KR20210082749A - Apparatus and method for indoor positioning - Google Patents

Abstract

The application relates to an indoor positioning device and an indoor positioning method. The indoor positioning method according to an embodiment of the present invention comprises the steps of: receiving a sample radio signal transmitted from a plurality of nodes included in a target space; generating a data set corresponding to each node by using the amplitude and phase difference of the channel state information (CSI) packet of the sample radio signal; generating a radio positioning model for the target space by machine learning on the basis of supervised learning on the dataset; and generating coordinate information for a radio signal output from an arbitrary point in a target space by using the radio positioning model.

Description

Indoor positioning device and indoor positioning method {Apparatus and method for indoor positioning}

The present application relates to an indoor positioning device and an indoor positioning method, and to an indoor positioning device and an indoor positioning method capable of accurately positioning an indoor location using a CSI (Channel State Information) packet of a wireless LAN.

Since it is difficult to receive a GPS signal indoors, techniques for measuring or estimating a location using a radio resource other than the GPS signal have been proposed.

However, recently, as a technology that can replace GPS, indoor positioning technology using Wi-Fi signals is being used. In addition, as technology that can produce an indoor map accurately and at a low cost is developed together, indoor location-based services are gradually becoming a reality.

However, the conventional indoor location measurement technology using a Wi-Fi signal estimates the distance to each WiFi access point (AP) based on how much the received WiFi signal is attenuated compared to the first, and estimates the distance to each WiFi access point already known. It was a method of calculating one's own location using the location and the calculated distance. However, this method has a problem in that accuracy is lowered because the WiFi signal is not constantly reduced according to the distance due to the nature of the indoor space, but is reduced irregularly by hitting a wall or various objects.

An object of the present application is to provide an indoor positioning device and an indoor positioning method capable of accurately positioning an indoor location using a CSI (Channel State Information) packet of a WLAN.

An object of the present application is to provide an indoor positioning device and an indoor positioning method that can increase the accuracy of indoor positioning despite distortion caused by radio characteristics of CSI packets by applying supervised learning-based machine learning.

An object of the present application is to provide an indoor positioning device and an indoor positioning method that can learn by independently reflecting the characteristics of each wireless receiving device and each antenna group using a partially connected neural network.

An indoor positioning method according to an embodiment of the present invention relates to an indoor positioning method using a plurality of wireless receiving devices including two or more antennas, and sample wireless signals transmitted from a plurality of nodes included in a target space. receiving; generating a data set corresponding to each node by using the amplitude and phase difference of the CSI (Channel State Information) packet of the sample radio signal; generating a radio positioning model for the target space by machine learning based on supervised learning on the dataset; and generating coordinate information for a radio signal output from an arbitrary point in the target space by using the radio positioning model.

Here, the CSI packet may include channel information for each subcarrier corresponding to a preset index among a plurality of subcarriers included in the sample radio signal.

Here, the generating of the data set includes selecting two antennas included in the wireless receiver to form a group, and for each antenna group, including the amplitude and phase difference at each antenna for the CSI packet. You can create a dataset.

Here, the step of generating the dataset includes at least |CSI| Create a dataset including _n,m,k,i , R _{n,(m, m+1),k,I} and I _{n,(m, m+1),k,I , where |CSI} | is the amplitude of the CSI packet, R is the real part of the complex number indicating the phase difference of the CSI packet, I is the imaginary part of the complex number indicating the phase difference, n is the CSI packet, m is the antenna, k is the radio receiver, i may be each index indicating a subcarrier.

Here, the generating of the radio positioning model may include learning by connecting the datasets with a partially connected neural network.

Here, the partially connected neural network is a plurality of first layers that generate an intermediate result value by fully connecting each dataset, and a final result value by completely connecting the intermediate result values output from each first layer may include a second layer for generating

Here, the generating of the coordinate information may include extracting a plurality of nodes corresponding to the radio signal by using the radio positioning model, and generating the coordinate information by using the extracted nodes.

