Biological information acquisition device and biological information acquisition method

Abstract

To realize a biological information acquisition device and biological information acquisition method that can improve the acquisition accuracy of biological information.SOLUTION: A biological information acquisition device includes: non-contact type observation sensors (first observation sensor 2_1 and second observation sensor 2_2) that detect displacements of multiple portions of a biological information acquisition object; a reference sensor 3 that detects vibrations in the environment where at least the biological information acquisition object exists; and a signal processing unit 6 that generates a time-frequency mask from a reference signal acquired by the reference sensor 3 and generates a separation signal obtained by separating vibration components from observation signals acquired by the observation sensors (first observation sensor 2_1 and second observation sensor 2_2) on the basis of the time-frequency mask.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a biological information acquisition device and a biological information acquisition method. [Background technology]

[0002] For example, a non-contact biosensor such as a radio wave sensor is disclosed to detect occupants in the driver's seat, passenger seat, and rear seat of a vehicle, and in the driver's seat and other areas, it not only detects occupants but also acquires biometric information (for example, Patent Document 1). In the occupant state detection system of Patent Document 1, a radio wave sensor installed inside the vehicle transmits radio waves, and the reflected waves are received to detect the distance from the reflecting object. The occupant state detection system checks whether the detected distance is changing by calculating the distance change from the distance from the reflecting object detected over time. As a result, if there is no distance change or the distance change is below a certain detection value, it is determined that there is no person (occupant) inside the vehicle. Also, if the distance is changing or the distance change is above a certain value, it is determined that there is a person inside the vehicle. However, if the radio wave sensor vibrates, or if an object such as a person vibrates, the radio wave sensor signal output by the radio wave sensor will contain vibration components caused by the vibration. If vibration components are included in the radio wave sensor signal, the detection accuracy of the signal processing system may decrease.

[0003] For example, a signal processing system and a sensor system are disclosed that can improve the accuracy of detecting the state of an object by attenuating the vibration component from a radio wave sensor signal (for example, Patent Document 2). The signal processing system of Patent Document 2 comprises a first receiving unit, a second receiving unit, and a signal processing unit. The first receiving unit receives a radio wave sensor signal from a radio wave sensor that receives radio waves reflected by an object. The second receiving unit receives a vibration sensor signal from a vibration sensor corresponding to the vibration of at least one of the radio wave sensor and the object. The signal processing unit detects information about the state of the object based on the radio wave sensor signal and the vibration sensor signal. [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2018-202921 [Patent Document 2] Japanese Patent Application Laid-Open No. 2021-71326 [Summary of the Invention] [Problems to be Solved by the Invention]

[0005] It is conceivable to apply time-frequency masking, which is a signal processing technique for the acoustic field, to the sensing field and separate unnecessary vibration components from the observed signal to obtain biological information. Time-frequency masking can separate the target signal component and the unnecessary vibration component with high precision, but artificial distortion peculiar to non-linear processing may occur, and the acquisition accuracy of biological information may decrease.

[0006] The present disclosure has been made in view of the above, and an object thereof is to realize a biological information acquisition device and a biological information acquisition method capable of improving the acquisition accuracy of biological information. [Means for Solving the Problems]

[0007] A biological information acquisition device according to one aspect of the present disclosure includes a non-contact observation sensor that detects displacements of a plurality of sites of a biological information acquisition target, a reference sensor that detects vibrations in an environment where at least the biological information acquisition target exists, and a signal processing unit that generates a time-frequency mask from a reference signal acquired by the reference sensor and generates a separated signal obtained by separating vibration components from an observation signal acquired by the observation sensor based on the time-frequency mask.

[0008] In this configuration, by applying a time-frequency mask as a signal source model for blind signal source separation based on time-frequency masking, linear (less distorted) signal source separation becomes possible. Thereby, the acquisition accuracy of biological information can be improved.

[0009] A biological information acquisition method according to one aspect of this disclosure includes: an observation signal acquisition step of detecting the displacement of multiple parts of a biological information acquisition target and acquiring an observation signal; a reference signal acquisition step of detecting vibrations in the environment in which the biological information acquisition target exists and acquiring a reference signal; a time frequency mask generation step of generating a time frequency mask from the reference signal; and a vibration component separation step of generating a separated signal by separating the vibration component from the observation signal based on the time frequency mask.

[0010] In this configuration, applying a time-frequency mask as the signal source model for blind source separation based on time-frequency masking enables linear (low-distortion) signal source separation. This improves the accuracy of acquiring biological information. [Effects of the Invention]

[0011] According to this disclosure, it is possible to realize a biological information acquisition device and a biological information acquisition method that can improve the accuracy of acquiring biological information. [Brief explanation of the drawing]

[0012] [Figure 1] Figure 1 is a block diagram showing the schematic configuration of a biological information acquisition device according to Embodiment 1. [Figure 2] Figure 2 is a flowchart showing an example of the biometric information acquisition process according to Embodiment 1. [Figure 3] Figure 3 is a schematic diagram illustrating the concept of independent component analysis. [Figure 4] Figure 4 shows an algorithm that illustrates a specific example of TFMBSS. [Figure 5A] Figure 5A is a subflowchart showing an example of the time-frequency mask generation process. [Figure 5B] Figure 5B is a subflowchart showing a first modified example of the time-frequency mask generation process. [Figure 5C] Figure 5C is a subflowchart showing a second modified example of the time-frequency mask generation process. [Figure 5D]Figure 5D is a subflowchart showing a third modified example of the time-frequency mask generation process. [Figure 6] Figure 6 is a flowchart showing an example of the biometric information acquisition process according to Embodiment 2. [Figure 7] Figure 7 is a block diagram showing the schematic configuration of a biological information acquisition device according to Embodiment 3. [Figure 8A] Figure 8A shows a first example of the antenna configuration of the observation sensor according to Embodiment 3. [Figure 8B] Figure 8B shows a second example of the antenna configuration of the observation sensor according to Embodiment 3. [Modes for carrying out the invention]

[0013] A biological information acquisition device and a biological information acquisition method according to an embodiment will be described in detail below with reference to the drawings. However, this disclosure is not limited by this embodiment.

