JP5406616B2 - Event detection device - Google Patents

Description

  In the present invention, in a predetermined area such as a room, a transmitter that transmits radio waves and a receiver that receives radio waves transmitted from the transmitter are arranged, and a person has moved based on the reception characteristics of the radio waves. More particularly, the present invention relates to an event detection device that can particularly emphasize or mask an event in a specific area.

  The present inventors have proposed the following event detection device (see Patent Document 1).

  A plurality of antennas for receiving radio waves transmitted by the transmitter, a correlation matrix calculating means for calculating a correlation matrix from the received vector using signals received by the plurality of antennas as reception vectors, and a calculation performed by the correlation matrix calculating means An event detection apparatus comprising: eigenvector computing means for computing eigenvectors that expand a signal subspace by expanding eigenvalues of a correlation matrix; and event detection means for detecting an event by detecting a temporal change of the eigenvector computed by the eigenvector computing means.

JP 2008-216152 A

  However, the above-described event detection device cannot detect, for example, misplacement of an object in a specific area, and cannot detect even a small change event in the specific area. In addition, for example, it is desired not to detect even if there is a pet, and it is not possible to prevent detection of a specific area even if there is a large change event. That is, it is not possible to set the sensitivity according to the area.

  In view of the above problems, the present invention can detect an event with a small change according to an area by emphasizing or masking (ignoring) an event to be detected according to the area, An object of the present invention is to provide an event detection device capable of masking an event.

Event detection apparatus of the present invention includes a plurality of antennas for receiving radio waves transmitted from the transmitter, a correlation matrix calculating means for calculating a correlation matrix from the received vector signals received by the plurality of antennas as reception vector, the calculating a mapping matrix for dividing the signal subspace based on the arrival direction of the combination of the radio waves received by the antenna, a filter means for applying said mapping matrix to the correlation matrix calculated by the correlation matrix calculating unit, said filter means An eigenvector computing means for computing eigenvectors related to eigenvalues obtained by eigenvalue expansion of the correlation matrix multiplied by the mapping matrix, and an event for detecting an event by detecting temporal changes of the eigenvectors computed by the eigenvector computing means And a detecting means.

  The plurality of antennas are array antennas, so that general-purpose array antennas can be used.

  Further, the eigenvector calculation means can increase the reliability by calculating the eigenvector corresponding to the maximum eigenvalue of the correlation matrix.

  The event detection means can simplify the calculation by comparing the inner product of the eigenvector during normal time and the eigenvector during observation with a predetermined threshold.

According to the present invention, the event detection apparatus can detect an event with a small change according to an area by emphasizing or masking an event to be detected according to the area, or an event with a large change. It can also be masked. Further, the filter means calculates the mapping matrix based on the combination of the arrival directions of the radio waves received by the antenna, so that an effective mapping matrix can be obtained with a simple calculation.

It is a figure which shows the structure of the event detection apparatus by one Example of this invention. It is a figure which shows the environment where the experiment of the present Example was conducted. It is a figure which shows the relationship between the angle spectrum by MUSIC method, and an estimated radio wave arrival direction. It is a figure which shows the example of the evaluation result at the time of emphasizing a static event. It is a figure which shows the example of the evaluation result at the time of masking a static event. It is a figure which shows the example of the evaluation result at the time of emphasizing a dynamic event. It is a figure which shows the example of the evaluation result at the time of masking a dynamic event.

  The best mode for carrying out the present invention will be described below in detail with reference to the accompanying drawings.

  FIG. 1 is a diagram showing a configuration of an event detection apparatus according to an embodiment of the present invention. The event detection apparatus of this embodiment includes a transmitter 10 and a receiver 20. The transmitter 10 and the receiver 20 are installed in a predetermined area in order to detect an event such as movement of a person or opening / closing of a door. A closed space such as a room is desirable, but it may be an open area. The transmitter 10 transmits radio waves. The receiver 20 includes an array antenna 21, a correlation matrix calculation unit 22, a filter unit 23, an eigenvector calculation unit 24, and an event detection unit 25. The array antenna 21 includes a plurality of antenna elements, and each antenna element receives a radio wave transmitted by the transmitter 10.

