CN114783004A - Method, apparatus, device and medium for fingerprint positioning and processing fingerprint data - Google Patents

Method, apparatus, device and medium for fingerprint positioning and processing fingerprint data Download PDF

Info

Publication number
CN114783004A
CN114783004A CN202210443922.4A CN202210443922A CN114783004A CN 114783004 A CN114783004 A CN 114783004A CN 202210443922 A CN202210443922 A CN 202210443922A CN 114783004 A CN114783004 A CN 114783004A
Authority
CN
China
Prior art keywords
fingerprint
positioning
optimizer
location
fingerprint data
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.)
Pending
Application number
CN202210443922.4A
Other languages
Chinese (zh)
Inventor
邓中亮
钱峻
谢娜
胡恩文
张耀
罗凯
任海龙
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Duwei Technology Co ltd
Beijing University of Posts and Telecommunications
Original Assignee
Beijing Duwei Technology Co ltd
Beijing University of Posts and Telecommunications
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Duwei Technology Co ltd, Beijing University of Posts and Telecommunications filed Critical Beijing Duwei Technology Co ltd
Priority to CN202210443922.4A priority Critical patent/CN114783004A/en
Publication of CN114783004A publication Critical patent/CN114783004A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/08—Testing, supervising or monitoring using real traffic
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
    • H04W64/006—Locating users or terminals or network equipment for network management purposes, e.g. mobility management with additional information processing, e.g. for direction or speed determination

Abstract

The present disclosure relates to a method, apparatus, device and medium for fingerprint positioning and processing fingerprint data, the method for processing fingerprint data comprising: fingerprint data of a terminal is collected, and interference of the environment in the collection process on positioning of the terminal is carried; inputting the fingerprint data into a trained fingerprint optimizer to perform feature extraction to obtain positioning features; the positioning feature is used for positioning the terminal, and the fingerprint optimizer is used for removing the interference in the fingerprint data. The fingerprint positioning method comprises the following steps: acquiring fingerprint data of a user to be positioned in a specified area and a pre-configured fingerprint database containing real positioning and fingerprint characteristics of different position points in the specified area; extracting features through a trained fingerprint optimizer to obtain positioning features; calculating the matching degree between the fingerprint characteristics of different position points and the positioning characteristics of the fingerprint data; and determining the real location of the position point corresponding to the fingerprint feature with the highest matching degree as the location of the user to be located in the specified area.