[0014] (Embodiment 1) Figure 1 is a block diagram illustrating the schematic configuration of a biological information acquisition device according to Embodiment 1. As shown in Figure 1, the biological information acquisition device 1 comprises a first observation sensor 2_1, a second observation sensor 2_2, a reference sensor 3, a first observation signal acquisition unit 4_1, a second observation signal acquisition unit 4_2, a reference signal acquisition unit 5, a signal processing unit 6, and a data storage unit 7. In this disclosure, the biological information acquisition device 1 is, for example, a driver monitoring system (DMS) installed in a vehicle.

[0015] The first observation sensor 2_1 and the second observation sensor 2_2 are so-called non-contact biosensors that transmit waves to different parts of the driver (hereinafter also simply referred to as the "target") whose biometric information is to be acquired, and detect displacement at each part based on the reflected waves of the transmitted waves. Examples of locations where the first observation sensor 2_1 and the second observation sensor 2_2 are installed include, for example, on the rearview mirror or inside the seat of the vehicle on which the biometric information acquisition device 1 is mounted.

[0016] For example, when the first observation sensor 2_1 and the second observation sensor 2_2 are installed inside the seat, one of the first observation sensor 2_1 and the second observation sensor 2_2 may detect the displacement of a part corresponding to the subject's back, and the other may detect the displacement of a part corresponding to the subject's buttocks. The parts where the first observation sensor 2_1 and the second observation sensor 2_2 detect displacement may be, for example, parts that are about 5 cm apart from each other.

[0017] More specifically, the first observation sensor 2_1 and the second observation sensor 2_2 are, for example, radio wave sensors. The first observation sensor 2_1 and the second observation sensor 2_2 generate, for example, a continuous wave (CW) electromagnetic wave as a transmitted wave and irradiate it toward the target. For example, in a configuration in which the Doppler method is adopted as the modulation method for the electromagnetic wave by the first observation sensor 2_1 and the second observation sensor 2_2, the first observation sensor 2_1 and the second observation sensor 2_2 output a signal corresponding to the difference in frequency between the transmitted and received transmitted wave and the reflected wave to the first observation signal acquisition unit 4_1 and the second observation signal acquisition unit 4_2. The electromagnetic wave may be, for example, a millimeter wave or a microwave.

[0018] Furthermore, the first observation sensor 2_1 and the second observation sensor 2_2 are not limited to Doppler radar. For example, the first observation sensor 2_1 and the second observation sensor 2_2 may be FMCW (Frequency Modulated Continuous Wave) radar, UWB (Ultra Wide Band) radar, pulse radar, etc. Also, the first observation sensor 2_1 and the second observation sensor 2_2 are not limited to emitting electromagnetic waves as transmission waves. The transmission waves handled by the first observation sensor 2_1 and the second observation sensor 2_2 can be any wave that can measure the displacement of the target, and include a wide range of waves such as sound waves and light waves.

[0019] The first observation signal acquisition unit 4_1 outputs the signal acquired from the first observation sensor 2_1 as the first observation signal to the signal processing unit 6.

[0020] The second observation signal acquisition unit 4_2 outputs the signal acquired from the second observation sensor 2_2 as the second observation signal to the signal processing unit 6.

[0021] The reference sensor 3 is installed at a predetermined location within the vehicle and detects vehicle vibrations. Examples of locations where the reference sensor 3 may be installed include the dashboard of the vehicle on which the biometric information acquisition device 1 is installed, the airbag, or inside the seat.

[0022] More specifically, the reference sensor 3 is, for example, an acceleration sensor. The reference sensor 3 outputs a signal corresponding to the detected vibration to the reference signal acquisition unit 5.

[0023] The reference sensor 3 is not limited to a type that detects vibrations at the installation location, such as an acceleration sensor. For example, the reference sensor 3 may be a non-contact type sensor similar to the first observation sensor 2_1 and the second observation sensor 2_2. In this case, the reference sensor 3 transmits a wave toward a predetermined location inside the vehicle (for example, the dashboard of the vehicle on which the biometric information acquisition device 1 is installed) and outputs a signal corresponding to the frequency difference between the transmitted and received wave and the reflected wave to the reference signal acquisition unit 5. The reference sensor 3 only needs to be capable of detecting vibrations in the environment where the subject (biometric information acquisition target) is present (specifically, for example, vibrations of the vehicle on which the biometric information acquisition device 1 is installed).

[0024] The reference signal acquisition unit 5 outputs the signal acquired from the reference sensor 3 as a reference signal to the signal processing unit 6.

[0025] The biometric information acquisition process described later is implemented by a computer comprising, for example, a signal processing unit 6 exemplified by a central processing unit (CPU), a data storage unit 7 exemplified by a memory device such as RAM for storing data, and a program memory for storing programs. In this case, an example is provided in which the central processing unit (CPU) reads a program stored in the program memory to realize the functions of the signal processing unit 6.

[0026] Alternatively, the signal processing unit 6 can be implemented by software control processing of a microcomputer, by hardware configuration of an electronic circuit, or by both software control processing of a microcomputer and hardware configuration of an electronic circuit. Specifically, the signal processing unit 6 may be configured as an IC (integrated circuit), for example.

