JP6763241B2 - Person detection method and person detection device - Google Patents

Description

The present invention relates to a person detection method and a person detection device, and more particularly to a person detection method and a person detection device that separately detects the presence of one or more people in a detection area, walking, work, and other actions.

With increasing expectations for the IoT (Internet of Things) society, there is an increasing need for human detection technology that can automatically detect the movements and behaviors of people. If advanced human detection technology is combined with IT technology, for example, the behavior of children and care recipients can be automatically collected by sensors and checked on a smartphone while on the go, the congestion situation and the flow of people can be grasped in public places, or suspicious. It will be possible to discover people, obtain information on how customers move and stay in the store, and use it for marketing. Therefore, it is expected to be applied in many places such as homes, offices, medical care, and stores.

Human detection technology includes human detection by images, human detection by infrared rays, human detection by electromagnetic wave and ultrasonic irradiation, and human detection by static electricity.

In human detection using images, for example, image information obtained by a surveillance camera is image-analyzed by software. Human detection by image has a problem that a blind spot (blind area) inevitably exists in the image information. Further, there is a problem that the system for image processing becomes expensive. Furthermore, there is also a privacy issue regarding the use of image information.

Human detection by infrared irradiation is to observe infrared rays emitted from the human body with an infrared sensor. This infrared irradiation human detection system is relatively low cost and does not cause privacy problems. However, with this method, although it is possible to know whether or not a person has entered the detection range, it is not possible to obtain more detailed information such as which position within the detection range, how many people have entered, and how the person has moved.

Human detection by electromagnetic wave and ultrasonic irradiation irradiates the detection area with signals that are harmless to the human body, such as near infrared rays, microwaves, and ultrasonic waves, and measures their reflection from the human body. This human detection by electromagnetic wave and ultrasonic irradiation has a smaller amount of data and is easier to analyze than human detection by images. In addition, since detailed information cannot be obtained, there are few privacy issues. However, the problem of blind spots is greater than in the case of surveillance cameras.

Human detection by static electricity measures changes in the electrostatic field caused by the movement of the human body. The detection of movement using an electrostatic field has an advantage that it is possible to visually obtain information on a blind spot as compared with other methods centered on image analysis.

A person detection by static electricity will be described. The human body has conductivity and is approximated to a good conductor. In addition, the human body is usually insulated from what can be regarded as an electrical ground such as the ground. Therefore, the human body can have various electric potentials by being charged for some reason. When the human body as a conductor with various potentials moves in space, the electrostatic field distribution in space fluctuates. As a result, the movement of the human body can be known by measuring changes in the electric field and the like. As an example of this, a device that enables individual identification by observing a change in an electric field accompanying a person's walking is disclosed (see, for example, Patent Document 1).

The change in the electrostatic field caused by not only the human body but also an electric charge is determined by the Poisson's equation represented by the following equation (1) when it is quasi-static, that is, when the rate of change is sufficiently small.

Here, V is an electric potential, ρ is a charge density, and ε is a permittivity, which are all functions of positions (x, y, z) in space. Equation (1) above shows that the distribution of electric potential in space changes as the spatial distribution of ρ and ε changes in space. Since the human body has an electric charge and a dielectric constant different from that of air, the spatial distribution of the electric charge and the dielectric constant changes as the human body moves, and the potential distribution changes accordingly. Therefore, it can be seen that some information on the movement of the human body can be obtained by placing a plurality of electrodes (hereinafter also referred to as sensors) in the space where the human body moves and observing the potential generated by the electrodes.

It should be noted that the spatial distribution of ρ and ε cannot be determined only by measuring the potential at a limited location. However, if we assume that the human body is a conductor sphere with a diameter of 1 m, and if we model it simply, for example, there is only one person in the detection range and nothing else moves, the movement of the human body can be performed with a small number of electrodes. Can be estimated.

Japanese Unexamined Patent Publication No. 2004-147793

However, the above-mentioned conventional example of detecting a person by static electricity has the following problems.

The first problem is that when multiple human bodies are moving, it is difficult to separate their movements and estimate their movements. Since the above equation (1) holds at all points in the space, when a plurality of human bodies are scattered within the detection range, the potential distribution measured by the sensor is a superposition of them. However, it is difficult to determine where in space the measured potential change is due to changes in ρ and ε due to the movement of the human body.

The solution to this problem is to install a large number of sensors. However, it becomes difficult to install sensors, and a complicated algorithm that separates the effects of each person's movement from the signals of many sensors is required, and it is expected that estimation errors will increase.

