KR101264306B1 - Apparatus of tracking user indoor using user motion model learning and recording media therefor - Google Patents

Abstract

PURPOSE: An indoor location tracking device and a recording medium therefor are provided to update a mobile model of a user by using static data by collecting the static data related to a moving path of each particle. CONSTITUTION: A user location tracking unit(104) samples a replacement particle again by using updated weighted values. The user location tracking unit repetitively executes a particle resampling process according to the movement of a user. The user location tracking unit estimates a real-location in which particles are finally converged by the location of the user. A mobile model learning unit(106) updates a mobile model by using static data for the movement path of the matched particle. [Reference numerals] (100) Sensor module; (102) Wireless communication unit; (104) User location tracking unit; (106) Mobile model learning unit

Description

Indoor location tracking device through moving model learning and recording medium therefor {APPARATUS OF TRACKING USER INDOOR USING USER MOTION MODEL LEARNING AND RECORDING MEDIA THEREFOR}

The present invention relates to an indoor location tracking apparatus and a recording medium therefor through a moving model learning, and more particularly, to a device and a recording medium capable of increasing the accuracy of user location tracking indoors.

Recently, Location Based Service (LBS), which provides services based on user's location information, has received much attention. In particular, in recent years, a service that provides LBS by tracking a user's location using a sensor mounted on a smart phone is increasing.

In general, it is possible to track a user's location relatively easily by using a GPS (Global Positioning System) in an exterior environment of a building, but a method for tracking a user's location in an indoor environment in which no GPS signal is received has not been presented. to be.

 Currently, a method of using a WiFi signal map in an indoor space has been proposed, but conventional methods using a WiFi signal map utilize previous user location information or a user movement model such as k-NN, neural network, and naive Bayesian filter for real-time location tracking. There are many ways you can't.

In addition, the Kalman filter, which is a probabilistic approach that can use the user's previous location information and movement model, needs to calculate likelihoods for all possible candidate locations each time, so that real-time location tracking can be performed when the indoor space is large and complex. The computational complexity required for this is very high, which limits the performance and efficiency of the position tracker.

Korean Patent Publication No. 2010-0124649 relates to a location measuring system and a method of a mobile terminal, and “a method of measuring a location of a mobile terminal measures reception intensity values of a plurality of radio signals transmitted from a plurality of access points to the mobile terminal. Doing; Calculating a first position of the mobile terminal at the first time point using an extended Kalman filter, from received strength values of the plurality of radio signals at a first time point; And calculating a second position of the mobile terminal at the second time point by using a particle filter, from positions of the mobile terminal at the first time point and reception intensity values of the plurality of radio signals received at the second time point. It includes the steps to perform. ”However, this uses a Kalman filter is still a problem of high complexity.

In addition, various researches have been conducted on indoor location tracking of a device such as a robot. Unlike a robot having a relatively stable and accurate moving distance and speed and a moving direction, in the case of a human, a moving distance and a speed and a direction of a human are determined. It is difficult to track the location, especially indoors, because it is not easy to measure accurately.

In order to solve the problems of the prior art as described above, an indoor location tracking device and a recording medium using a mobile model learning that can increase the accuracy of user location tracking in the room.

According to a preferred embodiment of the present invention to achieve the above object, in the particle filter-based user location tracking device, when the user moves in the indoor space, each of the plurality of particles in accordance with the movement model to the new location The likelihood of each of the plurality of particles is calculated using the radio signal information and the pre-produced radio signal map received from at least one wireless AP at the actual location of the user's movement, and using the calculated likelihood. Update weights of each of the plurality of particles, perform particle resampling to select particles to replace each of the plurality of particles using the updated weights, move the particles according to the movement of the user, calculate the likelihood, Particle resampling is repeatedly performed at the updated weight and the newly moved position. Carried out by the user location tracking unit for estimating the particles to eventually converge to by performing the iteration to the user's physical location; And a movement model learner for updating the movement model by using statistical data on the movement path of the converged particles.

Each of the plurality of particles may be defined as at least one of a position, a direction, and a movement state.