Here, the step of generating the coordinate information is

X = p(N1)*x ₁ + p(N2)*x ₂ + ... + p(Na)*x _a ,

Y = p(N1)*y ₁ + p(N2)*y ₂ + ... + p(Na)*y _a ,

The coordinate information (X, Y) is calculated using , where a is the number of extracted nodes, p(N) is the probability that each node corresponds to a node outputting the radio signal, N1, ... , Na may be each extracted node, x ₁ , ... , x _a may be an x-axis coordinate of each extracted node, and y ₁ , ... , y _a may be a y-axis coordinate of each extracted node.

According to an embodiment of the present invention, a computer program stored in the medium may exist in combination with hardware to execute the above-described indoor positioning method.

An indoor positioning device according to an embodiment of the present invention relates to an indoor positioning device using a plurality of wireless receiving devices including two or more antennas, and sample wireless signals transmitted from a plurality of nodes included in a target space. a receiving unit for receiving; a data processing unit for generating a data set corresponding to each node by using the amplitude and phase difference of the CSI (Channel State Information) packet of the sample radio signal; a model generator for generating a radio positioning model for the target space by machine learning based on supervised learning on the dataset; and a positioning unit generating coordinate information for a wireless signal output from an arbitrary point in the target space by using the radio positioning model.

Incidentally, the means for solving the above problems do not enumerate all the features of the present invention. Various features of the present invention and its advantages and effects may be understood in more detail with reference to the following specific embodiments.

According to the indoor positioning apparatus and the indoor positioning method according to an embodiment of the present invention, since a wireless positioning model is generated by machine learning CSI packets of a wireless LAN, accurate indoor positioning despite distortion of CSI packets transmitted wirelessly is possible

According to the indoor positioning apparatus and the indoor positioning method according to an embodiment of the present invention, it is possible to process data so that the CSI packet expressed in the form of a complex number is rearranged in a form that is easy to machine learning. Therefore, it is possible to easily perform machine learning on the CSI packet.

According to the indoor positioning device and the indoor positioning method according to an embodiment of the present invention, since a partially connected neural network is used, it is possible to learn by independently reflecting the characteristics of each wireless receiving device and each antenna group. Do. Therefore, more accurate indoor positioning can be performed compared to a general fully connected neural network.

However, the effects that can be achieved by the indoor positioning device and the indoor positioning method according to an embodiment of the present invention are not limited to those mentioned above, and other effects not mentioned are from the description below. It will be clearly understood by those of ordinary skill in the art.

1 is a schematic diagram showing an indoor positioning system according to an embodiment of the present invention.
2 is a block diagram illustrating an indoor positioning device according to an embodiment of the present invention.
3 is a schematic diagram illustrating a data set according to an embodiment of the present invention.
4 is a schematic diagram showing a learning process of a radio positioning model according to an embodiment of the present invention.
5 is a schematic diagram showing indoor positioning according to an embodiment of the present invention.
6 is a flowchart illustrating an indoor positioning method according to an embodiment of the present invention.

Hereinafter, preferred embodiments will be described in detail so that those of ordinary skill in the art can easily practice the present invention with reference to the accompanying drawings. However, in describing a preferred embodiment of the present invention in detail, if it is determined that a detailed description of a related known function or configuration may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted. In addition, the same reference numerals are used throughout the drawings for parts having similar functions and functions.

In addition, throughout the specification, when a part is 'connected' with another part, it is not only 'directly connected' but also 'indirectly connected' with another element interposed therebetween. include In addition, 'including' a certain component means that other components may be further included, rather than excluding other components, unless otherwise stated. In addition, terms such as "~ unit" and "module" described in the specification mean a unit that processes at least one function or operation, which may be implemented as hardware or software, or a combination of hardware and software.

1 is a schematic diagram showing an indoor positioning system according to an embodiment of the present invention.

Referring to FIG. 1 , an indoor positioning system according to an embodiment of the present invention may include an indoor positioning device 100 using a plurality of wireless receiving devices 11 and 12 located in a target space A. .

Hereinafter, an indoor positioning system according to an embodiment of the present invention will be described with reference to FIG. 1 .

The target space A may be an indoor space such as a house, a company, a school, or a shopping mall, and a plurality of nodes N may be set in the target space A. In this case, the coordinate information of each node N may be preset, and wireless transmitters (not shown) may be located in each node N to transmit a wireless signal.

