CN115988634A - Indoor positioning method and subspace feature extraction method - Google Patents
Indoor positioning method and subspace feature extraction method Download PDFInfo
- Publication number
- CN115988634A CN115988634A CN202211480461.4A CN202211480461A CN115988634A CN 115988634 A CN115988634 A CN 115988634A CN 202211480461 A CN202211480461 A CN 202211480461A CN 115988634 A CN115988634 A CN 115988634A
- Authority
- CN
- China
- Prior art keywords
- subspace
- sample
- sensors
- equal
- matched
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Position Fixing By Use Of Radio Waves (AREA)
Abstract
An indoor positioning method and a subspace feature extraction method are provided, wherein in the indoor positioning method, to-be-matched data of an object to be positioned are obtained, wherein the to-be-matched data comprise signal intensity values between a third sensor and D first sensors; and determining the region matched with the data to be matched according to the relation between the measured received signal strength value of the D first sensors and the K regions. In the subspace feature extraction method, data to be matched of an object to be extracted is obtained, wherein the data to be matched comprises signal intensity values between a third sensor and D first sensors; and extracting subspace characteristics corresponding to the data to be matched according to the subspace segmentation network. Because the positioning is finished according to the received signal strength value from the equipment, no additional equipment is needed, the sample sequence for subspace segmentation network training is acquired through K region sequences, and samples without time and position labels are segmented and clustered into K segments, so that the hardware configuration is reduced, and the data calibration work is reduced.
Description
Technical Field
The invention relates to the field of position location, in particular to an indoor positioning method and a subspace feature extraction method.
Background
Indoor positioning is widely applied in human life, and a three-point positioning method and a fingerprint positioning method are commonly used.
The three-point positioning method is a triangular positioning method based on arrival time, arrival time difference or arrival angle, and can realize sub-meter positioning accuracy. However, some of them require a complicated hardware configuration, for example, the triangulation method based on the angle of arrival requires the use of a multi-antenna system; while other methods may require expensive infrastructure, for example, time difference of arrival based positioning needs to be applied to access point networks with highly accurate time synchronization functionality. Furthermore, such methods mostly require line-of-sight propagation conditions, which are difficult to meet in complex indoor environments.
Fingerprint positioning methods are fingerprint-based methods, can provide meter-level accuracy, and they do not require special hardware or line-of-sight conditions, however, fingerprint-based positioning methods require construction of a fingerprint database, and the data collection and data on-site calibration processes require a large amount of manpower. Specifically, the fingerprint positioning is divided into two stages of collecting and establishing a database offline and matching online. In the off-line data collection stage, in order to improve the positioning accuracy, it is necessary to acquire data at very small intervals and calibrate coordinates on the spot for the data. The calibration work will grow linearly with increasing room area and increasing accuracy requirements.
Most of the existing positioning methods need expensive hardware configuration or heavy data on-site calibration work, and due to the complex indoor environment, signal propagation attenuation is complex and cannot be modeled, so that the requirements of the complex indoor environment cannot be met.
Disclosure of Invention
The invention mainly solves the technical problem of providing an indoor positioning method and a subspace feature extraction method, and has the characteristics of reducing hardware configuration and/or reducing heavy data calibration work requirements.
According to a first aspect, an embodiment provides an indoor positioning method, which includes the steps of obtaining data x to be matched of an object to be positioned, wherein the data x to be matched includes signal strength values between a third sensor and D first sensors, and D is larger than or equal to 2;
the method comprises the following steps that K areas needing to be positioned are divided indoors, K is larger than or equal to 2, the D first sensors are arranged indoors, and the areas matched with data to be matched are determined according to the relation between the measured received signal strength values of the D first sensors and the K areas, so that the positioning of the object to be positioned is realized.
In one embodiment, the determining, according to the relationship between the measured received signal strength values of the D first sensors and the K regions, a region to which the data to be matched is matched, so as to realize the positioning of the object to be positioned, includes:
based on subspace features
Using the distance relationship f (x) of the signal intensity value samples to the subspace between the D first sensors and the second sensor i ,U k ,b k ) One region for each subspace, so that the matching error f (x, U) k ,b k ) The minimum subspace k is the matched subspace, so that the region matched by the data to be matched is determined; wherein K is more than or equal to 1 and less than or equal to K>
Is a base of a subspace, <' > v>
Is the center of the subspace d k Is a dimension of a subspace;
the subspace characteristics
The obtaining method comprises the following steps:
constructing a sample set X, wherein the sample set X comprises N samples, and N is more than or equal to 2,x i For a sample in the sample set X, i is more than or equal to 1 and less than or equal to N, the sample comprises signal intensity values between D first sensors and D second sensors which are acquired when the sample passes through the K areas, and the sample sequence in the sample set is that the sample sequence sequentially passes through the K areas sequentially and not repeatedlyA collected sample;
sequentially dividing the sample set into K clusters through K-1 dividing points, wherein each cluster sequentially corresponds to the K subspaces one by one;
constructing the distance relation f (x) of each sample to any subspace i ,U k ,b k )；
Constructing a subspace division network according to a minimum principle from each sample to a subspace to which each sample belongs, the distance relation and a principle that N samples are sequentially divided into K clusters, wherein the subspace division network is used for solving the K-1 division points and the subspace characteristics
In an embodiment, the central position coordinate of the kth region corresponding to the subspace k is the positioning coordinate of the positioning object.
In one embodiment, the determining, according to the relationship between the measured received signal strength values of the D first sensors and the K regions, a region to which the data to be matched is matched, so as to realize the positioning of the object to be positioned, includes:
the position of the object to be positioned is obtained by inputting the data to be matched into a subspace matching neural network, and the training method of the subspace matching neural network comprises the following steps:
constructing a sample set X, wherein the sample set X comprises N samples, and N is more than or equal to 2,x i The sample is a sample in a sample set X, i is more than or equal to 1 and less than or equal to N, the sample comprises signal intensity values between D first sensors and D second sensors which are acquired when the sample passes through the K regions, and the sample sequence in the sample set is samples which are sequentially acquired when the sample passes through the K regions sequentially and not repeatedly;
sequentially dividing the sample set into K clusters through K-1 dividing points, wherein each cluster sequentially corresponds to the K regions one by one, and the K regions correspond to K subspaces l (i) one by one, so that a training set is obtained
Using the training set
And training to obtain the subspace matching neural network.