The second problem is that the sensor can only detect the movement of a person in the immediate vicinity, and that it is directional depending on the direction of the detected movement. The reason will be described with reference to FIG. FIG. 1 is a schematic diagram for explaining human detection by static electricity. In FIG. 1, four electric field sensors S1, S2, S3, and S4 are used as sensors.

It is assumed that the sensor S1 is located at the origin O in polar coordinates (r, θ, φ), and the sensor S2 and the sensor S3 are located at positions O ′ and O ″ on a plane of φ = 0, respectively. The human body to be measured is on the point P, and is considered to be a point on the space having a constant charge amount Q. It is assumed that the human body (point P) moves in a plane of φ = 0. The plane of φ = 0 corresponds to, for example, the floor surface. Assuming that the position of the point P is (L ₁ , 0, 0), the distance OP between the sensor S1 and the point P is L ₁ . Further, the distance O'P between the sensor S2 and the point P is L ₂ , the distance O'P between the sensor S3 and the point P is L ₃ , the angle formed by OP and O'P is Θ', and OP and O'P. Let the angle to be formed be Θ´´. At this time, the absolute value of the electric field measured by the sensors S1 to S3 is represented by the following equation (2), respectively.

Here, ε ₀ is the permittivity of the vacuum. When the charged human body moves, for example, the point P approaches or moves away from the sensor S1, the electric field strength felt by each sensor changes. The rate of change in the electric field strength with respect to L ₁ felt by each sensor is obtained by differentiating the above equation (2) with L ₁ , and is expressed by the following equation (3).

The fluctuations of the electric fields E _{1 to} E ₃ represented by the above equation (3) are measured by each sensor. Under the assumption that the electric charge Q of the human body is constant, Θ ′ and Θ ″ are uniquely determined if L _{1 to} L ₃ are determined. Therefore, there are three variables in the above equation (3), L _{1 to} L ₃ . In the above equation (3), L _{1 to} L ₃ can be determined because the number of equations and the variable are equal. That is, the position of the point P can be specified. Human detection by static electricity is basically based on this principle.

However, looking at the above equation (3), the change in electric field strength is inversely proportional to the cube of the distance from the sensor. This indicates that the detection of the movement of the human body becomes rapidly difficult as the distance between the sensor and the detection target increases. For example, in FIG. 1, it is assumed that the distance L ₂ between O'P is 10 times larger than the distance L ₁ between OPs. Then, the change in the electric field strength measured by the sensor S2 is given by the following equation (4).

As described above, the value obtained by the sensor S2 is smaller than 1/1000 of the value obtained by the sensor S1, and its detection becomes difficult. In the above equation (3), if the change of any of the sensors cannot be observed, the number of equations for determining the motion of the point P decreases, and as a result, the motion of the human body cannot be detected. As described above, in the conventional human detection by the electric field sensor, there is a problem that the sensor can detect only the movement of the human body in the immediate vicinity.

Next, the problem of directivity of the electric field sensor will be described. Regarding the movement of the point P, even if the point P is in the vicinity of the sensor S1, the electric field does not change in the sensor S1 in the movement of P in the θ and φ directions orthogonal to the OP. Therefore, there is a problem that information on the movement of the point P cannot be obtained from the sensor S1.

In order to solve this problem, it is necessary to observe the change in the electric field with the sensor S4 located off the OP axis like the sensors S2 and S3, and determine the motion of the point P from the three sensors S2, S3 and S4. is there. However, the sensor S4 is located further away from the point P than the sensors S2 and S3, and there is a high possibility that the change in the electric field cannot be observed. In order to avoid this, it is necessary to use highly sensitive sensors for each sensor or to narrow the distance between the sensors until sufficient observation is possible.

As described above, in human detection by observing changes in the electrostatic field caused by the movement of the human body, in order to capture large movements of the human body such as walking, a very sensitive sensor is used and the number of sensors is increased. Must be dense. Therefore, the static electricity detection system has been used only in limited applications such as a touch sensor that detects only a human body in close proximity.

The present invention has been made in view of the above-mentioned problems, and an object of the present invention is a method for detecting a person by static electricity, which can identify and visualize the movement of a plurality of people for each individual. And to provide a person detection system.

