JP6515709B2 - Biological information detection device - Google Patents

Description

  The present invention relates to a biological information detection apparatus.

  What is shown by patent document 1 as one form of a biometric information detection apparatus is known. As shown in FIG. 1 of Patent Document 1, the biological information detection device includes a measurement device 20 that measures a biological signal of a driver of a vehicle, an acceleration sensor 31 that detects an acceleration applied to the vehicle, and a control device 10. Is equipped. The control device 10 corrects the biosignal measured by the measuring device 20 based on the acceleration detected by the acceleration sensor 31 to generate biomedical information of the driver with high accuracy, and estimate the driver's state. doing. The control device 10 derives the correction amount of the biological signal by applying a map according to the acceleration. This map is derived in advance by experiment or the like.

JP, 2009-213636, A

  However, in the biological information detection apparatus disclosed in Patent Document 1, since the correction of the biological signal is performed using the map derived in advance, the correction amount becomes uniform. In addition, since the map is based only on the information on the vehicle side, the driver's individual difference can not be sufficiently addressed. On the other hand, there is a demand to further improve the estimation accuracy of biological information of the biological information detection apparatus.

  Therefore, the present invention has been made to solve the above-mentioned problems, and it is an object of the present invention to improve the estimation accuracy of biological information regardless of the individual difference of the occupant in the biological information detection apparatus.

  In order to solve the above problems, the biometric information detection apparatus according to claim 1 is attached to a moving body, detects a person's biometric information, and outputs an output signal according to the biometric information, and movement Process noise in the Kalman filter's equation of state is determined based on the motion state amount detection unit that detects the motion state amount of the body, the output signal output by the biological information sensor, and the motion state amount detected by the motion state amount detection unit. A process noise determination unit, and a biological information estimation unit that estimates biological information using a process noise determined by the process noise determination unit and a Kalman filter using an output signal output from the biological information sensor as an observation value; Is equipped.

  According to this, the process noise determination unit determines the process noise of the state equation used for the Kalman filter based on the output signal from the biological information sensor and the motion state quantity from the motion state quantity detection unit. Since the output signal of the biological information sensor differs depending on the occupant, the process noise of the Kalman filter is determined to correspond to the individual difference of the occupant. Then, the biological information estimation unit estimates biological information using a Kalman filter using this process noise. Therefore, the biological information detection apparatus can estimate the biological information so as to correspond to the individual difference of the occupant, and therefore the estimation accuracy of the biological information as compared to the case of estimating the biological information using only the information on the vehicle side Can be improved.

It is a schematic diagram showing composition of one embodiment of a living body information detecting device by the present invention. It is a block diagram of a living body information detecting device shown in FIG. It is a figure which shows the analysis result of the frequency analysis part shown in FIG. It is a flowchart of the program run by the control apparatus shown in FIG.

  An embodiment of a biological information detection apparatus 1 according to the present invention will be described with reference to the drawings. The living body information detection device 1 estimates living body information of a passenger of a moving body. The moving body is, for example, a vehicle or an aircraft. The biological information is, for example, a heart beat or a respiration. In the present embodiment, a moving body is described as a vehicle, and biological information as respiration. As shown in FIG. 1, the living body information detection apparatus 1 includes a living body information sensor 10, an exercise state quantity detection unit 20, a control device 30, and an output unit 40. The control device 30, the living body information sensor 10, the motion state amount detection unit 20, and the output unit 40 are communicably connected by wireless or wire.

  The biological information sensor 10 is attached to a mobile body, detects biological information of the occupant M, and outputs an output signal S according to the biological information. The biological information sensor 10 is, for example, a radio wave type sensor such as a radar Doppler sensor, a pressure sensor such as a piezoelectric element, or an acceleration sensor. In the present embodiment, the biological information sensor 10 is a piezoelectric element. The biological information sensor 10 is disposed, for example, on a cushion portion 2a that supports a body portion on the back side of the chest of the occupant M in a seat 2 of a vehicle. Since the chest of the occupant M is repeatedly expanded and contracted by the breathing of the occupant M, the cushion portion 2a is repeatedly pressed. The pressure generated by pressing the cushion portion 2a is transmitted to the biological information sensor 10, whereby the biological information sensor 10 is deformed. The biological information sensor 10 outputs an output signal S corresponding to the amount of deformation to the control device 30. In addition, the biological information sensor 10 is disposed so as to be deformed in accordance with the pressure in a predetermined direction W (for example, the longitudinal direction of the vehicle).