In one embodiment, the sequentially partitioning the sample set into K clusters through K-1 partition points includes:
constructing the distance relationship f (x) of each sample to any subspace i ,U k ,b k )；
Constructing a subspace division network according to a minimum principle from each sample to a subspace to which each sample belongs, the distance relation and a principle that N samples are sequentially divided into K clusters, wherein the subspace division network is used for solving the K-1 division points and subspace characteristics
Solving the K-1 segmentation points according to the subspace segmentation network, so that the sample set is sequentially segmented into K clusters through the K-1 segmentation points.
In one embodiment, the distance relationship f (x) of each sample to any subspace i ,U k ,b k ) Can be expressed as:
according to the subspace minimum principle from each sample to each sample and the distance relationship, the subspace partitioning network may be expressed as:
wherein, t k Is a division point of the sample set X, t is more than or equal to 1 1 <t 2 <…<t K-1 N is less than or equal to N; solving equation (3) by using an iterative optimization algorithm, namely, assuming a known segmentation point t 1 ，t 2 ，…，t K-1 To obtain subspace characteristics
Then it is assumed that a known subspace characteristic is>
Calculating a division point t 1 ，t 2 ，…，t K-1 The loop iterates until convergence. />
In one embodiment, the sample x i And also position coordinates z (i) ＝(z 1 (i) ,z 2 (i) ) Wherein
o q For the position coordinate of the qth first sensor, <' >>
Represents the jth first sensor, r is a signal intensity value, < '> or <' > is greater than>
For the signal intensity value of the jth first sensor in the ith sample, <' >>
The signal intensity value of the qth first sensor in the ith sample is obtained.
In one embodiment, the D first sensors are the same type of sensor, or the D first sensors are different types of sensors.
According to a second aspect, there is provided in an embodiment a subspace feature extraction method, comprising,
acquiring data x to be matched of an object to be extracted, wherein the data x to be matched comprises signal intensity values between a third sensor and D first sensors, and D is more than or equal to 2;
extracting subspace characteristics corresponding to data x to be matched according to subspace segmentation network
The subspace partitioning network is based on each sample x i Distance relationship f (x) to any subspace i ,U k ,b k ) The minimum principle from each sample to the subspace to which each sample belongs and the principle that N samples are sequentially divided into K clusters are constructed, wherein K is more than or equal to 2,1 and is less than or equal to K and/or greater than or equal to K>
Is the base of the subspace>
Is the center of the subspace d k Is a dimension of a subspace;
x i is one sample in N samples in the sample set X, N is not less than 2,1 and not more than i and not more than N, each sample X i The method comprises the signal intensity values between D first sensors and D second sensors which are acquired when the samples pass through K areas, and the sample sequence in the sample set is samples which are sequentially acquired when the samples pass through the K areas in sequence and not repeatedly; the D first sensors are sensors arranged in the K areas and/or outside the K areas;
the K regions correspond to K subspaces one by one, the sample set is sequentially divided into K clusters through K-1 dividing points, and each cluster corresponds to the K subspaces one by one.
In one embodiment, the subspace partitioning network is based on each sample x i Distance relationship f (x) to any subspace i ,U k ,b k ) The principle that each sample is minimum to the subspace to which each sample belongs and the principle that N samples are sequentially divided into K clusters are constructed, and the method comprises the following steps:
each sample x i Distance relationship f (x) to any subspace i ,U k ,b k ) Can be expressed as:
according to the subspace minimum principle from each sample to each sample and the distance relationship, the subspace partitioning network may be expressed as:
wherein, t k Is a division point of the sample set X, t is more than or equal to 1 1 <t 2 <…<t K-1 N is less than or equal to N; solving equation (3) by using an iterative optimization algorithm, namely, assuming a known segmentation point t 1 ，t 2 ，…，t K-1 To find subspace features
Then it is assumed that a known subspace characteristic is>
Calculating a division point t 1 ，t 2 ，…，t K-1 The loop iterates until convergence.
According to a third aspect, an embodiment provides a computer-readable storage medium having stored thereon a program executable by a processor to implement any of the above-described indoor positioning methods, and/or any of the above-described subspace feature extraction methods.
In the indoor positioning method and the subspace feature extraction method provided by the embodiment of the invention,
according to the relation between the signal intensity values of the D first sensors and the K areas, the areas matched with the data to be matched are determined, so that the object to be positioned is positioned, the positioning is completed according to the signal intensity values received from the equipment by adopting a positioning scheme, no additional equipment is required to be installed, the positioning precision can be improved along with the increase of the data of the indoor sensors, and the coordinates of the data do not need to be calibrated on the spot, so that the hardware configuration is reduced, and the heavy data calibration work is reduced.
The method comprises the steps of extracting subspace characteristics, wherein a sample sequence adopted in subspace segmentation network training is a sample which is sequentially collected when the sample passes through K regions in sequence and does not repeatedly pass through the K regions, segmenting and clustering sample data without time labels and position labels into K sections, collecting each section of data at one position, and completing data calibration work by the segmentation clustering algorithm instead of manpower.
Drawings
FIG. 1 is a schematic diagram of an indoor positioning scenario according to an embodiment;
FIG. 2 is a schematic flow chart of an indoor positioning method;
FIG. 3 is a schematic flow chart of subspace partitioning network construction according to an embodiment;
FIG. 4 is a schematic diagram of an embodiment of a subspace feature extraction process;
FIG. 5 is a schematic view of an exemplary indoor positioning apparatus;
fig. 6 is a comparison graph of indoor positioning by the scheme of the present invention and indoor positioning by the existing scheme.