In order to achieve the above-mentioned object, the person detection method of the present invention includes the following process. First, intensity signal sequences indicating signal strength are acquired from a plurality of sensors, a correlation coefficient is calculated between intensity signal sequences indicating signal intensities other than 0, and the correlation coefficient is compared with a predetermined threshold value. As a result of comparison, if it is equal to or higher than the threshold value, the event position is specified from the signal strength ratio. If it is smaller than the threshold value, multiple event determination is performed. The process of determining a plurality of events further includes the following process. First, assuming that the number of events ξ is 2, ξ event signal sequences I, a coefficient matrix A of ξ rows and M columns, and an intensity signal sequence S of M rows and 1 column are given from I = AS, and the event signal sequence I The initial coefficient matrix A that minimizes the correlation coefficient r (ξ) between them is determined. Next, by adding 1 to the number of events ξ, from I = AS given by ξ event signal strings I, a coefficient matrix A of ξ rows and M columns, and an intensity signal sequence S of M rows and 1 column, The coefficient matrix A that minimizes the correlation coefficient r (ξ) between the event signal sequences I is determined, and r (ξ) and r (ξ-1) are compared. When r (ξ) is less than or equal to r (ξ-1), the process of determining the coefficient matrix A and the process of comparing r (ξ) and r (ξ-1) are further performed. In other cases, the process is performed. The event position is determined from the coefficients of each row of the coefficient matrix A obtained when the value is ξ-1 or less.

Further, the human detection device of the present invention includes a plurality of electrostatic measuring instruments, a bandpass filter that transmits a high frequency component of the output of the plurality of electrostatic measuring instruments, and an AD that generates an intensity signal sequence from a signal transmitted through the bandpass filter. It includes a converter and a signal processing unit. The signal processing unit executes each procedure of the above-mentioned person detection method.

According to the human detection method and the human detection device of the present invention, an electric field fluctuation event that occurs in walking of a human body, contact with another object, etc. by extracting correlated components between signals from signals generated by a plurality of electric field sensors. Can be extracted, the position where the event occurred, and the position of the human body at a predetermined time can be estimated by assigning the event to the virtual person. Further, since a plurality of events can be separated for each event, even when a plurality of human bodies exist, their movements can be identified for each individual.

It is a schematic diagram for demonstrating the detection of a person by static electricity. It is a schematic block diagram of the 1st Embodiment of the person detection device of this invention. It is a processing flow diagram for demonstrating the processing in a signal processing unit. It is a schematic block diagram of the 2nd Embodiment of the person detection device of this invention.

Hereinafter, embodiments of the present invention will be described with reference to the drawings, but the shape, size, and arrangement of each component are only schematically shown to the extent that the present invention can be understood. Further, although a preferable configuration example of the present invention will be described below, the material and numerical conditions of each component are merely suitable examples. Therefore, the present invention is not limited to the following embodiments, and many modifications or modifications can be made that can achieve the effects of the present invention without departing from the scope of the constitution of the present invention.

(First Embodiment)
The first embodiment of the person detection device and the person detection method will be described with reference to FIG. FIG. 2 is a schematic configuration diagram of a first embodiment of the human detection device of the present invention. The person detection device is installed for security of houses and factories and marketing in stores.

The human detection device includes a sensor unit 1, a signal processing unit 2, a display unit 3, and a transmission unit 4.

The sensor unit 1 is configured to include a plurality of static electricity measuring instruments 11. The static electricity measuring device 11 can be configured by, for example, any of a potential sensor, an electric field sensor, and an induced current monitoring sensor. Each of the static electricity measuring instruments 11 is configured to include two parallel electrodes. When measuring the potential difference between two electrodes, it functions as a so-called electric field sensor. Further, when one electrode is grounded and the potential of the other electrode is measured, it functions as a so-called potential sensor. Further, if the induced current generated in the grounded electrode is measured at this time, it functions as a so-called induced current monitoring sensor. In the following description, an example in which the static electricity measuring instrument 11 is an electric field sensor will be described.

In the static electricity measuring instrument 11 which is an electric field sensor, a voltage signal having a magnitude proportional to the electric field is generated. This voltage signal is sent to the signal processing unit 2 via the voltage follower 12, the power supply frequency removal filter 13, the bandpass filter 14, the amplifier circuit 15, and the AD converter 16.

The power supply frequency removal filter 13 removes components of the power supply frequency (50 Hz or 60 Hz) and its integral multiples (100 Hz, 150 Hz, 200 Hz ..., Or 120 Hz, 180 Hz, 240 Hz ...).

