JP5969818B2 - Behavior detection system - Google Patents

Description

  The present invention relates to a behavior detection system, and more particularly, to a behavior detection system capable of distinguishing each target and detecting each behavior even when there are a plurality of targets, for example, people, in the region to be observed.

  A conventional behavior detection system identifies a target behavior based on a temporal change in Doppler frequency (see, for example, Non-Patent Document 1).

Yoshihisa Okamoto et al., “Position Estimation Algorithm Based on Space-Time Virtual Array Configuration Using Doppler Shift with Human Movement”, IEICE Technical Report, vol. 111, no. 180, RCS2011-135, pp.135-140 , August 2011

  However, when there are multiple targets in the observation area, each echo wave is synthesized and received, so the position coordinates cannot be estimated by distinguishing each target. The behavior cannot be detected.

  In view of the above-described problems, the present invention is capable of detecting a behavior by estimating a position coordinate by distinguishing each target and detecting each behavior by distinguishing each target even when there are a plurality of targets in the observation region. The purpose is to provide a system.

  The behavior detection system of the present invention includes n (n is an integer of 3 or more) sensors that transmit signals of frequencies different from each other and receive signals of all the frequencies, and each frequency from the signal received by the sensor. A filter that separates the Doppler signals of the signal, a correlator that correlates two Doppler signals that pass through the same path among the Doppler signals separated by the filter, and two correlation signals obtained by the correlator Distance sum calculation means for calculating the sum of the distances from a phase difference corresponding to the sum of the distances from the sensor to two or more and {C (n, 2) +1} / 2 or less targets; and each target An association means for associating each Doppler signal related to an arbitrary target between the sensors based on temporal characteristics of the Doppler signals of Relative speed calculation means for calculating the relative speed of each target with each of the sensors based on each Doppler signal associated by the stage, and the sum of the distances calculated by the distance sum calculation means and the relative speed calculation means And a position estimating means for estimating the position of each target based on the relative speed.

  The behavior detection system according to the present invention includes n (n is an integer of 3 or more) sensors that transmit signals of different frequencies and receive signals of all the frequencies, and signals received by the sensors. A filter that separates the Doppler signals at each frequency, a correlator that correlates two Doppler signals that pass through the same path among the Doppler signals separated by the filter, and 2 obtained by the correlator. Distance sum calculation means for calculating the sum of the distances from a phase difference corresponding to the sum of the distances from the sensors to two or more and {C (n, 2) +1} / 2 or less targets; One-dimensional space generation means for generating a one-dimensional space that is a candidate for each target based on the sum of the distances calculated by the distance sum calculation means; An association means for associating each Doppler signal for an arbitrary target between each of the sensors based on a temporal characteristic of the error signal, and each sensor of each target based on each Doppler signal associated by the association means A relative speed calculating means for calculating a relative speed; a position estimating means for estimating the position of each target based on the one-dimensional space generated by the one-dimensional space generating means and the relative speed calculated by the relative speed calculating means; It is characterized by providing.

  According to the present invention, even when there are a plurality of targets in the observation region, the position coordinates can be estimated by distinguishing each target, and the behavior can be detected by distinguishing each target.

It is a flowchart explaining operation | movement of the action detection system by one Example of this invention. It is a figure which shows the basic composition of the bistatic radar model used by a present Example. It is a figure which shows the structure of a bistatic 3 frequency CW radar. It is a figure explaining the case where there is no consistency between relative velocities. It is a figure explaining the case where there is consistency between relative velocities. It is a figure which shows the estimation result and true value of the Doppler spectrum regarding the sensors 1 and 2. FIG. It is a figure which shows the result of having estimated the position coordinate of the target.

  DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the accompanying drawings.

  FIG. 1 is a flowchart for explaining the operation of the behavior detection system according to the embodiment of the present invention. FIG. 2 is a diagram showing the basic configuration of the bistatic radar model used in this embodiment. As shown in FIG. 2A, the bistatic radar model transmits a signal from the transmission antenna Tx and receives the signal reflected by the target by the reception antenna Rx. At this time, the total distance from the transmission antenna Tx to the reception antenna Rx is obtained based on the time difference between the transmission signal and the reception signal, that is, the phase difference of the signal. Based on the frequency difference between the transmission signal and the reception signal, that is, the Doppler frequency, the sum of the relative speeds seen from the transmission antenna Tx and the reception antenna Rx is obtained. As shown in FIG. 2B, the relative speed is a line-of-sight direction component from each antenna of the absolute speed. Thus, with this basic configuration alone, it is not possible to estimate the position of the target or the speed thereof. Therefore, a configuration in which the number of antennas is increased is adopted.

FIG. 3 is a diagram showing a configuration of a bistatic three-frequency CW radar. A transmission / reception antenna is installed at the position of sensors 1 to 3, and signals of different frequencies are transmitted from the sensors 1 to 3. For example, a signal of f1 = 24.150 GHz from sensor 1, f2 = 24.155 GHz from sensor 2, and f3 = 24.160 GHz from sensor 3 is transmitted. When there are two persons A and B as targets, the sensors 1 to 3 receive sum signals of six paths (ie, paths) (step S1).