Detailed Description
The present invention will be described in further detail with reference to the following detailed description and accompanying drawings. Wherein like elements in different embodiments are numbered with like associated elements. In the following description, numerous details are set forth in order to provide a better understanding of the present application. However, one skilled in the art will readily recognize that some of the features may be omitted or replaced with other elements, materials, methods in different instances. In some instances, certain operations related to the present application have not been shown or described in detail in order to avoid obscuring the core of the present application from excessive description, and it is not necessary for those skilled in the art to describe these operations in detail, so that they may be fully understood from the description in the specification and the general knowledge in the art.
Furthermore, the features, operations, or characteristics described in the specification may be combined in any suitable manner to form various embodiments. Also, the various steps or actions in the method descriptions may be transposed or transposed in order, as will be apparent to one of ordinary skill in the art. Thus, the various sequences in the specification and drawings are for the purpose of describing certain embodiments only and are not intended to imply a required sequence unless otherwise indicated where such sequence must be followed.
The numbering of the components as such, e.g., "first", "second", etc., is used herein only to distinguish the objects as described, and does not have any sequential or technical meaning. The term "connected" and "coupled" as used herein includes both direct and indirect connections (couplings), unless otherwise specified.
In order to make the description more clear, please refer to fig. 1, taking fig. 1 as an example, fig. 1 is a distribution diagram of an indoor area that can be divided into a plurality of areas for positioning, in fig. 1, a diagonal filling part is divided into areas of interest that need to be positioned, and a black solid dot part is a first sensor that is installed.
In one embodiment, referring to fig. 2, the following indoor positioning method is adopted:
s101, obtaining data x to be matched of an object to be positioned, wherein the data x to be matched comprises signal intensity values between a third sensor and D first sensors, and D is larger than or equal to 2;
s102, determining the area matched with the data to be matched according to the measured relation between the received signal strength value of the D first sensors and the K areas, and accordingly positioning the object to be positioned. The indoor space is divided into K areas needing positioning, K is more than or equal to 2, and the D first sensors are arranged in the indoor space.
When an object needs to be located, under certain conditions, a sensor which can receive and/or send signals and is used for realizing location needs to be matched with the object, and the sensor is distinguished from other sensors which have functional functions, and is called as a third sensor here, but does not mean that the third sensor and other first and second sensors cannot be the same type of sensor; alternatively, the third sensor may be directly used as an object to be positioned. Typically, the third sensor may be a sensor capable of receiving and/or sending signals, such as a mobile phone and a bracelet, for positioning, and when the D first sensors are capable of receiving signals, the third sensor may also be a sensor having only a signal transmitting function, and of course, may also be a sensor having only a signal receiving function.
The area needing to be positioned in the room is divided into K areas, the K areas can be partial areas in the room or all areas in the room, as shown in fig. 1, the divided areas needing to be positioned are 9 areas, and only partial areas in all areas in the room, the division mode can be divided depending on obstacles or not depending on obstacles, for example, a relatively large open area can be divided into several areas, typically, a large production workshop without higher obstacle intervals can be divided into several areas according to the needs.
For the D first sensors, the D first sensors may be disposed only in the K regions, or may be partially disposed outside the K regions, for example, 20 first sensors in fig. 1 may be disposed just outside the K regions, and the specific disposition positions may be set according to region division, layout and actual requirements.
For D first sensors at fixed positions, the strength of the emitted signal decreases with increasing distance, and the farther the distance is, the faster the signal attenuation is, that is, for D first sensors at different positions and randomly arranged, a third sensor at a certain position, for D first sensors, there are received signal strength values with different sizes, that is, the signal strength value sets formed by the received signal strength values of the third sensors at different positions for the D first sensors are different, so i can determine the currently obtained far-near distance relationship between the received signal strength value of the D first sensors for the third sensors and the K regions, that is, the region to which the data to be matched is matched, according to the received signal strength value of the D first sensors for the third sensors and the measured relationship between the received signal strength value of the D first sensors and the K regions, thereby realizing the positioning of the object to be positioned.
Therefore, the invention provides a novel indoor positioning method, which comprises the steps of obtaining data x to be matched of an object to be positioned, wherein the data x to be matched comprises signal intensity values between a third sensor and D first sensors; and determining the region matched with the data to be matched according to the relation between the measured received signal strength value of the D first sensors and the K regions, thereby realizing the positioning of the object to be positioned.
With the wide application of smart homes, more and more devices with the function of transmitting signals are arranged indoors, and the rapid development of smart homes enables a plurality of wireless sensors, such as routers, intelligent switches, voice assistants, computers, smart televisions and the like, to be arranged indoors, so that the wireless sensors transmit signals in real time, and the current position of a user can be roughly presumed by using the signal intensity according to the proportional relation between the distance and the signal intensity. In the existing positioning schemes, for example, the UWB positioning scheme of triangulation positioning requires that the transceiver has a function of calculating the signal transceiving time, and each sensor needs to be able to complete the time coordination calibration. The positioning scheme adopted by the invention completes positioning according to the strength value of the received signal from the equipment, does not need a position label or a time label, does not need any additional equipment, improves the positioning precision along with the increase of the data of the indoor sensor, and does not need to calibrate coordinates on the spot, thereby reducing hardware configuration and simultaneously reducing heavy data calibration work.
In the application of the present invention, two methods are provided for determining the region matched with the data to be matched according to the measured relationship between the received signal strength value of the D first sensors and the K regions, so as to realize the positioning of the object to be positioned.