The bandpass filter 14 transmits a signal of 5 to 1000 Hz. The transmission band is set to 1000 Hz or less in order to avoid aliasing that may occur due to AD conversion by the AD converter 16 in the subsequent stage. On the other hand, in order to measure the electric field fluctuation that occurs when a person walks, the transmission band is set to 5 Hz or more.

When the object to be measured, including the human body, moves in space, the spatial distribution of electric charge and permittivity changes, and the potential distribution changes accordingly. However, the speed at which the human body moves is usually slow at about 1 m / s, and the change in the electric field due to this usually appears in a low frequency region of 1 Hz or higher. Therefore, in the signal after passing through the bandpass filter 14 having a transmission band of 5 to 1000 Hz, the change in the electric field due to the simple horizontal movement of the human body hardly appears.

However, it is known that, for example, when a person walks, the electric field fluctuates at a frequency of about 8 ± 2 Hz. Such fluctuations in the electric field occur when the pedestrian's shoes leave the floor and when they land.

The floor is laid on a ground that can be regarded as a good conductor, electrical ground, and usually has some electrical resistance, but in normal floor materials, the charge on the floor has a time constant less than 1 millisecond. Since it escapes to the ground, it can be regarded as an earth almost like the ground.

The change in electric field that occurs when a pedestrian's shoes leave the floor and land is due to the electrical capacity between the floor as an electrical ground and the human body as a charged conductor, which causes the shoes to take off and land. The main factor is that the electric potential of the human body increases or decreases as a result. That is, if the amount of charge of the human body is Q, the potential of the human body with respect to the ground (floor surface) is V, and the electric capacity between the ground and the human body is C, the following equation (5) is obtained.

At the moment when the shoe surface takes off or lands on the floor, C changes abruptly, so V fluctuates accordingly and the electric field also changes.

Another cause of the electric field generated from the human body at a speed higher than 5 Hz is a discharge from or to the human body. The human body is charged from several hundred V to several kV by walking or the like, but when it comes into contact with an object having another potential, the charge moves in the direction in which both potentials become equal. For example, when an exposed part such as a hand comes into contact, the electric potential of the human body suddenly drops to the earth potential, and the time constant is on the order of 10 nanoseconds. Also, when people come into contact with each other, their potentials become equal, but this phenomenon also occurs rapidly.

Therefore, in the signal generated by the static electricity measuring device 11, the signal passing through the bandpass filter 14 is the object to be measured such as a human body within the detection range and an object different from the measurement target such as a floor surface or another human body. It is caused by interaction. The signal generated by the static electricity measuring instrument 11 has some features as described below.

The first feature is that changes in the electric charge and potential of the human body are dynamic, and the amount of change in the electric field is also large, especially when walking or discharging the human body. The electric potential of the human body during takeoff and landing during walking is several times higher than the electric potential at rest. It is also known that even when the human body has a zero potential with respect to the earth, the potential of the human body rises due to the charging phenomenon on the shoes and the floor surface. The potential of the human body reaches a maximum of several kV, and it is possible to easily observe a signal even at a distant place by observing an induced current to an ordinary potential sensor, an electric field sensor, or an electrode.

The second feature is that the signal strength decays slowly with respect to the distance. On the other hand, the change in the electric field that occurs when the object to be measured moves in space becomes smaller in inverse proportion to the cube of the distance between the sensor and the object to be measured, as described with reference to FIG. However, the change in the electric field caused by the interaction between the object to be measured and an object different from it, such as the floor surface or another human body, is caused by the fluctuation of the potential of the object to be measured itself, so that the change in the electric field is caused. It is inversely proportional to the square of the distance between the sensor and the object to be measured. Therefore, it is possible to observe the signal even at a place away from the sensor.

The third feature is that the output from each sensor obtained for a certain event has a correlation coefficient of 1 even if the signal magnitude differs depending on the distance from the event occurrence position to the sensor. It is to be. Also, since electric field changes at sufficiently low frequencies can be considered to conduct space instantaneously (by quasi-static approximation), their correlation occurs without time lag. From the third feature, the spatial position of the event can be estimated by extracting and analyzing the correlated signals without a time difference from the signals obtained by each sensor. In the present embodiment, the estimation of the spatial position of the event is performed by the signal processing unit 2.

