing the heart rate variability signal based on the extremum

energy decomposition method according to claim 1, wherein a specific denoising

preprocessing method in step (1) is: filtering the ECG signal by using a 40 Hz zero-phase FIR low-pass filter to remove high-frequency noise, and then removing baseline drift by using a median filter.
4. The method for analyzing the heart rate variability signal based on the extremum

energy decomposition method according to claim 1, wherein the determining condition of the

extremum modal function in step (3) is as follows: (a) in an entire data sequence, a quantity of

extremum points and a quantity of zero crossing points are equal or differ from each other by

one; (b) at any moment, the upper and lower envelopes are symmetric for a time axis.
5. The method for analyzing the heart rate variability signal based on the extremum

energy decomposition method according to claim 1, wherein a formula of hk(t) in step (4)

meeting the stopping criterion is as follows:

SD = t-O[hk1(t)hk E, wherein , indicates a screening threshold, and ranges t= 0 hk- 1 (t M

between 0.2 and 0.3.
6. The method for analyzing the heart rate variability signal based on the extremum

energy decomposition method according to claim 1, wherein in step (7), the first standard

value is -0.15 and the second standard value is 0.08.

Amplitude (V) Amplitude (V) Amplitude (V) Amplitude p tude (V)

Loower envelopee

1/ 3 Fig. 4 Fig. 3 Fig. 2 Fig. 1

Mean n curve Uppper envelope

Staart

Input an origginal sequence x(t)

Obtaain upper and low wer envelopes emax and 2019101756

emin of o x(t)

Enveelope mean m((t) = (emax + emin)/2

Whether h(tt) meets an extremumm modal function coondition?

Whetheer hk(t) meets a stopping s criterrion?

where

Ennd

Fig. 5

2/ 3

3/ 3 Fig.
7 Fig. 6