As an embodiment, the method for determining the region matched with the data to be matched according to the measured relation between the received signal strength value of the D first sensors and the K regions so as to realize the positioning of the object to be positioned comprises the following steps of determining the region matched with the data to be matched according to the measured relation between the received signal strength value of the D first sensors and the K regions
Using the distance relationship f (x) of the signal intensity value samples to the subspace between the D first sensors and the second sensor i ,U k ,b k ) One region for each subspace, so that the matching error f (x, U) k ,b k ) The minimum subspace k is the matched subspace, so that the region matched by the data to be matched is determined; wherein K is more than or equal to 1 and less than or equal to K>
Is the base of the subspace>
Is the center of the subspace d k Is a dimension of a subspace; the subspace characteristic->
The obtaining method comprises the following steps:
constructing a sample set X, wherein the sample set X comprises N samples, and N is more than or equal to 2,x i The method comprises the steps that i is more than or equal to 1 and less than or equal to N, a sample in a sample set X comprises signal intensity values between D first sensors and second sensors which are acquired when the sample passes through K regions, and a sample sequence in the sample set is samples which are acquired when the sample passes through the K regions sequentially and non-repeatedly; sequentially dividing the sample set into K clusters through K-1 dividing points, wherein each cluster corresponds to K subspaces one by one; constructing the distance relationship f (x) of each sample to any subspace i ,U k ,b k ) (ii) a According to the minimum principle of each sample to the subspace of each sample and the distance relationship f (x) i ,U k ,b k ) And constructing a subspace division network according to the principle that N samples are sequentially divided into K clustersThe cutting network is used for solving the above-mentioned cutting points and subspace characteristics
If the D signal intensity values in each sample are regarded as a point in a D-dimensional space, the data at each position are collected on a subspace in the D-dimensional space, and therefore, the clustering principle of the data in positioning can be set to find K subspaces corresponding to K positions, so that the data are divided into K clusters, and each cluster is closest to a subspace.
The task of reducing data labeling is always the focus research object of indoor positioning researchers, such as fingerprint positioning method, and the accurate position is required to be known and the signal characteristics received by the accurate position are required. According to the scheme of the invention, the acquisition process of the sample set is known, and the second sensor sequentially passes through K regions without repetition, so that the sample set can be sequentially divided into K clusters through K-1 dividing points, and each cluster is sequentially corresponding to K subspaces one by one; this allows us to complete the data grouping with non-labeled data according to a certain rule, thus also allowing to complete the data targeting work instead of manual work. Therefore, only the signal strength values between the D first sensors and the second sensors, which are acquired when the second sensor passes through the K regions, need to be acquired, and the precise position of any sensor does not need to be known, and other signal characteristics except the received signal strength values do not need to be known.
It can be understood that, based on the structural relationship of the indoor K regions, we cannot completely pass through the K regions sequentially and without repetition, for example, as shown in fig. 1, the left region and the right region of the lower right corner are passed through first, and then reach the right region, but when the right region reaches the upper region of the right region, it is required to pass through the left region and then reach the upper region of the right region from the right region, and in this process, the sequential and non-repetitive operations cannot be completely achieved, and therefore, the sequential and non-repetitive operations in the present application are relatively sequential and non-repetitive meanings. Therefore, the relative relationship between the time of staying in the corresponding region when the sample data needs to be acquired and the time of staying in the corresponding region when the sample data does not need to be acquired temporarily passes in the process of acquiring the sample data needs to be controlled, so that the time of staying in the corresponding region when the sample data does not need to be acquired temporarily passes can be relatively ignored. For example, the ratio of the time spent in the corresponding region when the sample data needs to be acquired to the time spent in the corresponding region when the sample data does not need to be acquired in the process of acquiring the sample data is greater than 10, and the relativity can be controlled by a person skilled in the art according to the actual situation, which is not limited herein.
It will be appreciated that the second sensor is only intended to distinguish from a third sensor of the object to be positioned, and that the third sensor used to effect positioning may be the same type of sensor as the second sensor used to collect the sample, may be a different type of sensor, and may even be the same sensor.
Based on the sample collection mode and the segmentation of the sample set, only the distance relationship f (x) between one sample and any subspace needs to be constructed i ,U k ,b k ) (ii) a According to the minimum principle of each sample to the subspace of each sample and the distance relationship f (x) i ,U k ,b k ) And constructing a subspace division network according to the principle that N samples are sequentially divided into K clusters, wherein the subspace division network is used for solving the division points and subspace characteristics
Because each sample is obtained in different region sequences, the distance relationship between each sample and the subspace corresponding to any region is different, and the distance between each sample and the subspace to which each sample belongs is the smallest, so that the distance relationship f (x) between one sample and any subspace can be obtained according to the principle that each sample and the subspace to which each sample belongs are the smallest i ,U k ,b k ) And the principle that N samples are sequentially divided into K clusters to construct a subspace division network which can be used for solving the division intoDividing points of the samples of the K clusters and the characteristics of subspaces to which the samples belong, so that the matching errors f (x, U) can be realized according to the data needing to be matched k ,b k ) The minimum subspace k is the matched subspace, and the feature of the subspace is obtained through the subspace segmentation network, namely the corresponding region is known, so that the positioning is realized.
In one embodiment, the coordinates of the center position of the kth region corresponding to the subspace k are the positioning coordinates of the positioning object.
Next, another method for determining the region matched with the data to be matched according to the measured relationship between the received signal strength values of the D first sensors and the K regions, so as to realize the positioning of the object to be positioned is described. The method specifically comprises the following steps:
the position of an object to be positioned is obtained by inputting data to be matched into a subspace matching neural network, and the training method of the subspace matching neural network comprises the following steps:
constructing a sample set X, wherein the sample set X comprises N samples, and N is more than or equal to 2,x i The method comprises the steps that i is more than or equal to 1 and less than or equal to N, a sample in a sample set X comprises signal intensity values between D first sensors and second sensors which are acquired when the sample passes through K regions, and a sample sequence in the sample set is samples which are acquired when the sample passes through the K regions sequentially and non-repeatedly;
sequentially dividing the sample set into K clusters through K-1 dividing points, wherein each cluster sequentially corresponds to K regions one by one, and the K regions correspond to K subspaces l (i) one by one, so that a training set is obtained
Using the training set>
And training to obtain a subspace matching neural network.