The AD converter 16 generates an intensity signal sequence S from the signals obtained by the respective sensors at a sampling rate of 1 kHz. In this embodiment, the time window of the intensity signal sequence S is set to 0.1 milliseconds. Assuming that the time window is 0.1 ms, each intensity signal sequence S is represented as a sequence of 100 sequences (0.1 ms × 1000 Hz = 100) as shown in the following equation (6).

Here, the time window is the time given at the sampling rate, and the time frame is given by multiplying the time window by the number of elements of the intensity signal sequence S. Further, here, an example in which the number M of the sensors is 6 will be described.

The signal processing unit 2 detects the number of events ξ, the time of occurrence, and the location of occurrence of the event that occurred within the time frame from the correlation between the intensity signal sequences S of the above equation (6).

In particular, the processing in the signal processing unit 2 will be described with reference to FIG. FIG. 3 is a processing flow diagram for explaining the processing by the signal processing unit 2 and the processing for detecting the occurrence number ξ, the occurrence time, and the occurrence location of the event that occurred within the time frame.

The event sequence acquisition means 21 included in the signal processing unit 2 first determines whether or not an event has occurred (S10). For a certain time frame, the event sequence acquisition means 21 cannot obtain a signal from any of the acquired intensity signal sequences S1 to S6, that is, in the equation (6), all the elements of the intensity signal sequences S1 to S6 are 0. In this case, it is determined that no event has occurred in the time frame, that is, ξ = 0. In this case (No), nothing is performed and the determination in the next time frame is performed.

When any signal of the intensity signal sequences S1 to S6 is obtained as a result of the determination in S10 (Yes), the event sequence acquisition means 21 assumes that an event has occurred and is a signal of the intensity signal sequences S1 to S6. The correlation coefficient is calculated between the signals, except for those for which is not obtained (S20). The correlation coefficient r _nm of Sn and Sm is calculated by the following equation (7). The overlined Sn and Sm are the average values of Sn and Sm, respectively.

Next, it is determined whether or not this correlation function r _nm is larger than a predetermined threshold value (S30). For example, when the correlation function r _nm is 0.9 or more and all of the correlation functions r _nm can be regarded as 1, the number of events occurring in the time window (number of events) ξ is determined to be 1, that is, ξ = 1. To. In this case (Yes), all the intensity signal sequences S1 to S6 are represented by a proportional relationship with one signal sequence Ivent1 as shown in the following equation (8).

The elements c _{1 to} c ₆ of the coefficient matrix on the right side of the above equation (8) are all real numbers of 0 or more, and represent the intensity ratio between each signal. The position estimation means 22 included in the signal processing unit 2 obtains this proportionality constant and specifies the position of the event by the following equation (9) (S40).

Here, L _{1 to} L ₆ are distances from the event occurrence position to the sensors S _{1 to S 6} . However, the coefficients c _{1 to} c ₆ that are zero are excluded from the proportional relationship in the above equation (9). For example, when c ₅ = 0 and c ₆ = 0, the position of the event is specified from this equation (10) by expressing it as the following equation (10).

The above equations (9) and (10) utilize the fact that the intensity of the electric field change felt by the sensor is inversely proportional to the square of the distance between the sensor and the event. However, in this embodiment, it is assumed that L _{1 to} L ₆ are not inversely proportional to the square of the coefficients c _{1 to} c ₆ but inversely proportional to the 2.1 power. This is because the influence of the grounding of the floor surface is taken into consideration. The position estimation means 22 derives the coordinates (x, y) of the event occurrence position based on the information regarding L _{1 to} L ₆ , and records them in the first memory 23 together with the event occurrence time t.

Further, the virtual person allocating means 24 included in the signal processing unit 2 allocates the coordinates (x, y) of the event occurrence position and the event occurrence time t as virtual person data already defined or defined at that time. (S200), recording is performed in the second memory 25. When assigning a new event (position, occurrence time) to an already defined virtual person, the event that has already been assigned is overwritten. This indicates that the position of the virtual person has been updated.

The location of the virtual person is visualized and displayed by the display unit 3 according to the specification mode, or is transmitted on the Internet by the transmission unit 4.

The above is the case where the correlation coefficient between the intensity signal sequences S is 1, and it is determined that only one event has occurred in the detection area during the time frame.

