Estimation device, estimation method, and program recording medium

Abstract

The sensor (1) includes: a complex transfer function calculation unit (30) that uses received signals of radio waves transmitted from N transmitting antenna elements to the space where one or more organisms exist and received by M receiving antenna elements to calculate a complex transfer function representing the propagation characteristics between the transmitting antenna elements and the receiving antenna elements, where N is a natural number of 2 or more and M is a natural number of 2 or more; a spectrum calculation unit (70) that (a) uses each of a plurality of different values ​​as the number of organisms to calculate a likelihood spectrum representing the likelihood of the existence of the organisms derived from the organism information by using an estimation algorithm that estimates the existence of the organisms, where the organism information is the component of the complex transfer function corresponding to the organisms; (b) calculates a comprehensive spectrum by combining the calculated multiple likelihood spectra; and an estimation unit (80) that estimates and outputs organism information representing at least the number of organisms existing in the space based on the comprehensive spectrum.

Description

Technical Field

[0001] This disclosure relates to estimation apparatus, estimation methods, and procedures. Background Technology

[0002] Technology for detecting objects using wirelessly transmitted signals is being developed (see, for example, Patent Document 1).

[0003] Patent document 1 discloses the following technology: using Fourier transform to analyze the feature values ​​containing Doppler frequency shift components of a wirelessly received signal, thereby enabling the determination of the number or location of the organism being detected.

[0004] Prior art literature

[0005] Patent documents

[0006] Patent Document 1: Japanese Patent Application Publication No. 2015-117972

[0007] Patent Document 2: Japanese Patent Application Publication No. 2014-228291

[0008] Patent Document 3: Japanese Patent No. 5047002

[0009] Patent Document 4: Japanese Patent No. 5025170 Summary of the Invention

[0010] The problem that the invention aims to solve

[0011] The algorithm used to detect objects sometimes requires inputting the number of objects to be detected. In this case, the following problem arises: when the number of objects to be detected is unknown, it is impossible to detect them.

[0012] The purpose of this disclosure is to provide an estimation device, etc., that can estimate information related to an organism even when the number of organisms being detected is unknown.

[0013] Methods for solving problems

[0014] The estimation apparatus of this disclosure comprises: a complex transfer function calculation unit that calculates a complex transfer function representing the propagation characteristics between the transmitting antenna elements and the receiving antenna elements, using received signals of radio waves transmitted from N (N is a natural number of 2 or more) transmitting antenna elements to a space where one or more organisms exist and received by M (M is a natural number of 2 or more) receiving antenna elements; a spectrum calculation unit that (a) uses each of a plurality of different values ​​as the number of organisms, calculates a likelihood spectrum representing the likelihood of the existence of the organisms derived from organism information by using an estimation algorithm that estimates the existence of the organisms, the organism information being a component of the complex transfer function corresponding to the organisms, and (b) calculates a composite spectrum formed by combining the calculated plurality of likelihood spectra; and an estimation unit that estimates and outputs organism information representing at least the number of organisms existing in the space based on the composite spectrum.

[0015] Furthermore, these general or specific methods can also be implemented through systems, methods, integrated circuits, computer programs, or computer-readable recording media such as CD-ROMs, or through any combination of systems, methods, integrated circuits, computer programs, and recording media.

[0016] Invention Effects

[0017] According to the estimation device disclosed herein, information related to organisms can be estimated even when the number of organisms being detected is unknown. Attached Figure Description

[0018] Figure 1 This is a block diagram showing the configuration of the sensor in Implementation Method 1.

[0019] Figure 2 This is a conceptual diagram representing the sensor's estimation of the direction of arrival in Implementation 1.

[0020] Figure 3 This is a block diagram showing the configuration of the estimation unit in Implementation Method 1.

[0021] Figure 4 This is a conceptual diagram illustrating the operation of the peak exploration unit in Implementation Method 1.

[0022] Figure 5 This is a conceptual diagram illustrating the operation of the inspection unit in Implementation Method 1.

[0023] Figure 6 This is a flowchart illustrating the processing of the sensor in Implementation Method 1.

[0024] Figure 7 This is a flowchart illustrating the calculation and processing of human information by the sensor in Implementation Method 1.

[0025] Figure 8 This is a block diagram illustrating the structure of the estimation unit in Implementation Method 2.

[0026] Figure 9 This is a conceptual diagram illustrating the operation of the block detection unit in Implementation Method 2.

[0027] Figure 10 This is a block diagram illustrating the structure of the estimation unit in Implementation Method 3. Detailed Implementation

[0028] (The knowledge upon which this disclosure is based)

[0029] The inventors have discovered the following problems with the detection-related techniques described in the "Background Art" section.

[0030] Previously, technologies for detecting objects using wirelessly transmitted signals were being developed (see, for example, Patent Documents 1-4).

[0031] For example, Patent Document 1 discloses a technique that uses Fourier transform to analyze eigenvalues ​​containing Doppler frequency shift components, thereby estimating the number or location of people being detected. Specifically, the processing apparatus in Patent Document 1 performs a Fourier transform on the received signal, calculates the autocorrelation matrix of the waveform obtained by extracting specific frequency components, and performs eigenvalue decomposition on the autocorrelation matrix to obtain eigenvalues. Generally, eigenvalues ​​and eigenvectors each represent the propagation path of the radio wave from the transmitting antenna to the receiving antenna, i.e., one path. However, in the technique of Patent Document 1, components that do not contain biological information are removed. Therefore, only the path corresponding to the signal reflected by the biological body and the path corresponding to its secondary reflection and noise appear in the eigenvalues ​​and eigenvectors. Here, the value of the eigenvalue corresponding to noise is smaller than the value of the eigenvalue corresponding to the biological body. Therefore, by listing the number of eigenvalues ​​that are larger than a predetermined threshold, the number of biological bodies can be estimated.

[0032] However, in the technology disclosed in Patent Document 1, when the organisms being detected are located far from the detection device, or when there are many organisms, the following problems exist: the difference between the characteristic value corresponding to the organism and the characteristic value corresponding to the noise decreases, reducing the accuracy of the number of people estimated. This is because, when the Doppler effect is very weak, it is difficult to detect weak signals that have undergone Doppler frequency shift due to the influence of internal noise in the receiver, interference waves transmitted from outside the detection object, and the presence of objects other than the detection object that generate Doppler frequency shift. In addition, the organisms being measured have a certain size, and the components of the organisms are distributed across multiple characteristic values. Therefore, when there are many organisms, it is impossible to completely separate the characteristic values ​​of the organisms, making it difficult to estimate the number of people.

[0033] Patent Document 2 discloses a technique for estimating the position of an object using orientation estimation algorithms such as MUSIC (Multiple Signal Classification). Specifically, a receiving station that receives a signal transmitted from a transmitting station performs a Fourier transform on the received signal, calculates the autocorrelation matrix of the waveform obtained by extracting specific frequency components, and then applies an orientation estimation algorithm such as MUSIC. This allows for high-precision orientation estimation. However, the MUSIC method used in Patent Document 2 requires the number of organisms to be detected; therefore, when using the technique in Patent Document 2 for detection, the number of people needs to be estimated in advance.

[0034] In addition, for example, Patent Document 3 discloses a technique that estimates the number of arriving waves, or even the number of transmitters such as portable telephones, based on the correlation between the characteristic vector of the received signal received by multiple antennas and the steering vector of the range to which the radio waves may reach.