The wireless signal information is an intensity vector of a signal received from one or more wireless APs, and the wireless signal map is received from the one or more wireless APs for each region based on a phase map relating to dividing the indoor space into a plurality of regions. The generated signal strength data can be produced based on the acquisition.

The phase map may include at least one of a vertex set for defining each region, an edge set for defining a connection relationship between two regions adjacent to each other, and a set of the one or more wireless APs.

The wireless signal map may be defined as an average matrix and a standard deviation matrix generated by using the average and standard deviation of the signal strength data collected for each region.

When moving, the user is assumed to be located on the vertex and may be assumed to move continuously on the edge.

The user location tracking unit may randomly select a plurality of candidate locations of a user at a start point of user location tracking.

The user position tracking unit may randomly select a new particle to replace each particle by reflecting the weights calculated for each of the plurality of particles using a roulette wheel algorithm for the particle resampling.

The movement model learner may update the movement model using a Monte Carlo Expectation Maximization (MCEM) algorithm.

According to another aspect of the present invention, initially selecting a plurality of candidate positions (particles) of the user at random, when the user moves, each of the plurality of initially selected particles according to the moving model, the user moves The likelihood of each of the plurality of particles is calculated using a radio signal received from at least one wireless AP at a real location and a pre-made map of a radio signal, and the weight of each of the plurality of particles is set using the calculated likelihood. And a particle resampling unit for selecting particles to replace each of the plurality of particles using the set weights, and moving the particles, calculating the likelihood, updating the weights, and newly moving positions according to a user's movement. Particle resampling is iteratively performed and finally converges. User positioning portion of particles to estimate a user's physical location; And a movement model learner for updating the movement model by using statistical data on the movement path of the converged particles.

According to another aspect of the present invention, there is tangibly implemented a program of instructions that can be executed in a mobile terminal for user location tracking, and which can be read by the mobile terminal as (a) a user in an indoor space. Moving a particle, moving each of the plurality of particles from a previous position to a new position according to a moving model, wherein the plurality of particles are candidate positions of a user; A likelihood calculation step of calculating a likelihood of each of the plurality of particles using wireless signal information received from at least one wireless AP and a pre-produced wireless signal map at a user's actual location; (b) a weight updating step of updating weights of each of the plurality of particles using the calculated likelihood; (c) particle resampling to replace each of the plurality of particles using the updated weights; (d) repeating the particle movement, the likelihood calculation, the update of the weight, and the resampling of the particle at the newly moved position according to the movement of the user to estimate the final converged particle as the actual position of the user; And (e) updating the movement model using statistical data on the movement path of the converging particles.

According to the present invention, while tracking the location of the user based on the particle filter collects statistical data about the movement path of each particle, and through this update the user's movement model can improve the location tracking accuracy of the user indoors There is an advantage.

1 is a block diagram of a user location tracking device according to an embodiment of the present invention.
2 is a block diagram of a user location tracking unit according to an embodiment of the present invention.
3 is a view showing positions of particles randomly selected according to the present embodiment.
4 is a diagram illustrating a moving state of particles according to the present embodiment.
5 is a diagram illustrating a method for producing a wireless signal map according to an embodiment of the present invention.
6 illustrates a phase map according to an embodiment of the present invention.

While the invention is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and will herein be described in detail. It should be understood, however, that the invention is not intended to be limited to the particular embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of a user location tracking device according to an embodiment of the present invention.

As shown in FIG. 1, the user location tracking apparatus according to the present embodiment may include a sensor module 100, a wireless communication unit 102, a user location tracking unit 104, and a moving model learning unit 106. .

The user location tracking device according to the present embodiment is a mobile communication terminal in which the sensor module 100 is built as described above and may receive a wireless signal from a wireless access point (AP) located at a short distance such as a WiFi signal. Can be.

The sensor module 100 may detect movement of a user and determine a moving direction, and may include an acceleration sensor and a magnetic field sensor.

The acceleration sensor is a sensor for measuring acceleration or impact strength of a moving object. The acceleration sensor detects whether the user moves, and also calculates the moving distance of the user using the change information of the acceleration.

