JPH11128185A - Heart rate variability analysis method and apparatus - Google Patents

Abstract

(57) [Summary] [PROBLEMS] To obtain the power of frequency components included in heart rate variability, which is an unsteady signal, with high time resolution. A heartbeat interval calculation unit obtains heartbeat intervals from biological signals measured by a biological signal measurement unit, thereby obtaining time-series heartbeat interval data that are unequally spaced in time. The equal-interval data generation unit 103 generates temporally-equivalent heartbeat-interval time-series data by interpolating and resampling the temporally incorrect heartbeat-interval time-series data. The discrete binary wavelet transform unit 104
The heartbeat interval time-series data at equal intervals in time is divided into wavelet coefficients in a plurality of frequency bands by discrete binary wavelet transform. Each power calculation unit 105a to 105
f calculates, for each section of the minimum time resolution corresponding to each frequency band, 1/2 of the square of the maximum value of the absolute value of the wavelet coefficient as the power of the section. Thus, the time variation of the power of the frequency component included in the heart rate variation is obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a method and apparatus for analyzing heart rate variability, and more particularly to a method and apparatus for analyzing heart rate variability that can be used for testing autonomic nervous function and evaluating stress.

[0002]

2. Description of the Related Art As is well known, the heart is antagonistically controlled by both autonomic nerves, sympathetic nerves and parasympathetic nerves.
Sympathetic nerves act in a stimulatory manner to make the body more active, while parasympathetic nerves act in a depressing way to make the body more relaxed. In addition, these activities of the autonomic nervous system are closely related to mental activities.

[0003] When frequency analysis of heart rate variability, two main frequency components are observed. One is low frequency (LF: 0.
04-0.15 Hz), which mainly reflects the activity of the sympathetic and parasympathetic nerves. The other is a high frequency (HF: 0.15 to 0.45 Hz) component, which mainly reflects the activity state of the parasympathetic nerve. LF / H
F has been proposed as an indicator of sympathetic activity. Therefore, the power of these frequency components can be determined by examining autonomic nervous function (see “Circulatory Disease and Autonomic Nervous Function”, Medical School) and evaluating mental stress (Japanese Patent Laid-Open No. 54940/1990).
No. 1).

FIG. 3 is a view for explaining a method for obtaining heartbeat interval time series data using an electrocardiogram R wave. FIG.
FIG. 4 is a diagram for explaining a method of obtaining the power of the frequency component by the conventional heart rate variability analysis method. Hereinafter, a conventional method for obtaining the power of each frequency component of heart rate variability will be described with reference to FIGS.

First, as shown in FIG. 3A, a time at which an R wave is generated is detected, and a heartbeat interval (hereinafter, abbreviated as RRI) is measured. Next, as shown in FIG.
The RI data is plotted at the temporal position of the rear R wave.
Next, as shown in FIG. 3 (3), each plotted RR
By interpolating between I and resampling at equal intervals, time-series heartbeat interval time-series data is generated at equal intervals in time.

FIG. 14A shows time series data of heartbeat intervals generated at the same time intervals as described above, and this data is cut out in a fixed section as shown in the figure to obtain an FFT (Fast Fourier Transform). ) Or the power of the LF component and the HF component is obtained by frequency analysis such as AR (autoregressive model). FIG. 14 (2) shows the power spectrum density obtained by the FFT. In this case, the LF component,
The power of the HF component is obtained by integrating the power spectrum density in the corresponding frequency domain.

[0007]

SUMMARY OF THE INVENTION However, FFT
And the conventional analysis method using AR presupposes that the time-series data to be processed has a stationarity (frequency components do not change abruptly), and the power of each frequency component is also cut out. It only estimates the average of For this reason, it is not possible to check the fluctuation of the power of each frequency component in the cut section. There have been attempts to shorten the section to be cut out or analyze the section to be cut out little by little, but at least about 1 or 2 minutes of data is required, and it is necessary to capture rapid autonomic nervous reaction with high time resolution. Has limitations.

Therefore, as a method for capturing temporal changes in the autonomic nerves, it is conceivable to use a wavelet transform, which is a non-stationary state analysis method, instead of the conventional steady state analysis method. The discrete wavelet transform is a time-frequency analysis method as shown in, for example, "Wavelet Beginner's Guide" by Susumu Sakakibara published by Tokyo Denki University Press on May 20, 1995. It can be divided into wavelet coefficients of a plurality of frequency bands. In the discrete wavelet transform, each frequency band has a different time resolution. That is, in the discrete wavelet transform, the higher the frequency band, the higher the minimum time resolution. This discrete wavelet transform is used for signal noise removal and image compression.

In the following, a method of obtaining the power of each frequency component of heart rate variability using the discrete wavelet transform will be considered. The discrete wavelet transform is realized using a filter bank including a discrete time high-pass filter, a discrete time low-pass filter, and a downsampler as shown in FIG. 15, and converts the digital signal S ₀ (n) into a plurality of frequency bands. It can be divided into wavelet coefficients. FIG. 15 shows six frequency bands D ₁ , D ₂ , D ₃ , D
₄ shows a configuration example of a filter bank in the case of dividing the D _5, D _6. The wavelet coefficients of each frequency band obtained in this manner are obtained with the minimum time resolution according to each frequency. Therefore, the heartbeat interval time series data is input to this filter bank, divided into wavelet coefficients of a plurality of frequency bands by discrete wavelet transform, and the power of each divided frequency band is calculated by squaring the wavelet coefficients. Can be.

FIG. 16 shows a result obtained by analyzing a digital signal obtained by sampling a sin wave having a frequency of 0.2 Hz at 5 Hz using the discrete wavelet transform method. When the input digital signal is 5 Hz, each frequency band of the wavelet coefficient divided by the discrete wavelet transform is as follows. _{D 1: 1.25~2.5Hz D 2: 0.625~1.25Hz} D 3: 0.3125~0.625Hz D 4: 0.15625~0.3125Hz D 5: 0.07812~0.15625Hz D _6: 0.03906~0.07812Hz

Referring to FIG. 16, but a large power to the frequency band of the D ₄ corresponding to the frequency 0.2Hz input signal appearing, the value of the power it can be seen that the variation with time. The input signal has a constant amplitude and frequency s.
Since it is an in-wave, a constant power should appear in the time direction. However, the wavelet coefficients in each frequency band of the discrete wavelet transform include:
Since the amplitude value of the input signal at the time interval corresponding to the time resolution of each frequency band is reflected, if the time interval at each frequency band and the phase of the input signal are shifted, the power value will be It fluctuates with time, and the power of the input signal cannot be obtained accurately.

FIG. 17 shows a result obtained by analyzing a digital signal obtained by sampling a sine wave whose frequency increases in proportion to time at 5 Hz using the discrete wavelet transform method. Along with the frequency of the input signal becomes higher continuously changing, it must be smoothly changed to D ₁ of the higher frequency band from the D ₆ values even lower frequency band power of each frequency band However, as shown, the analysis result does not show a smooth variation. As described above, when the discrete wavelet transform is used, the power of the frequency component included in the input signal cannot be accurately obtained.

[0013] Therefore, an object of the present invention is to provide a method and apparatus for analyzing heart rate variability in which temporal changes in autonomic nerves can be captured with high temporal resolution and the power of each frequency component can be accurately obtained. It is to provide.

[0014]

Means for Solving the Problems and Effects of the Invention A first invention is a method for electrically detecting and analyzing heartbeat fluctuations, comprising a first step of measuring a biological signal related to the heartbeat. A second step of detecting a heartbeat interval based on the biological signal measured in the first step, and temporally unequally spaced heartbeat interval time-series data obtained from the heartbeat interval detected in the second step A third step of generating time-series heartbeat interval time-series data at the same time interval by interpolation, and converting the time-series heartbeat interval time-series data at the same time interval generated in the third step to discrete binary A fourth step of dividing into wavelet coefficients for each of a plurality of frequency bands by a wavelet transform, and in each of the plurality of frequency bands, for each time section corresponding to the minimum time resolution in the frequency band, A fifth step of calculating the time variation of the power of the frequency component included in the heart rate variation by calculating the power of the frequency component in each time interval with reference to a plurality of wavelet coefficients obtained in the time interval. .

As described above, according to the first aspect of the present invention, a discrete binary wavelet transform, which is a technique for analyzing a non-stationary signal, is applied, and in each frequency band, a plurality of wavelet coefficients existing in each time section are provided. , The power of each time interval is calculated, so that even if the interval between the time intervals and the phase of the input signal are shifted in each frequency band, the power of the input signal can be obtained accurately, and rapid autonomous It becomes possible to capture neural responses with high temporal resolution.

According to a second aspect, in the first aspect, the fifth aspect is provided.
Is characterized in that half of the square of the maximum value of the absolute values of the plurality of wavelet coefficients obtained in each time interval is calculated as the power in each time interval.

As described above, according to the second aspect, 1/2 of the square of the maximum value of the absolute value of the plurality of wavelet coefficients obtained in each time interval is calculated as the power in each time interval. In particular, the input signal is si
This is effective when the waveform is close to an n-wave waveform.

According to a third aspect, in the first aspect, the fifth aspect is provided.
Is characterized in that an average value of squares of a plurality of wavelet coefficients obtained in each time section is calculated as power in each time section.

As described above, according to the third aspect, the average value of the square of a plurality of wavelet coefficients obtained in each time interval is calculated as the power in each time interval. Is different from the waveform of the sine wave.

According to a fourth aspect of the present invention, there is provided an apparatus for electrically detecting and analyzing heartbeat fluctuations, comprising a biological signal measuring means for measuring a biological signal related to the heartbeat, and data obtained by the biological signal measuring means. Heartbeat interval calculation means for obtaining a heartbeat interval from
Equal interval data generating means for generating temporally equal heart rate interval time series data by interpolating temporally non-equidistant heart rate interval time series data obtained by the heart rate interval calculating means; Discrete binary wavelet transforming means for performing discrete binary wavelet transform on the time-series heartbeat interval time series data obtained by the generating means and dividing the data into wavelet coefficients in a plurality of frequency bands; By calculating the power of the frequency component in each time section with reference to a plurality of wavelet coefficients obtained in each time section for each time section corresponding to the minimum time resolution in the frequency band,
Power calculation means for calculating a time variation of the power of the frequency component included in the heartbeat variation.

As described above, according to the fourth aspect, the discrete binary wavelet transform, which is a technique for analyzing a non-stationary signal, is applied, and in each frequency band, a plurality of wavelet coefficients existing in each time section are provided. , The power of each time interval is calculated, so that even if the interval between the time intervals and the phase of the input signal are shifted in each frequency band, the power of the input signal can be obtained accurately, and rapid autonomous It becomes possible to capture neural responses with high temporal resolution.

In a fifth aspect based on the fourth aspect, the power calculating means calculates half of the square of the maximum value of the absolute values of the plurality of wavelet coefficients obtained in each time interval as power in each time interval. It is characterized by doing.

As described above, according to the fifth aspect, 1/2 of the square of the maximum value of the absolute value of the plurality of wavelet coefficients obtained in each time interval is calculated as the power in each time interval. In particular, the input signal is si
This is effective when the waveform is close to an n-wave waveform.

In a sixth aspect based on the fourth aspect, the power calculation means calculates an average value of squares of a plurality of wavelet coefficients obtained in each time section as power in each time section.

As described above, according to the sixth aspect, the average value of the square of a plurality of wavelet coefficients obtained in each time interval is calculated as the power in each time interval. Is different from the waveform of the sine wave.

[0026]

BEST MODE FOR CARRYING OUT THE INVENTION

(First Embodiment) FIG. 1 is a block diagram showing a configuration of a heart rate variability analyzer according to a first embodiment of the present invention.
The heart rate variability analysis apparatus of the present embodiment interpolates a biological signal measurement unit 101 that performs an electrocardiogram measurement, a heartbeat interval calculation unit 102 that obtains a heartbeat interval from an electrocardiogram waveform, and heartbeat interval time-series data that is unequal in time. An equally-spaced data generation unit 103 for generating time-series heartbeat-interval time-series data, and dividing the time-series heartbeat-interval time-series data into wavelet coefficients in a plurality of frequency bands by discrete binary wavelet transform Binary wavelet transform unit 104
And power calculation units 105a to 105f for calculating the power of each frequency band from the wavelet coefficients.

FIG. 2 is a flowchart showing a processing procedure of the heart rate variability analyzer shown in FIG. Hereinafter, the operation of the heart rate variability analyzer shown in FIG. 1 will be described with reference to FIG.

First, the electrocardiogram data measured by the biological signal measuring unit 101 is read into the heartbeat interval calculating unit 102 (step S1). Next, the heartbeat interval calculation unit 102 calculates a heartbeat interval (steps S2 and S3). That is, the heartbeat interval calculation unit 102 detects the time at which the R wave occurs as shown in FIG. 3A (step S2), and
As shown in (2), a heartbeat interval (hereinafter abbreviated as RRI) is calculated (step S3), and each RRI data is plotted at a time position where the rear R wave is generated, so that the time is unequal. Obtain the heartbeat interval time series data of the interval. next,
The equal-interval data generation unit 103 generates time-series data at equal intervals in time from heart-interval time-series data at irregular intervals in time (step S4). That is, as shown in FIG. 3 (3), the equal interval data generation unit 103 interpolates the time-series heartbeat interval time data at unequal intervals using linear interpolation or the like, and performs resampling at equal intervals. Generate heartbeat interval time series data at equal intervals in time. In this embodiment, as an example, a case where resampling is performed at 5 Hz will be described. The time-series heartbeat interval data obtained at equal intervals in time obtained in this way is supplied to the discrete binary wavelet transform unit 104, and is divided into wavelet coefficients of a plurality of frequency bands (step S5). Discrete binary wavelet transform is first performed by Stephane
Researched by Mallat for image compression and feature extraction, the following references are available. (1) S. Mallat "Zero-crossin
gs of Wavelet Transform "
IEEE TRANSACTIONS ONINFOR
MATHION THEORY, VOL. 37, N
O. 4, JULY 1991 (2) S.M. Mallat and S.M. Zhongng
"Characterization of Sign
als from Multiscale Edge
s "IEEE TRANSACTIONS ON P
ATTERN ANALYSIS AND MACHI
NE INTELIGENCE, VOL. 14, N
O. 7, JULY 1992 (3) Makoto Nakashizu et al., "ECG by Multi-Resolution Peak Analysis"
Data Compression ”IEICE Transactions D-II Vo
l. J79-D-II NO. 8 1412-1421
Page August 1996

Here, the processing in step S5 for performing the discrete binary wavelet transform will be described with reference to FIG. FIG. 4 is a diagram illustrating a configuration example of a filter bank of a discrete binary wavelet transform when a digital input signal is divided into six frequency bands. Discrete binary wavelet transform is performed by calculating the inner product of a function system Ψ _i (x) (see the following equation (1)) and a signal f (x) obtained by a power-of-two scale transform of the basic wavelet Ψ (x). (See the following equation (2)). Ψ _i (x) = 2 ⁻ⁱ Ψ (2 ^−ix ) (1) _Wi (n) = <（ _i (x−n), f (x)> (2)

Now, the Fourier transform of Ψ _i (x) ψ _i (ω)
For a scaling function Φ that satisfies the following equation (3):
(X) is defined. Here, φ (ω) is a Fourier transform of φ (x).

(Equation 1)

Further, a smoothed signal obtained by the convolution operation between the scale converted scaling function [Phi (x) and the signal f (x) in the second i squared and S _i (t), S _i
A signal obtained by discretizing (t) is defined as S _i (n). The discrete-time signal S ₀ (n) is obtained by dividing a discrete binary wavelet transform W _i (x) (where 1 ≦ i ≦ m) into m bands,
It can be reconstructed with the discrete time signal S _m (n). Further, the discrete binary wavelet transform W _i (n) is based on the discrete time signal S ₀ (n), and the discrete time low-pass filter G (ω) satisfying the following equations (4), (5), and (6). ) And the discrete-time high-pass filter H (ω). φ (2ω) = H (ω) · φ (ω) (4) ψ (2ω) = G (ω) · φ (ω) (5) | G (ω) | ² + | H (ω) | ² = 1 ... (6)

The filter bank of FIG. 4 passes a digital (discrete) input signal S ₀ (n) 200 through a discrete-time high-pass filter H (ω) 211 and a discrete-time low-pass filter G (ω) 221. , The wavelet transform coefficient W ₁ (n) 231 and the smoothed signal S ₁ (n) 24
I have one. Similarly, by passing the smoothed signal S ₁ (n) 241 through the discrete-time high-pass filter 212 and the discrete-time low-pass filter 222, the wavelet transform coefficient W ₂ (n) 232 and the smoothed signal S ₂ (n ) 2
42. Further, the same processing is performed for i = 3, 4,
By executing up to 5,6, thereby obtaining a wavelet transform coefficient _{W i (n) 233~W i (} n) 236, and a smoothed signal _{S i (n) 243~S i (} n) 246.

As described above, in step S5, the wavelet coefficients W ₁ (n) to W ₆ divided into six frequency bands by the discrete binary wavelet transform.
(N) is obtained. Digital (discrete) input signal S ₀
(N) When the sampling frequency of 200 is 5 Hz, the frequency band of each wavelet coefficient is as follows. W ₁ (n): 1.25 to 2.5 Hz W ₂ (n): 0.625 to 1.25 Hz W ₃ (n): 0.3125 to 0.625 Hz W ₄ (n): 0.15625 to 0 _{.3125Hz W 5 (n): 0.07812~0.15625Hz} W 6 (n): 0.03906~0.07812Hz

Next, the power calculation units 105a to 105a to
5f calculates the power for each section of the minimum time resolution determined by the uncertainty relation for each frequency band (step S6). The minimum time resolution of each frequency band is as follows. W ₁ (n): 2 sample points (0.4 seconds) W ₂ (n): 4 sample points (0.8 seconds) W ₃ (n): 8 sample points (1.6 seconds) W ₄ (n) : 16 sample points (3.2 seconds) W ₅ (n): 32 sample points (6.4 seconds) W ₆ (n): 64 sample points (12.8 seconds)

Each of the power calculators 105a to 105f obtains the maximum value of the absolute value of the coefficient of each frequency band for each section of the time resolution, and calculates a half of the square of the maximum value.
To calculate the power in that section. Assuming that the power of each frequency band is P _m (l), the following equation (7)
Is used to calculate the power.

(Equation 2)

FIG. 5 shows that the power calculator 105c
FIG. 9 illustrates a case where the power of W ₃ (n) in the _third frequency band is obtained. Since the time of the frequency band of the W ₃ resolution is 8 sample points, first, n = 0 to 7 interval 0
, The maximum value of the absolute value of W ₃ (n) is obtained. Assuming that the maximum value is W ₃ (1), the power P ₃ (0) in section 0
Is calculated by の of the square of W ₃ (1). Then, the power is sequentially calculated for section 1, section 2,....

As described above, in the first embodiment,
The power for each frequency band can be obtained. In addition,
The two main frequency components of heart rate variability, the LF component,
When determining the power of the HF component, step S7 is executed. In step S7, the power of the HF component is obtained by adding the power P ₃ and P ₄ in W ₃ and W ₄ of the frequency band corresponding to the frequency of the HF component, corresponding to the frequency of the LF component W ₅ and W power of the LF component is obtained by adding the power P ₅ and P ₆ in _6.

FIG. 6 shows a sine wave having a frequency of 0.2 Hz
FIG. 5 shows a result when a digital signal sampled at Hz is analyzed using the device of the first embodiment.
Referring to FIG. 6, it can be seen that the constant power (about 0.5) appears in the P ₄ is a power of the frequency band W ₄ including the frequency 0.2Hz input signal. Since the amplitude value of the input signal is 1, and the power of the input signal at this time is 0.5, it can be seen that the correct power has been calculated.

FIG. 7 shows a result obtained by analyzing a digital signal obtained by sampling a sine wave whose frequency increases in proportion to time at 5 Hz using the apparatus of the first embodiment. Referring to FIG. 7, as the frequency of the input signal changes continuously, the power of each frequency band is increased from P _{6 in the} low frequency band to P _{1 in the} high frequency band.
It can be seen that the situation changes smoothly.

FIG. 8 shows an electrocardiogram measured by exercising with an ergometer for 10 minutes and then sitting and resting for 20 minutes. The power of the HF component, the LF component and the LF / HF were measured by the apparatus of the first embodiment. The results obtained are shown.
Referring to FIG. 8, it can be seen that the power of the HF component, which reflects the activity state of the parasympathetic nerve, decreases with exercise and gradually recovers after exercise cessation, and the actual state is faithfully captured. . In addition, LF / HF, which is an index of the activity state of the sympathetic nerve, shows a high value during exercise and for about 10 minutes after exercise cessation, and the state in which the sympathetic nerve is dominant continues for a while after exercise. Recognize.

(Second Embodiment) Next, a heart rate variability analyzer according to a second embodiment of the present invention will be described with reference to the drawings. The configuration of the heart rate variability analyzer of the present embodiment is the same as that of FIG.

FIG. 9 is a flowchart showing a processing procedure of the heart rate variability analysis device according to the second embodiment. Less than,
The operation of the heart rate variability analyzer according to the second embodiment will be described with reference to FIGS.

In the operation of the heart rate variability analysis apparatus of the second embodiment, the difference from the operation of the first embodiment described above is the operation of the power calculation units 105a to 105f and FIG.
Step S8. Instead of step S6 in FIG.
In FIG. 9, in step S8, power is calculated for each frequency band for each minimum time resolution determined by the uncertainty relationship. The minimum time resolution of each frequency band is as follows. W ₁ (n): 2 sample points (0.4 seconds) W ₂ (n): 4 sample points (0.8 seconds) W ₃ (n): 8 sample points (1.6 seconds) W ₄ (n) : 16 sample points (3.2 seconds) W ₅ (n): 32 sample points (6.4 seconds) W ₆ (n): 64 sample points (12.8 seconds)

Each of the power calculators 105a to 105f calculates, for each section of the time resolution, the average value of the square of the coefficient of each frequency band, thereby calculating the power of that section. Assuming that the power of each frequency band is P _m (l), the power is calculated by the following equation (8).

(Equation 3)

FIG. 10 shows a case where the power of the W ₃ (n) in the _third frequency band is obtained by the power calculator 105c. Since the time of the frequency band of the W ₃ resolution is 8 sample points, first, n = the interval 0 0-7, by obtaining the average of the squares of W ₃ (n), the power P ₃ of the section 0 (0 ) Is calculated. Then, power is sequentially calculated for section 1, section 2,....

As described above, in the second embodiment,
The power for each frequency band can be obtained. In addition,
The two main frequency components of heart rate variability, the LF component,
When determining the power of the HF component, step S7 is executed. In step S7, the power of the HF component is obtained by adding the power P ₃ and P ₄ in W ₃ and W ₄ of the frequency band corresponding to the frequency of the HF component, corresponding to the frequency of the LF component W ₅ and W power of the LF component is obtained by adding the power P ₅ and P ₆ in _6.

FIG. 11 shows a result obtained by analyzing a digital signal obtained by sampling a sin wave having a frequency of 0.2 Hz at 5 Hz using the apparatus of the second embodiment. Referring to FIG. 11, the frequency of the input signal is 0.2
Hz It can be seen that has appeared substantially constant power (about 0.5) to P ₄ is a power of the frequency band W ₄ including. Since the amplitude value of the input signal is 1, and the power of the input signal at this time is 0.5, it can be seen that the correct power has been calculated. If the input signal has a sinusoidal waveform, the first
The embodiment can more accurately obtain a constant power. However, when the waveform of the input signal is different from the waveform of the sine wave, the power calculated in the first embodiment is different from the power of the input signal. In such a case, the method using the average value in the second embodiment can obtain a power closer to the actual value.

FIG. 12 shows a result obtained by analyzing a digital signal obtained by sampling a sine wave whose frequency increases in proportion to time at 5 Hz using the apparatus of the second embodiment. Referring to FIG. 12, as the frequency of the input signal changes continuously, the power of each frequency band is increased from P _{6 in the} lower frequency band to P _{6 in the} higher frequency band.
_It can be seen that the transition to ₁ is captured smoothly.

FIG. 13 shows an electrocardiogram obtained by exercising with an ergometer for 10 minutes and then sitting and resting for 20 minutes. The power of the HF component, the LF component and the LF / HF were measured by the apparatus of the second embodiment. The results obtained are shown.
Referring to FIG. 13, similarly to the first embodiment, the power of the HF component reflecting the activity state of the parasympathetic nerve decreases with exercise, and gradually recovers after exercise cessation. Being faithfully captured. In addition, LF / HF, which is an index of the activity state of the sympathetic nerve, shows a high value during exercise and for about 10 minutes after exercise cessation, and the state in which the sympathetic nerve is dominant continues for a while after exercise. Recognize.

In the first and second embodiments described above, an electrocardiogram is measured, and a heartbeat interval is calculated from the R wave. However, the present invention is not limited to this. Instead, it is also possible to use other biological signals that can calculate a heartbeat interval such as a pulse wave or a peak value of a heart sound. Also, the case has been described where the sampling rate of time-series heartbeat interval data at equal intervals of time is set to 5 Hz and divided into six frequency bands by discrete binary wavelet transform, but both the sampling rate and the number of divided frequency bands are limited to this. It is not something to be done.

In the above description, the first and second embodiments have been described as examples of the calculation method for obtaining the power of the frequency component in a section from a plurality of wavelet coefficients obtained in the section corresponding to the minimum time resolution. However, the present invention is not limited to these, and the power in each section may be obtained by other calculation methods.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a heart rate variability analysis device according to a first embodiment of the present invention.

FIG. 2 is a flowchart showing a processing procedure of the heart rate variability analyzer shown in FIG.

FIG. 3 is a diagram for explaining a method of obtaining heartbeat interval time series data using an electrocardiogram R wave.

FIG. 4 is a diagram illustrating a configuration example of a filter bank of a discrete binary wavelet transform when a digital input signal is divided into six frequency bands.

FIG. 5 is a diagram for describing a method of calculating power from discrete binary wavelet coefficients in the first embodiment.

FIG. 6 is a diagram showing a result of analyzing a sine wave having a constant frequency using the heart rate variability analysis device of the first embodiment.

FIG. 7 is a diagram showing a result of analyzing a sine wave whose frequency increases in proportion to time using the heart rate variability analysis device of the first embodiment.

FIG. 8 is a diagram showing a result of analyzing actual electrocardiogram data using the heart rate variability analyzer of the first embodiment.

FIG. 9 is a flowchart illustrating a processing procedure of the heart rate variability analyzer according to the second embodiment.

FIG. 10 is a diagram for explaining a method of calculating power from discrete binary wavelet coefficients in the second embodiment.

FIG. 11 is a diagram showing a result of analyzing a sine wave of a constant frequency using the heart rate variability analysis device of the second embodiment.

FIG. 12 is a diagram illustrating a result of analyzing a sine wave whose frequency increases in proportion to time using the heart rate variability analysis device of the second embodiment.

FIG. 13 is a diagram showing a result of analyzing actual electrocardiogram data using the heart rate variability analyzer of the second embodiment.

FIG. 14 is a diagram for explaining a method of obtaining power of a frequency component by a conventional heart rate variability analysis method.

FIG. 15 shows a conventional discrete wavelet transform,
FIG. 4 is a diagram illustrating a configuration example of a filter bank when dividing into two frequency bands.

FIG. 16 is a diagram showing a result of analyzing a sine wave having a constant frequency using a conventional analysis method based on discrete wavelet transform.

FIG. 17 is a diagram showing a result of analyzing a sine wave whose frequency increases in proportion to time using a conventional analysis method based on discrete wavelet transform.

[Explanation of symbols]

 Reference Signs List 101: biological signal measuring unit 102: heartbeat interval calculating unit 103: equal interval data generating unit 104: discrete binary wavelet transform unit 105a to 105f: power calculating unit 200: digital input signal 211 to 216: discrete time high-pass filter 221 ... 226 discrete time low-pass filter 231 to 236 wavelet transform coefficients 241 to 246 smoothed signal

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Shuichi Ojima 1006 Kadoma Kadoma, Osaka Prefecture Matsushita Electric Industrial Co., Ltd. (72) Inventor Naoshi Inoue 1006 Odaka Kadoma Kadoma City, Osaka Matsushita Electric Industrial Co., Ltd. ( 72) Inventor Takushi Katsura 1006 Kazuma Kadoma, Kadoma, Osaka Prefecture Inside Matsushita Electric Industrial Co., Ltd.

Claims (6)

[Claims] 1. A method for electrically detecting and analyzing heartbeat fluctuations, comprising: a first step of measuring a biological signal related to a heartbeat; and a biological signal measured in the first step. A second step of detecting a heartbeat interval by using the above-described second step, and interpolating temporally unequally spaced heartbeat interval time-series data obtained from the heartbeat interval detected in the second step to obtain a temporally equal heartbeat interval. A third step of generating heartbeat interval time-series data, and a wavelet for each of a plurality of frequency bands by discrete binary wavelet transform of the temporally equal heartbeat interval time-series data generated in the third step. A fourth step of dividing into a coefficient, and in each of the plurality of frequency bands, for each time section corresponding to a minimum time resolution in the frequency band, a plurality of obtained in each time section. By referring to Eburetto coefficient for calculating the power of the frequency components for each time interval, and a fifth step of determining a time variation of the power of the frequency components contained in the HRV, heart rate variability analysis method. 2. The method according to claim 1, wherein the fifth step calculates, as a power in each time section, a half of a square of a maximum value of absolute values of a plurality of wavelet coefficients obtained in each time section. Item 7. A heart rate variability analysis method according to Item 1. 3. The heart rate variability according to claim 1, wherein in the fifth step, an average value of squares of a plurality of wavelet coefficients obtained in each time interval is calculated as power in each time interval. analysis method. 4. An apparatus for electrically detecting and analyzing heartbeat fluctuations, comprising: a biological signal measuring means for measuring a biological signal related to a heartbeat; and a heartbeat interval from data obtained by the biological signal measuring means. Heart rate interval time-series data that is temporally equally spaced by interpolating the temporally unequally spaced heart rate interval time-series data obtained by the heart rate interval calculating means. Interval data generating means; discrete binary wavelet transform for performing discrete binary wavelet transform on temporally equally spaced heartbeat interval time series data obtained by the uniform interval data generating means, and dividing the data into wavelet coefficients of a plurality of frequency bands Means, for each of the plurality of frequency bands, for each time interval corresponding to the minimum time resolution in that frequency band, Of by calculating the power of the frequency components in referring to wavelet coefficients each time period, and a power calculating means for calculating the time variation of the power of the frequency components contained in the HRV, heart rate variability analysis apparatus. 5. The power calculation unit according to claim 1, wherein the power calculation unit calculates half of the square of the maximum value of the absolute values of the plurality of wavelet coefficients obtained in each time interval as the power in each time interval. The heart rate variability analyzer according to claim 4. 6. The heart rate variability analysis according to claim 4, wherein said power calculation means calculates an average value of squares of a plurality of wavelet coefficients obtained in each time section as power in each time section. apparatus.