Here, the Doppler signal is separated from the signal of each frequency by a filter (step S2). Specifically, this filter includes a mixer and a low-pass filter corresponding to each of the frequencies f1 to f3. Focusing on the path between sensors 1 and 2, the signal of frequency f1 is via A (Doppler frequency, phase delay) + via B (Doppler frequency, phase delay).
The sum of the two signals is received, and the signal of frequency f2 also passes through A (Doppler frequency, phase delay) + B (Doppler frequency, phase delay)
The sum of the two signals is received as follows. These are both the sum of the signal of the path via A and the signal of the path via B between the sensors 1 and 2, and these sums are two signals that pass through the same plurality of paths. The two signals are correlated (step S3), that is, when these signals are multiplied, the Doppler frequency via A and the Doppler frequency via B are considered to be uncorrelated. The Doppler frequency signals are multiplied together, and the phase delay is obtained because the Doppler frequency is almost equal regardless of the carrier frequency,
Cross-correlation value via A + cross-correlation value via B =
exp {j2π (f2−f1) (R1A + R2A) / c} + exp {j2π (f2−f1) (R1B + R2B) / c}
= Expφ1 + expφ2
(1)
R1A: Distance between sensor 1 and A
R2A: Distance between sensor 2 and A
R1B: Distance between sensor 1 and B
R2B: Distance between sensor 2 and B
(sinφ1 + sinφ2) / (cosφ1 + cosφ2) = tan {(φ1 + φ2) / 2}
(2)
(Φ1 + φ2) is obtained (step S4), and (R1A + R2A + R1B + R2B) is obtained (step S5). Thus, it can be seen that (the sum of the distances between the sensors 1 and 2 and the targets A and B) is obtained. Similarly, (the sum of the distances between the sensors 1 and 3 and the targets A and B) and (the sum of the distances between the sensors 2 and 3 and the targets A and B) can be obtained, so that three equations can be established. it can. In order to obtain the X and Y coordinates of the targets A and B, since there is only one equation for the four variables, a one-dimensional space can be generated as a candidate for the position where the targets A and B exist. (Step S6).

Here, the Doppler spectrum acquired by each sensor contains frequency shift information related to multiple human movements, but it is arbitrary between each sensor based on the temporal spectrum behavior related to the rotational motion of arms and legs. The frequency shift information regarding the target can be associated (step S7). For this reason, the Doppler signal when the signal transmitted by the sensor 1 is received by the sensor 2 can be divided into the targets A and B separately. Then, for target A,
fd12 = (fd1 + fd2) / 2
fd23 = (fd2 + fd3) / 2
fd31 = (fd3 + fd1) / 2
(3)
Where fd12 is the Doppler frequency between the sensors 1 and 2
Since fd1 and the like are Doppler frequencies for the sensor 1,
fd1 = (fd31 + fd12−fd23)
fd2 = (fd12 + fd23−fd31)
fd3 = (fd23 + fd31-fd12)
(4)
Therefore, the relative speed with respect to each sensor can be calculated from these Doppler frequencies, and the relative speed can be similarly calculated for the target B (step S8).

  Here, three relative velocities were obtained for each sensor with respect to the one-dimensional space as a candidate for the position where the targets A and B exist and the targets A and B, respectively. Therefore, while checking whether there is consistency between the three relative velocities, the one-dimensional space is scanned, that is, an absolute value generated by two relative velocities with different combinations among the three relative velocities. Check if the speeds match each other. A position where there is consistency between the three relative velocities is estimated as a position where each of the targets A and B exists (step S9).

  FIG. 4 is a diagram illustrating a case where there is no consistency between relative velocities. In the one-dimensional space, the absolute speed generated by the relative speed with respect to the sensor 1 and the relative speed with respect to the sensor 2 is different from the absolute speed generated by the relative speed with respect to the sensor 1 and the relative speed with respect to the sensor 3. It can be determined that there is no target at this position.

  FIG. 5 is a diagram illustrating a case where there is consistency between relative velocities. In the one-dimensional space, the absolute speed generated by the relative speed with respect to the sensor 1 and the relative speed with respect to the sensor 2 and the absolute speed generated by the relative speed with respect to the sensor 2 and the relative speed with respect to the sensor 3 match. It can be estimated that the target (A or B) exists at the position.

  In this way, the positions of the targets A and B can be estimated. Further, based on the Doppler frequency / time characteristics of the targets A and B, how the targets A and B behave at those positions can be determined. Can be estimated.

  Next, the result of computer simulation was shown. Human position estimation and action identification were performed using a three-frequency CW radar model. Three sensors were placed in the corner of the room to radiate radio waves. A 10 m × 10 m room was assumed. It was a model in which two people were walking in the room at the same time. The human body was reproduced by configuring it with 80 scatters. The scatter is a virtual scatterer (point) for simulation having a position of infinitely small size and a predetermined reflection cross section. Each scatter has a three-dimensional position vector and velocity vector. A reflection cross section was taken into account by adding a weight to each scatter. Table 1 shows the specifications of the simulation.