Since the way of collecting samples is the same as that of the previous method for positioning the object to be positioned, we do not need to describe here specifically, and differently, the sample set is sequentially divided into K clusters through K-1 dividing points, and each cluster corresponds to K regions one by oneThe domain, K regions are one-to-one corresponding to K subspaces l (i), so that each sample data x can be obtained i And each sample data x i Training set composed of subspaces l (i) to which they belong
We use this training set->
The subspace matching neural network belonging to the CNN may be obtained through machine learning, so that the subspace matching neural network can be utilized to obtain a matched subspace according to the acquired data to be matched, thereby realizing the positioning of the object to be positioned.
As an embodiment, the sequential partitioning of the sample set into K clusters by K-1 partitioning points includes:
constructing the distance relationship f (x) of each sample to any subspace i ,U k ,b k )；
According to the minimum principle of each sample to the subspace of each sample and the distance relationship f (x) i ,U k ,b k ) And constructing a subspace division network by the principle that the N samples are sequentially divided into K clusters, wherein the subspace division network is used for solving the K-1 division points and the subspace characteristics
The K-1 segmentation points are solved according to a subspace segmentation network, so that the sample set is sequentially segmented into K clusters through the K-1 segmentation points.
In this embodiment, we can also solve the K-1 division points according to the subspace division network constructed according to the subspace minimum rule, the constructed distance relationship and the K clusters to which each sample belongs, and the constructed subspace division network related to the division points and the subspace characteristics.
In one embodiment, referring to fig. 4, we can choose to obtain the subspace partitioning network by:
s201, relating the distance relationship f (x) of each sample to any subspace i ,U k ,b k ) Expressed as:
s202, according to the minimum principle and the distance relation f (x) from each sample to the subspace of each sample i ,U k ,b k ) The subspace partitioning network may be expressed as:
wherein, t k Is a division point of the sample set X, t is more than or equal to 1 1 <t 2 <…<t K-1 ≤N；
S203, solving the formula (3) by adopting an iterative optimization algorithm, namely, assuming that the known segmentation point t is known 1 ，t 2 ，…，t K-1 To obtain subspace characteristics
It is then assumed that a known subspace characteristic->
Calculating a division point t 1 ，t 2 ，…，t K-1 The loop iterates until convergence.
Since each sample x i The relationship with any subspace feature can be modeled as
x i = Uy + b + epsilon formula (1)
Where y is the coordinate of the sample in subspace and ε is the zero mean Gaussian error, so each sample x i Distance relationship f (x) to any subspace i ,U k ,b k ) Can be expressed as:
and then according to the minimum principle and distance relationship f (x) from each sample to the subspace of each sample i ,U k ,b k ) Then, the subspace segmentation network represented by the above formula (3) is obtained, and according to the characteristics of the subspace network, we can pair K-1 segmentation points t according to the trained subspace network 1 ，t 2 ，…，t K-1 And subspace characterization
And (6) solving.
In one embodiment, where the positions of the D first sensors are unknown, the vector data x in equation (1) is used i The construction method of (2) may be:
for each time point i in the time series i =1,2, …, N, the signal intensity value between the second sensor and the D first sensors is obtained
Sample x i Can be constructed as a D-dimensional vector
Here, it should be noted that x i Is not unique, when the positions of U and y in equation (1) are interchanged, then->
In one embodiment, sample x i May also include a position coordinate z (i) ＝(z 1 (i) ,z 2 (i) ) In which
o q For the position coordinates of the qth first sensor, in the device>
Represents the jth first sensor, r is a signal intensity value>
For the signal intensity value of the jth first sensor in the ith sample, <' >>
The signal intensity value of the qth first sensor in the ith sample is obtained.
With the known positions of the D first sensors, for the vector data x in equation (1) i Can be constructed as a D + 2-dimensional vector
Here, it should be noted that x i Is not unique, when the positions of U and y in equation (1) are interchanged, then->
Based on the above, the present invention also provides a new subspace feature extraction method, please refer to fig. 5, which includes,
s301, obtaining data x to be matched of an object to be extracted, wherein the data x to be matched comprises signal intensity values between a third sensor and D first sensors, and D is larger than or equal to 2;
s302, extracting subspace characteristics corresponding to the data x to be matched according to the subspace segmentation network
Subspace partitioning network from each sample x i Distance relationship f (x) to any subspace i ,U k ,b k ) Constructing a principle that each sample to the subspace to which each sample belongs is minimum and a principle that N samples are sequentially divided into K clusters, wherein K is more than or equal to 2,1 and is more than or equal to K and less than or equal to K,
is the base of the subspace>
Is the center of the subspace d k Is a dimension of a subspace;
x i is one sample in N samples in the sample set X, N is more than or equal to 2,1 and is more than or equal to i and less than or equal to N, and each sample X i The method comprises the steps of acquiring signal intensity values between D first sensors and D second sensors when the signals pass through K regions, wherein a sample sequence in a sample set is samples acquired sequentially when the signals pass through the K regions sequentially and non-repeatedly; the system comprises K areas to be positioned, D first sensors and a positioning module, wherein the K areas are divided into K areas to be positioned, and the D first sensors are sensors arranged in the K areas and/or outside the K areas;
the K regions correspond to K subspaces one by one, the sample set is sequentially divided into K clusters through K-1 dividing points, and each cluster corresponds to the K subspaces one by one.
It is understood that the subspace feature extraction method may be used in an indoor positioning method, and may also be used in other positioning applications that may occur to those skilled in the art in other environments. It should also be understood that the sequential and non-repeated passing through the K regions is also the sequential and non-repeated passing through the K regions, which is not described in detail herein, and in addition, it should also be understood that the D first sensors may be sensors of the same type or different types, and likewise, the first sensor and the second sensor may be sensors of the same type or different types.