[0035] In addition, for example, Patent Document 4 discloses the following technique: various arrival wavenumbers are assumed for received signals received by multiple antennas, an evaluation function of the steering vector is used to calculate each arrival wavenumber, and the arrival wavenumber with the largest evaluation function is estimated as the true arrival wavenumber.

[0036] However, the technology disclosed in Patent Documents 3-4 is a technology for estimating the number of transmitters that emit radio waves, and cannot estimate the number of organisms.

[0037] Therefore, in view of the above circumstances, the inventors have invented an estimation device that can estimate the number of organisms more accurately and more extensively using wireless signals without requiring the organism to hold a special device such as a transmitter, and thus completed this disclosure.

[0038] One aspect of the estimation apparatus disclosed herein includes: a complex transfer function calculation unit that calculates a complex transfer function representing the propagation characteristics between the transmitting antenna elements and the receiving antenna elements, using received signals of radio waves transmitted from N (N being a natural number of 2 or more) transmitting antenna elements to a space where one or more organisms exist and received by M (M being a natural number of 2 or more) receiving antenna elements; a spectrum calculation unit that (a) uses each of a plurality of different values ​​as the number of organisms, calculates a likelihood spectrum representing the likelihood of the existence of the organisms derived from organism information by using an estimation algorithm that estimates the existence of the organisms, the organism information being a component of the complex transfer function corresponding to the organisms, and (b) calculates a composite spectrum formed by combining the calculated plurality of likelihood spectra; and an estimation unit that estimates and outputs organism information representing at least the number of organisms existing in the space based on the composite spectrum.

[0039] According to the above method, the estimation device uses a composite spectrum, calculated by combining multiple likelihood spectra that are different from each other and are used as the number of organisms to be detected, to output information related to the organisms present in space. Therefore, it is not necessary to input the number of organisms to be detected. Thus, the estimation device can estimate information related to organisms even when the number of organisms to be detected is unknown.

[0040] For example, the estimation unit may estimate and output the organism information, which also represents the location of organisms in the space, based on the comprehensive spectrum.

[0041] According to the above method, the estimation device estimates information not only representing the number of organisms but also their location, as information related to the organisms. Therefore, even when the number of organisms being detected is unknown, the estimation device can estimate more information related to the organisms.

[0042] For example, the spectral calculation unit may use multiple natural numbers less than (N × M-1), multiple natural numbers less than N, or multiple natural numbers less than M as the multiple values ​​to calculate the likelihood spectrum.

[0043] According to the above method, the estimation device calculates multiple likelihood spectra using at least one of the number of transmitting antenna elements and the number of receiving antenna elements. When using biological information with a complex transfer function, the biological information is determined with high accuracy when the estimated number of biological organisms is less than or equal to the product of the number of transmitting and receiving antenna elements; furthermore, the biological information is determined with even higher accuracy when the estimated number of biological organisms is less than or equal to either the number of transmitting or receiving antenna elements. Therefore, the estimation device can more easily and accurately estimate information related to biological organisms even when the number of organisms being detected is unknown.

[0044] For example, the spectral calculation unit may use multiple natural numbers as the multiple values ​​to calculate the likelihood spectrum, wherein each of the multiple natural numbers is less than or equal to a number determined as the maximum number of organisms that may exist in the space.

[0045] According to the above method, the estimation device calculates multiple likelihood spectra using a number determined as the maximum number of organisms that may exist in the space. The maximum number of organisms that may exist in the space is sometimes predetermined, for example, based on the size of the space (area or volume). In this case, it is assumed that a number of organisms below this maximum number exist in the space; in other words, it is not necessary to assume that a number of organisms exceeding this maximum number exist. Therefore, by calculating multiple likelihood spectra using multiple natural numbers, each less than or equal to the maximum number, the computational processing can be suppressed to a necessary and sufficient amount, and computational processing based on an assumed number of organisms exceeding the necessary amount can be avoided in advance. Thus, the estimation device, through necessary and sufficient computational processing, can estimate information related to organisms even when the number of organisms being detected is unknown.

[0046] For example, it may also include a storage unit that stores the organism information estimated by the estimation unit in the past, and the spectrum calculation unit uses a plurality of natural numbers within the range of the number of organisms represented by the organism information stored in the storage unit as the plurality of values ​​to calculate the likelihood spectrum.

[0047] According to the above method, the estimation device uses the number of organisms that have existed in the space in the past to calculate the likelihood spectrum. Therefore, it is easier to calculate multiple likelihood spectra in a space conceived as containing the same number of organisms as those that have existed in the space in the past. Thus, the estimation device can more easily estimate information related to organisms even when the number of organisms being detected is unknown.

[0048] For example, the estimation unit may obtain one or more maxima among the multiple maxima of the likelihood spectrum, where the maxima is the largest within a specified range including the maxima, and determine a first maxima among the obtained one or more maxima, wherein the first maxima is the maxima with the largest difference between it and the second maxima that follow the first maxima in descending order, and the estimation unit will estimate the number of organisms as the number of the determined first maxima that is the largest among the one or more maxima.

[0049] According to the above method, the estimation device can use a ratio method to exclude virtual image-based peaks from multiple peaks in the likelihood spectrum, outputting the number of organism-based peaks. The inventors of this application discovered that virtual image-based peaks in the likelihood spectrum have characteristics of low or relatively flat peak values. Based on this insight, they conceived of a technique using a ratio method to exclude virtual image-based peaks in the likelihood spectrum. The estimation device processes multiple peaks in the likelihood spectrum; in other words, it does not require setting a threshold for the likelihood, thus avoiding the impact of the threshold setting on the processing. Furthermore, since no machine learning model is used, preparation work such as preparing teacher data and prior learning processing is eliminated. Therefore, the estimation device can more easily estimate organism-related information even when the number of organisms being detected is unknown.

[0050] For example, the estimation unit may use only one or more of the third maximum values ​​among the one or more maximum values ​​as the one or more maximum values ​​to determine the first maximum value, wherein the difference between the third maximum value and the value obtained by multiplying the value contained in the specified range including the third maximum value by a specified ratio is above a threshold.

[0051] According to the above method, the estimation device can more appropriately exclude virtual image-based peaks among the peaks in the likelihood spectrum. Virtual image-based peaks in the likelihood spectrum are relatively flat, and therefore can be distinguished based on the magnitude of the difference between the maximum value and the value obtained by multiplying the value within a specified range containing that maximum value by a specified ratio. Thus, by excluding the influence of virtual images, the estimation device can more easily estimate information related to the organism even when the number of organisms being detected is unknown.

[0052] For example, the estimation unit may estimate the number of intervals in the likelihood spectrum with a likelihood of more than a threshold as the number of organisms.

[0053] According to the above method, the estimation device can use intervals determined based on the magnitude of likelihood and threshold in the likelihood spectrum to output the number of organism-based peaks after excluding image-based peaks among the multiple peaks in the likelihood spectrum. Based on the above understanding, the inventors of this application conceived of a technique for excluding image-based peaks among the peaks in the likelihood spectrum using the aforementioned interval method. By using the interval method, the estimation device eliminates the need for comparison of differences between multiple peaks, thus simplifying the process. Furthermore, since no machine learning model is used, preparation work such as preparing teacher data and prior learning processing is eliminated. Therefore, the estimation device can more easily estimate organism-related information even when the number of organisms being detected is unknown.