Description

Method, apparatus, device and medium for fingerprint positioning and processing fingerprint data
Technical Field
The present disclosure relates to the field of artificial intelligence and communications technologies, and in particular, to a method, an apparatus, a device, and a medium for fingerprint positioning and processing fingerprint data.
Background
With the advent of 5G communication systems and the application of indoor positioning technologies in the technical fields of artificial intelligence, the internet of things, 5G, and the like, the 5G-based fingerprint positioning has a good application prospect, and wireless signals acquired by a user terminal accessing a 5G base station are generally used as collected crowdsourcing fingerprint data for positioning, however, in the crowdsourcing data collection process, the wireless fingerprint signals are greatly changed due to the influence of factors such as an indoor environment, and in order to reduce the influence, in the related art, a method of establishing a mapping relationship between fingerprint database data and crowdsourcing data is generally adopted, and combining PDR (Pedestrian dead reckoning) and crowd-sourced fingerprint positioning to carry out PDR according to the initial position of the user, and calibrating the position of the user by acquiring accurate motion sensors such as an accelerometer, a gyroscope, a geomagnetic meter and the like after obtaining the estimated PDR position.
In the process of implementing the invention, the inventor finds that at least the following technical problems exist in the related art:
in the existing method for positioning by fusing the PDR and the crowdsourced fingerprint, the requirement on the accuracy of a sensor used by mobile equipment participating in crowdsourced sourcing is high, an accurate motion sensor is needed to calibrate the position of a user, the error of the PDR is large, and the positioning accuracy is difficult to guarantee.
Disclosure of Invention
To solve the above technical problems, or to at least partially solve the above technical problems, embodiments of the present disclosure provide a method, apparatus, device, and medium for locating and processing fingerprint data.
In a first aspect, an embodiment of the present disclosure provides a method for processing fingerprint data, including: collecting fingerprint data of a terminal, wherein the fingerprint data carries interference generated by a source of a fingerprint in a collecting process to the positioning of the terminal; inputting the fingerprint data into a trained fingerprint optimizer to perform feature extraction to obtain the positioning features of the fingerprint data; wherein the positioning feature is configured to position the terminal, and the fingerprint optimizer is configured to remove the interference from the fingerprint data.
In a possible embodiment, the fingerprint optimizer is trained by: acquiring a training sample, inputting the training sample into the fingerprint optimizer for feature extraction to obtain features of the training sample, wherein the training sample comprises a preset fingerprint sample and a crowdsourcing fingerprint sample, and the features of the training sample comprise actual position information of the training sample and actual source information representing that the training sample belongs to the preset fingerprint sample or the crowdsourcing fingerprint sample; inputting the characteristics of the training samples into a preset state discriminator and a positioning decision maker, and respectively extracting the source characteristics and the positioning characteristics of the training samples; determining a loss function of the state discriminator and the positioning decision maker based on the source feature and the positioning feature; and iteratively optimizing parameters of the state discriminator, the positioning decision-maker and the fingerprint optimizer based on the loss functions of the state discriminator and the positioning decision-maker, performing gradient inversion on the optimized loss function of the state discriminator, and updating the parameters of the fingerprint optimizer to obtain the trained fingerprint optimizer.
In one possible embodiment, determining the loss function of the state discriminator and the location decider based on the source feature and the location feature includes: calculating a loss function of the state discriminator according to the actual source information and the source characteristics of the training samples; and calculating a loss function of the positioning decision device according to the actual position information and the positioning characteristics of the training samples.
In one possible embodiment, the method for optimizing iteratively parameters of the state discriminator, the location decider and the fingerprint optimizer based on the loss functions of the state discriminator and the location decider, and performing gradient inversion on the loss function of the state discriminator after iterative optimization and updating the parameters of the fingerprint optimizer to obtain a trained fingerprint optimizer includes: iteratively optimizing parameters of the state discriminator, the positioning decider and the fingerprint optimizer based on the loss functions of the state discriminator and the positioning decider; stopping the iterative optimization when the iterative optimization reaches the saddle point of the parameter, and performing gradient inversion on the loss function of the state discriminator; and updating the parameters of the fingerprint optimizer, and taking the fingerprint optimizer after the parameters are updated as a trained fingerprint optimizer.
In one possible embodiment, the iteratively optimizing the parameters of the state discriminator, the location decider and the fingerprint optimizer based on the loss functions of the state discriminator and the location decider includes: constructing a loss function of a network based on the state discriminator and the loss function of the positioning decider, wherein the network comprises the state discriminator, the positioning decider and the fingerprint optimizer; and performing iterative optimization on the parameters of the state discriminator, the positioning decision maker and the fingerprint optimizer by iteratively optimizing the loss function of the network.
In one possible embodiment, the loss function of the network is determined by:
Figure BDA0003615087190000021
wherein, thetafFor the above-mentioned parameters of the fingerprint optimizer, θdAs a parameter of the above-mentioned state discriminator, θyAs a parameter of the location decider, LyIs a loss function of the location decider, LdFor the loss function of the state discriminator, λ is a scaling coefficient, i represents a sequence number of a training sample, 1 st to nth training samples are preset fingerprint samples, N +1 th to N th training samples are crowdsourcing fingerprint samples, and N are positive integers.
In a possible embodiment, iteratively optimizing the parameters of the state discriminator, the location decider and the fingerprint optimizer by iteratively optimizing a loss function of the network includes: and (4) performing iterative optimization on the loss function of the network by a gradient descent method.
In a possible embodiment, the iteratively optimizing the loss function of the network by the gradient descent method includes iteratively optimizing by:
Figure BDA0003615087190000031
Figure BDA0003615087190000032
Figure BDA0003615087190000033
where μ is the learning rate, θfFor the above-mentioned parameters of the fingerprint optimizer, θdAs a parameter of the above-mentioned state discriminator, θyAs a parameter of the location decider, LyIs a loss function of the location decider, LdFor the loss function of the state discriminator, λ is a scaling coefficient, i represents a serial number of a training sample, the 1 st to nth training samples are preset fingerprint samples, the N +1 st to nth training samples are crowdsourcing fingerprint samples, and N are positive integers.
In one possible embodiment, the penalty function of the state arbiter and the penalty function of the position decider are cross entropy penalty functions.
In one possible embodiment, the loss function of the state discriminator and the loss function of the location decider are cross entropy loss functions expressed by the following formula:
Figure BDA0003615087190000041
where y is the true probability distribution and y' is the predicted probability distribution.
In a second aspect, an embodiment of the present disclosure provides a method for fingerprint positioning, including: acquiring fingerprint data of a user to be positioned in a designated area and a pre-configured fingerprint database, wherein the fingerprint database comprises real positioning and fingerprint characteristics of different position points in the designated area; inputting the fingerprint data into a trained fingerprint optimizer for feature extraction to obtain the positioning features of the fingerprint data; calculating the matching degree between the fingerprint characteristics of the different position points and the positioning characteristics of the fingerprint data; and determining the real positioning of the position point corresponding to the fingerprint feature with the highest matching degree as the positioning of the user to be positioned in the specified area.
In one possible embodiment, the fingerprint optimizer trains by: acquiring a training sample, inputting the training sample into the fingerprint optimizer for feature extraction to obtain features of the training sample, wherein the training sample comprises a preset fingerprint sample and a crowdsourcing fingerprint sample, and the features of the training sample comprise actual position information of the training sample and actual source information representing that the training sample belongs to the preset fingerprint sample or the crowdsourcing fingerprint sample; inputting the characteristics of the training sample into a preset state discriminator and a positioning decision device, and respectively extracting the source characteristics and the positioning characteristics of the training sample; determining a loss function of the state discriminator and the location decider based on the source feature and the location feature; and iteratively optimizing parameters of the state discriminator, the positioning decision device and the fingerprint optimizer based on the loss functions of the state discriminator and the positioning decision device, performing gradient inversion on the optimized loss function of the state discriminator, and updating the parameters of the fingerprint optimizer to obtain the trained fingerprint optimizer.
In one possible embodiment, the calculating the degree of matching between the fingerprint features of the different location points and the localization features of the fingerprint data includes calculating the confidence of the matching relationship between the different location points in the designated area and the fingerprint data by the following formula:
Z＝softmax(ReLU(XW0)W1)
Figure BDA0003615087190000042
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0003615087190000043
between 0 and 1, Z represents the confidence of the matching relationship between different position points in the designated area and the fingerprint data, X represents the positioning characteristic of the fingerprint data, and W represents the positioning characteristic of the fingerprint data0And W1Representing a learnable weight matrix, where xiIs the ith characteristic value.
In a possible implementation manner, the determining the true location of the location point corresponding to the fingerprint feature with the highest matching degree as the location of the user to be located in the designated area includes: and updating and optimizing the confidence coefficient of the matching relationship between different position points in the designated area and the fingerprint data by adopting a gradient descent method, wherein the loss function is a cross entropy loss function represented by the following formula:
Figure BDA0003615087190000051