In the scheme of the invention, a sample data acquisition scheme is set to sequentially pass through each area desired to be positioned, so that a common clustering problem is changed into a segmentation clustering problem, namely, data is divided into continuous K segments, each segment of data is acquired at one position, then an optimization problem is designed by using a clustering criterion, and the solution of segmentation clustering is completed.
Based on the above, in one embodiment, there is provided an indoor positioning device, as shown in fig. 6, including,
the matching data acquisition module 01 is used for acquiring data x to be matched of an object to be positioned, wherein the data x to be matched comprises signal intensity values between the third sensor and D first sensors, and D is more than or equal to 2; and
the position determining module 02 is used for determining the region matched with the data to be matched according to the relation between the measured received signal strength value of the D first sensors and the K regions, so that the positioning of the object to be positioned is realized; k areas needing to be positioned are divided indoors, K is larger than or equal to 2, and D first sensors are arranged indoors.
It can be understood that the indoor positioning device may be one intelligent device, or may be two or more devices, and when the indoor positioning device is one intelligent device, the intelligent device may communicate with the third sensor and/or the D first sensors to obtain data to be matched, and a program for implementing the positioning method is stored in a memory of the intelligent device, and a processor of the intelligent device may execute the program for implementing the positioning method, and the positioning device may further include a display to display a positioning result. When the number of the devices is more than two, at least one device can communicate with the third sensor and/or the D first sensors to obtain data to be matched, and at least one device can execute a positioning program according to the data to be matched to obtain a positioning result.
Based on the foregoing, one embodiment provides a computer-readable storage medium having a program stored thereon, where the program can be executed by a processor to implement any of the above-mentioned indoor positioning methods, and/or any of the above-mentioned subspace feature extraction methods.
Based on the application scenario shown in fig. 1, the data sets of multiple unlabeled sequences are obtained by sequentially passing through 9 regions in the figure for multiple times, the data sets are divided into a training set, a test set 1 and a test set 2, and a design method is evaluated by using region classification errors and region positioning errors. As shown in fig. 6, the subspace matching algorithm and the DNN classification algorithm proposed by the present application are superior to the conventional signal maximum matching algorithm and the weight center algorithm. In the traditional method, original data are often directly utilized to carry out indoor positioning, the original data often have noise and interference, and particularly in a complex indoor environment, the positioning accuracy of the traditional scheme is low. The scheme provided by the invention extracts the subspace characteristics of each region, and the subspace characteristics can more effectively represent each region, so that the positioning accuracy can be improved by completing the positioning by utilizing the characteristic matching.
It will be appreciated that the processor may be an integrated circuit chip having signal processing capabilities. The Processor may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like. The various methods, steps, and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The storage medium may be, but is not limited to, a Random Access Memory (RAM), a Read Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Read-Only Memory (EPROM), an electrically Erasable Read-Only Memory (EEPROM), and the like.
Reference is made herein to various exemplary embodiments. However, those skilled in the art will recognize that changes and modifications may be made to the exemplary embodiments without departing from the scope hereof. For example, the various operational steps, as well as the components for performing the operational steps, may be implemented in differing ways depending upon the particular application or consideration of any number of cost functions associated with the operation of the system (e.g., one or more steps may be deleted, modified, or combined with other steps).
Additionally, as will be appreciated by one skilled in the art, the principles herein may be reflected in a computer program product on a computer readable storage medium, which is pre-loaded with computer readable program code. Any tangible, non-transitory computer-readable storage medium may be used, including magnetic storage devices (hard disks, floppy disks, etc.), optical storage devices (CD-ROMs, DVDs, blu Ray disks, etc.), flash memory, and/or the like. These computer program instructions may be loaded onto a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions which execute on the computer or other programmable data processing apparatus create means for implementing the functions specified. These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including means for implementing the function specified. The computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified.
The present invention has been described in terms of specific examples, which are provided to aid understanding of the invention and are not intended to be limiting. For a person skilled in the art to which the invention pertains, several simple deductions, modifications or substitutions may be made according to the idea of the invention.
Claims (11)
1. An indoor positioning method is characterized by comprising the following steps,
acquiring data x to be matched of an object to be positioned, wherein the data x to be matched comprises signal intensity values between a third sensor and D first sensors, and D is more than or equal to 2;
the method comprises the following steps that K areas needing to be positioned are divided indoors, K is larger than or equal to 2, the D first sensors are arranged indoors, and the areas matched with data to be matched are determined according to the relation between the measured received signal strength values of the D first sensors and the K areas, so that the positioning of the object to be positioned is realized.
2. The indoor positioning method of claim 1, wherein the determining the region matched with the data to be matched according to the relationship between the measured received signal strength values of the D first sensors and the K regions, so as to realize the positioning of the object to be positioned, comprises:
based on subspace features
Using the distance relationship f (x) of the signal intensity value samples to the subspace between the D first sensors and the second sensor i ,U k ,b k ) One region for each subspace, so that the matching error f (x, U) k ,b k ) The minimum subspace k is the matched subspace, so as to determine the matched region of the data to be matched; wherein K is more than or equal to 1 and less than or equal to K>
Is a base of a subspace, <' > v>
Is the center of the subspace d k Is a dimension of a subspace;
the subspace characteristics
The obtaining method comprises the following steps:
constructing a sample set X, wherein the sample set X comprises N samples, and N is more than or equal to 2,x i The sample is a sample in a sample set X, i is more than or equal to 1 and less than or equal to N, the sample comprises signal intensity values between D first sensors and D second sensors which are acquired when the sample passes through the K regions, and the sample sequence in the sample set is samples which are sequentially acquired when the sample passes through the K regions sequentially and not repeatedly;
sequentially dividing the sample set into K clusters through K-1 dividing points, wherein each cluster sequentially corresponds to the K subspaces one by one;
constructing the distance relationship f (x) of each sample to any subspace i ,U k ,b k )；
According to the minimum principle from each sample to the subspace of each sample, the distance relation and the N samples, sequentially dividingConstructing a subspace division network according to the principle of dividing the K clusters, wherein the subspace division network is used for solving the K-1 division points and the subspace characteristics
3. The indoor positioning method according to claim 2, wherein a central position coordinate of a kth region corresponding to a subspace k is a positioning coordinate of the positioning object.