[0054] For example, the estimation unit may estimate the number of organisms by inputting the comprehensive spectrum calculated by the spectrum calculation unit into a pre-made model, thereby outputting the number of organisms. The pre-made model is a model pre-made by machine learning using an image representing the likelihood spectrum of the likelihood of the existence of organisms in the space and the number of organisms as teacher data.

[0055] According to the above method, the estimation device can use a model pre-built through machine learning to output the number of organism-based peaks after excluding image-based peaks. Based on the above understanding, the inventors of this application conceived of a technique for excluding image-based peaks from the peaks in a likelihood spectrum using a model built through machine learning. Since the estimation device uses a model built through machine learning, in other words, it does not require comparison of differences between multiple peaks, thus simplifying the process. It does not require setting a threshold for the likelihood, thus avoiding the impact of the threshold setting on the processing. Therefore, the estimation device can more easily estimate organism-related information even when the number of organisms being detected is unknown.

[0056] For example, the estimation unit may use a convolutional neural network model as the model to output the organism information.

[0057] According to the above method, the estimation device uses a convolutional neural network, which can more appropriately estimate information related to organisms even when the number of organisms being detected is unknown.

[0058] For example, the spectral calculation unit may use an estimation algorithm that estimates the number of organisms present in the space when the number of organisms present in the space is input, as the estimation algorithm, to calculate the likelihood spectrum.

[0059] According to the above method, the estimation device can use an estimation algorithm based on the number of organisms present in the input space, and obtain information related to the organisms present in the space without requesting the number of organisms present in the input space. Therefore, the estimation device can estimate information related to organisms even when the number of organisms being detected is unknown.

[0060] For example, the spectrum calculation unit may use the MUSIC (Multiple Signal Classification) method as the estimation algorithm to calculate the likelihood spectrum.

[0061] According to the above method, the estimation device uses the MUSIC method to estimate information related to organisms even when the number of organisms being detected is unknown.

[0062] In addition, one aspect of this disclosure involves an estimation method that uses received signals of radio waves transmitted from N (N being a natural number greater than 2) transmitting antenna elements to a space where one or more organisms exist and received by M (M being a natural number greater than 2) receiving antenna elements. The method calculates a complex transfer function representing the propagation characteristics between the transmitting and receiving antenna elements, uses each of a plurality of distinct values ​​as the number of organisms, calculates a likelihood spectrum representing the likelihood of the existence of the organisms derived from the organism information using an estimation algorithm that estimates the existence of the organisms. This organism information is the component of the complex transfer function corresponding to the organisms. The estimation method then calculates a composite spectrum by combining the calculated plurality of likelihood spectra, estimates organism information representing at least the number of organisms present in the space based on the composite spectrum, and outputs this information.

[0063] The method described above achieves the same effect as the estimation device described above.

[0064] In addition, one aspect of this disclosure relates to a program that enables a computer to perform the above-described estimation method.

[0065] The method described above achieves the same effect as the estimation device described above.

[0066] Furthermore, this disclosure can be implemented not only as an apparatus, but also as an integrated circuit having the processing mechanism of such an apparatus, as a method comprising the processing mechanism of the apparatus as steps, as a program for causing a computer to execute these steps, or as information, data, or signals representing the program. Additionally, these programs, information, data, and signals can be distributed via recording media such as CD-ROMs or communication media such as the Internet.

[0067] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. Furthermore, the embodiments described below illustrate preferred examples of the present disclosure. The numerical values, shapes, materials, constituent elements, arrangement and connection methods of constituent elements, steps, and order of steps shown in the following embodiments are examples and are not intended to limit the present disclosure. Additionally, constituent elements in the following embodiments that are not described in the independent claims representing the highest-level concept of the present disclosure are described as arbitrary constituent elements constituting a more preferred embodiment. Furthermore, in this specification and the accompanying drawings, constituent elements that have substantially the same functional configuration are given the same reference numerals to avoid repetitive descriptions.

[0068] (Implementation Method 1)

[0069] Hereinafter, the method for estimating the number of people using sensor 1 in Embodiment 1 will be described with reference to the accompanying drawings. Sensor 1 is an example of an estimation device that can estimate information related to organisms even when the number of organisms to be detected is unknown.

[0070] [Structure of Sensor 1]

[0071] Figure 1 This is a block diagram showing the configuration of sensor 1 in embodiment 1. Figure 2 This is a conceptual diagram representing the estimation of the arrival direction by sensor 1 in Implementation 1.

[0072] Figure 1 The sensor 1 shown includes a complex transfer function calculation unit 30, a biological component extraction unit 40, a correlation matrix calculation unit 50, a spectrum calculation unit 70, and an estimation unit 80. The sensor 1 is connected to a transmitter 10 and a receiver 20. Alternatively, the sensor 1 may include one or both of the transmitter 10 and the receiver 20. Furthermore, the transmitter 10 and the receiver 20 may be housed in the same casing.

[0073] [Sender 10]

[0074] The transmitter 10 includes a transmitting unit 11 and a transmitting antenna unit 12. The transmitter 10 transmits radio waves into space S. Imagine that a living organism 200 exists in space S. The living organism 200 is, for example, a human being (that is, a human body), and this will be used as an example for explanation.

[0075] The transmitting antenna section 12 consists of M T One transmitting antenna element #1 to #M T The antenna array is configured such that the transmitting antenna section 12 is, for example, a 4-element patch array antenna with an element spacing of half a wavelength.

[0076] The transmitting unit 11 generates a high-frequency signal. The high-frequency signal generated by the transmitting unit 11 can be used to estimate the presence or absence, location, or number of organisms 200. For example, the transmitting unit 11 generates a 2.4 GHz CW (Continuous Wave) and transmits the generated CW as a radio wave, i.e., a transmission wave, from the transmitting antenna unit 12. Furthermore, the transmitted signal is not limited to CW; it can also be a modulated signal.

[0077] [Receiver 20]

[0078] The receiver 20 includes a receiving antenna section 21 and a receiving section 22. The receiver 20 receives radio waves from the space S in which the transmitter 10 transmits radio waves. The received radio waves may include reflected or scattered waves, which are part of the transmitted waves transmitted from the transmitting antenna section 12 and are reflected or scattered by the organism 200.

[0079] The receiving antenna section 21 consists of M R One receiving antenna element #1 to #M R The antenna array is configured as follows. For example, it is a 4-element patch array antenna with an element spacing of half a wavelength. The receiving antenna section 21 receives high-frequency signals from the array antenna.

[0080] The receiving unit 22 uses, for example, a down-converter to convert the high-frequency signal received by the receiving antenna unit 21 into a low-frequency signal that can be processed. Furthermore, when the transmitter 10 transmits a modulated signal, the receiving unit 22 also demodulates the received modulated signal. The receiving unit 22 then transmits the converted low-frequency signal to the complex transfer function calculation unit 30.

[0081] Furthermore, while the frequency used in this embodiment is 2.4 GHz as an example, frequencies such as 5 GHz or millimeter wave bands can also be used.