In general, the acceleration sensor may be a three-axis acceleration sensor for measuring the three-axis acceleration, the present invention can be used without any acceleration sensor without limitation, inertial, gyro, silicon semiconductor.

A magnetic field sensor is a sensor that detects a change in a magnetic field according to a movement of an object. The magnetic field sensor mainly uses a hall effect in which a voltage changes when a magnetic field is vertically applied to a current flowing through a semiconductor.

The magnetic field sensor may be used in various terms such as a digital compass, a geomagnetic sensor, a geomagnetic sensor, and a magnetic sensor. In the present embodiment, all sensors for detecting a change in the magnetic field are defined as magnetic field sensors.

The wireless communication unit 102 receives a wireless signal from one or more wireless APs located in the vicinity.

The user location tracking unit 104 may detect sensing information sensed by the sensor module 100 and wireless signals received through the wireless communication unit 102 (strength data of wireless signals received from one or more wireless APs) as the user moves. To track your location.

According to the present embodiment, the user location tracking unit 104 tracks the location of the user based on the particle filter, and at this time, tracks the location of the user using a pre-produced wireless signal map for the indoor space where the user is located.

The user location tracking unit 104 tracks the user location according to the user's moving model. According to an exemplary embodiment of the present invention, the moving model learning unit 106 moves the particle at the same time as the location tracking process of the user. Collect the statistical data for and update the user's movement model.

The user location tracking unit 104 continuously performs the user location tracking process using the updated movement model.

2 is a block diagram of a user location tracking unit according to an exemplary embodiment of the present invention.

As shown in FIG. 2, the user location tracking unit 104 according to the present embodiment includes a particle initialization unit 200, a particle moving unit 202, a likelihood calculator 204, a weight setting unit 206, and a particle. The resampling unit 208 may be included.

As illustrated in FIG. 3, the particle initialization unit 200 randomly selects a plurality of candidate positions (particles) of the user on the indoor space when the user's location tracking is started.

Here, each of the plurality of particles may further include information regarding a direction and a movement state as well as a position as shown in Equation 1 below.

는 사용자의 위치로서, 본 발명에서 사용자의 위치는 무선 신호 지도의 각 지역(

)을 의미하며, 지역에 대해서는 하기에서 상세하게 설명한다. here,

Is the location of the user, and in the present invention, the location of the user is defined in each region of the wireless signal map (

), And the area will be described in detail below.

는 방향이며, 본 발명의 일 실시예에 따르면 계산의 복잡도를 낮추기 위해 동서남북의 4개의 방향으로 파티클을 정의할 수 있다.

Is a direction, and according to an embodiment of the present invention, particles may be defined in four directions of east, west, north, and south to reduce the complexity of the calculation.

는 사용자의 이동 상태에 관한 정보로서, 정지 또는 이동으로 정의될 수 있다. Meanwhile,

Is information about a movement state of a user and may be defined as a stop or a movement.

According to an embodiment of the present invention, the direction and the movement state may be determined through the data sensed by the sensor module 100.

That is, the direction of the current user (direction to move) and whether or not the movement may be determined using the acceleration sensor and the magnetic field sensor.

As shown in FIG. 4, when the user moves, the particle moving unit 202 moves each of a plurality of particles randomly selected according to a previously set movement model to a new position.

에 있던 사용자가 다음 순간 어느 위치로 이동하는지, 즉 다음 예상 위치

를 결정하는데 이용되는 것으로서, 보행중인 사용자의 움직임에 대한 비-결정성과 불확실성을 반영하여 일반적으로 확률

로서 표현된다. Your movement model is in the previous location

Where did the user at go to the next moment?

Is used to determine the probability of a user, typically reflecting non-determinism and uncertainty

.

When the user moves, the particle moving unit 202 moves each of the plurality of particles from the previous position to the new position using the movement model as described above.

The likelihood calculator 204 calculates the likelihood of each of a plurality of particles using wireless signal information received from one or more wireless APs and a pre-produced wireless signal map at the actual location of the user's movement.