As a result of the determination in S30, when it is determined that the correlation coefficient between the intensity signal sequences S is equal to or less than a predetermined threshold value, that is, it cannot be regarded as 1 (No), the event ξ generated within the time frame is increased by one. Let ξ = 2 (S50) and try to explain the signal sequences S1 to S6. Therefore, the event column acquisition means 21 searches for the event signal sequence, and is obtained as the calculation result of the following equation (11) in the matrix A of M rows ξ columns (here, 6 rows and 2 columns). The coefficient of the matrix A is determined so that the correlation coefficient between the two event signal sequences I ₁ and I ₂ is minimized (S60). As a method for determining the coefficients of the matrix A, any suitable conventionally known method may be used, including the case where those coefficients are complex numbers.

Here, the 12 elements of the matrix A can all be limited to 0 or a real number greater than 0. Such a limitation is possible because the changes in the event signal sequences I ₁ and I ₂ are transmitted to the intensity signal sequences S1 to S6 without a time difference, and the influence thereof increases or decreases depending on the distance between the intensity signal sequences S1 to S6 and the signal source. , This is because the sign is not reversed. This limitation allows the elements of matrix A to be determined with less computational complexity.

After the elements of the matrix A and the event signal sequences I ₁ and I ₂ at that time are determined, the event sequence acquisition means 21 further determines the event signal sequences I ₁ and I ₂ at that time and the correlation coefficient r between them ( 2) is stored (S70). The event signal sequences I ₁ and I ₂ are signal sequences that are supposed to represent the respective events.

The threshold value in S30 can be determined in advance by using the samples when the number of events ξ is 1 and 2.

Subsequently, the event sequence acquisition means 21 increases the number of events ξ that occurred within the time frame by one so that ξ = 3 (S80). Then, the event signal sequence is searched, and in the matrix A of M rows and ξ columns (here, 6 rows and 3 columns), the three event signal sequences I ₁ and I obtained as the calculation results of the following equation (12) are obtained. _2. Determine the elements of the matrix A such that the average value of the absolute values of the correlation coefficients between I ₃ is minimized (S90).

That is, from the event signal sequences I ₁ , I ₂ , I ₃ obtained by the above equation (12), the correlation coefficient between I ₁ and I ₂ , the correlation coefficient between I ₂ and I ₃ , I ₃ and I ₁ The arithmetic mean of the absolute values of the three correlation coefficients between them is calculated, and the matrix A having the smallest value is determined. After the elements of such a matrix A and the event signal sequences I _{1 to} I ₃ at that time are determined, the event sequence acquisition means 21 has the event signal sequences I _{1 to} I ₃ at that time and the correlation coefficient between them. Record r (3) in the first memory 23 (S100).

Next, the event sequence acquisition means 21 compares r (ξ) and r (ξ-1) (S110). Here, when r (ξ)> r (ξ-1), that is, when r (3) is greater than r (2) (Yes), ξ = 2, that is, the number of events that occurred within the time frame is 2. Judge as an individual. When r (3) is smaller than r (2) (No), the number of assumed events occurring in the time frame is further increased to ξ = 4 (S80), and in an arbitrary 6-by-4 matrix A, the following equation is used. The coefficient of the matrix A is determined so that the average value of the absolute values of the correlation coefficients between the _four signal sequences I _{1 to} I ₄ obtained as the calculation result of (13) is minimized (S90).

The event sequence acquisition means 21 records the minimum arithmetic mean r (4) in the first memory 23 (S100). Also, record I _{1 to} I ₄ at that time. Comparing r (ξ) and r (ξ-1) (S110), when r (4) is greater than r (3) (Yes), ξ = 3, that is, the number of events occurring in the time frame is 3. Judge as an individual. When r (4) is r (3) or less (No), the number of assumed events is increased by one.

Similarly, in S80 to S110, the average value r (ξ) of the absolute value of the correlation function of the number of events ξ is larger than the average value r (ξ-1) of the absolute values of the correlation function when the number of events ξ is 1 less. Repeat until it becomes. The number of events is determined when r (ξ) > r (ξ-1). That is, in this embodiment, when the average of the absolute values of the correlation coefficients takes the minimum value with respect to the increase in the number of events, the number of events is determined as the number of events ξ in the time frame.

Through the above process , the number of events ξ that occurred in the time frame and the event signal sequence I indicating each event are obtained. When the number of events ξ is n, the following equation (14) is obtained.

From the above equation (14), the equation (15) when the correlation of the event signal sequence is minimized can be obtained.

Further, a 6-row, n-column matrix that reconstructs the original signals S1, S2, S3, S4, S5, and S6 is obtained. That is, the matrix A'that satisfies the following equation (16) is obtained (S120).