In the present invention, the shape of the array antenna 21 is arbitrary, but here, an equally spaced circular array used in the experiment is used as a model. Since the element used in the experiment is a dipole antenna and a linear element, the elevation angle is considered only in the 0 degree direction. The incoming wave s (t) assumes a far field and becomes a plane wave on the receiver side. The noise n (t) is assumed to be additive white Gaussian noise (AWGN) having an average of 0 and a variance σ ² . When the phase reference of the incoming wave is set at the center of the array circle and θ is taken clockwise with respect to the direction from the # 1 antenna element toward the center, reception of the array antenna 21 with respect to the plane wave coming from the direction of θ The signal is represented by a reception vector x → (t) having reception signals of the antenna elements of the array antenna 21 as elements. Here, “→” indicates that the left character in the sentence is a vector (or matrix).
x → (t) = a → (θ) s (t) + n → (t) (1)
Where a → (θ): L-dimensional vector when the number of antenna elements is L s (t): received signal at the reference point at the center of the array circle n → (t): noise

  Here, a → (θ) is called a steering vector, which is a complex vector having as its component the phase difference between the incoming wave received at the reference point and the incoming wave received at each antenna element. It can be expressed as follows.

（２）
ただし、ν：電波の波長
ｒ：アレイ円の半径
Ｔ：転置

(2)
Where ν: wavelength of radio wave r: radius of array circle T: transposition

  Here, when M arriving waves arrive as plane waves from the direction of θi (i = 1 to M), the received vector x → (t) is

（３）
ただし、αi：ｉ番目の到来波の振幅及び位相を表す複素数

(3)
Where αi is a complex number representing the amplitude and phase of the i-th incoming wave

  a → ′ is a new steering vector, and is composed of a linear combination of the steering vectors of each incoming wave, and depends on the arrival direction, power, and phase of the incoming wave.

  For the radio wave propagation analysis, the spatial correlation between the antenna elements of the array antenna 21 is examined. For this purpose, the correlation matrix calculation means 22 generates a correlation matrix from the received vector x → (t). The correlation matrix R → xx can be expressed as Equation (4).

（４）
ただし、Ｅ［・］：アンサンブル平均
Ｈ：複素共役転置（エルミート転置）
Ｓ→：Ｅ［ｓ→(ｔ)ｓ→(ｔ)^H］
ｓ→(ｔ)：基準点において複数信号源を想定した信号源数の次元の受信信号ベクトル
Ｉ→：単位行列

(4)
Where E [•]: ensemble average H: complex conjugate transpose (Hermitian transpose)
S →: E [s → (t) s → (t) ^H ]
s → (t): received signal vector having the number of signal sources assuming a plurality of signal sources at the reference point I →: unit matrix

  Noise is assumed to be independent between each element. When actually obtaining the correlation matrix, the ensemble average in equation (4) is replaced with the time average based on ergodicity. The correlation matrix R → xx is estimated. When sampled at time t = t1, t2,..., TN, the correlation matrix estimate R → ^ xx can be expressed by equation (5). Here, “^” represents that the left character in the sentence is an estimated value.

（５）
ここで、Ｎ：スナップショット数

(5)
Where N: number of snapshots

  As the number of snapshots increases, the estimation accuracy of R → ^ xx increases. Thereafter, R → ^ xx is treated as R → xx.

  The subspace method is a radio wave propagation analysis method focusing on the properties of eigenvalues and eigenvectors obtained from eigenvalue expansion of the correlation matrix. The correlation matrix is separated into a signal subspace and a noise subspace that are orthogonal to each other. In the model of the present embodiment, the eigenvector spanning the signal subspace is proportional to a → ′ in Equation (3).

  The eigenvalue expansion of the correlation matrix R → xx is calculated as follows.

（６）
ただし、diag：対角行列

(6)
Where diag: diagonal matrix

  Further, λi and v → i are an eigenvalue and an eigenvector, respectively, and satisfy Expression (7).

（７）

(7)

  Since Rxx is a positive definite Hermitian matrix, the eigenvalue is a positive real number. In addition, it is assumed that λ1 ≧ λ2 ≧... ≧ λL (> 0) and the data is sorted in descending order. Then, from equation (7)

（８）
と書ける。ここで rank［ａ→′Ｓ→ａ→′^H］＝１なので、λ'1＞λ'2＝・・・＝λ'L＝０となる。式（８）から、相関行列の固有値は式（９）のようになる。

(8)
Can be written. Here, rank [a → 'S → a →' ^H ] = 1, so that λ′1> λ′2 =... Λ′L = 0. From equation (8), the eigenvalue of the correlation matrix is as in equation (9).

（９）

(9)

  Therefore, the diagonal matrix Λ → is separated into a signal eigenvalue and a noise eigenvalue from the magnitude of the eigenvalue. However, since the eigenvalue is actually obtained from the estimated value R → xx of the correlation matrix, an error occurs, and the values do not match as in equation (9).