Here, the wireless transmission devices may transmit a wireless signal using a wireless LAN. That is, a wireless signal may be transmitted using an 802.11n channel, and a CSI (Channel State Information) packet may be included in the wireless signal. The wireless transmitter may use a bandwidth of 20Mhz or 40Mhz, and may include 64 or 128 Orthogonal Frequency Division Multiplexing (OFDM) subcarriers.

The indoor positioning device 100 may receive a wireless signal transmitted by the wireless transmitter using a plurality of wireless receivers 11 and 12, respectively, and the location of the wireless transmitter using CSI packets included in the wireless signal. can be positioned. According to the embodiment, the indoor positioning device 100, among the sub-carriers included in the wireless signal, [-28, -26, -24, -22, -20, -18, -16, -14, -12, -10, -8, -6, -4, -2, -1, 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 28] CSI packets for a total of 30 subcarriers having

In general, it is possible to extract and analyze a CSI packet during Wi-Fi communication, and it is possible to perform radio positioning through analysis of the CSI packet. That is, radio positioning may be performed using a plurality of radio receiving apparatuses 11 and 12 that receive the CSI packet. Specifically, two or more antennas m1 and m2 may be mounted in the wireless receiving devices 11 and 12, and it is possible to obtain a phase difference between CSI packets received by each antenna m1 and m2. . The phase difference obtained at this time is closely related to the angle of incidence between the antennas m1 and m2 of the currently arranged wireless transmitter and the wireless receiver 11 and 12 .

Therefore, if the radio receiving apparatuses 11 and 12 equipped with two or more antennas are used, the angle of incidence at each of the radio receiving apparatuses 11 and 12 can be obtained, and the From each angle of incidence, it is possible to obtain the position of the wireless transmitter using triangulation.

However, in the case of the CSI packet received by the wireless receivers 11 and 12, distortion may occur severely due to the characteristics of radio propagation, and thus, an error may be included in the phase difference obtained from the CSI packet. Therefore, the position of the wireless transmitter calculated by the triangulation method has a large error from the actual position, so it may be difficult to accurately measure the position.

In order to solve this problem, the indoor positioning device 100 according to an embodiment of the present invention may increase the accuracy of positioning by using a machine learning technique. Hereinafter, the indoor positioning device 100 according to an embodiment of the present invention will be described.

2 is a block diagram illustrating an indoor positioning device according to an embodiment of the present invention.

Referring to FIG. 2 , the indoor positioning device 100 according to an embodiment of the present invention may include a receiving unit 110 , a data processing unit 120 , a model generating unit 130 , and a positioning unit 140 .

The receiver 110 may receive a sample radio signal transmitted from a plurality of nodes N included in the target space A. The receiver 110 may include a plurality of wireless receivers 11 and 12 , and may receive sample radio signals input through the plurality of wireless receivers 11 and 12 , respectively. Here, the sample wireless signal may be any wireless signal transmitted by the wireless transmitter for machine learning.

Meanwhile, a wireless transmitter (not shown) may be located at each node N, and a plurality of wireless receivers 11 and 12 may also be fixed to a preset specific location. Thereafter, when the wireless transmitter transmits a sample wireless signal at a specific node, the wireless receivers 11 and 12 may receive the sample wireless signal transmitted by the corresponding wireless transmitter, respectively, and provide it to the receiver 110 .

According to an embodiment, the wireless transmitter may be sequentially positioned to each node (N) and then a sample wireless signal may be transmitted, and the receiver 110 may transmit a wireless signal transmitted from each node (N) to a specific The input may be individually received through a plurality of wireless receiving devices 11 and 12 fixed at positions.

Here, the sample radio signal may include a CSI packet, and the CSI packet may include channel information of each subcarrier corresponding to a preset index among a plurality of subcarriers included in the radio signal.

The data processing unit 120 may generate a data set from the sample radio signal to apply the received sample radio signal to machine learning. That is, the data processing unit 120 may generate a data set corresponding to each node N by using the amplitude and phase difference of the CSI packet of the sample radio signal.