  The motion state amount detection unit 20 detects a motion state amount of a moving object. In the present embodiment, the motion state quantity is acceleration. Specifically, the motion state quantity detection unit 20 is an acceleration sensor that detects the acceleration of the vehicle. The motion state quantity detection unit 20 detects an acceleration in a predetermined direction W. The motion state quantity detection unit 20 is disposed, for example, in a frame portion (not shown) of the seat 2 of the vehicle. The motion state quantity detection unit 20 outputs the acceleration of the vehicle to the control device 30 as a detection signal (hereinafter referred to as an acceleration signal A).

  The control device 30 centrally controls the biological information detection device 1. As shown in FIG. 2, the control device 30 includes a vibration noise removal unit 31, a frequency analysis unit 32, a subtraction unit 33, a biological information frequency derivation unit 34, a process noise determination unit 35, and a biological information estimation unit 36.

  The vibration noise removal unit 31 removes vibration noise of the vehicle included in the output signal S from the output signal S output by the biological information sensor 10. The vibration noise is determined (estimated) from the acceleration signal A outputted by the motion state quantity detection unit 20 based on the transfer function model of the biological information sensor 10. That is, by inputting the acceleration signal A to the transfer function model, vibration noise is estimated and output. The transfer function model is, for example, an external autoregressive model of ARX type, and is stored in the controller 30. Then, by subtracting the vibration noise from the output signal S, most of the vibration noise included in the output signal S is removed from the output signal S. The output signal S from which the vibration noise has been removed is output to the biological information estimation unit 36.

  The frequency analysis unit 32 performs frequency analysis on the output signal S output by the biological information sensor 10 and the acceleration signal A output by the motion state quantity detection unit 20. Specifically, the frequency analysis unit 32 performs fast Fourier transform operations on the output signal S and the acceleration signal A, respectively. Thereby, the biological body amplitude Sa which is the amplitude of the output signal S and the acceleration amplitude Aa which is the amplitude of the acceleration signal A are derived for each predetermined frequency (see FIG. 3). The predetermined frequency is, for example, 0.05 Hz.

  The subtracting unit 33 subtracts the intensity of the acceleration signal A from the intensity of the output signal S based on the frequency analysis result of the frequency analysis unit 32. Specifically, the subtraction unit 33 subtracts the acceleration amplitude Aa, which is the strength of the acceleration signal A, from the biological amplitude Sa, which is the strength of the output signal S, for each predetermined frequency, and calculates the difference in amplitude as a subtraction result. It derives for every predetermined frequency.

  Based on the output signal S output by the biological information sensor 10 and the acceleration signal A output by the motion state quantity detection unit 20, the biological information frequency deriving unit 34 is biological information that is a frequency including biological information in the output signal S. The frequency fb is derived. Specifically, the biological information frequency deriving unit 34 derives the biological information frequency fb based on the amplitude difference which is the subtraction result of the subtracting unit 33. Here, since the biological information sensor 10 detects not only the biological information but also the acceleration, the output signal S includes the biological information and the acceleration. On the other hand, the motion state quantity detection unit 20 detects only the acceleration. Therefore, when the calculation result of the fast Fourier transform of the output signal S and the acceleration signal A is compared, the frequency having a relatively large amplitude difference which is the difference between the biological amplitude Sa of the output signal S and the acceleration amplitude Aa of the acceleration signal A , It can be estimated that it is a frequency including biological information in the output signal S. Therefore, a frequency with a relatively large amplitude difference corresponds to the biological information frequency fb (see FIG. 3). Specifically, the biological information frequency deriving unit 34 derives, as the biological information frequency fb, a frequency whose amplitude difference, which is the subtraction result of the subtracting unit 33, is equal to or greater than a predetermined determination value. The predetermined judgment value is measured and derived by experiment or the like.

  Further, the biological information frequency deriving unit 34 calculates a biological frequency width fbh and a biological frequency center value fbc (see FIG. 3) of the derived biological information frequency fb. The biological frequency width fbh is a difference between the maximum value and the minimum value of the biological information frequency fb. The biological frequency center value fbc is a value at the center of the biological information frequency fb.