  From Equation (7) and Equation (9),

（１０）
すなわち、

(10)
That is,

（１１）
となる。それゆえに、行列Ｖ→によって張られる空間は、固有値展開によって、互いが直交する信号部分空間と雑音部分空間を張る固有ベクトルに分離できる。本実施例のモデルでは、第１固有ベクトルｖ→１のみが信号部分空間を張っていて、固有ベクトル同士が互いに直交していることから、信号部分空間を張る固有ベクトルはａ→′に比例する。

(11)
It becomes. Therefore, the space spanned by the matrix V → can be separated into eigenvectors spanning a signal subspace and a noise subspace that are orthogonal to each other by eigenvalue expansion. In the model of the present embodiment, only the first eigenvector v → 1 extends the signal subspace, and the eigenvectors are orthogonal to each other. Therefore, the eigenvector extending the signal subspace is proportional to a → ′.

  As described above, it has been shown that the eigenvector spanning the signal subspace obtained by eigenvalue expansion of the correlation matrix is proportional to a → ′. Therefore, the eigenvector spanning the signal subspace is composed of a linear combination of steering vectors from the multipath, and can be expressed as in Expression (12).

（１２）
ただし、span{・}：{・}内のベクトルの線形結合

(12)
However, span {•}: linear combination of vectors in {•}

  The weighting coefficient at this time depends on the phase and power of each incoming wave signal. Therefore, the eigenvector spanning the signal subspace represents the propagation environment, and event detection can be performed using the eigenvector spanning this signal subspace.

  In this embodiment, events are filtered by filtering the correlation matrix using the mapping matrix of the divided signal subspace as a filter. In order to divide the signal subspace, the base of the signal subspace is estimated. As a basis of the signal subspace, a steering vector having an arrival direction estimated by the MUSIC method is used. The MUSIC method uses the fact that the signal subspace and the noise subspace are orthogonal, and the angle spectrum can be expressed as shown in Equation (13).

（１３）
ただし、ａ→(θ)：式（２）のステアリングベクトル
ＥN：雑音部分空間

(13)
Where a → (θ): steering vector of equation (2) EN: noise subspace

  Since there are a plurality of estimated arrival directions, a plurality of steering vectors are obtained as a basis. As the divided signal subspace, a steering matrix in which the basis of the steering vector is combined is used. For example, a steering matrix A → composed of a → (θ1) and a → (θ2) is A → = [a → (θ1) a → (θ2)]. A plurality of divided signal subspaces are obtained by changing the combination of the steering vectors. The mapping matrix P → of the steering matrix used as a filter can be expressed as shown in Equation (14).

（１４）
ただし、†：疑似逆行列
フィルタ手段２３は、受信信号から得られる相関行列Ｒ→xxに対して写像行列Ｐ→のフィルタをかけ、式（１５）に示す新しい相関行列Ｒ￣→xxを得る。
Ｒ￣→xx＝Ｐ→Ｒ→xxＰ→^H （１５）

(14)
However, †: pseudo inverse matrix filter means 23 filters mapping matrix P → on correlation matrix R → xx obtained from the received signal to obtain a new correlation matrix R￣ → xx shown in equation (15).
R￣ → xx = P → R → xxP → ^H (15)

  This operation emphasizes the signal subspace components divided from the correlation matrix. Furthermore, since it is orthogonal to the noise subspace, the influence of noise is reduced. The eigenvector computing means 24 performs eigenvalue expansion on this new correlation matrix R￣ → xx, and computes the first eigenvector related to the maximum eigenvalue obtained. The event detection means 25 detects an event using this first eigenvector.

  Here, the evaluation function P (t) for event detection is obtained by using the first eigenvector v → none acquired in advance when no event has occurred and the first eigenvector v → acquired when observing the event detection. inner product with ob

（１６）
とする。ただし固有ベクトルの大きさはどちらも１に正規化しておく。

(16)
And However, both eigenvector sizes are normalized to 1.

  Since the propagation environment does not change at the observation time when no event occurs, v → ob (tnone) shows a value very close to v → none, and is close to 1. On the other hand, at the observation time t = tevent when the event occurs, the propagation environment changes, and v → ob (tevent) shows a value different from v → none, and thus becomes a value smaller than 1. Therefore,

（１７）
となる閾値Ｐthを適切に設定することによりイベント検出が可能となる。そこで、イベント検出手段２５は、この式（１８）の判断を行ってイベントを検出する。

(17)
The event can be detected by appropriately setting the threshold value Pth. Therefore, the event detection unit 25 detects the event by performing the determination of the equation (18).