Specifically, the data processing unit 120 may form a group by selecting two of the plurality of antennas included in the wireless receiving devices 11 and 12, and each antenna group includes an amplitude and a phase difference in each antenna. You can create a dataset. At this time, since the amplitude and phase difference are generated for each subcarrier included in the CSI packet, when the number of subcarriers is 30, 30 amplitudes and phase differences corresponding to each antenna can be generated.

For example, when three antennas are included in each of the radio receivers 11 and 12, as shown in FIG. 3, four data sets can be generated. Here, FIG. 3A is a data set corresponding to the first antenna ant.1 and the second antenna ant.2 of the first radio receiver Rx1, and FIG. 3B is the first radio receiver Rx1. It corresponds to a data set corresponding to the second antenna ant.2 and the third antenna ant.3 of the receiver Rx1.

In addition, FIG. 3(c) is a data set corresponding to the first antenna ant.1 and the second antenna ant.2 of the second radio receiving device Rx2, and FIG. 3(d) is the second radio receiving apparatus. They correspond to data sets corresponding to the second antenna ant.2 and the third antenna ant.3 of the device Rx2, respectively.

Here, referring to FIG. 3( a ), in the dataset, the amplitude Amp of the first antenna ant.1 and the second antenna ant.2 included in the first radio receiver Rx1 and the second Each real part (Real) and an imaginary part (Image) of the phase difference (Phase 1-2) between the first antenna and the second antenna may be included. That is, the data generation unit 120 |CSI| A dataset can be created by extracting _n,m,k,i , R _{n,(m, m+1),k,I} and I _{n,(m, m+1),k,I respectively, where} , |CSI| is the amplitude of the CSI packet, R is the real part of the complex number indicating the phase difference of the CSI packet, I is the imaginary part of the complex number indicating the phase difference, n is the CSI packet, m is the antenna, k is the radio receiver, i corresponds to each index indicating a subcarrier. Thereafter, the datasets of FIGS. 3(b), 3(c) and 3(d) may be generated in the same manner.

Here, since the data set generator 120 generates a real part and an imaginary part of the amplitude and the phase difference of each antenna for each subcarrier, in the case of including 30 subcarriers, each data set 120 pieces of data can be generated. Accordingly, a total of 480 pieces of data can be generated if all data sets are included.

On the other hand, although the case in which three antennas are included in the wireless receiving devices 11 and 12 has been described as an example, the number of antennas included in the wireless receiving devices 11 and 12 may vary depending on the embodiment, and accordingly The number of generated data sets may also be different. For example, when each of the radio receivers 11 and 12 includes two antennas, a total of two data sets are generated, and when four antennas are included, a total of six data sets can be generated. .

The model generator 130 may generate a radio positioning model for the target space A by machine learning the dataset based on supervised learning. Here, in order to generate a radiolocation model, data sets may first be connected with a partially connected neural network.

That is, as shown in FIG. 4 , a plurality of first layers L1 that generate an intermediate result node O1 by fully connecting each dataset, and each of the first layers L1 By completely connecting the output intermediate result nodes O1 to form a second layer L2 that generates a final result node R, a partially connected neural network can be implemented. Here, the partially connected neural network creates a separate first layer (L1) for each dataset, and then completely connects the intermediate result nodes in each first layer (L1) again, so each wireless receiving device 11 , 12) and a structure that independently reflects dependencies on antenna groups can be implemented.

First, the first layer L1 may generate an intermediate result node O1 by applying a weight and a bias to each input dataset. At this time, as shown in FIG. 4 , the first layer L1 may receive 120 pieces of data from each dataset, and transmit each data to 50 intermediate result nodes through 200 hidden nodes. (O1) can be connected.

Thereafter, the intermediate result nodes O1 generated for each dataset may be collected, and the collected intermediate result nodes O1 may be applied as an input of the second layer L2. In this case, the second layer L2 may completely connect the intermediate result nodes O1 with 300 hidden nodes to reflect their respective weights and tendencies. Through this, the second layer L2 may be finally connected to the final result node R corresponding to the number of each node N included in the target space A. As shown in FIG. That is, when 16 nodes N are included in the target space A, 16 final result nodes R may be included.

Thereafter, the model generator 130 may perform supervised learning-based machine learning based on the partially connected neural network connected to the dataset. That is, each node of the partially connected neural network can be trained so that the final result node determined as the outgoing node that transmits a wireless signal in the partially connected neural network matches the correct answer set during supervised learning. For example, when a radio signal is transmitted from the 10th node among 16 nodes (N) included in the target space (A), the final result node (R) corresponding to the correct answer is [0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0]. At this time, the inference value (matrix of [16,1]) calculated in the partially connected neural network minimizes the MSE (Mean Square Error) for each matrix element with the final result node R, the first layer (L1) and weights and tendencies of the second layer L2 may be learned.

Through this, the model generating unit 130 may finally generate a radio positioning model for the target space (A).

The positioning unit 140 may generate coordinate information for a wireless signal output from an arbitrary point in the target space A by using the generated wireless positioning model. That is, after extracting a plurality of nodes corresponding to the radio signal using the radio positioning model, coordinate information corresponding to the corresponding point may be generated from the extracted nodes.

Specifically, as shown in FIG. 5 , a radio signal may be transmitted at an arbitrary point P in the target space A. As shown in FIG. In this case, the receiving unit 110 may receive the corresponding radio signal, and the data processing unit 120 may generate the corresponding radio signal as a data set. Thereafter, the positioning unit 140 may adapt the corresponding dataset to the wireless positioning model, and extract a plurality of nodes N1 , N2 , N3 , and N4 as candidate nodes for transmitting a wireless signal. Here, since the radio positioning model calculates the probability that the input data sets correspond to each node, when a radio signal is transmitted from an arbitrary point P, the probability of nodes adjacent to the corresponding point P is high. do. Accordingly, the positioning unit 140 may select a set number (eg, a number) of nodes from among the plurality of nodes in the order of increasing the probability value, or select all nodes having a probability value greater than or equal to the set value as candidate nodes.

after,

X = p(N1)*x ₁ + p(N2)*x ₂ + ... + p(Na)*x _a ,

Y = p(N1)*y ₁ + p(N2)*y ₂ + ... + p(Na)*y _a ,

Coordinate information (X, Y) corresponding to an arbitrary point (P) that transmits a radio signal can be calculated using . Here, a is the total number of extracted candidate nodes, p(N) is the probability that each node corresponds to a node outputting a radio signal, N1, ... , Na are each candidate node, x ₁ , ... , x _a corresponds to the x-axis coordinates of each candidate node, and y ₁ , ... , y _a corresponds to the y-axis coordinates of each candidate node. Accordingly, the positioning unit 140 may generate accurate coordinate information for an arbitrary point P outputting a radio signal in the target space A. As shown in FIG.

6 is a flowchart illustrating an indoor positioning method according to an embodiment of the present invention. Here, each step may be performed by the indoor positioning device according to an embodiment of the present invention. Hereinafter, an indoor positioning method according to an embodiment of the present invention will be described with reference to FIG. 6 .

First, the indoor positioning device may receive a sample radio signal transmitted from a plurality of nodes included in the target space (S110). Here, the indoor positioning device may include a plurality of wireless receiving devices, and may receive respective sample wireless signals through the plurality of wireless receiving devices. Here, the sample wireless signal may be any wireless signal transmitted by the wireless transmitter for machine learning.

Meanwhile, the wireless transmitter may be located in any one of the plurality of nodes, and the plurality of wireless receivers may also be fixed to a preset specific location. Thereafter, when the wireless transmitter transmits a sample wireless signal at a specific node, the wireless receiver may receive sample wireless signals transmitted by the corresponding wireless transmitter, respectively.

According to an embodiment, a sample wireless signal may be transmitted after sequentially moving the location of the wireless transmitter to each node, and the indoor positioning device may receive a wireless signal transmitted from each node.

Here, the sample radio signal may include a CSI packet, and the CSI packet may include channel information of each subcarrier corresponding to a preset index among a plurality of subcarriers included in the radio signal.

Thereafter, the indoor positioning device may generate a data set corresponding to each node by using the amplitude and phase difference of the CSI packet of the sample radio signal ( S120 ). That is, in order to apply the received sample radio signal to machine learning, a dataset may be generated from the sample radio signal.

Specifically, a group may be formed by selecting a plurality of antennas included in the wireless receiving device by two, and a data set including an amplitude and a phase difference in each antenna may be generated for each antenna group. At this time, since the amplitude and phase difference are generated for each subcarrier included in the CSI packet, when the number of subcarriers is 30, 30 amplitudes and phase differences corresponding to each antenna can be generated.

For example, the indoor positioning device is differentiated by each radio receiving device and each antenna group, and |CSI| A dataset can be created by extracting _n,m,k,i , R _{n,(m, m+1),k,I} and I _{n,(m, m+1),k,I respectively.} Here, |CSI| is the amplitude of the CSI packet, R is the real part of the complex number indicating the phase difference of the CSI packet, I is the imaginary part of the complex number indicating the phase difference, n is the CSI packet, m is the antenna, k is the radio receiver, i corresponds to each index indicating a subcarrier.

When the dataset is generated, the indoor positioning device may machine learning the dataset based on supervised learning, and through this, may generate a radio positioning model for the target space (S130). Here, in order to generate the radio positioning model, the indoor positioning device may first connect the datasets with a partially connected neural network.

That is, a plurality of first layers for generating intermediate result nodes by fully connecting each dataset, and a second layer for generating final result nodes by completely connecting intermediate result nodes output from each first layer By forming , a partially connected neural network can be implemented. Here, the partially connected neural network creates a separate first layer for each dataset and then completely connects the intermediate result nodes in each first layer again, so the dependency ( dependency) can be reflected independently.

Thereafter, the indoor positioning device may perform supervised learning-based machine learning based on the partially connected neural network connected to the dataset. That is, each node of the partially connected neural network can be trained so that the final result node determined as the outgoing node that transmits a wireless signal in the partially connected neural network matches the correct answer set during supervised learning. Through this, the indoor positioning device can finally generate a wireless positioning model for the target space.

When the wireless positioning model is generated, the indoor positioning device may generate coordinate information for a wireless signal output from an arbitrary point in the target space by using the wireless positioning model (S140). That is, after extracting a plurality of nodes corresponding to the radio signal using the radio positioning model, coordinate information corresponding to the corresponding point may be generated from the extracted nodes.

For example, when a wireless signal is transmitted from an arbitrary point in the target space, the indoor positioning device may receive the corresponding wireless signal and generate a corresponding data set. Thereafter, by adapting the data set to the radio positioning model, candidate nodes that transmit the radio signal from among the plurality of nodes can be extracted. Since the radio positioning model can calculate the probability that the input data sets correspond to each node, the probability of nodes adjacent to an arbitrary point that transmits a radio signal is high. Accordingly, the indoor positioning device may select a set number (eg, a number) of nodes from among the plurality of nodes in the order in which the probability value is high, or select all nodes having a probability value equal to or greater than the set value as candidate nodes.

after,

X = p(N1)*x ₁ + p(N2)*x ₂ + ... + p(Na)*x _a ,

Y = p(N1)*y ₁ + p(N2)*y ₂ + ... + p(Na)*y _a ,

Coordinate information (X, Y) corresponding to an arbitrary point (P) that transmits a radio signal can be calculated using . Here, a is the total number of extracted candidate nodes, p(N) is the probability that each node corresponds to a node outputting a radio signal, N1, ... , Na are each candidate node, x ₁ , ... , x _a corresponds to the x-axis coordinates of each candidate node, and y ₁ , ... , y _a corresponds to the y-axis coordinates of each candidate node. Accordingly, it is possible for the indoor positioning device to generate accurate coordinate information for an arbitrary point outputting a radio signal in the target space.