The position where the first event occurs is obtained from the ratio of the coefficients in the first column of the matrix A', the position where the second event occurs is obtained from the ratio of the coefficients in the second column, and so on, for each of the n events. (S130).

Further, the virtual person allocating means 24 included in the signal processing unit 2 allocates the coordinates (x, y) of the event occurrence position and the event occurrence time t as virtual person data already defined or defined at that time. (S200), recording is performed in the second memory 25. When assigning a new event (position, occurrence time) to an already defined virtual person, the already assigned event is overwritten. This indicates that the position of the virtual person has been updated.

As described above, in the first embodiment, by extracting the correlated components between the signals from the signals generated by the plurality of electric field sensors, the fluctuation event of the electric field generated in the walking of the human body, the contact with other objects, etc. is generated. By extracting, estimating the position where the event occurred, and assigning the event to the virtual person, the position of the human body at a predetermined time can be estimated. Further, since a plurality of events can be separated for each event, even when a plurality of human bodies exist, their movements can be identified for each individual.

Further, the signal obtained here is an electric field fluctuation event caused by human walking, contact with another object, or the like, and the fluctuation amount is larger than the electric field fluctuation event caused by the parallel movement of the human body. Therefore, since the decrease in the electric field fluctuation due to the distance between the event occurrence location and the electric field sensor is small, it is possible to detect a distant event. Therefore, the number of sensors used in the human detection device can be small, or specially. There is an effect that it is not necessary to use a high-sensitivity sensor.

In addition, if the electric field sensor is installed so as to detect the electric field in the direction perpendicular to the floor, the sensor detects the electric field equally regardless of the direction in which the electric field fluctuates with respect to the sensor. That is, there is no directionality. Therefore, there is an effect that the position estimation of the event by the signal processing circuit becomes more precise.

(Second Embodiment)
A second embodiment of the human detection device will be described with reference to FIG. FIG. 4 is a schematic configuration diagram of a second embodiment of the human detection device of the present invention.

The second embodiment is different from the first embodiment in that low frequency components are used together. Since other configurations and operations are the same as those in the first embodiment, duplicate description and illustration may be omitted.

In the second embodiment, the bandpass filter 14 separates the high frequency component larger than 5 Hz and the low frequency component smaller than 5 Hz, and performs different signal processing for each. The high frequency component is processed in the same manner as in the first embodiment. Further, regarding the low frequency component, the configuration and operation until it is sent to the low frequency signal processing unit 122 via the amplifier circuit 115 and the AD converter 116 are the same as those of the high frequency component, except that the frequency band is low frequency. Is.

The operation of the high frequency signal processing unit 120 is the same as that of the signal processing unit 2 of the first embodiment. When the presence of a person is recognized by the high frequency component, for example, at points P1, P2, and P3, the total charge and potential of each person from the low frequency components (that is, DC components) of three nearby sensors. Can be estimated. The electric fields E ₁ , E ₂ , and E ₃ at points P1, P2, and P3 are represented by the following equations (17), respectively.

Here _{Q 1,} Q 2, _{Q 3} is the charge of the person in the P1, P2, P3 _{point, L ij} is the distance from the sensor _{S i} until _{P j.} L _ij is because it is known by the signal analysis of the high-frequency component, by solving the simultaneous equations of the equation _(17), it can be obtained _{Q 1, Q 2, Q 3} . Moreover, the electric potential of each person can be estimated by giving the volume component of the human body. The electric charge and electric potential of the human body can be used as information about each person by linking them to each person to identify each person, or as a barometer of each person's activity such as when an exposed part such as a hand comes into contact with something. it can. Further, in the first embodiment, even when a person's walking or the like is missed or misidentified, the information due to the high frequency can be corrected by capturing the decrease in the electric field in the low frequency component.

According to the second embodiment, the output of the electric field sensor is filtered, and the signal of the low frequency component and the location information of the person provided by the high frequency component are used together, so that the total charge, potential, etc. of each person are obtained and each person is obtained. It can be used for identification of. In addition, it is possible to correct the error of the location information of the person caused by the high frequency component and to detect the person at a higher level.

(Other embodiments)
In the first embodiment and the second embodiment, an example in which an electric field sensor is used to measure a change in an electric potential (electric field) has been described, but if the change in the electric potential (electric field) can be detected, the type is It doesn't stop there. For example, it is possible to measure the induced current flowing into the electrode by grounding the electrode, and not only the potential of the potential sensor or the electric field sensor, the output proportional to the electric field, but also the current and voltage generated by the operation of those sensors. It can also be a sensor that measures.