  FIG. 2 is a diagram illustrating an environment in which the experiment of the present example was performed. The experimental environment is a room surrounded by reinforced concrete, and most of the radio waves are reflected without being transmitted. Tx and Rx in FIG. 2 represent a transmitter and a receiver, respectively. The transceiver is in an unforeseen environment. Table 1 shows the experimental specifications.

  A box or a person will be placed as an event for this experiment. Since the box does not move, it is a static event, and because a person moves, it is a dynamic event. In addition, data was acquired by changing the location of each event. The person moves around the specified location. In order to distinguish the place of arrangement, numbers are assigned to FIG. In each event, the evaluation function for each place where the event occurred is compared with and without filtering.

  In the direction-of-arrival estimation for estimating the base of the signal subspace, it is better that the number of directions of arrival to be estimated is large in order to create as many filters as possible. Therefore, the noise subspace EN

（１９）
と設定する。ここで、ｖ→Lは式（６）の最小固有値に対応する固有ベクトルである。

(19)
And set. Here, v → L is an eigenvector corresponding to the minimum eigenvalue of Equation (6).

FIG. 3 is a diagram showing the relationship between the angle spectrum by the MUSIC method and the estimated radio wave arrival direction. The vertical axis represents the spectrum intensity, and the horizontal axis represents the angle of arrival direction. When the direction of arrival of radio waves is estimated by the MUSIC method, the spectrum is as shown in FIG. Among these, a portion that is hardly peaked is also selected as the arrival direction component as the basis of the signal subspace. This is because the current direction-of-arrival estimation is for obtaining a signal subspace orthogonal to the noise subspace of the obtained correlation matrix, and does not require strict accuracy. This is also because many candidates are needed to find an appropriate filter. The direction of arrival estimated in the environment of this example is 76.3 °, −98.8 °, 144.9 °, −37.3 °, −8.5 °, 170.9 °, and −135.9 ° in descending order of peak. Considering the combination, 2 ⁷ −1 divided signal subspaces can be formed.

  FIG. 4 is a diagram illustrating an example of an evaluation result when a static event is emphasized. In the static event of sequentially placing boxes at each numbered position in FIG. 2, the average value of the evaluation function when using a signal subspace partitioning filter that emphasizes the change in the static event at each position Standard deviations are shown in FIG. 4 and Table 2, respectively. An average value and a standard deviation are obtained from 60 evaluation functions. In this evaluation, a divided signal subspace consisting of 76.3 °, 144.9 ° and -8.5 ° was used. In FIG. 4, the numbers on the horizontal axis indicate the locations where they are arranged, and correspond to the numbers in FIG. 2, and “No event” means observation in a state where nothing has occurred. The vertical axis represents the average value of the evaluation function. The bold letters in Table 2 are the same as the numbers on the horizontal axis in FIG. In FIG. 4 and Table 2, the case where the filter which is a present Example is applied and the case where the conventional filter is not applied are compared.

  From Table 2, it can be said that the standard deviation of the evaluation function is small, and each value of the evaluation function is almost equal to the average value of the evaluation function. Therefore, the change of the average value of the evaluation function in FIG. 4 is almost equal to the change of each evaluation function. From Fig. 4, when using the filter, the value of the evaluation function changes little from 1 in the state where nothing has happened, but changes greatly compared to the case where no filter is used in all the observed static events. ing. From the above, it can be seen that an event can be emphasized according to an area by selecting an appropriate signal subspace.

  FIG. 5 is a diagram illustrating an example of an evaluation result when a static event is masked. The average value of the evaluation function when using a signal subspace partitioning filter that masks changes in static events at some positions in a static event in which boxes are sequentially placed at each numbered position in FIG. And standard deviation are shown in FIG. 5 and Table 3, respectively. In this evaluation, a divided signal subspace consisting of 144.9 ° and -135.9 ° was used. The horizontal and vertical axes in FIG. 5 are the same as those in FIG.

  From Table 3, as in Table 2, the standard deviation of the evaluation function is small, the value of each evaluation function is almost equal to the average value of the evaluation function, the change in the average value of the evaluation function in FIG. One change is almost equal. From FIG. 5, it can be seen that when the box is placed at # 1, 2, 3, 4 in FIG. 2 and the filter is applied, the value of the evaluation function is almost the same as 1, and the event is masked. I understand. When the box is placed elsewhere, the change in the value of the evaluation function is small compared to when no filter is applied, but it is sufficiently large compared to the state where nothing has happened, so event detection is possible. is there. From the above, it can be seen that the event can be masked according to the area by selecting an appropriate signal subspace.