[0082] [Complex Transfer Function Calculation Section 30]

[0083] The complex transfer function calculation unit 30 calculates the complex transfer function representing the propagation characteristics between the transmitting antenna section 12 and the receiving antenna section 21 of the transmitter 10 based on the received signal received by the array antenna of the receiving antenna section 21. More specifically, the complex transfer function calculation unit 30 calculates the M-value representing the propagation characteristics of the transmitting antenna section 12 based on the low-frequency signal transmitted by the receiving section 22. T The M-shaped antenna elements and receiving antenna sections 21 each have M-shaped antenna elements. R The complex transfer function of the propagation characteristics between each receiving antenna element.

[0084] Furthermore, the complex transfer function calculated by the complex transfer function calculation unit 30 sometimes includes components corresponding to reflected or scattered waves (also called biological components) of signals that are part of the transmitted wave transmitted from the transmitting antenna unit 12 and are reflected or scattered by the biological body 200. Additionally, the complex transfer function calculated by the complex transfer function calculation unit 30 sometimes includes components corresponding to direct waves from the transmitting antenna unit 12 and reflected waves from stationary objects that are not reflected by the biological body 200. Furthermore, the amplitude and phase of the signals reflected or scattered by the biological body 200, i.e., the reflected and scattered waves transmitted through the biological body 200, frequently change according to biological activities such as respiration and heart rate of the biological body 200.

[0085] Hereinafter, it will be explained that the complex transfer function calculated by the complex transfer function calculation unit 30 includes biological components corresponding to the reflected and scattered waves that are signals reflected or scattered by the organism 200.

[0086] In addition, Figure 1 The diagram illustrates a configuration where transmitter 10 and receiver 20 are arranged adjacently, but the configuration of transmitter 10 and receiver 20 is not limited to this; for example, it could also be configured as follows: Figure 2 The antennas are configured separately as shown. Additionally, the transmitting antenna can also be used as a receiving antenna. Furthermore, the transmitting and receiving antennas can be shared as hardware in wireless devices such as Wi-Fi routers or handsets.

[0087] [Biological Component Extraction Section 40]

[0088] The biological component extraction unit 40 obtains the signal (also called the received signal) received by the receiving array antenna of the receiving antenna unit 21 from the complex transfer function calculation unit 30. Then, the biological component extraction unit 40 extracts the biological components contained in the received signal, that is, the signal components transmitted from the transmitting antenna unit 12 and reflected or scattered by one or more biological bodies 200.

[0089] More specifically, the organism component extraction unit 40 records the complex transfer function calculated by the complex transfer function calculation unit 30 in the time sequence of signal reception. Then, the organism component extraction unit 40 extracts the variable components of the complex transfer function recorded in the time sequence that are caused by the influence of the organism 200. The variable components of the complex transfer function caused by the influence of the organism 200 extracted in this way correspond to the organism components.

[0090] Methods for extracting biological components include, for example, transforming the changes in the complex transfer function to the frequency domain using Fourier transform and then extracting components at frequencies corresponding to the biological components, or extracting components by calculating the difference between the complex transfer functions at two different times. These methods remove components of direct waves and reflected waves from stationary objects contained in the complex transfer function, retaining the biological components transmitted through the organism 200. For example, by using a 5-second complex transfer function, extracting components from 0.3 Hz to 3 Hz as frequencies corresponding to the biological components, it is possible to extract the respiratory components of the organism 200 that exist even when the organism 200 is stationary.

[0091] Furthermore, in this embodiment, an example of extracting components from 0.3Hz to 3Hz was given as an example. However, if it is desired to extract slower or faster movements, it is obviously possible to make changes to extract frequency components corresponding to the desired movements.

[0092] Furthermore, in this embodiment, the transmitting antenna elements constituting the transmitting array antenna include M T In addition, the receiving antenna elements constituting the receiving array antenna include M. R Therefore, the complex transfer function corresponding to the transmitting array antenna and the receiving array antenna also contains multiple biological components via the biological body 200.

[0093] The multiple biological components of organism 200 are represented as an M-row N-column matrix (also known as the biological component channel matrix F(f)), as shown in Equation 1.

[0094] [Number 1]

[0095]

[0096] Furthermore, the elements F of the complex transfer function matrix of organism components, i.e., the channel matrix F(f) of organism components, are... ij It is from the elements h of the complex transfer function ij The elements are obtained by extracting the variable components. Furthermore, the complex transfer function matrix of the organism components, i.e., the channel matrix F(f) of the organism components, is a function of frequency or a frequency-like difference period, containing information corresponding to multiple frequencies. In addition, the difference period (differential period) refers to the time difference between two complex transfer functions in the method of extracting organism components by calculating the difference between the complex transfer functions at two different times.

[0097] [Relevance Matrix Calculation Section 50]

[0098] The correlation matrix calculation unit 50 rearranges the elements of the organism component channel matrix (M rows and N columns) calculated by the organism component extraction unit 40, thereby transforming it into an (M×N) row and 1 column organism component channel vector F. vec (f). As for the arrangement of elements, there are methods such as (Equation 2), but any operation that rearranges the matrix is ​​acceptable, and the order of the elements is not limited.

[0099] [Number 2]

[0100] F vec (f)=vec[F(f)]=[F 11 (f)…F M1 (f)F 12 (f)…F M2 (f)…F 1N (f)…F MN (f)] T (Equation 2)

[0101] Subsequently, the correlation matrix calculation unit 50 calculates the biological component channel vector F. vec (f) Calculate the correlation matrix. More specifically, the correlation matrix calculation unit 50 calculates the organism component channel vector F, which is composed of multiple variable components generated by the organism 200, according to (Equation 3). vec The correlation matrix R of (f)

[0102] [Number 3]

[0103] R = E[F vec (f)F vec (f) H (Equation 3)

[0104] In Equation 3, E[] represents the averaging operation, and the operator H represents the complex conjugate transpose. Here, in the correlation matrix calculation, the correlation matrix calculation unit 50 calculates the biological component channel vector F containing multiple frequency components in the frequency direction. vec (f) Take the average, so that the information contained in each frequency can be used for sensing at the same time.

[0105] [Spectral Calculation Department 70]

[0106] The spectrum calculation unit 70 calculates a likelihood spectrum representing the likelihood of the presence of organisms 200 in space S, and further calculates a comprehensive spectrum using the calculated likelihood spectrum. The spectrum calculation unit 70 uses an estimation algorithm that estimates the number of organisms present in the space when the number of organisms present in the space is input, as the estimation algorithm, to calculate the likelihood spectrum. The likelihood spectrum is calculated, for example, using the MUSIC method; this will be used as an example for explanation. The likelihood spectrum calculated using the MUSIC method is also referred to as the MUSIC spectrum.

[0107] Generally, to calculate the likelihood spectrum, the wavenumber of arrivals is needed as the number of arriving waves. To calculate the MUSIC spectrum using the MUSIC method, the wavenumber of arrivals is needed. The wavenumber of arrivals corresponds to the number of organisms 200 present in space S in this embodiment.

[0108] The spectrum calculation unit 70 does not use a specific value as the number of organisms, but instead uses multiple different values ​​as the number of organisms in sequence to calculate the MUSIC spectrum.

[0109] That is, the spectrum calculation unit 70, while making the variable L change from the initial value L... start Change to L end The MUSIC spectrum is calculated using the variable L as the number of organisms. Then, the spectrum calculation unit 70 calculates a composite MUSIC spectrum by combining multiple MUSIC spectra calculated using multiple different variables L. The operation of the MUSIC spectrum calculation unit 70 is explained below using mathematical formulas.