Here, the wireless signal information is a strength vector of a signal received by the user location tracking device from one or more wireless APs.

The wireless signal map is produced based on collecting received signal strength data from the one or more wireless APs for each region based on a phase map relating to dividing the indoor space into a plurality of regions.

According to an embodiment of the present invention, the wireless signal map may be produced through the following process.

5 is a diagram illustrating a method for producing a wireless signal map according to an embodiment of the present invention.

The manufacturing process of the wireless signal map shown in FIG. 5 may be performed by a mobile terminal capable of receiving a wireless signal from a plurality of wireless access points (APs) located nearby.

In addition, the wireless AP may be a device for transmitting a WiFi signal for enabling Internet service to a mobile terminal in a short distance, but is not limited thereto.

Referring to FIG. 5, the mobile terminal (wireless signal mapping apparatus) according to the present embodiment stores a topological map of an indoor space that is a target for producing a wireless signal map (step 500).

As shown in FIG. 6, the phase map according to the present embodiment is generated based on a plurality of regions determined to be important points in tracking the location of the user, and includes a vertex set and an edge set. And a wireless AP set.

That is, the phase map may be expressed as Equation 2 below.

은 위상 지도, V는 정점(v∈V는 실내 특정 지역), E는 간선(e ∈E는 두 지역 간 연결관계), A는 무선 AP(a ∈A는 무선 AP)이다. here,

Is the phase map, V is the vertex (v ∈ V is the specific area of the room), E is the trunk (e ∈ E is the connection between the two regions), and A is the wireless AP (a ∈ A is the wireless AP).

In general, location tracking of a user is not an open space with no compartments, but an interior space that includes moving passageways (eg, corridors) and functional zones (eg, classrooms, seminar rooms, etc.), as shown in FIG. 6. In consideration of the fact that is mainly performed in, the wireless signal map according to the present embodiment, as described above, to represent the interior space as a graph-based phase map.

Assuming the interior space of the university, individual areas such as lecture rooms, server rooms, seminar rooms, etc. can be defined as one vertex, and moving passageways, such as corridors, are divided and divided based on the entrance location of the functional area described above. Each of these regions may be defined as one vertex (L1 to L11 in FIG. 6).

In the actual user location tracking, the user is assumed to be located on the vertex of the indoor space, and is also assumed to continuously move along the trunk line when moving.

Next, the apparatus for mapping radio signals according to the present embodiment collects radio signals transmitted from the plurality of radio APs based on the phase map (step 502).

According to an exemplary embodiment of the present invention, the apparatus for producing a wireless signal map collects signal strength data wirelessly received at different time periods of the day so that time-phase variation of the wireless signal may be reflected in the signal map.

It also collects signal strength data from all areas (peaks) of a given indoor space.

Further, in order to reflect the variation of the wireless signal received in each region, the wireless signal production apparatus according to the present embodiment collects the signal strength data for the peripheral point in addition to the vertex (center point) for the region definition.

In collecting the signal strength data as described above, the signal strength data may be collected while changing the direction of the wireless signal mapping apparatus for one point.

In general short-range wireless communication, since the user's body may interfere with the reception of the signal from the wireless AP, the signal strength data is collected in different directions as described above.

 According to an embodiment of the present invention, in mapping a radio signal, the strength of a radio signal received in each region is based on the assumption that the Gaussian probability distribution has a Gaussian probability distribution, where one Gaussian probability distribution is a mean (m, m). And the standard deviation (d).

Based on this assumption, the signal mapping apparatus then performs a step of learning using the collected wireless signal strength data as described above (step 504).

In more detail, the apparatus for mapping a wireless signal calculates an average and a standard deviation of the signal strength received by each wireless AP for each region by using the collected signal strength data.

에서

번째 무선 AP로부터 수집된 신호 세기 데이터 집합에서 가우시안 확률 분포 함수의 매개변수인 평균

과 표준편차

를 각각 계산한다. For example, region

in

The average of the parameters of the Gaussian probability distribution function in the signal strength dataset collected from the first wireless AP

And standard deviation

Respectively.