For example, when an electric field sensor is used, when a potential sensor is used, each potential sensor has two outputs, an output proportional to the potential and an output proportional to the time change of the potential. The output proportional to the change in potential with time passes through a bandpass filter that transmits frequencies between 5 Hz and 1000 Hz, and then, as in the first and second embodiments, human walking, contact with other objects, etc. Extract the events that occur in. The output proportional to the potential is used to capture and correct the information regarding the staffing obtained in the former, as in the low frequency output of the second embodiment.

In this case, there are the following differences because the fluctuation of the potential is observed instead of the electric field. That is, as shown in the following equation (18), the electric field decreases in inverse proportion to the square of the distance L, but the potential decreases in inverse proportion to L.

Therefore, when the potential sensor is used, the output difference of the sensor does not differ significantly from the place away from the sensor and the place close to the sensor as in the first embodiment and the second embodiment, and the position estimation becomes more precise in many cases. There is an advantage.

Regarding the process of acquiring the event signal sequence, many methods of estimating the original signal that brought them from a plurality of signals such as multi-component analysis and independent component analysis have been proposed. In the present invention, the method is not limited, and any suitable conventionally known method can be used.

In addition, there are many conceivable rules for assigning an event that occurs at a certain point in time to a virtual person, and for determining whether a certain event and another event are due to the same person. For example, when a signal related to walking occurs continuously within a certain distance, it may be an event related to the same person. Any suitable method may be used for this specific algorithm.

1 Sensor unit 2 Signal processing unit 3 Display unit 4 Transmitter unit 11 Electrostatic measuring instrument 12 Voltage follower 13 Power frequency removal filter 14 Bandpass filter 15, 115 Amplifier circuit 16, 116 AD converter 120 High frequency signal processing unit 122 Low frequency signal processing Department

Claims (6)

The process of acquiring an intensity signal sequence indicating the signal intensity from multiple sensors,
The process of calculating the correlation coefficient between intensity signal sequences and
The process of comparing the correlation coefficient with a predetermined threshold and
As a result of the above comparison, the process of specifying the event position from the signal intensity ratio, which is performed when the value is equal to or higher than the threshold value,
It includes a process for determining multiple events, which is performed when the value is smaller than the threshold value.
The process of determining multiple events is
Let the number of events ξ be 2
Correlation coefficient r (ξ) between I = AS given by ξ event signal strings I, coefficient matrix A of ξ rows and M columns, and intensity signal sequence S of M rows and 1 column. The first process of determining the coefficient matrix A that minimizes
From I = AS, which is given by adding 1 to the number of events ξ, ξ event signal strings I, the coefficient matrix A of ξ rows and M columns, and the intensity signal sequence S of M rows and 1 column, the event signal sequence. The second process of determining the coefficient matrix A that minimizes the correlation coefficient r (ξ) between I and
With the process of comparing r (ξ) and r (ξ-1),
When r (ξ) is r (ξ-1) or less , the second process and the process of comparing r (ξ) and r (ξ-1) are further performed.
In other cases, the person detection method is characterized in that the event position is specified from the coefficient of each row of the coefficient matrix A obtained when the coefficient is ξ-1 or less. The person detection method according to claim 1, wherein after the event position is specified, the specified event is assigned to a virtual person. With multiple static electricity measuring instruments
A bandpass filter that transmits high-frequency components of the outputs of the plurality of static electricity measuring instruments,
An AD converter that generates an intensity signal sequence from a signal that has passed through the bandpass filter,
Equipped with a signal processing unit
The signal processing unit
A person detection device comprising means for executing each procedure of the person detection method according to claim 1 or 2. The person detection device according to claim 3, wherein the bandpass filter transmits 5 to 1000 Hz. Equipped with an AD converter for low frequencies and a signal processing unit
The bandpass filter transmits 5 to 1000 Hz and separates and outputs low frequency components of less than 5 Hz.
The AD converter for low frequencies generates an intensity signal sequence of low frequency components.
The person detection device according to claim 3, wherein the low-frequency signal processing unit acquires the total charge amount of the signal source from the intensity signal sequence of the low-frequency component. The person detection device according to any one of claims 3 to 5, wherein the static electricity measuring device is any one or a combination of an electric field sensor, a potential sensor, and an induced current monitoring sensor.

Images