[0110] If we perform eigenvalue decomposition on the correlation matrix R calculated by the correlation matrix calculation unit 50, it can be written as:

[0111] [Number 4]

[0112] R=U∧U H

[0113] Here it is:

[0114] [Number 5]

[0115] ∪=[u1,…,u L u L+1 , ...; u MR ]

[0116] [Number 6]

[0117] ∧=diag[λ1,…,λ L λ L+1 , …, λ MR ]

[0118] Here,

[0119] [Number 7]

[0120] u1,…,u MR

[0121] It has M elements R 1 eigenvector,

[0122] [Number 8]

[0123] λ1, …, λ MR

[0124] These are the eigenvalues ​​corresponding to the eigenvectors.

[0125] [Number 9]

[0126] λ1≥λ2≥…≥λ L ≥λ L+1 ≥…≥λ MR

[0127] Let's assume it's true. Additionally, L is a loop variable used to represent the number of organisms, i.e., the number of people.

[0128] In addition, the steering vector (direction vector) of the transmitting array antenna is defined as:

[0129] [Number 10]

[0130]

[0131] The steering vector (direction vector) of the receiving array antenna is defined as:

[0132] [Number 11]

[0133]

[0134] Furthermore, when the antenna elements used do not have uniform complex directivity, the transmit and receive steering vectors can also be vectors created based on measured complex directivity data. Here, k is the wave number.

[0135] Furthermore, the steering vector resulting from multiplying these steering vectors and taking into account the angular information of both the transmitting and receiving array antennas is defined as:

[0136] [Number 12]

[0137]

[0138] The MUSIC method is applied while subjecting variable L to various changes.

[0139] That is, the spectrum calculation unit 70, based on the MUSIC method, uses the steering vector obtained by multiplication to calculate the evaluation function P, which is a synthesis of multiple MUSIC spectra, as expressed in Equation 4 below. music (θ T θ R This evaluation function is called the comprehensive MUSIC spectrum, or simply the comprehensive spectrum.

[0140] [Number 13]

[0141]

[0142] Furthermore, while the comprehensive operation in Equation 4 uses summation, it can also be replaced by product (multiplication). That is, in Equation 4, the summation notation can also be used.

[0143] [Number 14]

[0144]

[0145] Replace with quadrature notation

[0146] [Number 15]

[0147]

[0148] And use the mathematical expression after the substitution.

[0149] In addition, the minimum value L of variable L start and maximum value L end A predetermined value needs to be set. For example, the minimum value L. start Set to 1, or this number if the minimum number of organisms existing in the space S to be measured is known. Additionally, the maximum value L... end When the maximum number of organisms existing in the space S to be measured is known, it can be set to that number or a number that is 1 to 3 times larger than that number.

[0150] Additionally, for example, the maximum value L end Alternatively, it can be set to a number approximately one less than the product of the number of transmitting antenna elements and the number of receiving antenna elements. This is because the maximum number of objects that can be detected by the MUSIC method is one less than the product of the number of transmitting antenna elements and the number of receiving antenna elements. Furthermore, the maximum number L... end It can also be the number of transmitting antenna elements or the number of receiving antenna elements.

[0151] In other words, the spectrum calculation unit 70 can, for example, use multiple natural numbers less than (number of transmitting antenna elements N × number of receiving antenna elements M-1), multiple natural numbers less than N, or multiple natural numbers less than M as variables L to calculate the likelihood spectrum. This is because when the estimated number of organisms is less than or equal to the product of the number of transmitting antenna elements and the number of receiving antenna elements, the organism information is determined with high accuracy. Furthermore, when the estimated number of organisms is less than or equal to either the number of transmitting antenna elements or the number of receiving antenna elements, the organism information is determined with even higher accuracy.

[0152] In addition, the spectrum calculation unit 70 can use multiple natural numbers as variables L to calculate the likelihood spectrum, wherein each of the multiple natural numbers is less than or equal to the number determined as the maximum number of organisms that may exist in space S.

[0153] Furthermore, the spectrum calculation unit 70 can use multiple natural numbers within the range of organism numbers represented by the organism number information stored in the storage unit as variables L to calculate the likelihood spectrum. Here, the storage unit is a storage device (not shown) that stores organism number information previously estimated by the estimation unit 80.

[0154] Furthermore, in the example above, if variable L is incremented by 1 sequentially, but not necessarily at equal intervals, variable L can also be changed according to a different pattern than sequential incrementing. The pattern can be predetermined or randomly selected as the process progresses.

[0155] Furthermore, the MUSIC spectrum can also be replaced by spectra based on the Beamformer or Capon methods. However, it should be noted that the Beamformer or Capon methods have lower accuracy compared to the MUSIC method and cannot achieve high-precision estimation when used alone. In other words, the MUSIC method has the advantage of being able to achieve higher-precision estimation when used alone compared to the Beamformer or Capon methods.

[0156] [Estimation Section 80]

[0157] The estimation unit 80 estimates, based on the comprehensive spectrum calculated by the MUSIC spectrum calculation unit 70, biological information that at least indicates the number of organisms 200 present in the space S, which is the object of measurement, and human information that at least indicates the number of people present in the space S, and outputs this information. Additionally, the estimation unit 80 can also estimate, based on the comprehensive spectrum, biological information that also indicates the location of organisms present in the space S, and human information that also indicates the location of people present in the space S, and output this information.

[0158] Originally, when the correct number of people (that is, the number of people actually existing in space S) is input and the calculated MUSIC spectrum appears, the number of peaks is the same as the number of people input. However, in this embodiment, multiple MUSIC spectra obtained by inputting various values ​​as the number of people are combined, so sometimes virtual images appear in the combined spectrum (that is, peaks appearing in positions where there are actually no people).

[0159] In the estimation unit 80, by identifying non-virtual peaks among the peaks appearing in the comprehensive spectrum, and using these non-virtual peaks as objects to calculate the number of people, the estimation unit 80 estimates person information representing the number of people existing in space S. Alternatively, the estimation unit 80 can also estimate person information representing the location of people existing in space S by calculating the position of the non-virtual peaks among the aforementioned peaks.

[0160] When calculating the number or location of people, there are methods such as using the ratio method for the peak values ​​of the spectrum; counting the number of intervals in the MUSIC spectrum with a likelihood of a specified threshold or higher (also called blocks); or treating the MUSIC spectrum as an image and using machine learning methods such as convolutional neural networks. In this embodiment, as an example, a method for calculating person information using the ratio method will be described.

[0161] Figure 3 This is a detailed block diagram of the estimation unit 80 in Embodiment 1.

[0162] Figure 3 The estimation unit 80 shown includes a peak exploration unit 81, a false peak determination unit 82, a peak sorting unit 83, and a verification unit 84.

[0163] <Peak Exploration Department 81>

[0164] The peak exploration unit 81 explores the peaks that take the maximum value in the comprehensive spectrum. The set of peaks discovered through exploration is set as the first peak set. Furthermore, in order to exclude small peaks caused by noise, the first peak set is preferably limited to peaks whose peak value is the maximum value within a specified range x.

[0165] Figure 4 This is a conceptual diagram illustrating the operation of the peak exploration unit 81 in Implementation Method 1. (Refer to...) Figure 4 The processing of peak exploration section 81 is explained using a 1D comprehensive spectrum of 1000.

[0166] exist Figure 4 The diagram shows four peaks included in the comprehensive spectrum 1000: peaks 1001-A, 1001-B, 1001-C, and 1001-D. For each of the four peaks, within a range of 0.5m from that peak (i.e., ranges 1002-A, 1002-B, 1002-C, and 1002-D), the peaks with the highest value are peaks 1001-A, 1001-B, and 1001-D. The peak exploration unit 81 extracts these three peaks from the comprehensive spectrum 1000 and obtains the extracted peaks as the first peak set.

[0167] The first set of peaks corresponds to one or more maxima of the likelihood spectrum that have the maximum value within a specified range that includes the maximum value of that maxima.

[0168] <Peak Detection Section 82>

[0169] The false peak determination unit 82 excludes relatively flat peaks from the peaks contained in the first peak set. This is because virtual images in the composite spectrum 1000 appear as relatively flat peaks, so by excluding relatively flat peaks, peaks based on virtual images are also excluded.

[0170] Specifically, the false peak determination unit 82 calculates the y% value of the values ​​contained within a range of a predetermined distance x from each peak in the first peak set. The false peak determination unit 82 extracts peaks whose difference between the peak value and the y% value is a predetermined threshold z or higher, and obtains the extracted peaks as the second peak set. The difference between the peak value and the y% value can be either the difference between the peak value and the y% value (i.e., peak value - y% value) or the ratio between the peak value and the y% value (i.e., y% value / peak value). Furthermore, the "values ​​contained within a range of a predetermined distance x" can use any value contained within that range, the average value, the maximum value, or the minimum value contained within that range, etc.

[0171] Therefore, the false peak determination unit 82 can exclude relatively flat peaks from the peaks contained in the first peak set. For example, when the specified distance x is set to 0.5m, y is set to 70%, and z is set to 0.4dB, the false peak determination unit 82 extracts peaks that are 0.4dB or more greater than 70% of the values ​​contained within 0.5m of each peak contained in the first peak set.

[0172] The second set of peaks, after excluding peaks based on virtual images from the first set of peaks by the false peak determination unit 82, is equivalent to one or more third maxima. The difference between this third maxima and the value obtained by multiplying a value within a specified range containing the third maxima by a specified ratio is greater than or equal to a threshold. Here, the specified ratio is a specified value that is greater than 0 and less than 1.

[0173] <Peak Sequencing Section 83>

[0174] The peak sorting unit 83 sorts the values ​​of the multiple peaks contained in the second peak set in descending order. Furthermore, the peak sorting unit 83 can also add a value w smaller than the smallest value among the peaks in the second peak set to serve as a virtual peak. In the process of comparing each peak in the second peak set with the peaks following it in descending order, the virtual peak can be used as the peak following the smallest peak in descending order. For example, when w is set to 3.4 dB, and the smallest peak is -3 dB relative to the largest peak, the added virtual peak is -6.4 dB relative to the largest peak.

[0175] <Inspection Department 84>

[0176] The inspection unit 84 calculates the differences between adjacent peaks in the second peak set sorted by the peak sorting unit 83, thereby estimating the number of people. More specifically, it calculates the ratio or difference between the i-th peak and the (i+1)-th peak in the descending-order second peak set, and outputs the i-th peak with the largest difference or ratio as the number of people. Here, i is an integer greater than or equal to 1 and less than or equal to the number of elements in the second peak set.

[0177] The following explanation will use the case where the difference is used as an example.

[0178] Figure 5 This is a conceptual diagram illustrating the operation of the inspection unit 84 in Implementation Method 1.

[0179] exist Figure 5 In the second peak set, peaks 1101-A, 1101-B, 1101-C, and 1102 are represented in descending order according to their peak values. In addition, peak 1102 is a virtual peak added by the peak sorting unit 83.

[0180] The peak sorting unit 83 calculates the differences 1103-A, 1103-B and 1103-C between adjacent peaks in the second peak set, and finds the combination of peaks with the largest calculated differences.

[0181] exist Figure 5 In the example shown, the difference 1103-B, that is, the difference between the second peak 1101-B and the third peak 1101-C, is the largest, so i is 2, and the number of people calculated is 2.

[0182] As described above, the inspection unit 84 obtains the first maximum value among the more than one maximum values ​​obtained by the peak exploration unit 81. This first maximum value is the maximum value with the largest difference between itself and the second maximum value that follows it in descending order. The inspection unit 84 obtains a quantity indicating which of the more than one maximum values ​​the first maximum value is. Then, the estimation unit 80 estimates the quantity obtained by the inspection unit 84 as the number of people existing in space S and outputs it.

[0183] In addition, the inspection unit 84 can directly use one or more maxima obtained by the peak exploration unit 81 and output human information as described above, or it can use the false peak determination unit 82 to exclude one or more third maxima after the peak based on the virtual image from one or more maxima obtained by the peak exploration unit 81, and output human information as one or more maxima as described above.

[0184] Furthermore, the above example illustrates the case where sensor 1 outputs information about the number of people, but the MUSIC spectrum can also be used to estimate the location of people and output information about their location.

[0185] Furthermore, this embodiment describes an example where both the transmitting and receiving antennas are configured with multiple MIMO (Multiple-Input Multiple-Output) antennas, but either the transmitting or receiving antenna can also be configured with a single antenna. In this case, the synthesized spectrum output by the spectrum calculation unit 70 is one-dimensional, and human information can be estimated by exploring peaks, just as in the two-dimensional case.

[0186] Furthermore, the determination can be made only when no one is detected in space S, i.e., 0 people, based on the magnitude of the largest eigenvalue, the power of the variation component of the complex transfer function, or the magnitude of the correlation with the absence of people. The likelihood spectrum and composite spectrum can only be calculated by the spectrum calculation unit 70 when people are present. Therefore, when no one is detected in space S, the processing required to calculate the likelihood spectrum and composite spectrum can be omitted, which helps reduce power consumption.

[0187] [Action of Sensor 1]

[0188] The process of estimating the number of organisms using the sensor 1 constructed as described above will be explained.

[0189] Figure 6 This is a flowchart illustrating the processing of sensor 1 in implementation method 1.

[0190] like Figure 6 As shown, in step S10, sensor 1 receives signals in receiver 20 for a predetermined period.

[0191] In step S20, sensor 1 calculates the complex transfer function based on the received signal.

[0192] In step S30, sensor 1 records the calculated complex transfer function in time sequence, extracts the variable components caused by the influence of organisms from the recorded time sequence complex transfer function, and then calculates the organism component channel matrix.

[0193] In step S40, sensor 1 calculates the correlation matrix of the extracted biological component channel matrix.

[0194] In step S50, sensor 1 sets an initial value L for variable L. start .

[0195] In step S60, sensor 1 calculates the likelihood spectrum using the MUSIC method based on the variable L set in step S50 or S75 and the correlation matrix calculated in step S40.

[0196] In step S70, sensor 1 determines the relationship between variables L and L. end Are they consistent? (The question is incomplete and requires further context.) endIf the conditions are met (step S70: yes), proceed to step S80; otherwise (step S70: no), proceed to step S75.

[0197] In step S75, sensor 1 increments variable L by 1. Then, sensor 1 executes step S60 again.

[0198] In step S80, sensor 1 calculates a comprehensive spectrum by synthesizing the likelihood spectrum. The synthesized likelihood spectrum is obtained by processing the variable L through steps S50, S60, S70, and S75, which changes the value of L from LL. start Change one by one to L end One side is the likelihood spectrum calculated by sensor 1.