FIG. 6A is a diagram illustrating the Doppler spectrum estimation results for the sensors 1 and 2. The human walking interval and relative velocity are estimated from the Doppler spectrum, and the human movement can be identified from the temporal shift of the Doppler frequency related to the periodicity of the movement and the translational movement. FIG. 6 (b) shows the true value of the relative velocity when the same parameters as the Doppler spectrum shown in FIG. 6 (a) are used. The Doppler spectrum can estimate the relative velocity for each sensor almost accurately. In addition, since there are two continuous frequencies with high received power values in the Doppler spectrum, it can be estimated that there are two targets in the observation region. Further, by looking at the peak interval of the spectrum, it can be seen that both are moving periodically, and it can be estimated that each movement period is 1.2 seconds and 1.1 seconds, respectively. The term “periodic exercise” as used herein refers to a repetitive motion such as walking or running, and does not belong to a behavior while sitting on a chair or a peculiar behavior. Also, since the maximum value of the Doppler frequency for translational motion is 200 Hz,
fd / fc = 2vr / c [Doppler frequency / carrier frequency = 2 × speed / light speed]
(5)
Since it can be estimated that the relative speed of the person is 1.3 m / s using the relational expression, it can be estimated here that the movement is closer to walking than to running.

  FIG. 7 is a diagram showing a result of estimating the position coordinates (flow line) of a target existing in the observation region using two kinds of evaluation methods. Regarding the evaluation method, the case where two targets exist in the observation area is type 1, and the case where only one target exists is type 2. As can be seen from FIG. 7, in both types, the position coordinate estimation error is suppressed to within 1 m, and for type 1, the position coordinates can be estimated by separating a plurality of targets. In Type 2, since the distance estimation value includes only information on one target, the position coordinates of the target can be estimated directly without scanning on the one-dimensional candidate space by the relative velocity vector, and almost accurate tracking can be performed. On the other hand, in Type 1, since the position is estimated by scanning the one-dimensional candidate space using a plurality of relative velocity vectors, unlike the conventional method, the movement of each target and its position coordinates are estimated in complete synchronization. It became possible to do. By using the present invention, it is possible to identify when, where, and what a plurality of people are doing by using only radio waves by completely synchronizing the action identification result and the position coordinate estimation result.

In addition, this invention is not limited to the said Example.
If the number of sensors is increased, the target position is uniquely determined. Therefore, it is not always necessary to create a one-dimensional space. Even in this case, in order to identify which target is located at which position among a plurality of targets, the above-mentioned association and relative speed consistency determination (steps S7 to S9) are required.

On the other hand, if the configuration is such that a one-dimensional space is generated, the number of sensors can be reduced. In this case, if the number of sensors is n (n is an integer of 3 or more), the number of independent paths is the number of combinations C (n, 2), and the same number as the combination C (n, 2). When the number of targets is m (m is an integer of 2 or more), the number of independent variables to be calculated is 2m, so the number of sensors n is C (n, 2) ≧ (2m− 1)
{C (n, 2) +1} / 2 ≧ m
(6)
Any number that satisfies this requirement is sufficient.

Claims (2)

N sensors (n is an integer of 3 or more) for transmitting signals of frequencies different from each other and receiving signals of all those frequencies;
A filter that separates the Doppler signal of each frequency from the signal received by the sensor;
Correlation means for correlating the Doppler signals of two signals passing through the same plurality of paths among the Doppler signals separated by the filter;
From the phase difference corresponding to the sum of the distances from the two sensors to two or more and {C (n, 2) +1} / 2 or less targets obtained by the correlation means, the sum of the distances is calculated. A distance sum calculating means for calculating;
An association means for associating each Doppler signal for an arbitrary target between each of the sensors based on temporal characteristics of the Doppler signal for each target;
A relative speed calculation means for calculating a relative speed of each target with each sensor based on each Doppler signal associated by the association means;
A behavior detection system comprising: position estimation means for estimating the position of each target based on the sum of distances calculated by the distance sum calculation means and the relative speed calculated by the relative speed calculation means. N sensors (n is an integer of 3 or more) for transmitting signals of frequencies different from each other and receiving signals of all those frequencies;
A filter that separates the Doppler signal of each frequency from the signal received by the sensor;
Correlation means for correlating the Doppler signals of two signals passing through the same plurality of paths among the Doppler signals separated by the filter;
From the phase difference corresponding to the sum of the distances from the two sensors to two or more and {C (n, 2) +1} / 2 or less targets obtained by the correlation means, the sum of the distances is calculated. A distance sum calculating means for calculating;
One-dimensional space generation means for generating a one-dimensional space that is a candidate in which each target exists based on the sum of distances calculated by the distance sum calculation means;
An association means for associating each Doppler signal for an arbitrary target between each of the sensors based on temporal characteristics of the Doppler signal for each target;
A relative speed calculation means for calculating a relative speed of each target with each sensor based on each Doppler signal associated by the association means;
A behavior detecting system comprising: a one-dimensional space generated by the one-dimensional space generating means; and a position estimating means for estimating the position of each target based on the relative speed calculated by the relative speed calculating means.