  The process noise determination unit 35 determines the process noise v in the equation of state of the Kalman filter based on the output signal S output by the biological information sensor 10 and the acceleration signal A output by the motion state quantity detection unit 20. . Specifically, the process noise determination unit 35 determines the process noise v from the biological information frequency fb derived by the biological information frequency deriving unit 34 based on the process noise map. The process noise map is stored in the control device 30, and indicates the correlation between the process noise v and the biological information frequency fb. Specifically, the process noise map indicates the correlation between the process noise v, the biological frequency width fbh of the biological information frequency fb, and the biological frequency center value fbc. The correlation is a positive correlation that the process noise v increases as the biological frequency width fbh increases. The correlation is measured and derived by experiment or the like. That is, the process noise determination unit 35 determines the process noise v from the biological frequency width fbh of the biological information frequency fb derived by the biological information frequency derivation unit 34 and the biological frequency center value fbc based on the process noise map. The process noise v determined by the process noise determination unit 35 is output to the biological information estimation unit 36.

  The biological information estimation unit 36 estimates biological information by using a process noise v determined by the process noise determination unit 35 and a Kalman filter using an output signal S output by the biological information sensor 10 as an observation value. is there. Specifically, the biological information estimation unit 36 uses the output signal S from which the vibration noise has been removed by the vibration noise removal unit 31 as an observation value.

Next, use of a Kalman filter for estimation of biological information by the biological information estimation unit 36 will be shown. In the present embodiment, the Kalman filter is an extended Kalman filter.
Here, the output signal S is modeled as the sum of harmonic components of a sinusoidal wave with slowly varying amplitude and phase elements.

From this equation, a vector indicating the following condition is given.

The change over time of the amplitude a _{k, i} of the sine component is modeled by the additional white Gaussian noise as follows.
(Number 3)
a _{k + 1, i} = a _{k, i} + v _{k, i} ^a

The temporal change of the instantaneous basic pulse ω _k is also modeled by the additional white Gaussian noise as follows.
(Number 4)
ω _{k + 1, i} = ω _{k, i} + v _k ^ω

The same applies to the phase difference θ _i (t) between the basic pulse and the harmonic component, and the temporal change of the instantaneous phase φ _{k + 1, i} (t) is obtained by the following equation.
(Number 5)
φ _{k + 1, i} = i · φ _{k, i} + v _{k, i} ^φ

Such a choice implies that ω _k is expressed as the ratio of the actual pulse to the sampling frequency of the signal. As a result, the state equation indicating the state transition is linear and is given by the following equation. v _k is process noise v.

This can also be written as:
(Number 7)
x _{k + 1} = A · x _k + v _k (1)
here,

The estimation of the variance of the component of the process noise v _k affects the rate of change of the estimated parameters (pulse, amplitude component and phase component) and the algorithm convergence speed.

観測を示す式は非線形であるため、拡張カルマンフィルタの使用が裏付けられる。ｙ_ｋは観測値である。観測ノイズであるｎ_ｋの分散は、出力信号Ｓで観測されるノイズの分散に関連する。 Considering equation (1), the equation showing the fundamentally expected observations is given by the following equation.

The equation that indicates the observation is non-linear, supporting the use of an extended Kalman filter. y _k is an observation value. The variance of the observation noise _nk is related to the variance of the noise observed in the output signal S.

The extended Kalman filter algorithm is implemented as follows.
The initialization steps are as follows. Also, the value of the process noise v determined by the process noise determination unit 35 is given as an initial value of v _k .

更新される式は、次のようになる。

ここで、

ここで、Ｑ_ωとＲ_ηは、それぞれｖ_ｋとｎ_ｋの共分散行列である。また、Ｉは、単位行列である。従って、これは、拡張カルマンフィルタの標準アルゴリズムである。 If k∈ {1, 2,...,}}, the prediction formula of the extended Kalman filter is as follows.

The updated expression is as follows.

here,

Here, Q _ω and R _η are covariance matrices of v _k and n _k , respectively. Also, I is an identity matrix. Thus, this is the standard algorithm of the extended Kalman filter.

In the case of the present invention, this algorithm is adapted as follows. The value of the next observation noise may be assumed to be equal to zero.

ここで、Ａは、式（２）から得られる。 Since it is known that Formula (1) which shows a state transition is linear, it becomes as follows.

Here, A is obtained from equation (2).

および

Considering Equation (1) and Equation (3), the following is obtained.

and

Finally, equation (3) is as follows.

  The output unit 40 is an interface for outputting a signal indicating the biological information estimated by the biological information estimation unit 36 to the outside.