4. The indoor positioning method of claim 1, wherein the determining the region matched with the data to be matched according to the relationship between the measured received signal strength values of the D first sensors and the K regions, so as to realize the positioning of the object to be positioned, comprises:
the position of the object to be positioned is obtained by inputting the data to be matched into a subspace matching neural network, and the training method of the subspace matching neural network comprises the following steps:
constructing a sample set X, wherein the sample set X comprises N samples, and N is more than or equal to 2,x i The sample is a sample in a sample set X, i is more than or equal to 1 and less than or equal to N, the sample comprises signal intensity values between D first sensors and D second sensors which are acquired when the sample passes through the K regions, and the sample sequence in the sample set is samples which are sequentially acquired when the sample passes through the K regions sequentially and not repeatedly;
sequentially dividing the sample set into K clusters through K-1 dividing points, wherein each cluster sequentially corresponds to the K regions one by one, and the K regions correspond to K subspaces l (i) one by one, so that a training set is obtained
Using the training set
And training to obtain the subspace matching neural network. />
5. The indoor positioning method as claimed in claim 4, wherein the sequentially dividing the sample set into K clusters by K-1 division points comprises:
constructing the distance relation f (x) of each sample to any subspace i ,U k ,b k )；
Constructing a subspace segmentation network according to a rule that each sample is minimum to a subspace to which each sample belongs, the distance relation and a rule that N samples are sequentially segmented into K clusters, wherein the subspace segmentation network is used for solving the K-1 segmentation points and subspace characteristics
Solving the K-1 segmentation points according to the subspace segmentation network, so that the sample set is sequentially segmented into K clusters through the K-1 segmentation points.
6. Indoor positioning method according to claim 2 or 5, characterized in that the distance relation f (x) of each sample to any subspace i ,U k ,b k ) Can be expressed as:
according to the subspace minimum rule and the distance relationship from each sample to each sample, the subspace partitioning network can be expressed as:
wherein, t k Is a division point of the sample set X, t is more than or equal to 1 1 <t 2 <…<t K-1 N is less than or equal to N; solving equation (3) by using an iterative optimization algorithm, namely, assuming a known segmentation point t 1 ，t 2 ，…，t K-1 To findSubspace characterization
Then, the known subspace characteristics are assumed
Calculating a division point t 1 ，t 2 ，…，t K-1 The loop iterates until convergence.
7. The indoor positioning method according to claim 2 or 4, wherein the sample x i And also position coordinates z (i) ＝(z 1 (i) ,z 2 (i) ) In which
1≤q≤D，o q For the position coordinates of the qth first sensor,
j is more than or equal to 1 and less than or equal to D represents the jth first sensor, r is the signal intensity value, and is greater than or equal to>
For the signal intensity value of the jth first sensor in the ith sample, <' >>
The signal intensity value of the qth first sensor in the ith sample is obtained.
8. The indoor positioning method according to claim 1, wherein the D first sensors are the same type of sensor or the D first sensors are different types of sensors.
9. A subspace feature extraction method is characterized by comprising the following steps,
acquiring data x to be matched of an object to be extracted, wherein the data x to be matched comprises signal intensity values between a third sensor and D first sensors, and D is more than or equal to 2;
extracting subspace characteristics corresponding to the data x to be matched according to a subspace segmentation network
The subspace partitioning network is based on each sample x i Distance relationship f (x) to any subspace i ,U k ,b k ) The minimum principle from each sample to the subspace to which each sample belongs and the principle that N samples are sequentially divided into K clusters are constructed, wherein K is more than or equal to 2,1 and is less than or equal to K and/or greater than or equal to K>
Is the base of the subspace>
Is the center of the subspace d k Is a dimension of a subspace;
x i is one sample in N samples in the sample set X, N is more than or equal to 2,1 and is more than or equal to i and less than or equal to N, and each sample X i The method comprises the signal intensity values between D first sensors and D second sensors which are acquired when the samples pass through K areas, and the sample sequence in the sample set is samples which are sequentially acquired when the samples pass through the K areas in sequence and not repeatedly; the D first sensors are sensors arranged in the K areas and/or outside the K areas;
the K regions are in one-to-one correspondence to K subspaces, the sample set is sequentially divided into K clusters through K-1 division points, and each cluster is in one-to-one correspondence to the K subspaces.
10. The subspace feature extraction method of claim 9, wherein the subspace partitioning network is based on each sample x i Distance relationship f (x) to any subspace i ,U k ,b k ) From sample to sampleThe subspace minimum principle and the principle construction that N samples are sequentially divided into K clusters comprise:
each sample x i Distance relationship f (x) to any subspace i ,U k ,b k ) Can be expressed as:
according to the subspace minimum principle from each sample to each sample and the distance relationship, the subspace partitioning network may be expressed as:
wherein, t k Is a division point of the sample set X, t is more than or equal to 1 1 <t 2 <…<t K-1 N is less than or equal to N; solving equation (3) by using an iterative optimization algorithm, namely, assuming a known segmentation point t 1 ，t 2 ，…，t K-1 To find subspace features
Then, the known subspace characteristics are assumed
Calculating a division point t 1 ，t 2 ，…，t K-1 The loop iterates until convergence.