[0027] In this disclosure, the signal processing unit 6 applies blind source separation (BBS), a signal processing technique for the acoustic field, and separates unwanted vibration components from the observed signal using time-frequency masking-based BSS (TFMBSS).

[0028] In BBS, at least the same number of observed signals as the number of signal sources N to be separated are required, but in TFMBSS, the number of signal sources N and the number of observed signals must be the same. In this disclosure, the displacement of the subject's body surface is defined as the first signal source, and the vibration component is defined as the second signal source. That is, in this disclosure, the number of signal sources N is 2 (N=2), and at least two observation sensors are required. In other words, the number of observation sensors may be greater than the number of signal sources N. In this case, for example, displacements of more parts than the number of signal sources N can be acquired by observation sensors corresponding to each part, and principal component analysis of the signals acquired by each observation sensor can be performed to generate the same number of observed signals as the number of signal sources N.

[0029] Figure 2 is a flowchart showing an example of the biometric information acquisition process according to Embodiment 1.

[0030] In the biological information acquisition process according to Embodiment 1 shown in Figure 2, the signal processing unit 6 acquires a first observation signal, a second observation signal, and a reference signal for a predetermined period and converts them into digital signals, and sequentially stores the first observation signal, the second observation signal, and the reference signal for the predetermined period that have been converted into digital signals into the data storage unit 7 (step S100).

[0031] The signal processing unit 6 reads out the first observation signal, the second observation signal, and the reference signal stored in the data storage unit 7 and converts them into time-frequency domain signals. Specifically, the signal processing unit 6 performs a Short-Time Fourier Transform (STFT) on each signal to generate a complex spectrogram (step S200).

[0032] In this disclosure, the number of frequency bins I and the number of time frames (hereinafter also simply referred to as "number of frames") J of the complex spectrogram of each signal after STFT processing are assumed to be the same. In this disclosure, the number of channels M, which indicates the number of observed signals, is 2 (M=2). The first observed signal acquired by the first observation sensor 2_1 can also be referred to as the "first channel signal". The second observed signal acquired by the second observation sensor 2_2 can also be referred to as the "second channel signal".

[0033] BBS is a technique for estimating the separated signals from a mixed observation signal. BBS employs methods based on the assumption of statistical independence of the signal sources, such as Independent Component Analysis (ICA). The separation accuracy in Independent Component Analysis is characterized by the validity of the signal source model provided as prior information. Figure 3 is a schematic diagram illustrating the concept of Independent Component Analysis.

[0034] Observation signal x acquired by the first observation sensor 2_1 and the second observation sensor 2_2 ij and complex spectrogram X m This is shown by equation (1) below. In equation (1) below, i represents the frequency index (i is an integer from 1 to I), j represents the frame index (j is an integer from 1 to J), and m represents the channel index (m is an integer from 1 to M). Complex spectrogram X m The elements of the matrix that show this are x ijm It is said that...

[0035]

number

[0036] Observed signal x ij contains an unknown signal source s ij and a complex spectrogram S n is represented by the following equation (2). In equation (2) below, n represents the signal source index (n is an integer from 1 to N). The elements of the matrix representing the complex spectrogram S n are s ijn .

[0037]

Equation

[0038] Separation signal y ij and a complex spectrogram Y n are represented by the following equation (3). The elements of the matrix representing the complex spectrogram Y n are y ijn .

[0039]

Equation

[0040] Observed signal x ij is regarded as a mixed signal of a plurality of unknown signal sources s ij . The observed signal xij can be expressed by the following equation (5) using the complex mixing matrix A i shown in the following equation (4). The complex mixing matrix A<​​​​​​​​​​​​​​​​​​​​​​​​

[0043] The number of channels M and the number of signal sources N are equal (M=N), and the complex mixture matrix A i When is an invertible matrix, the separating signal y ij The separable matrix W shown in equation (6) below is the separable matrix W. i (=A i -1 Using ), it can be expressed by equation (7) below. Separable matrix W i is the separation vector w in (=(w in1 ,···,w inM ) T It consists of ).

[0044]

number

[0045]

number

[0046] The signal processing unit 6 uses TFMBSS to determine the separation matrix W for all frequency bins. i This is estimated. Figure 4 shows an algorithm that illustrates a specific example of TFMBSS. In the algorithm shown in Figure 4, the number of iterations in TFMBSS is denoted as K.

[0047] Furthermore, in the algorithm shown in Figure 4, X is the complex spectrogram of the observed signal X1,···,X M It is a complex matrix composed of each element, and is expressed by equations (8) to (11) below. Also, w is the separation matrix W1,···W for all frequency bins. I This is a complex vector obtained by vectorizing it, and is represented as an I × N × M x 1 column vector as shown in equation (12) below.

[0048]

number

[0049]

number

[0050]

number

[0051]

number

[0052]

number

[0053] The χ shown in equation (10) above i This is a complex spectrogram X1,···,X M Each element x ijm This shows a J x M matrix formed by arranging the elements. The matrix χ shown in equation (10) above. i The block diagonal matrix operator (equation (9) above), obtained by repeating this process M times, is further block diagonalized to obtain the complex matrix X (equation (8) above).

[0054] The function M(z) shown in Figure 4 represents a time-frequency mask that separates biological information from vibrational components, with the intermediate variable z as an argument. The signal processing unit 6 generates the function M(z) representing the time-frequency mask from the complex spectrogram of the reference signal (step S300). Figure 5A is a subflowchart showing an example of the time-frequency mask generation process.