  Next, estimation control for estimating the biological information of the occupant M of the vehicle in the biological information detection apparatus 1 described above will be described along the flowchart shown in FIG. For example, when a switch (not shown) for starting estimation of biometric information is turned on, the biometric information detection device 1 starts estimation control.

  Control device 30 obtains output signal S and acceleration signal A at the current time (step S102), and performs fast Fourier transform operation on output signal S and acceleration signal A as described above (step S104; Frequency analysis unit 32). As described above, control device 30 subtracts acceleration amplitude Aa from living body amplitude Sa for each predetermined frequency (step S106; subtraction unit 33), and based on the difference between the amplitudes, the living body frequency width of biological information frequency fb fbh and a biological frequency center value fbc are calculated (step S108; biological information frequency deriving unit 34). Then, as described above, control device 30 determines process noise v from biological frequency width fbh of biological information frequency fb and biological frequency center value fbc based on the process noise map (step S110; process noise determination unit 35). ).

  As described above, the control device 30 executes the initialization step in the Kalman filter algorithm (step S112; biological information estimation unit 36). Furthermore, as described above, control device 30 removes vibration noise from output signal S (step S114; vibration noise removal unit 31), and estimates biological information at the current time based on the algorithm of the Kalman filter (step S116; Biological information estimation unit 36). Control device 30 outputs the biological information estimated at the current time (step S118; output unit 40).

  Then, control device 30 determines whether or not predetermined time Ts has elapsed (step S120). The predetermined time Ts is, for example, one second. If the predetermined time Ts has not elapsed, the control device 30 repeatedly executes step S120. On the other hand, when the predetermined time Ts has elapsed, the control device 30 advances the program to step S122. Control device 30 obtains output signal S and acceleration signal A at the current time in step S122, and returns the program to step S114. That is, the control device 30 repeatedly executes estimation of biological information at predetermined time intervals Ts.

According to the present embodiment, the biological information detection device 1 is attached to the vehicle, detects the biological information (respiration) of the occupant M, and outputs the output signal S according to the biological information; The process noise v in the state equation of the Kalman filter is determined based on the motion state quantity detection unit 20 that detects the acceleration of the target, the output signal S output by the biological information sensor 10 and the acceleration detected by the movement state quantity detection unit 20 A living body which estimates biological information by using a process noise determination unit 35 and a process noise v determined by the process noise determination unit 35, and using an output signal S output from the biological information sensor 10 as an observation value And an information estimation unit 36.
According to this, the process noise determination unit 35 determines the process noise v of the state equation used for the Kalman filter based on the output signal S from the biological information sensor 10 and the acceleration signal A from the motion state quantity detection unit 20. . Since the output signal S of the biological information sensor 10 differs depending on the occupant M, the process noise v of the Kalman filter is determined to correspond to the individual difference of the occupant M. Then, the biological information estimation unit 36 estimates biological information by a Kalman filter using the process noise v. Therefore, the biological information detection device 1 can estimate the biological information so as to correspond to the individual difference of the occupant M. Therefore, compared with the case where the biological information is estimated using only the information on the vehicle side, The estimation accuracy can be improved.

A biological information frequency deriving unit 34 for deriving a biological information frequency fb which is a frequency of biological information based on the output signal S outputted by the biological information sensor 10 and the acceleration detected by the motion state quantity detecting unit 20; The process noise determination unit 35 further includes a process noise map indicating the correlation between the process noise v and the biological information frequency fb, and the process noise determination unit 35 calculates the biological information frequency derived by the biological information frequency deriving unit 34 based on the process noise map. Determine the process noise v from fb.
According to this, the biological information frequency deriving unit 34 derives the biological information frequency fb corresponding to the frequency of the biological information. Then, the process noise determination unit 35 determines the process noise v from the biological information frequency fb based on the process noise map. Therefore, the biological information detection device 1 reliably improves the estimation accuracy of the biological information as compared with the case where the process noise v is determined only by the information on the vehicle side regardless of the biological information of the occupant M as in the prior art. can do.