In step 504, the mean and standard deviation are calculated for all regions, and the mean matrix M and the standard deviation matrix D can be obtained through the above calculation.

Here, the Gaussian probability distribution of the received radio signal strength in each region is obtained by the average matrix and the standard deviation matrix, and the user's location tracking is performed using the average matrix and the standard deviation matrix.

The likelihood calculator 204 according to the present exemplary embodiment calculates the likelihood of each particle using the wireless signal map obtained through the above process and the strength vector of the currently received wireless signal.

Likelihood calculation using the signal strength vector may be performed through Equation 3 below.

는 신호 세기 벡터,

는 각 파티클의 위치(지역),

는 특정 무선 AP로부터 수신되는 신호 세기,

는

지역에서

번째 무선 AP로부터 수신되는 신호 세기의 표준편차,

는

지역에서

번째 무선 AP로부터 수신되는 신호 세기의 평균이다. here,

Signal strength vector,

Is the location (region) of each particle,

Signal strength received from a specific wireless AP,

The

In the area

Standard deviation of the signal strength received from the first wireless AP,

The

In the area

The average of the signal strength received from the first wireless AP.

Referring back to FIG. 2, the weight setting unit 206 according to the present embodiment performs a process of setting and updating weights of each of the plurality of particles using the likelihood calculated for each particle.

)에서 가중치(weight)를 설정한다. In the initial process, the weight setting unit 206 is the position where each particle is moved (

Set the weight in.

According to an exemplary embodiment of the present invention, the particle resampling unit 208 newly selects a particle to replace each of the plurality of particles by using a weight set for each particle.

In this case, the particle resampling unit 208 repeats the particle resampling process until the user's position is finally determined using the roulette wheel algorithm to which the weight of each particle is applied. Can be defined as a process.

For example, assume that three particles are initially set randomly in an indoor space.

When the user moves, assuming that the weights of the first to third particles are set to 0.5, 0.7, and 0.9 through the likelihood calculation, when particle resampling is performed, the third particles having the highest weight are the first and the third. It is most likely to replace the second particle.

However, since the roulette wheel algorithm is applied, the above-described particle replacement is only a possibility, and the first particle may be replaced by the second particle, or may be replaced by itself (the first particle). Also, the third particle may be replaced with the first or second particle.

As described above, after the particles are replaced through particle resampling, when the user moves again, the movement and likelihood calculation of the particles is performed again, and the weight updating process of each replaced particle is performed through the recalculated likelihood. .

That is, the weight setting unit 206 performs the process of updating the weight of each particle through the iteration as shown in the following equation.

The particle resampling unit 208 performs the replacement process of each particle by using the updated weight of each particle.

When particle movement, likelihood calculation, weight updating, and particle resampling are repeatedly performed, a finally converged particle appears, and the user location tracking apparatus estimates the finally converged particle as the user's location.

Meanwhile, as described above, at the same time as the location tracking of the user is performed based on the particle filter, the movement model learner 106 according to the present embodiment collects statistical data about the movement path of each particle and based on the statistical data. Update the movement model.

Here, the statistical data on the movement path of the particle may be a movement path regarding the particles converged through the iterative process as described above.

According to an embodiment of the present invention, a Monte Carlo Expectation Maximization (MCEM) algorithm may be used to train a moving model.

When the movement model is updated, the particle moving unit 202 moves the particles based on the updated movement model.

Since the updated movement model is based on the movement path of the particle which has already been tracked, it is possible to predict the next movement path of the particle more effectively than the initial movement model, so that the user's location tracking process can be more accurate.

The user location tracking can be performed by a computer readable recording medium.

The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be those specially designed and constructed for the present invention or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks, and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Examples of program instructions, such as magneto-optical and ROM, RAM, flash memory and the like, can be executed by a computer using an interpreter or the like, as well as machine code, Includes a high-level language code. The hardware device described above may be configured to operate as at least one software module to perform the operations of one embodiment of the present invention, and vice versa.