  FIG. 6 is a diagram illustrating an example of an evaluation result when a dynamic event is emphasized. In the dynamic event of placing a person at each numbered position in FIG. 2, the average value and standard deviation of the evaluation function when using a signal subspace partitioning filter that emphasizes the change of the dynamic event at each position are shown. They are shown in FIG. 6 and Table 4, respectively. In this evaluation, the same signal subspace consisting of 76.3 °, 144.9 °, and -8.5 ° as the enhancement filter used in the static event was used. The horizontal and vertical axes in FIG. 6 are almost the same as those in FIG. 3, and it is not a box but a person that is arranged.

  From Table 4, the standard deviation of the evaluation function in the dynamic event is larger than the standard deviation of the evaluation function in the static event, and each evaluation function has a different value. However, it can be seen from FIG. 6 that when the filter is used, the average value of the evaluation function changes more in all dynamic events than when the filter is not used. From the above, in this experiment, it can be seen that the average value of the dynamic event evaluation function can be increased according to the area by selecting an appropriate signal subspace.

  FIG. 7 is a diagram illustrating an example of an evaluation result when a dynamic event is masked. In a dynamic event in which a person is placed at each numbered position in FIG. 2, the average value and standard deviation of the evaluation function when using a signal subspace division filter that masks the change of the dynamic event at some positions Are shown in FIG. 7 and Table 5, respectively. In this evaluation, a divided signal subspace consisting of -37.3 ° and -98.8 ° was used. The horizontal and vertical axes in FIG. 7 are almost the same as those in FIG. 3, and it is not a box but a person that is arranged.

  From FIG. 7, it can be seen that by using a filter, an event when a person is placed other than # 6 of FIG. 2 is masked. In addition, it can be seen from Table 5 that by using the filter, the standard deviation of the evaluation function becomes smaller at the masked event than when the filter is not used. Therefore, it can be said that each value of the evaluation function of the masked event is a value close to 1. From the above, in this experiment, it can be seen that by selecting an appropriate signal subspace, it is possible to mask each change in the dynamic event evaluation function according to the area.

  The signal subspace partition event filtering has been described above. Experiments have shown that events can be enhanced or masked according to area using a filter that divides the signal subspace according to the present invention. By selecting an appropriate filter, various demands and problems in home / office security can be met.

In addition, this invention is not limited to the said Example.
As long as the transmitter can generate radio waves and can be received by the array antenna, the transmitter used in other systems can be used in combination. For example, it corresponds to a wireless LAN base station. The signal may be a wide band or a narrow band.

  The mapping matrix may not be based on the combination of the arrival directions of radio waves, but may be an appropriate mapping matrix that changes the correlation matrix.

  The antenna may be an antenna composed of a plurality of elements, and is not necessarily an array antenna.

  The eigenvector computing means may compute a plurality of eigenvectors, and is not necessarily limited to computing only the eigenvector corresponding to the maximum eigenvalue of the correlation matrix.

  The event detection means is not limited to an inner product of eigenvectors, and may be, for example, a difference or a ratio as long as it can detect a change with time.

DESCRIPTION OF SYMBOLS 10 Transmitter 20 Receiver 21 Array antenna 22 Correlation matrix calculation means 23 Filter means 24 Eigenvector calculation means 25 Event detection means

Claims (4)

Multiple antennas to receive the radio waves transmitted by the transmitter,
Correlation matrix computing means for computing a correlation matrix from the received vector using signals received by the plurality of antennas as received vectors;
A filter means for subjecting said antenna calculates the mapping matrix for dividing the signal subspace based on the direction of arrival of a combination of radio waves received by the mapping matrix to the correlation matrix calculated by the correlation matrix calculating unit,
Eigenvector computing means for computing eigenvectors related to eigenvalues obtained by eigenvalue expansion of the correlation matrix multiplied by the mapping matrix by the filter means;
An event detection apparatus comprising: event detection means for detecting an event by detecting a change with time of the eigenvector calculated by the eigenvector calculation means. The event detection apparatus according to claim 1 , wherein the plurality of antennas are array antennas. 3. The event detection apparatus according to claim 1 , wherein the eigenvector calculation means calculates an eigenvector corresponding to a maximum eigenvalue of the correlation matrix multiplied by the mapping matrix. 4. The event detection apparatus according to claim 1 , wherein the event detection unit compares an inner product of the eigenvector during normal time and the eigenvector during observation with a predetermined threshold value.

Images