Also, based on the frequency analysis result of the frequency analysis unit 32 that performs frequency analysis on the output signal S output by the biological information sensor 10 and the acceleration detected by the motion state quantity detection unit 20, respectively. The biological information frequency deriving unit 34 further includes a subtracting unit 33 that subtracts the intensity of the acceleration from the intensity of the output signal S, and the biological information frequency deriving unit 34 determines the frequency at which the subtraction result of the subtracting unit 33 is equal to or higher than a predetermined determination value. It derives as frequency fb.
According to this, based on the analysis result of the frequency analysis unit 32, the subtraction unit 33 subtracts the acceleration amplitude Aa, which is the strength of the acceleration signal A, from the biological amplitude Sa, which is the strength of the output signal S, for each predetermined frequency. The biological information frequency deriving unit 34 derives the biological information frequency fb based on the subtraction result. Therefore, the living body information detection device 1 can set the living body information frequency fb, which is the frequency of the living body information of the occupant M, relatively easily.

The biological information is the breathing of the occupant M.
According to this, since there is an individual difference in the breathing of the occupant M, the process noise v is determined to correspond to the individual difference of the occupant M with certainty. Therefore, the living body information detection apparatus 1 can estimate living body information so as to correspond reliably to the individual difference of the occupant M. In addition, the breathing of the occupant M has periodic fluctuation. Even in such a case, since the process noise determination unit 35 determines the process noise v based on the output signal S from the biological information sensor 10 and the acceleration signal A from the motion state quantity detection unit 20, the process noise v is , According to the fluctuation of the breathing cycle of the occupant M. Therefore, the biological information detection device 1 can estimate the biological information so as to correspond to the individual difference of the occupant M and the fluctuation of the respiration, and therefore, compared to the case where the biological information is estimated using only the information on the vehicle side. The estimation accuracy of biological information can be more reliably improved.

In the embodiment described above, an example of the biological information detection apparatus is shown, but the present invention is not limited to this, and other configurations can be adopted. For example, in each of the embodiments described above, the motion state amount is the acceleration of the vehicle, but instead, it may be the velocity of the vehicle.
Also, in the embodiment described above, the Kalman filter is a non-linear system, but may be a linear system instead.

In the above-described embodiment, the process noise v determined by the process noise determination unit 35 is used as an initial value of the process noise v _k in the Kalman filter algorithm, but instead, for example, a predetermined time Ts or more The process noise v determined by the process noise determination unit 35 may be used for each update time set to the length of. Specifically, after setting the initial value of the process noise v, the process noise determination unit 35 determines the process noise v at every update time, and updates the process noise v _k in the Kalman filter algorithm using the process noise v Do.
In addition, the numbers and positions of the living body information sensor 10 and the motion state quantity detection unit 20, directions of detectable acceleration and the like may be changed without departing from the scope of the present invention.

  DESCRIPTION OF SYMBOLS 1 ... biological information detection apparatus, 10 ... biological information sensor, 20 ... motion state quantity detection part, 30 ... control apparatus, 31 ... vibration noise removal part, storage part, 32 ... frequency analysis part, 33 ... subtraction part, 34 ... living body Information frequency derivation unit, 35: Process noise determination unit, 36: Biological information estimation unit, 40: Output unit, A: Acceleration signal, fb: Biological information frequency, M: Crew, S: Output signal, v: Process noise.

Claims (3)

A biological information sensor attached to a mobile body, detecting an occupant's biological information, and outputting an output signal according to the biological information;
A motion state amount detection unit that detects a motion state amount of the moving body;
A process noise determination unit that determines process noise in a state equation of a Kalman filter based on the output signal output by the biological information sensor and the movement state amount detected by the movement state amount detection unit;
And a biological information estimation unit for estimating the biological information by a Kalman filter using the process noise determined by the process noise determination unit and using the output signal output from the biological information sensor as an observation value. Biological information detection device. A biological information frequency deriving unit that derives a biological information frequency that is a frequency of the biological information based on the output signal output by the biological information sensor and the exercise state amount detected by the exercise state amount detection unit;
And a process noise map indicating correlation between the process noise and the biological information frequency,
The biological information detection apparatus according to claim 1, wherein the process noise determination unit determines the process noise from the biological information frequency derived by the biological information frequency deriving unit based on the process noise map. Based on the frequency analysis result of frequency analysis performed on the output signal output by the biological information sensor and the motion state amount detected by the motion state amount detection unit, and the frequency analysis result of the frequency analysis unit And a subtracting unit that subtracts the strength of the motional state quantity from the strength of the output signal;
The biological information detection apparatus according to claim 2, wherein the biological information frequency deriving unit derives a frequency at which the subtraction result of the subtracting unit is equal to or higher than a predetermined determination value as the biological information frequency.

Images