Preferred embodiments of the present invention described above are disclosed for purposes of illustration, and those skilled in the art will be able to make various modifications, changes, and additions within the spirit and scope of the present invention. Additions should be considered to be within the scope of the following claims.

Claims (11)

In the particle filter-based user location tracking device,
When the user moves in an indoor space, the plurality of particles are moved from the previous position to the new position according to the moving model, and the radio signal information and the pre-produced radio signal received from one or more wireless APs at the actual position where the user moves. The likelihood of each of the plurality of particles is calculated using a map, the weight of each of the plurality of particles is updated using the calculated likelihood, and the particles to replace each of the plurality of particles are updated using the updated weight. Particle resampling is selected, and particles are finally converged by performing the repetition by repeatedly performing particle resampling at the movement of the particle, the likelihood calculation, the updating of the weight, and the newly moved position according to the movement of the user. A user location tracking unit for estimating a user's actual location; And
And a moving model learner for updating the moving model using statistical data on the moving path of the converged particles.
The method of claim 1,
Wherein each of the plurality of particles is defined by at least one of a position, a direction, and a movement state. The method of claim 1,
The wireless signal information is an intensity vector of a signal received from one or more wireless APs, and the wireless signal map is received from the one or more wireless APs for each region based on a phase map relating to dividing the indoor space into a plurality of regions. And a user position tracking device produced based on collecting the collected signal strength data. Claim 4 has been abandoned due to the setting registration fee. The method of claim 3,
The topological map includes at least one of a vertex set for each region definition, an edge set for defining a connection relationship between two regions adjacent to each other, and a set of the one or more wireless APs. Claim 5 was abandoned upon payment of a set-up fee. The method of claim 3,
The wireless signal map is a user location tracking device defined by the average matrix and the standard deviation matrix generated using the average and standard deviation of the signal strength data collected for each region. Claim 6 has been abandoned due to the setting registration fee. 5. The method of claim 4,
When moving, the user is assumed to be located on the vertex, it is assumed to move continuously on the edge. The method of claim 1,
And the user location tracking unit randomly selects a plurality of candidate locations of a user at a start point of user location tracking.
The method of claim 1,
And the user position tracking unit randomly selects a new particle to replace each particle by reflecting the weights calculated for each of the plurality of particles using a roulette wheel algorithm for the particle resampling. The method of claim 1,
The movement model learner updates the movement model by using a Monte Carlo Expectation Maximization (MCEM) algorithm. Initially randomly select a plurality of candidate positions (particles) of the user, and when the user moves, each of the plurality of initially selected particles are moved according to the moving model, and at least one wireless AP at the actual position moved by the user The likelihood of each of the plurality of particles is calculated using a radio signal received from the radio signal map and a pre-produced radio signal map, the weight of each of the plurality of particles is set using the calculated likelihood, and the set weight is used. Particle resampling unit for selecting a particle to replace each of the plurality of particles, Particle resampling is repeatedly performed at the movement of the particle, the likelihood calculation, the update of the weight and the newly moved position according to the movement of the user, The final converging particle by the iteration is the user's actual position A user location tracking unit estimated to be; And
And a moving model learner for updating the moving model using statistical data on the moving path of the converged particles.
A program of instructions that can be executed in a mobile terminal for user location tracking is tangibly embodied and can be read by the mobile terminal.
(a) a particle moving step of moving each of a plurality of particles from an old position to a new position according to a moving model when the user moves in the indoor space, wherein the plurality of particles are candidate positions of the user;
A likelihood calculation step of calculating a likelihood of each of the plurality of particles using wireless signal information received from at least one wireless AP and a pre-produced wireless signal map at a user's actual location;
(b) a weight updating step of updating weights of each of the plurality of particles using the calculated likelihood;
(c) particle resampling to replace each of the plurality of particles using the updated weights;
(d) repeating the particle movement, the likelihood calculation, the update of the weight, and the resampling of the particle at the newly moved position according to the movement of the user to estimate the final converged particle as the actual position of the user; And
(e) updating the movement model using statistical data on the movement path of the converging particles.

Images