[0055] The complex spectrogram R of the reference signal acquired by the reference sensor 3 is given by equation (13) below. The elements of the matrix representing the complex spectrogram R are r ij It is said that...

[0056]

number

[0057] The signal processing unit 6 processes each element r of the complex spectrogram R of the reference signal. ij The absolute value is applied (step S321), and furthermore, each element of the complex spectrogram after the absolute value application |r ij The | is normalized (step S324), and a matrix representing the complex spectrogram after normalization is generated as an oscillatory component mask (step S326). Specifically, the signal processing unit 6 processes each element |r of the complex spectrogram after absolute value processing. ij Extract the maximum value rmax of |, and then use this maximum value rmax to perform absolute value processing on each element of the complex spectrogram |r ij Divide by |. Each element m of the complex spectrogram showing the vibrational component mask. ij This can be expressed by equation (14) below. The complex spectrogram M2, which shows the vibrational component mask, is given by equation (15) below.

[0058]

number

[0059]

number

[0060] Next, the signal processing unit 6 processes each element m of the complex spectrogram of the vibration component mask. ij The complex spectrogram after the inversion process is reversed (step S327), and a matrix showing the complex spectrogram after the inversion process is generated as a displacement component mask (step S328). Specifically, the signal processing unit 6 processes each element m of the complex spectrogram of the vibration component mask. ij Subtract this from 1. The complex spectrogram M1, which shows the displacement component mask, is given by equation (16) below.

[0061]

number

[0062] The signal processing unit 6 then generates a function M(z) representing the time-frequency mask composed of each element of the complex spectrogram M1 of the displacement component mask and the complex spectrogram M2 of the vibration component mask (step S329), and applies this time-frequency mask function M(z) to the algorithm shown in Figure 4 to execute the iterative processing in TFMBSS (step S400, Figure 2). The time-frequency mask function M(z) is represented by an I × J × M x 1 vector, as shown in equation (17) below.

[0063]

number

[0064] TFMBSS applies the primal-dual splitting method, a type of proximity separation method, as its optimization algorithm, and further utilizes a time-frequency mask as a signal source model, taking an intermediate variable z as an argument to enhance separation.

[0065] In this disclosure, a time-frequency mask is generated from a reference signal acquired simultaneously with the observed signal, and the function M(z) representing this time-frequency mask is used as the signal source model in TFMBSS. This enables linear (low-distortion) signal source separation.

[0066] The iterative process in TFMBSS (step S400, Figure 2) yields the I × J × M x 1 vector y shown in equation (18) below.

[0067]

number

[0068] Each element of the vector y shown in equation (18) above ij1 This shows the elements of a complex spectrogram representing the time-frequency domain separated signal, in which at least the oscillation components contained in the reference signal have been separated from the observed signal.

[0069] Furthermore, in the time-frequency mask shown in equation (17) above, the element m of the complex spectrogram M1 of the displacement component mask ij1 And the elements m of the complex spectrogram M2 of the vibration component mask ij2 The function M(z) that represents the time-frequency mask when and are swapped is given by (19) below. In this case, each element of the vector y shown in equation (18) above y ij2 This will show each element of the complex spectrogram, which represents the separated signal in the time-frequency domain, where at least the oscillation components contained in the reference signal have been separated from the observed signal.

[0070]

number

[0071] Next, the signal processing unit 6 processes the time-frequency domain separated signal after TFMBSS processing, in other words, each element of the vector y shown in equation (18) above y ij1 The separated signal is shown in a complex spectrogram that includes (or each element of the vector y shown in equation (18) above) ij2 The separated signal (shown as a complex spectrogram including the TFMBSS) is converted into a time-domain signal. Specifically, the signal processing unit 6 performs an inverse short-time Fourier transform (ISTFT) on the time-frequency domain separated signal after TFMBSS processing to generate a time-domain separated signal (step S500).

[0072] The signal processing unit 6 then acquires vital signs such as the subject's heart rate, heart rate variability, respiratory rate, and respiratory depth as biological information from the time-domain separated signals after ISTFT processing (step S600). A detailed explanation of the method for acquiring vital signs is omitted here. This disclosure is not limited by the method for acquiring vital signs.

[0073] (First variation) Figure 5B is a subflowchart showing a first modified example of the time-frequency mask generation process. In the first modified example, the signal processing unit 6 uses equation (20) below to process each element |r of the complex spectrogram after absolute value processing. ij The frequency response correction process is performed (step S322).

[0074]

number

[0075] Specifically, in the first modified example, the signal processing unit 6 processes each element |r of the complex spectrogram after absolute value processing. ij | Correction factor c i Multiply by .

[0076] Equation (20) above shows each element of the complex spectrogram after frequency response correction processing, and the correction coefficient c for each frequency bin. i A column vector (c1,···,c) whose elements are I ) T It can be expressed as the element-wise product of the complex spectrogram of row I and column J after absolute value processing. Correction coefficient c i This can be estimated by solving the optimization problem shown in equation (21) below.

[0077]

number

[0078] The solution to the optimization problem shown in equation (21) above is derived by equation (22) below. Equation (22) below shows " * " represents the complex conjugate.

[0079]

number

[0080] Correction coefficient c i This can be expressed by equation (23) below as the optimal solution for which the derivative of the function J shown in equation (22) above is 0.

[0081]

number

[0082] In the biometric information acquisition process shown in Figure 2, the observed signal and reference signal acquired at the same time contain the same vibration components. By applying each element of the complex spectrogram of the observed signal and reference signal acquired in this biometric information acquisition process to equation (23) above, the correction coefficient c for each frequency bin is obtained. i It is possible to calculate this.

[0083] When the sensors that acquire the observation signal and the sensors that acquire the reference signal are different sensors, it is expected that the frequency characteristics of the signals acquired by each sensor will differ. Specifically, if the first observation sensor 2_1 and the second observation sensor 2_2 are radio wave sensors and the reference sensor 3 is an accelerometer, it is expected that the reference signal acquired by the accelerometer will have a stronger emphasis on high-frequency characteristics than the observation signal acquired by the radio wave sensor.

[0084] In the first modified example, a correction coefficient is used to compensate for the difference in frequency characteristics due to differences in sensors. This suppresses the difference in characteristics between the observed signal and the reference signal due to differences in sensors, thereby improving the signal separation accuracy of the subsequent TFMBSS processing.

[0085] Note that the correction coefficient c shown in equation (20) above i This may be a configuration that is pre-set based on the difference in frequency characteristics between sensors, or it may be a configuration in which the data is acquired before the execution of the biometric information acquisition process, stored in the data storage unit 7, and read from the data storage unit 7 and applied when the frequency characteristic correction process is executed in the biometric information acquisition process.

[0086] (Second variation) Figure 5C is a subflowchart showing a second modified example of the time-frequency mask generation process. In the second modified example, the signal processing unit 6 uses equation (24) below to process each element |r of the complex spectrogram after absolute value processing.ij The threshold processing is performed (step S323).

[0087]

number

[0088] Specifically, in the second modified example, the signal processing unit 6 processes each element |r of the complex spectrogram after absolute value processing. ij Compare | with the threshold rth, and the element |r that is less than the threshold rth ij Replace | with 0. In this disclosure, the threshold value rth is stored in the data storage unit 7 in advance.

[0089] Equation (24) above shows each element of the complex spectrogram after thresholding.

[0090] Depending on the installation location of the reference sensor 3, the reference signal may include a component of the subject's body surface displacement. Specifically, for example, the reference sensor 3 may be installed inside the seat where the subject (e.g., a driver) sits.

[0091] In the second modification, elements that do not meet the threshold rth are removed by the threshold processing described above. This suppresses the subject's body surface displacement component included in the reference signal, thereby improving the signal separation accuracy of the subsequent TFMBSS processing.

[0092] The thresholding process may also be performed on each element of the matrix representing the complex spectrogram after normalization. Furthermore, the thresholding process may be used in combination with the frequency response correction process (step S322) of the first modified example.

[0093] (Third variation) Figure 5D is a subflowchart showing a third modified example of the time-frequency mask generation process. In the third modified example, the signal processing unit 6 uses equation (25) below to process each element m of the matrix representing the complex spectrogram after normalization. ij Perform the power calculation (step S325).

[0094]

Number

[0095] Specifically, in the third modification example, the range in which the power value e can be taken is 0 < e < 1. In the present disclosure, the power value e is stored in the data storage unit 7 in advance.

[0096] The above equation (25) shows each element of the complex spectrogram after the power processing.

[0097] By the power processing, each element of the matrix showing the complex spectrogram becomes a value approaching 1. As a result, minute vibration components included in the reference signal after the normalization processing are emphasized, and the signal separation accuracy by the subsequent TFMBSS processing can be improved.

[0098] Note that the power processing may be used in combination with the frequency characteristic correction processing (step S322) of the first modification example and the threshold processing (step S323) of the second modification example.

[0099] (Embodiment 2) FIG. 6 is a flowchart showing an example of the biological information acquisition process according to Embodiment 2. The configuration of the biological information acquisition device according to Embodiment 2 is the same as the configuration of the biological information acquisition device according to Embodiment 1, and thus detailed description thereof is omitted here. Further, in Embodiment 2, the processes different from the biological information acquisition process according to Embodiment 1 will be described in detail, and the processes similar to the biological information acquisition process according to Embodiment 1 may be omitted from the description.

[0100] In the biological information acquisition process according to Embodiment 2, the signal processing unit 6 performs filtering processing on the first observation signal, the second observation signal, and the reference signal before the STFT processing (step S110).

[0101] The filtering process is implemented using digital filtering, such as IIR (Infinite Impulse Response) filters or FIR (Finite Impulse Response) filters. The number of taps (filter order) and coefficient values ​​for the filtering process are stored in the data storage unit 7 beforehand.

[0102] For example, if the subject's heart rate is used as the target of biometric information acquisition, the displacement component due to respiration, which is relatively large as a body surface displacement included in the observed signal, may mask the displacement component due to the heart rate, potentially reducing the acquisition accuracy.

[0103] In Embodiment 2, the filtering process described above removes components below 1 Hz, for example. This suppresses displacement components due to respiration (e.g., components around 0.2 Hz) and improves the accuracy of acquiring displacement components due to heartbeat (e.g., components above 1 Hz).

[0104] The filtering process may also be used in combination with the frequency characteristic correction process (step S322) of the first modified example of the time-frequency mask generation process in Embodiment 1 (step S322), the threshold processing (step S323) of the second modified example, and the power processing (step S325) of the third modified example.

[0105] (Embodiment 3) Figure 7 is a block diagram showing the schematic configuration of a biological information acquisition device according to Embodiment 3. In Embodiment 3, components that differ from the configuration of the biological information acquisition device according to Embodiments 1 and 2 will be described in detail, while components that are the same as those in the biological information acquisition device according to Embodiments 1 and 2 may be omitted from the description. Furthermore, since the biological information acquisition process according to Embodiment 3 is the same as the biological information acquisition process according to Embodiments 1 and 2, a detailed explanation will be omitted here.

[0106] The biological information acquisition device 1a according to Embodiment 3 is equipped with an observation sensor 2 in place of the first observation sensor 2_1 and the second observation sensor 2_2 of the biological information acquisition device 1, and is equipped with an observation signal acquisition unit 4 in place of the first observation signal acquisition unit 4_1 and the second observation signal acquisition unit 4_2.

[0107] The observation sensor 2 is, for example, a radio wave sensor. Figure 8A shows a first example of the antenna configuration of the observation sensor according to Embodiment 3. Figure 8B shows a second example of the antenna configuration of the observation sensor according to Embodiment 3.

[0108] In the first example of the antenna configuration shown in Figure 8A, the observation sensor 2 comprises at least two transmitting antennas Tx1 and Tx2 and one receiving antenna Rx. In the second example of the antenna configuration shown in Figure 8B, the observation sensor 2 comprises at least one transmitting antenna Tx and two receiving antennas Rx1 and Rx2.

[0109] The antenna configuration of the observation sensor 2 is not limited to the first example shown in Figure 8A and the second example shown in Figure 8B. For example, in the configuration of the first example shown in Figure 8A, there may be multiple transmitting antennas Tx1, Tx2 and receiving antenna Rx, and in the configuration of the second example shown in Figure 8B, there may be multiple transmitting antenna Tx and receiving antennas Rx1, Rx2.

[0110] In the biological information acquisition device 1a according to Embodiment 3, a single observation sensor 2 is configured to acquire first and second observation signals from different parts of a subject using various methods such as antenna diversity, analog beamforming, and digital beamforming. In Embodiment 3 as well, for example, it is possible to acquire displacements from more parts than the number of signal sources N using a single observation sensor, and generate the same number of observation signals as the number of signal sources N by performing principal component analysis of the signals corresponding to each part.

[0111] The embodiments described above are provided to facilitate understanding of this disclosure and are not intended to limit the invention. This disclosure may be modified or improved without departing from its spirit, and equivalents thereof are included.

[0112] This disclosure may take the following configuration, as described above, or alternatively.

[0113] (1) A biological information acquisition device according to one aspect of the present disclosure comprises: a non-contact observation sensor that detects the displacement of multiple parts of a biological information acquisition target; a reference sensor that detects vibrations in the environment in which at least the biological information acquisition target exists; and a signal processing unit that generates a time-frequency mask from a reference signal acquired by the reference sensor and generates a separated signal by separating the vibration component from the observation signal acquired by the observation sensor based on the time-frequency mask.

[0114] (2) In the biological information acquisition device described in (1) above, the signal processing unit applies the time-frequency mask as a signal source model for blind signal source separation based on time-frequency masking.

[0115] This configuration enables linear (low-distortion) signal source separation, improving the accuracy of acquiring biological information.

[0116] (3) In the biological information acquisition device described in (2) above, the signal processing unit converts the time-domain observation signal acquired by the observation sensor and the time-domain reference signal acquired by the reference sensor into time-frequency domain signals, generates the time-frequency mask from the time-frequency domain reference signal, and separates the vibration component from the time-frequency domain observation signal to generate the separated signal.

[0117] (4) In the biological information acquisition device described in (3) above, the signal processing unit generates an vibration component mask by performing absolute value processing and normalization processing on each element of the reference signal in the time-frequency domain, generates a displacement component mask by performing inversion processing on each element of the vibration component mask, and generates a function as the time-frequency mask composed of each element of the vibration component mask and the displacement component mask.

[0118] (5) In the biological information acquisition device described in (3) or (4) above, the signal processing unit performs frequency characteristic correction processing of the reference signal in the time-frequency domain.

[0119] In this configuration, differences in the characteristics of the observed signal and the reference signal due to differences in sensors are suppressed. This improves the signal separation accuracy of the subsequent TFMBSS processing.

[0120] (6) In the biological information acquisition device described in (3) to (5) above, the signal processing unit performs a predetermined threshold processing on each element of the reference signal in the time-frequency domain.

[0121] In this configuration, the body surface displacement component of the biological information acquisition target included in the reference signal is suppressed. This improves the signal separation accuracy of the subsequent TFMBSS processing.

[0122] (7) In the biological information acquisition device described in (3) to (6) above, the signal processing unit performs a predetermined power-up process on each element of the time-frequency domain reference signal after the normalization process.

[0123] In this configuration, minute oscillation components in the reference signal are amplified. This improves the signal separation accuracy of the subsequent TFMBSS processing.

[0124] (8) In the biological information acquisition device described in (3) to (7) above, the signal processing unit performs a predetermined filtering process on the time-domain observation signal acquired by the observation sensor and the time-domain reference signal acquired by the reference sensor.

[0125] This configuration allows for improved accuracy in acquiring desired biological information.

[0126] (9) In the biological information acquisition device described in (3) to (8) above, the signal processing unit converts the time-frequency domain separated signal into a time-domain signal and acquires the biological information of the target of biological information acquisition from the time-domain separated signal.

[0127] (10) In the biological information acquisition device described in (1) to (9) above, the observation sensor includes a first observation sensor that detects the displacement of a first part of the biological information acquisition target, and a second observation sensor that detects the displacement of a second part different from the first part of the biological information acquisition target.

[0128] (11) In the biological information acquisition device described in (1) to (9) above, the observation sensor comprises at least two transmitting antennas and one receiving antenna, and uses one of the following methods to detect the displacement of the first part of the biological information acquisition target and the displacement of a second part different from the first part of the biological information acquisition target.

[0129] (12) In the biological information acquisition device described in (1) to (9) above, the observation sensor comprises at least one transmitting antenna and two receiving antennas, and uses one of the following methods to detect the displacement of the first part of the biological information acquisition target and the displacement of a second part of the biological information acquisition target that is different from the first part.

[0130] (13) A method for acquiring biological information according to one aspect of the present disclosure includes: a displacement detection step for detecting the displacement of a plurality of parts of a biological information acquisition target; a vibration detection step for detecting vibrations in the environment in which at least the biological information acquisition target exists; a time-frequency mask generation step for generating a time-frequency mask from a reference signal acquired in the vibration detection step; and a separation signal generation step for generating a separation signal by separating vibration components from an observation signal acquired in the displacement detection step based on the time-frequency mask.

[0131] (14) In the method for acquiring biological information described in (13) above, in the separation signal generation step, the time-frequency mask is applied as a signal source model for blind signal source separation based on time-frequency masking.

[0132] This configuration enables linear (low-distortion) signal source separation, improving the accuracy of acquiring biological information.

[0133] (15) The method for acquiring biological information described in (14) above further comprises a time-frequency domain signal generation step in which the observed signal acquired in the displacement detection step and the reference signal acquired in the vibration detection step are converted into signals in the time-frequency domain, wherein in the time-frequency mask generation step a time-frequency mask is generated from the reference signal in the time-frequency domain, and in the separation signal generation step vibration components are separated from the observed signal in the time-frequency domain to generate the separation signal.

[0134] (16) In the method for acquiring biological information described in (15) above, the time-frequency mask generation step includes: a vibration component mask generation step that generates a vibration component mask by performing absolute value processing and normalization processing on each element of a reference signal in the time-frequency domain; a displacement component mask generation step that generates a displacement component mask by performing inversion processing on each element of the vibration component mask; and a function generation step that generates a function composed of each element of the vibration component mask and the displacement component mask as the time-frequency mask.

[0135] (17) In the method for acquiring biological information described in (15) or (16) above, the time-frequency mask generation step is performed to correct the frequency characteristics of the reference signal in the time-frequency domain.

[0136] In this configuration, differences in the characteristics of the observed signal and the reference signal due to differences in sensors are suppressed. This improves the signal separation accuracy in the vibration component separation step.

[0137] (18) In the biological information acquisition method described in (15) to (17) above, in the time-frequency mask generation step, a predetermined thresholding process is performed on each element of the reference signal in the time-frequency domain.

[0138] In this configuration, the body surface displacement component of the biological information acquisition target included in the reference signal is suppressed. This improves the signal separation accuracy in the vibration component separation step.

[0139] (19) In the biological information acquisition method described in (15) to (18) above, in the time-frequency mask generation step, a predetermined power-up process is performed on each element of the reference signal in the time-frequency domain after the normalization process.

[0140] In this configuration, minute vibrational components contained in the reference signal are amplified. This improves the signal separation accuracy in the vibrational component separation step.

[0141] (20) In the biological information acquisition method described in (15) to (19) above, a predetermined filtering process is performed on the time-domain observation signal acquired in the displacement detection step and the time-domain reference signal acquired in the vibration detection step.

[0142] This configuration allows for improved accuracy in acquiring desired biological information.

[0143] (21) The biological information acquisition method described in (13) to (20) above further comprises a time-domain signal generation step of converting a time-frequency domain separated signal into a time-domain signal, and a biological information acquisition step of acquiring the biological information to be acquired from the time-domain separated signal.

[0144] This disclosure enables the realization of a biological information acquisition device and a biological information acquisition method that can improve the accuracy of acquiring biological information. [Explanation of symbols]

[0145] 1,1a Biometric information acquisition device 2 Observation sensors 2_1 First observation sensor 2_2 Second Observation Sensor 3. Reference Sensor 4. Observation signal acquisition unit 4_1 First Observation Signal Acquisition Unit 4_2 Second Observation Signal Acquisition Unit 5 Reference signal acquisition section 6. Signal Processing Unit 7. Data Storage Unit

Claims

1. A non-contact observation sensor that detects the displacement of multiple body parts targeted for biometric information acquisition, A reference sensor that detects vibrations in the environment where the object for acquiring biological information is located, A signal processing unit generates a time-frequency mask from a reference signal acquired by the reference sensor, and generates a separated signal by separating the vibration component from the observation signal acquired by the observation sensor based on the time-frequency mask. Equipped with, The aforementioned observation sensor is A first observation sensor that detects the displacement of the first part of the biological information acquisition target, A second observation sensor that detects the displacement of a second part different from the first part of the biological information acquisition target, Includes, The signal processing unit, As a signal source model for blind source separation based on time-frequency masking, the time-frequency mask is applied, The time-domain observation signals acquired by the first observation sensor and the second observation sensor, and the time-domain reference signal acquired by the reference sensor are converted into time-frequency domain signals. The time-frequency mask is generated from a reference signal in the time-frequency domain. The oscillatory component is separated from the observed signal in the time-frequency domain to generate the separated signal. A device for acquiring biological information.

2. A non-contact observation sensor that detects the displacement of multiple body parts targeted for biometric information acquisition, A reference sensor that detects vibrations in the environment where the object for acquiring biological information is located, A signal processing unit generates a time-frequency mask from a reference signal acquired by the reference sensor, and generates a separated signal by separating the vibration component from the observation signal acquired by the observation sensor based on the time-frequency mask. Equipped with, The aforementioned observation sensor is It comprises at least two transmitting antennas and one receiving antenna, Using one of the following methods: antenna diversity, analog beamforming, or digital beamforming, the displacement of the first part of the biological information acquisition target and the displacement of a second part different from the first part of the biological information acquisition target are detected. The signal processing unit, As a signal source model for blind source separation based on time-frequency masking, the time-frequency mask is applied, The time-domain observation signal acquired by the observation sensor and the time-domain reference signal acquired by the reference sensor are converted into time-frequency domain signals. The time-frequency mask is generated from a reference signal in the time-frequency domain. The oscillatory component is separated from the observed signal in the time-frequency domain to generate the separated signal. A device for acquiring biological information.