11. A computer-readable storage medium, characterized in that said medium has stored thereon a program executable by a processor for implementing the indoor positioning method of any one of claims 1-8, and/or the subspace feature extraction method of any one of claims 9-10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211480461.4A CN115988634B (en) | 2022-11-23 | 2022-11-23 | Indoor positioning method and subspace feature extraction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211480461.4A CN115988634B (en) | 2022-11-23 | 2022-11-23 | Indoor positioning method and subspace feature extraction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115988634A true CN115988634A (en) | 2023-04-18 |
| CN115988634B CN115988634B (en) | 2023-08-18 |
Family
ID=85956931
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211480461.4A Active CN115988634B (en) | 2022-11-23 | 2022-11-23 | Indoor positioning method and subspace feature extraction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115988634B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102279383A (en) * | 2011-04-22 | 2011-12-14 | 华南理工大学 | Indoor positioning method based on active RFID |
| CN104053129A (en) * | 2014-06-19 | 2014-09-17 | 北京芯同汇科技有限公司 | Wireless sensor network indoor positioning method and device based on sparse RF fingerprint interpolations |
| CN107360548A (en) * | 2017-08-21 | 2017-11-17 | 山东师范大学 | Indoor article localization method and system based on RFID |
| CN110703189A (en) * | 2019-09-30 | 2020-01-17 | 天津大学 | Passive positioning method based on wireless signal RSS subspace feature vector matching |
| US20200096598A1 (en) * | 2018-09-20 | 2020-03-26 | International Business Machines Corporation | Dynamic, cognitive hybrid method and system for indoor sensing and positioning |
| CN111263295A (en) * | 2020-01-15 | 2020-06-09 | 广东工业大学 | WLAN indoor positioning method and device |
| WO2021077218A1 (en) * | 2019-10-21 | 2021-04-29 | Innerspace Technology Inc. | Stable and accurate indoor localization based on predictive hyperbolic location fingerprinting |
-
2022
- 2022-11-23 CN CN202211480461.4A patent/CN115988634B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102279383A (en) * | 2011-04-22 | 2011-12-14 | 华南理工大学 | Indoor positioning method based on active RFID |
| CN104053129A (en) * | 2014-06-19 | 2014-09-17 | 北京芯同汇科技有限公司 | Wireless sensor network indoor positioning method and device based on sparse RF fingerprint interpolations |
| CN107360548A (en) * | 2017-08-21 | 2017-11-17 | 山东师范大学 | Indoor article localization method and system based on RFID |
| US20200096598A1 (en) * | 2018-09-20 | 2020-03-26 | International Business Machines Corporation | Dynamic, cognitive hybrid method and system for indoor sensing and positioning |
| US20210341563A1 (en) * | 2018-09-20 | 2021-11-04 | International Business Machines Corporation | Dynamic, cognitive hybrid method and system for indoor sensing and positioning |
| CN110703189A (en) * | 2019-09-30 | 2020-01-17 | 天津大学 | Passive positioning method based on wireless signal RSS subspace feature vector matching |
| WO2021077218A1 (en) * | 2019-10-21 | 2021-04-29 | Innerspace Technology Inc. | Stable and accurate indoor localization based on predictive hyperbolic location fingerprinting |
| CN111263295A (en) * | 2020-01-15 | 2020-06-09 | 广东工业大学 | WLAN indoor positioning method and device |
Non-Patent Citations (2)
| Title |
|---|
| ACHOUR ACHROUFENE;ET AL: "RSS-Based indoor localization using belief function theory", 《IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING》 * |
| 黄威；张玲华: "基于RSSI的无线传感器网络室内定位算法研究", 《计算机技术与发展》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115988634B (en) | 2023-08-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hsieh et al. | Deep learning-based indoor localization using received signal strength and channel state information | |
| CN104703143B (en) | A kind of indoor orientation method based on WIFI signal intensity | |
| CN109444813B (en) | RFID indoor positioning method based on BP and DNN double neural networks | |
| CN105101408B (en) | Indoor orientation method based on distributed AP selection strategy | |
| CN107703480B (en) | Mixed kernel function indoor positioning method based on machine learning | |
| CN106851571B (en) | Decision tree-based rapid KNN indoor WiFi positioning method | |
| EP3404439A1 (en) | Cluster-based magnetic positioning method, device and system | |
| CN112135248B (en) | WIFI fingerprint positioning method based on K-means optimal estimation | |
| Aranda et al. | Performance analysis of fingerprinting indoor positioning methods with BLE | |
| CN109348416B (en) | Fingerprint indoor positioning method based on binary k-means | |
| CN112748397A (en) | UWB positioning method based on self-adaptive BP neural network under non-line-of-sight condition | |
| Zhong et al. | WiFi indoor localization based on K-means | |
| CN108225332B (en) | Indoor positioning fingerprint map dimension reduction method based on supervision | |
| Yang et al. | DeepWiPos: A deep learning-based wireless positioning framework to address fingerprint instability | |
| Wang et al. | Study on improvement of fingerprint matching algorithm in wireless LAN based indoor positioning system | |
| Wei et al. | RSSI-based location fingerprint method for RFID indoor positioning: a review | |
| Sun et al. | A novel GCN based indoor localization system with multiple access points | |
| CN108919182B (en) | Target positioning method based on support set and expectation maximization in WIFI environment | |
| CN111757257B (en) | Dynamic fuzzy matching indoor positioning method for overcoming equipment difference | |
| Torres-Sospedra et al. | Exploiting different combinations of complementary sensor’s data for fingerprint-based indoor positioning in industrial environments | |
| CN115988634A (en) | Indoor positioning method and subspace feature extraction method | |
| CN108445443B (en) | KNN-based fingerprint point clustering method | |
| Yuan et al. | MFMCF: A novel indoor location method combining multiple fingerprints and multiple classifiers | |
| JP5450703B2 (en) | Method and apparatus for determining a spatial area in which a target is located | |
| CN113194401B (en) | Millimeter wave indoor positioning method and system based on generative countermeasure network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |