JPH11276448A - Signal extraction device and signal extraction method - Google Patents

Abstract

(57) Abstract: A signal x (pulse wave signal) including a desired component p (pulsation component) and an unnecessary component m (body motion component), and the unnecessary component m
Then, the desired component p is extracted with a simple configuration and with high reliability based on the signal q (body motion signal) correlated with. SOLUTION: A predicted value M of an unnecessary component m based on a signal q.
, A subtractor 11 for calculating a difference e between the signal x and the predicted value M, and a coefficient calculator 10 for determining a constant of the adaptive filter 9 so that the average of the square of the difference e is small. And outputs the difference e as the extraction result of the desired component p.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a signal extraction device and a signal extraction method suitable for pulse wave measurement.

[0002]

2. Description of the Related Art In recent years, for health management during exercise,
It is desired to detect a pulse wave, an electrocardiogram, and various other biological states of a subject. However, during exercise, noise due to body motion is superimposed on the detection result of the biological condition. For example, a body motion component is superimposed on the pulse wave signal in addition to the original pulse component, and a myoelectric is superimposed on the measurement result of the electrocardiogram.

For example, in JP-A-7-227383, an FFT process is performed on a pulse wave signal, a body motion signal is detected by an acceleration sensor or the like, and a frequency component corresponding to the body motion signal is removed from the pulse wave signal. Techniques are disclosed.
In addition, Japanese Patent Application Laid-Open No. 5-184548 discloses a technique for obtaining a pulse rate by removing a body motion component based on a correlation between a plurality of pulse wave signals.

[0004]

Problems to be Solved by the Invention
In the technique disclosed in Japanese Unexamined Patent Publication No. 3 (Kokai) No. 3, the relationship between the body motion signal and the body motion component on the pulse wave signal must be determined. However, since such a relationship differs depending on the type of exercise, it is necessary to obtain a correspondence relationship in advance for each type of exercise. In addition, since there is an individual difference in such a relationship, it is necessary to seek a corresponding relationship for each subject, which is extremely complicated.

Further, in the technique disclosed in Japanese Patent Application Laid-Open No. 5-184548, it is difficult to obtain an accurate pulse rate because a body motion signal is not used for extracting a pulsation component. The present invention has been made in view of the above circumstances, and has as its object to provide a signal extraction device and a signal extraction method capable of extracting a desired signal with a very simple configuration and high reliability.

[0006]

According to a first aspect of the present invention, a signal x including a desired component p and an unnecessary component m, and a signal q correlating with the unnecessary component m are provided.
A signal extraction device for extracting the desired component p based on the signal x, the adaptive filter for obtaining the predicted value M of the unnecessary component m based on the signal q, the signal x and the predicted value M
And a coefficient calculating means for determining a constant of the adaptive filter so that the average of the square value of the difference e becomes small.
Is output as an extraction result. further,
In the configuration according to claim 2, in the signal extraction device according to claim 1, the signal x is a pulse wave signal of a living body,
The desired component p is a pulsation component included in the pulse wave signal, the unnecessary component m is a body motion component included in the pulse wave signal, the signal q is a body motion signal of the living body, A pulse wave sensor that measures a pulse wave of a living body and outputs the pulse wave signal, and a body movement sensor that measures the body movement of the living body and outputs the body movement signal. In the configuration according to the third aspect, the desired component p and the unnecessary component m
A signal extraction method for extracting the desired component p based on a signal x including the following and a signal q correlated with the unnecessary component m, wherein a prediction value of the unnecessary component m is based on the signal q by an adaptive filter. M; obtaining a difference e between the signal x and the predicted value M;
A step of determining the constant of the adaptive filter so that the average of the power values becomes small; and a step of outputting the difference e as a result of extracting the desired component p.

[0007]

BEST MODE FOR CARRYING OUT THE INVENTION Next, the configuration of a pulse meter according to an embodiment of the present invention will be described with reference to FIG. In the figure, reference numeral 1 denotes a pulse wave sensor, which detects a pulse wave signal of a subject. This pulse wave signal is supplied to the AD converter 3 via the amplifier 2 and is sampled.
Stored in the buffer 4. Note that the sampling frequency in the present embodiment is “8 Hz”.

Here, the pulse wave signal in the n-th sample is represented as "x (n)". This pulse wave signal x (n) is
The sum of the original pulsation component p (n) and the body motion component m (n) based on the body motion is obtained. Reference numeral 5 denotes a body motion sensor, which includes an acceleration sensor and the like, and measures a body motion signal of the subject. This body motion signal is supplied to the AD converter 7 via the amplifier 6,
After being sampled, it is stored in the buffer 8. The body motion signal in the n-th sample is represented as “q (n)”.

Next, reference numeral 9 denotes an adaptive filter constituted by a "17th-order" FIR filter, which calculates a predicted value M (n) of the body motion component m (n) based on the following equation. In the following equation, N is the order (= 17) of the FIR filter, and h _k
Is a constant.

(Equation 1)

Here, the detailed configuration of the adaptive filter 9 is shown in FIG.
FIG. In the figure, reference numerals 401 to 40n denote delay circuits which delay the body motion signal q (n) by one sampling period. 410~41n are multipliers, body motion signal q (n) and its delayed signal q (n-1), q (n-2), with respect ...... q (nk), each constant h _0, h ₁ , ……, h _N-1 are multiplied. Reference numeral 420 denotes an adder, which calculates the sum of the multiplication results. Thereby, the body motion signal q (n) and its delay signals q (n-1), q (n
-2),... Q (nk) is weighted, and the predicted value M
(n) will be required.

Returning to FIG. 1, reference numeral 11 denotes a subtractor, which subtracts a predicted value M (n) from the pulse wave signal x (n). That is, the residual signal e (n) is equal to "p (n) + m (n) -M (n)". Next, reference numeral 10 denotes a coefficient calculation unit employing the sequential least squares method. That is, in the coefficient calculation unit 10, the past "12
8 ”, a constant h _k (k = 1, 2,..., So that the root mean square of the residual signal e (n) is minimized.
N) are sequentially calculated, and the calculated constant h _k is set in the adaptive filter 9.

Reference numeral 12 denotes a frequency analysis unit that performs an FFT (fast Fourier transform) process on a series of residual signals e (n) and obtains the frequency components. Reference numeral 13 denotes a pulse wave extraction unit, which extracts a component having the highest level from these frequency components as a pulse wave component. Reference numeral 14 denotes a pulse rate calculator, which calculates a pulse rate per minute based on the frequency value related to the pulse wave component. A display unit 15 displays the calculated pulse rate. As described above, since the pulse rate is obtained by applying the adaptive filter to each sampling data (128 points in the present embodiment), it is possible to continuously obtain the correct pulse rate in daily life.

In the above configuration, the residual signal e (n) is "p
(n) + m (n) -M (n) ", of which the pulsation component p (n)
And the body motion component m (n) have a small correlation, so that the influence of the pulsation component p (n) in determining the constant h _k in the coefficient calculation unit 10 is small. Therefore, the predicted value M (n) is mostly determined based on the body motion component m (n), and the predicted value M (n)
Becomes a value close to the body motion component m (n). Accordingly, the residual signal e (n) approximates the pulsation component p (n), and the frequency analysis unit 12
An accurate pulse rate can be obtained via the pulse wave extraction unit 13 and the pulse rate calculation unit 14.

FIGS. 2 to 6 show examples in which a pulsation component is extracted from a pulse wave signal in the prior art and this embodiment.
This will be described with reference to FIG. FIG. 2 shows the waveform of the pulse wave signal detected by the pulse wave sensor 1 and the result of the FFT analysis, in which the pulsation component is buried in the body motion component and the detection is difficult. FIG. 3 shows a waveform of a body motion signal detected by the body motion sensor 5 and an FFT analysis result. FIG. 4 simply shows the waveform of the difference between the two and the result of FFT analysis. The level of the body motion component is higher than the pulsation component, and it is difficult to accurately measure the pulse rate.

Next, FIG. 5 shows a waveform and an FFT analysis result when a body motion component is removed by obtaining an autocorrelation from the pulse wave signal. In FIG. 5, although the pulsation component is slightly different from FIG. 4, a malfunction may occur because a body motion component at a higher level than the pulsation component still exists.

FIG. 6 shows the waveform of the residual signal e obtained by the present embodiment and the result of FFT analysis. According to the figure, since the level of the spectrum of the pulsation component is higher than the spectrum of the body motion component, the FF of the residual signal e
If the maximum spectrum is simply obtained from the T analysis result, the pulse rate can be obtained immediately.

As described above, according to the pulsimeter of the present embodiment, the predicted value M (n) can be approximated to the body motion component m (n) regardless of the type of exercise. Dynamic component p (n)
Can be extracted, and an accurate pulse rate can be measured.

2. Configuration of sensor part 2.1. Pulse wave sensor 1 and amplifier 2 Next, a detailed configuration of each unit in the present embodiment will be described.
First, a circuit diagram of the pulse wave sensor 1 and the amplifier 2 is shown in FIG. In the figure, LED 31, resistor 351, filter 32
The pulse wave sensor 1 is constituted by the photodiode 33 and the amplifier 2 is constituted by the OP amplifier 34 and the resistor 352.

When the power supply + V is supplied to the LED 31, a current determined by the value of the resistor 351 flows through the LED 31,
Light is irradiated. This irradiation light is irradiated to the light of the wrist via the filter 32, and the oxyhemoglobin HbO
^The light that has been absorbed by ² and has escaped absorption is reflected by the tissue of the living body. The reflected light is again incident on the photodiode 33 via the filter 32.

FIG. 11 shows the light emission characteristics of the LED 31 for pulse wave measurement. In the illustrated example, the LED 31 ′
Has a light emission characteristic of a half value width of 40 nm with a peak wavelength of 525 nm. By the way, in the wavelength region of 600 nm or more, absorption of irradiation light by oxyhemoglobin HbO ² is hardly performed, and in the wavelength region of less than 600 nm, the absorption coefficient of oxyhemoglobin HbO ² increases. In addition, since oxyhemoglobin HbO ² moves in accordance with the pulsation of the blood flow, a change in the light absorption characteristics due to oxyhemoglobin HbO ² corresponds to a pulse wave.

The cathode of the photodiode 33 is connected to a power supply +
V, and the anode is connected to the negative input terminal of the OP amplifier 34. Since the positive input terminal of the OP amplifier 34 is grounded, the anode of the photodiode 33 is imaginarily short-circuited to ground. Therefore, the photodiode 33 is reverse-biased, and when light enters there, a current flows according to the amount of light. The OP amplifier 34 and the resistor 352 convert the current from the photodiode 33 into a voltage and amplify it. Therefore, O
The output signal of the P amplifier 34 fluctuates according to the amount of incident light. FIG. 12 shows the spectral sensitivity characteristics of the photodiode 33.

2.2. Next, a circuit diagram of the body motion sensor 5 and the amplifier 6 is shown in FIG. The configuration of FIG. 8 is the same as that of FIG.
7, a resistor 351, a filter 38, and a photodiode 39 constitute the body movement sensor 5, and the OP amplifier 34
The amplifier 6 includes the resistor 352 and the resistor 352.

However, the LED 37 in the body movement sensor 5
FIG. 13 shows a light emission characteristic having a peak wavelength at 660 nm and a half value width of 40 nm as shown in FIG. When the wavelength region used for measurement is set to 600 nm or more in this manner, the absorption of irradiation light by oxyhemoglobin HbO ² is hardly absorbed, and the spectrum intensity of the spectrum Sm related to the pulse wave component is greatly reduced.

On the other hand, when the tissue is vibrated due to the body movement, the absorption rate of the irradiation light changes, so that the light incident on the photodiode 39 fluctuates according to the body movement. Thus, an output signal of a level corresponding to the body movement is output from the photodiode 39. The detailed configurations of the pulse wave sensor 1 and the body motion sensor 5 are disclosed in Japanese Patent Application No. 9-308 filed by the present applicant.
913 and the like.

3. FIG. 14 is a diagram showing an external configuration of the pulse meter according to the present embodiment. In the figure, in the pulsimeter of the present example, an apparatus main body 110 having a wristwatch structure, a cable 120 pulled out from the apparatus main body, a body movement sensor unit 130 provided at a distal end side of the cable 120, and a body movement A sensor fixing band 140 for attaching the sensor unit 130 to a finger, a wrist, or the like is roughly configured. The body motion sensor unit 130 is constituted by a reflection type optical sensor.

The main body 110 includes a watch case 211 having a built-in clock function, and a wristband 212 for attaching the watch case 211 to an arm. On the front side of the watch case 211, the liquid crystal display device 1 that displays, in addition to the current time and date, body movement information (biological information) based on the detection result of the body movement sensor unit 130, and the like.
13 are constituted. Further, inside the watch case 211, a data processing circuit 50 to which a body movement signal Vt as a detection result by the body movement sensor unit 130 is supplied is incorporated. The data processing circuit 50 applies FFT to the body motion signal Vt.
The pitch P is calculated by performing processing and analyzing the processing result. Button switches 111 and 112 for adjusting the time and switching the display mode are provided on the outer surface of the watch case 211.

The power source of the pulsimeter is a battery built in the watch case 211, and the cable 120 supplies electric power from the battery to the body movement sensor unit 130 and outputs the detection result of the body movement sensor unit 130 to the watch case 211. 211
The data can be input to the data processing circuit 50 in the inside. Velcro is stretched over the sensor fixing band 140 of this example, and as shown in FIG.
Can be attached with the body motion sensor unit 130 in close contact with the base of the finger.

A disc-shaped body motion sensor unit 130 is fixed to the inner surface of the sensor fixing band 140, and includes LEDs 31, 37, as schematically shown in FIG.
The photodiodes 33, 39 are pointed at the finger. L
When the ED 31 irradiates light toward the finger, the irradiating light is absorbed by the oxyhemoglobin HbO ² , the irradiating light that has escaped absorption is reflected by the tissue, and the reflected light is received by the photodiode 33.

When the LED 37 emits light toward the finger, the emitted light is absorbed by the tissue, the irradiated light that has escaped absorption is reflected by the tissue, and the reflected light is received by the photodiode 39. Sensor fixing band 1
Forty materials are selected that do not transmit light. Therefore, even when the pulse meter is used outdoors, natural light does not directly enter the photodiodes 33 and 39.

4. Modifications The present invention is not limited to the embodiments described above,
For example, various modifications are possible as follows. (1) The coefficient calculation unit 10 in the above embodiment sequentially has a minimum of 2
Although the multiplication method is adopted, an arbitrary adaptive algorithm such as a learning identification method can be used instead.

(2) In the above embodiment, the pulse rate is obtained based on the pulsating component of the pulse wave signal. For example, a detailed frequency analysis of the pulsating component is performed to diagnose the health condition of the subject. It may be configured as follows.

(3) Further, in the above embodiment, the case where the body motion component is removed from the pulse wave signal has been described. However, the present invention is not limited to such an application, and the desired component p and the unnecessary component m can be separated. The present invention can be applied to any use for extracting a desired component p based on the included signal x and the signal q correlated with the unnecessary component m.

For example, when the electrocardiogram is measured during the exercise of the subject, myoelectrics are superimposed on the electrocardiogram in addition to the original electrocardiogram.
Assuming that the original electrocardiogram is a desired component p, the myoelectric is an unnecessary component m, and the body motion signal is a signal q, the original electrocardiogram can be extracted by the same configuration as in the above embodiment.

(4) In the above embodiment, a photoelectric sensor is used as the body movement sensor 5, but various types of body movement sensors are conceivable. For example, as shown in FIG. 9, a PZT having a length of about 8 mm and a diameter of about 3 mm
The PZT 500 can be mounted on a substrate, and acceleration, that is, body movement can be detected based on the strain applied to the PZT 500. FIG. 10 shows a configuration example of the amplifier 6 in such a case.

(5) Similarly, in the above embodiment, a photoelectric sensor is used as the pulse wave sensor 1, but various known pulse wave sensors such as a pressure pulse wave sensor may be used instead. .

(6) In the above embodiment, the frequency analyzing unit 12, the pulse wave extracting unit 13, and the pulse rate calculating unit 14
The pulse rate was determined by
A conversion circuit for converting (n) into a rectangular wave may be provided, and the pulse rate may be obtained based on the period of the rectangular wave.

(7) In the above embodiment, the pulse wave and the body movement are measured by the body movement sensor unit 130 attached to the finger. However, the pulse wave and the body movement can be measured at any part of the subject. it can. As an example,
FIG. 16 shows an example in which these sensors are provided on the back surface of the watch case 211. As shown, a pulse wave sensor unit 300 corresponding to the above-described body motion sensor unit 130
Are formed integrally with the main body on the back side of the watch case 211. The watch case 211 is provided with a wrist band 212 for wearing the watch case on the wrist. When the wrist band 212 is wound around the wrist and worn, the back side of the watch case 211 comes into close contact with the back of the wrist.

On the back side of the watch case 211, a back cover 54 is provided.
There is provided a transparent glass 137 fixed by. The transparent glass 137 protects the pulse wave sensor unit 300 and transmits the irradiation light of the LEDs 31 and 37 and the reflected light obtained through the living body.

On the front side of the watch case 211, there is provided a liquid crystal display device 113 for displaying biological information such as a pulse rate HR based on the detection result of the pulse wave sensor unit 300 in addition to the current time and date. In addition, the watch case 211
Inside the main board 51, various I / Os such as a CPU
A C circuit is provided, and the data processing circuit 50 is configured by the C circuit.

On the back side of the main board 51, the battery 5
2 is provided, and the liquid crystal display device 11
3. Power is supplied to the main board 51 and the pulse wave sensor unit 300. The main board 51 and the pulse wave sensor unit 300 are connected by a heat seal 53, and power is supplied from the main board 51 to the pulse wave sensor unit 300 by wiring formed in the heat seal 53, and the pulse wave sensor unit The pulse wave signal Vm is supplied from 300 to the main board 51.