The present invention described above can be implemented as computer-readable code on a medium in which a program is recorded. The computer-readable medium may continuously store a computer-executable program, or may be temporarily stored for execution or download. In addition, the medium may be various recording means or storage means in the form of a single or several hardware combined, it is not limited to a medium directly connected to any computer system, and may exist distributed on a network. Examples of the medium include a hard disk, a magnetic medium such as a floppy disk and a magnetic tape, an optical recording medium such as CD-ROM and DVD, a magneto-optical medium such as a floppy disk, and those configured to store program instructions, including ROM, RAM, flash memory, and the like. In addition, examples of other media may include recording media or storage media managed by an app store that distributes applications, sites that supply or distribute various other software, and servers. Accordingly, the above detailed description should not be construed as restrictive in all respects but as exemplary. The scope of the present invention should be determined by a reasonable interpretation of the appended claims, and all modifications within the equivalent scope of the present invention are included in the scope of the present invention.

The present invention is not limited by the above embodiments and the accompanying drawings. For those of ordinary skill in the art to which the present invention pertains, it will be apparent that the components according to the present invention can be substituted, modified and changed without departing from the technical spirit of the present invention.

100: indoor positioning device 110: receiver
120: data processing unit 130: model generation unit
140: positioning unit

Claims (10)

In the indoor positioning method using a plurality of wireless receiving devices including two or more antennas,
Receiving a sample radio signal transmitted from a plurality of nodes (node) included in the target space;
generating a data set corresponding to each node by using the amplitude and phase difference of the CSI (Channel State Information) packet of the sample radio signal;
generating a radio positioning model for the target space by machine learning based on supervised learning on the dataset; and
and generating coordinate information for a radio signal output from an arbitrary point in the target space by using the radio positioning model.
The method of claim 1, wherein the CSI packet is
and channel information for each subcarrier corresponding to a preset index among a plurality of subcarriers included in the sample radio signal.
The method of claim 2, wherein the generating of the dataset comprises:
Indoors, characterized in that by selecting two antennas included in the radio receiver to form a group, and for each antenna group, generating the data set including the amplitude and phase difference at each antenna for the CSI packet positioning method.
The method of claim 3, wherein the generating of the dataset comprises:
At least |CSI| Create a dataset including _n,m,k,i , R _{n,(m, m+1),k,I} and I _{n,(m, m+1),k,I , where |CSI} | is the amplitude of the CSI packet, R is the real part of the complex number indicating the phase difference of the CSI packet, I is the imaginary part of the complex number indicating the phase difference, n is the CSI packet, m is the antenna, k is the radio receiver, i is each index representing a subcarrier.
The method of claim 1, wherein generating the radio positioning model comprises:
Indoor positioning method, characterized in that the learning by connecting the datasets with a partially connected neural network (Partially Connected Neural Network).
The method of claim 5, wherein the partially connected neural network is
A plurality of first layers that generate an intermediate result node by fully connecting each dataset, and a second layer that creates a final result node by completely connecting the intermediate result nodes output from each first layer Indoor positioning method comprising the.
The method of claim 1, wherein the generating of the coordinate information comprises:
An indoor positioning method, characterized in that by using the wireless positioning model, extracting a plurality of nodes corresponding to the wireless signal, and generating the coordinate information by using the extracted nodes.
The method of claim 7, wherein the generating of the coordinate information comprises:
X = p(N1)*x ₁ + p(N2)*x ₂ + ... + p(Na)*x _a ,
Y = p(N1)*y ₁ + p(N2)*y ₂ + ... + p(Na)*y _a ,
The coordinate information (X, Y) is calculated using , where a is the number of extracted nodes, p(N) is the probability that each node corresponds to a node outputting the radio signal, N1, ... , Na is each extracted node, x ₁ , ... , x _a is the x-axis coordinate of each extracted node, y ₁ , ... , y _a is the y-axis coordinate of each extracted node Indoor positioning method.
A computer program stored in a medium in combination with hardware to execute the indoor positioning method of any one of claims 1 to 8.
An indoor positioning device using a plurality of wireless receiving devices including two or more antennas,
a receiver for receiving sample radio signals transmitted from a plurality of nodes included in the target space;
a data processing unit for generating a data set corresponding to each node by using the amplitude and phase difference of the CSI (Channel State Information) packet of the sample radio signal;
a model generator for generating a radio positioning model for the target space by machine learning based on supervised learning on the dataset; and
and a positioning unit generating coordinate information for a wireless signal output from an arbitrary point in the target space by using the wireless positioning model.

Images