[0199] In step S90, sensor 1 calculates the number of people based on the composite spectrum calculated in step S80, estimates the person information, and outputs it. The processing in step S90 is performed, for example, by using a ratio method for the peak value of the composite spectrum; a method for counting the number of consecutive blocks in the range above a specified value in the composite spectrum; or, treating the composite spectrum as an image and using machine learning methods such as convolutional neural networks.

[0200] Figure 7 This is a flowchart illustrating the calculation process of human information by sensor 1 in Implementation Method 1. Figure 7 The process shown is an example of the process in which step S90 is performed using the ratio method.

[0201] like Figure 7 As shown, in step S110, sensor 1 extracts the peaks in the comprehensive spectrum that have the maximum value within a specified range, and obtains the extracted peaks as the first peak set.

[0202] In step S120, sensor 1 calculates the y% value of the values ​​contained in the range of a specified distance from each peak in the first set of peaks.

[0203] In step S130, sensor 1 extracts peaks whose difference between the peak value and the y% value calculated in step S120 is above a specified threshold, based on the peak extracted in step S110, and obtains the extracted peaks as the second peak set.

[0204] In step S140, sensor 1 sorts the peaks contained in the second peak set in descending order of peak value.

[0205] In step S150, sensor 1 calculates the difference between the i-th peak and the (i+1)-th peak in the second peak set, estimates the i-th peak with the largest difference as the number of people represented, and outputs it. Here, i is an integer greater than 1 and less than the number of elements in the second peak set.

[0206] [Effects, etc.]

[0207] According to the sensor 1 of this embodiment, the number of organisms 200 present in space S can be estimated with high accuracy using wireless signals.

[0208] In existing methods for estimating the number of organisms 200 present in space S, it is sometimes necessary to assign a number of organisms present in space S.

[0209] According to the sensor 1 of this embodiment, the number of organisms in space S is estimated using a comprehensive spectrum, which is calculated by combining multiple likelihood spectra using multiple values ​​as the number of organisms present in space S. Therefore, even when the number of organisms present in space S is unknown, it is possible to estimate organism information representing the number of organisms present in space S.

[0210] (Implementation Method 2)

[0211] In Embodiment 1, a method for estimating organism information (i.e., human information) using a ratio method based on a comprehensive spectrum is described. In Embodiment 2, a method for estimating organism information is described using a method of counting the number of intervals, i.e., blocks, with a likelihood of a predetermined threshold or higher based on a comprehensive spectrum.

[0212] The sensor in this embodiment has the same configuration as the sensor 1 in Embodiment 1, except that the estimation unit 80 of the sensor 1 in Embodiment 1 is replaced by an estimation unit 2080. The configuration other than the estimation unit 2080 is the same as in Embodiment 1, and therefore, description is omitted here.

[0213] Figure 8 This is a block diagram illustrating the configuration of the estimation unit 2080 in Embodiment 2. Figure 9 This is a conceptual diagram illustrating the operation of the block detection unit 2082 in Embodiment 2. Figure 9 The composite spectrum 2100 shown is an example of the composite spectrum calculated by the spectrum calculation unit 70.

[0214] Estimated Department 2080 Figure 8 As shown, it includes a threshold setting unit 2081 and a block detection unit 2082.

[0215] The threshold setting unit 2081 sets a threshold 2101 that is v [dB] smaller than the maximum value of the comprehensive spectrum 2100. Furthermore, v and the threshold 2101 can be preset fixed values, or v and the threshold 2101 can be varied beforehand, and the accuracy of the population estimation can be evaluated, with the threshold 2101 showing the highest accuracy being used as the optimal value. For example, when using an unmodulated continuous wave at 2.47125 GHz and sensing a 4-element patch array antenna with element spacing of half a wavelength in a 4m square room, v can be set to 3.9 dB.

[0216] The block detection unit 2082 detects the intervals in the comprehensive spectrum 2100 with a likelihood of 2101 or higher as blocks, and obtains the number of detected blocks.

[0217] The estimation unit 2080 estimates the number of blocks obtained by the block detection unit 2082 as the number of people existing in space S.

[0218] exist Figure 9 In the example shown, as the comprehensive spectrum 2100 is the interval above the threshold 2101, two blocks, 2102-A and 2102-B, are detected. The block detection unit 2082 calculates information representing the number of people, which is 2.

[0219] [Effects, etc.]

[0220] According to the sensor of Embodiment 2, the computational load in the estimation unit 2080 can be reduced compared to sensor 1 in Embodiment 1. Therefore, the capability requirements of the processing device needed for real-time processing can be lowered, enabling the estimation of human-related information at a lower cost.

[0221] (Implementation Method 3)

[0222] In Implementation 1, a method for estimating organism information (i.e., human information) based on a comprehensive spectrum using a ratio method was described. In Implementation 3, a method for estimating organism information based on a comprehensive spectrum using a machine learning model (e.g., a convolutional neural network) was described.

[0223] The sensor in this embodiment has the same configuration as the sensor 1 in Embodiment 1, except that the estimation unit 80 of the sensor 1 in Embodiment 1 is replaced by an estimation unit 3080. The configuration other than the estimation unit 3080 is the same as in Embodiment 1, and therefore will not be described here.

[0224] Figure 10 This is a block diagram illustrating the configuration of the estimation unit 3080 in Embodiment 3.

[0225] Estimated Department 3080 Figure 10As shown, it includes a teacher data production unit 3081, a learning unit 3082, a network storage unit 3083, an image transformation unit 3084, and a judgment unit 3085.

[0226] The teacher data production unit 3081, the learning unit 3082, and the network storage unit 3083 perform pre-learning of machine learning models. The image transformation unit 3084 and the judgment unit 3085 use the pre-learned machine learning model to calculate human information for the test data.

[0227] The teacher data production department 3081 pre-acquires multiple images of music spectra representing known numbers of people and saves them as teacher data images. Here, in the teacher data images, multiple images representing music spectra are included for each number of people assumed to exist in space S. For example, in the case where the upper limit of the number of people existing in space S, which is the object of measurement, is 3, multiple teacher data images, for example, more than 100, are included for 0 people, 1 person, 2 people, and 3 people respectively.

[0228] Learning unit 3082 uses teacher data images as input to learn a machine learning model. The machine learning model is, for example, a convolutional neural network model. The teacher data images used as input are teacher data images stored in teacher data production unit 3081. Furthermore, methods to improve the learning efficiency of the neural network, such as transfer learning, can also be used here.

[0229] The network storage unit 3083 stores the convolutional neural network generated by the learning unit 3082 through learning to a recording medium such as a computer memory or CD-ROM, or to a server external to the sensor. When stored on a server external to the sensor, the data of the convolutional neural network is sent to the server via network communication.

[0230] The image transformation unit 3084 transforms the comprehensive spectrum calculated by the spectrum calculation unit 70 into a form that can be processed by a convolutional neural network, thereby generating input data. An image in a form that can be processed by a convolutional neural network is, for example, a heatmap image corresponding to the values ​​of each pixel and the comprehensive spectrum.

[0231] The determination unit 3085 obtains human information by inputting the input data generated by the image transformation unit 3084 into the convolutional neural network stored in the network storage unit 3083.

[0232] The estimation unit 3080 will estimate the person information obtained by the determination unit 3085 as person information representing the person existing in the space S.

[0233] The sensor involved in one aspect of this disclosure has been described above based on an implementation method, but this disclosure is not limited to these implementation methods. Any modifications to this embodiment that would be conceived by those skilled in the art, or combinations of constituent elements from different embodiments, that are constructed without departing from the spirit of this disclosure are all included within the scope of this disclosure.

[0234] Furthermore, this disclosure can be implemented not only as a sensor possessing such characteristic components, but also as an estimation method for the characteristic components contained in the sensor as steps. It can also be implemented as a computer program that causes a computer to execute the characteristic steps included in such a method. Moreover, it is evident that such a computer program can be distributed via a non-volatile recording medium readable by a computer, such as a CD-ROM, or via a communication network such as the Internet.

[0235] [Effects, etc.]

[0236] By utilizing the sensor of Implementation 3 and employing machine learning based on convolutional neural networks, various parameters, such as thresholds, that need to be adjusted to correspond to changes in various environments in which the sensor is set can be automatically adjusted. Furthermore, by continuously updating the learned network, further improvements in the accuracy of people estimation can be expected.

[0237] Industrial applicability

[0238] This disclosure enables the use of measuring devices for determining the number and location of organisms, household appliances for controlling the number and location of organisms accordingly, and monitoring devices for detecting the intrusion of organisms.

[0239] Explanation of reference numerals in the attached figures:

[0240] 1. Sensor

[0241] 10 transmitters

[0242] 11. Sending Department

[0243] 12 Transmitting Antenna Section

[0244] 20 Receivers

[0245] 21 Receiving Antenna Section

[0246] 22 Receiving Section

[0247] 30 Complex Transfer Function Calculation Section

[0248] 40. Biological Component Extraction Department

[0249] 50 Correlation Matrix Calculation Department

[0250] 70 Spectrum Calculation Department

[0251] 80, 2080, 3080 Estimation Department

[0252] 81 Peak Exploration Department

[0253] 82 Peak Error Judgment Department

[0254] 83 Peak Sequencing Section

[0255] 84 Inspection Department

[0256] 200 organisms

[0257] 1000, 2100 combined spectrum

[0258] 1001-A, 1001-B, 1001-C, 1001-D, 1101-A, 1101-B, 1101-C, 1102 peak

[0259] Ranges 1002-A, 1002-B, 1002-C, and 1002-D

[0260] Differences between 1103-A, 1103-B, and 1103-C

[0261] 2081 Threshold Setting Section

[0262] 2082 Block Testing Department

[0263] 2101 Threshold

[0264] Blocks 2102-A and 2102-B

[0265] 3081 Teacher Data Production Department

[0266] 3082 Study Department

[0267] 3083 Network Storage Division

[0268] 3084 Image Transformation Unit

[0269] 3085 Judgment Department

[0270] S Space

Claims

1. An estimation device comprising: The complex transfer function calculation unit uses the received signal of radio waves transmitted from N transmitting antenna elements to the space where one or more organisms exist and received by M receiving antenna elements to calculate the complex transfer function representing the propagation characteristics between the transmitting antenna elements and the receiving antenna elements, where N is a natural number of 2 or more and M is a natural number of 2 or more. The spectral calculation unit (a) uses each of a plurality of distinct values ​​as the number of organisms, calculates a likelihood spectrum representing the likelihood of the existence of the organisms derived from organism information, which is the organism-corresponding component contained in the complex transfer function, using an estimation algorithm that estimates the existence of the organisms; and (b) calculates a composite spectrum formed by combining the calculated plurality of likelihood spectra; and The estimation unit estimates and outputs biological information, at least representing the number of organisms present in the space, based on the comprehensive spectrum.

2. The estimation device as described in claim 1, The estimation unit estimates and outputs the organism information, which also represents the location of organisms existing in the space, based on the comprehensive spectrum.

3. The estimation device as described in claim 1, The spectral calculation unit uses multiple natural numbers less than or equal to N × M-1, multiple natural numbers less than N, or multiple natural numbers less than M as the multiple values ​​to calculate the likelihood spectrum.

4. The estimation device as described in claim 1, The spectral calculation unit uses multiple natural numbers as the multiple values ​​to calculate the likelihood spectrum, wherein each of the multiple natural numbers is less than or equal to a number determined as the maximum number of organisms that may exist in the space.

5. The estimation apparatus as claimed in claim 1, further comprising: The storage unit stores information about the organism that was previously estimated by the estimation unit. The spectral calculation unit uses a plurality of natural numbers within the range of the number of organisms represented by the organism information stored in the storage unit as the plurality of values ​​to calculate the likelihood spectrum.

6. The estimation apparatus as described in claim 1, The estimation unit performs: Obtain one or more maxima from among the multiple maxima of the likelihood spectrum, where the maximum value of the maximum value is within a specified range that includes the maximum value of the maximum value. Determine the first maximum among the obtained one or more maximum values, wherein the first maximum value is the maximum value with the largest difference between it and the second maximum values ​​that follow the first maximum value in descending order. The number of organisms is estimated to be the largest among the more than one maxima determined by the first maximum value.

7. The estimation apparatus as described in claim 6, The estimation unit uses only one or more of the third maximum values ​​among the one or more maximum values ​​as the one or more maximum values ​​to determine the first maximum value, wherein the difference between the one or more third maximum values ​​and the value obtained by multiplying the value contained in the specified range including the third maximum value by a specified ratio is above a threshold.

8. The estimation device as claimed in claim 1, The estimation unit estimates the number of intervals in the likelihood spectrum with a likelihood of above the threshold as the number of organisms.

9. The estimation device as claimed in claim 1, The estimation unit estimates the number of organisms by inputting the comprehensive spectrum calculated by the spectrum calculation unit into a pre-made model. The pre-made model is a model pre-made by machine learning using an image representing the likelihood spectrum of the likelihood of the existence of organisms in the space and the number of organisms as teacher data.

10. The estimation apparatus as described in claim 9, The estimation unit uses a convolutional neural network model as the model to output the organism information.

11. The estimation apparatus as claimed in claim 1, The spectral calculation unit uses an estimation algorithm that estimates the number of organisms present in the space when the number of organisms present in the space is input as the estimation algorithm to calculate the likelihood spectrum.

12. The estimating apparatus as claimed in any one of claims 1 to 11, The spectral calculation unit uses MUSIC, or multi-signal classification, as the estimation algorithm to calculate the likelihood spectrum.

13. An estimation method, Using the received signals of radio waves transmitted from N transmitting antenna elements to the space where one or more organisms exist and received by M receiving antenna elements, calculate the complex transfer function representing the propagation characteristics between the transmitting antenna elements and the receiving antenna elements, where N is a natural number greater than 2 and M is a natural number greater than 2. Using each of a plurality of distinct values ​​as the number of organisms, a likelihood spectrum representing the likelihood of the existence of the organisms is calculated, derived from the organism information using an estimation algorithm that estimates the existence of the organisms. This organism information is the organism-corresponding component contained in the complex transfer function. The calculation involves synthesizing a composite spectrum by combining the calculated multiple likelihood spectra. Based on the comprehensive spectrum, estimate and output biological information that at least represents the number of organisms present in the space.

14. A program recording medium that enables a computer to perform the estimation method as described in claim 13.

Images