The data processing circuit 50 outputs the pulse wave signal Vm to F
By performing FT processing and analyzing the processing result,
The pulse rate HR is calculated. Note that button switches 111 and 112 (not shown) for adjusting the time and switching the display mode are provided on the outer surface of the watch case 211 in the same manner as the pulse meter shown in FIG.

The wristband 12 of this biological information measuring device
Is wound around the wrist and the back side of the watch case 211 is turned to the back of the wrist. Therefore, the LED 31,
Light from 37 is applied to the back of the wrist via transparent glass 37, and the reflected light is received by photodiodes 33 and 39.

[0043]

As described above, according to the present invention,
Since the constant of the adaptive filter is determined such that the average of the square value of the difference e between the signal x and the predicted value M becomes small, a desired signal can be extracted with a very simple configuration and with high reliability.

[Brief description of the drawings]

FIG. 1 is a block diagram of one embodiment of the present invention.

FIG. 2 is a diagram showing a waveform of a pulse wave signal and an FFT analysis result.

FIG. 3 is a diagram showing a waveform of a body motion signal and an FFT analysis result.

FIG. 4 is a diagram showing a waveform of a beat component extracted by a conventional technique and an FFT analysis result.

FIG. 5 is a diagram showing a waveform of a pulsation component extracted by another conventional technique and an FFT analysis result.

FIG. 6 is a diagram showing a waveform of a beat component extracted by the present embodiment and an FFT analysis result.

FIG. 7 is a circuit diagram of the pulse wave sensor 1 and the amplifier 2.

FIG. 8 is a circuit diagram of the body motion sensor 5 and the amplifier 6.

FIG. 9 is a view showing a modification of the body motion sensor 5.

FIG. 10 is a circuit diagram of a modified example of the amplifier 6.

FIG. 11 is a light emission characteristic diagram of the LED 31 in the pulse wave sensor 1.

FIG. 12 is a spectral sensitivity characteristic diagram of photodiodes 33 and 39.

FIG. 13 is a light emission characteristic diagram of the LED 37 in the body motion sensor 5.

FIG. 14 is a diagram illustrating an external configuration of a pulse meter according to the present embodiment.

15 is a sectional view of a main part of FIG.

FIG. 16 is a sectional view of a modification of the above embodiment.

17 is a circuit diagram of the adaptive filter 9. FIG.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 pulse wave sensor 2 amplifier 3 A / D converter 4 buffer 5 body motion sensor 6 amplifier 7 A / D converter 8 buffer 9 adaptive filter 10 coefficient calculation unit 11 subtractor 12 frequency analysis unit 13 pulse wave extraction unit 14 pulse rate calculation unit 15 display Unit 31, 37 LED 32 filter 33, 39 photodiode 34 OP amplifier 37 LED 38 filter 50 data processing circuit 51 main board 52 battery 53 heat seal 54 back cover 101 pulse meter 110 device main body 120 cable 130 body movement sensor unit 137 transparent glass 140 Sensor fixing band 211 Watch case 212 Wrist band 300 Pulse wave sensor unit 351 Resistance 352 Resistance 401 to 40n Delay circuit 410 to 41n Multiplier 420 Adder

Claims (3)

[Claims] 1. A signal x including a desired component p and an unnecessary component m.
A signal extraction device for extracting the desired component p based on a signal q correlated with the unnecessary component m, and an adaptive filter for obtaining a predicted value M of the unnecessary component m based on the signal q; A subtractor for calculating a difference e between the signal x and the predicted value M; and coefficient calculating means for determining a constant of the adaptive filter so that an average of squares of the difference e is small. A signal extracting device for outputting e as an extraction result of the desired component p. 2. The signal x is a pulse wave signal of a living body, the desired component p is a pulsation component included in the pulse wave signal,
The unnecessary component m is a body movement component included in the pulse wave signal, the signal q is a body movement signal of the living body, and a pulse wave sensor that measures the pulse wave of the living body and outputs the pulse wave signal The signal extraction device according to claim 1, further comprising: a body movement sensor that measures the body movement of the living body and outputs the body movement signal. 3. A signal x including a desired component p and an unnecessary component m.
And a signal extraction method for extracting the desired component p based on the signal q correlated with the unnecessary component m, wherein an adaptive filter determines a predicted value M of the unnecessary component m based on the signal q. Obtaining a difference e between the signal x and the predicted value M; determining a constant of the adaptive filter so that the average of the squares of the difference e is small; Outputting the result as an extraction result of the component p.