3. A non-contact observation sensor that detects the displacement of multiple body parts targeted for biometric information acquisition, A reference sensor that detects vibrations in the environment where the object for acquiring biological information is located, A signal processing unit generates a time-frequency mask from a reference signal acquired by the reference sensor, and generates a separated signal by separating the vibration component from the observation signal acquired by the observation sensor based on the time-frequency mask. Equipped with, The aforementioned observation sensor is It comprises at least one transmitting antenna and two receiving antennas, Using one of the following methods: antenna diversity, analog beamforming, or digital beamforming, the displacement of the first part of the biological information acquisition target and the displacement of a second part different from the first part of the biological information acquisition target are detected. The signal processing unit, As a signal source model for blind source separation based on time-frequency masking, the time-frequency mask is applied, The time-domain observation signal acquired by the observation sensor and the time-domain reference signal acquired by the reference sensor are converted into time-frequency domain signals. The time-frequency mask is generated from a reference signal in the time-frequency domain. The oscillatory component is separated from the observed signal in the time-frequency domain to generate the separated signal. A device for acquiring biological information.

4. A biological information acquisition device according to any one of claims 1 to 3, The signal processing unit, An absolute value processing and normalization processing are performed on each element of the reference signal in the time-frequency domain to generate an oscillatory component mask. The displacement component mask is generated by performing an inversion process on each element of the vibration component mask. As the time-frequency mask, a function is generated that is composed of the elements of the displacement component mask and the vibration component mask. A device for acquiring biological information.

5. A biological information acquisition device according to any one of claims 1 to 3, The signal processing unit performs frequency characteristic correction processing on the reference signal in the time-frequency domain. A device for acquiring biological information.

6. A biological information acquisition device according to any one of claims 1 to 3, The signal processing unit performs a predetermined thresholding process on each element of the reference signal in the time-frequency domain. A device for acquiring biological information.

7. A biological information acquisition device according to any one of claims 1 to 3, The signal processing unit performs a predetermined power-up operation on each element of the time-frequency domain reference signal after normalization. A device for acquiring biological information.

8. A biological information acquisition device according to any one of claims 1 to 3, The signal processing unit performs a predetermined filtering process on the time-domain observation signal acquired by the observation sensor and the time-domain reference signal acquired by the reference sensor. A device for acquiring biological information.

9. A biological information acquisition device according to any one of claims 1 to 3, The signal processing unit, Convert a time-frequency separated signal into a time-domain signal, From the time-domain separated signal, the biological information of the target of biological information acquisition is obtained. A device for acquiring biological information.

10. A displacement detection step that detects the displacement of multiple body parts from which biological information is to be acquired, A vibration detection step that detects vibrations in the environment in which the object for acquiring biological information exists, A time-frequency mask generation step, which generates a time-frequency mask from a reference signal acquired in the vibration detection step, A separation signal generation step generates a separation signal obtained by separating the vibration component from the observation signal acquired in the displacement detection step based on the time-frequency mask, It has, In the displacement detection step, the displacement of the first part of the body from which biological information is to be acquired, and the displacement of a second part of the body from which biological information is to be acquired, different from the first part, In the separated signal generation step, the time-frequency mask is applied as the signal source model for blind signal source separation based on time-frequency masking. The system further includes a time-frequency domain signal generation step that converts the observation signal acquired in the displacement detection step and the reference signal acquired in the vibration detection step into a time-frequency domain signal, In the time-frequency mask generation step, a time-frequency mask is generated from a reference signal in the time-frequency domain. In the separated signal generation step, the oscillation component is separated from the observed signal in the time-frequency domain to generate the separated signal. Methods for acquiring biometric information.

11. A method for acquiring biological information according to claim 10, The aforementioned time-frequency mask generation step is: A vibration component mask generation step involves generating a vibration component mask by performing absolute value processing and normalization processing on each element of a reference signal in the time-frequency domain, A displacement component mask generation step is performed to generate a displacement component mask by performing an inversion process on each element of the vibration component mask, A function generation step of generating a function composed of the elements of the displacement component mask and the vibration component mask as the time-frequency mask, including, Methods for acquiring biometric information.

12. A method for acquiring biological information according to claim 10, In the time-frequency mask generation step, a frequency characteristic correction process is performed on the reference signal in the time-frequency domain. Methods for acquiring biometric information.

13. A method for acquiring biological information according to claim 10, In the time-frequency mask generation step, a predetermined thresholding process is performed on each element of the reference signal in the time-frequency domain. Methods for acquiring biometric information.

14. A method for acquiring biological information according to claim 10, In the time-frequency mask generation step, a predetermined power-up process is performed on each element of the reference signal in the time-frequency domain after the normalization process. Methods for acquiring biometric information.

15. A method for acquiring biological information according to claim 10, A predetermined filtering process is performed on the time-domain observation signal acquired in the displacement detection step and the time-domain reference signal acquired in the vibration detection step. Methods for acquiring biometric information.

16. A method for acquiring biological information according to any one of claims 10 to 15, A time-domain signal generation step that converts a time-frequency domain separated signal into a time-domain signal, A biological information acquisition step in which biological information to be acquired is acquired from a time-domain separated signal, It further possesses, Methods for acquiring biometric information.

Images