wherein N represents the number of location points in the fingerprint database, YifFor the sign function, take 1 when the real position of the sample i is f, otherwise take 0, ZifProbability that the fingerprint data i belongs to the position point f; and determining the real positioning of the position point with the highest probability as the positioning of the user to be positioned in the specified area.
In a third aspect, an embodiment of the present disclosure provides an apparatus for processing fingerprint data, including: the terminal comprises a fingerprint acquisition module, a fingerprint acquisition module and a fingerprint processing module, wherein the fingerprint acquisition module is used for acquiring fingerprint data of the terminal, and the fingerprint data carries interference generated by a fingerprint source in an acquisition process on positioning of the terminal; the characteristic extraction module is used for inputting the fingerprint data into a trained fingerprint optimizer for characteristic extraction to obtain the positioning characteristics of the fingerprint data; wherein the positioning feature is configured to position the terminal, and the fingerprint optimizer is configured to remove the interference from the fingerprint data.
In a possible implementation manner, the apparatus for processing fingerprint data further includes: the training module is used for training the fingerprint optimizer in advance, and the training module comprises: the first sub-module is used for acquiring a training sample, inputting the training sample into the fingerprint optimizer and performing feature extraction to obtain features of the training sample, wherein the training sample comprises a preset fingerprint sample and a crowdsourcing fingerprint sample, and the features of the training sample comprise actual position information of the training sample and actual source information representing that the training sample belongs to the preset fingerprint sample or the crowdsourcing fingerprint sample; the second sub-module is used for inputting the characteristics of the training sample into a preset state discriminator and a positioning decision-making device and respectively extracting the source characteristics and the positioning characteristics of the training sample; a third sub-module for determining a loss function of the state discriminator and the location decider based on the source feature and the location feature; and the fourth sub-module is used for iteratively optimizing the parameters of the state discriminator, the positioning decider and the fingerprint optimizer based on the loss functions of the state discriminator and the positioning decider, performing gradient inversion on the optimized loss function of the state discriminator and updating the parameters of the fingerprint optimizer to obtain the trained fingerprint optimizer.
In a fourth aspect, an embodiment of the present disclosure provides an apparatus for locating a fingerprint, including: the system comprises an acquisition module, a positioning module and a positioning module, wherein the acquisition module is used for acquiring fingerprint data of a user to be positioned in a specified area and a pre-configured fingerprint database, and the fingerprint database comprises real positioning and fingerprint characteristics of different position points in the specified area; the optimization module is used for inputting the fingerprint data into a trained fingerprint optimizer to perform feature extraction so as to obtain the positioning features of the fingerprint data; the matching module is used for calculating the matching degree between the fingerprint characteristics of the different position points and the positioning characteristics of the fingerprint data; and the determining module is used for determining the real positioning of the position point corresponding to the fingerprint feature with the highest matching degree as the positioning of the user to be positioned in the specified area.
In a fifth aspect, an embodiment of the present disclosure provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, where the processor, the communication interface, and the memory complete communication with each other through the communication bus; a memory for storing a computer program; the processor is used for realizing the method for processing the fingerprint data or the method for positioning the fingerprint when executing the program stored in the memory.
In a sixth aspect, an embodiment of the present disclosure provides a computer-readable storage medium, on which a computer program is stored, wherein the computer program is implemented, when being executed by a processor, to implement the method for processing fingerprint data or the method for fingerprint location as described above.
Compared with the prior art, the technical scheme provided by the embodiment of the disclosure at least has part or all of the following advantages:
the method for processing the fingerprint data of the embodiment of the disclosure obtains the positioning characteristic by inputting the fingerprint data of the acquired terminal into a trained fingerprint optimizer for characteristic extraction, wherein the fingerprint optimizer is used for removing the interference of the source of the fingerprint carried in the fingerprint data in the acquisition process on the positioning of the terminal, and the positioning characteristic is used for positioning the terminal, thereby solving the problem that the positioning error caused by obvious disturbance and interference of the fingerprint data due to the influence of the source factors such as the environment and the like in the fingerprint data acquisition process cannot be reduced in the scene of dynamic change of the environment, the method comprises the steps of combining a pre-configured fingerprint database containing real positioning and fingerprint characteristics of different position points in an appointed area, calculating the matching degree between the fingerprint characteristics of the different position points in the fingerprint database and the positioning characteristics of user fingerprint data, determining the real positioning of the position point corresponding to the fingerprint characteristic with the highest matching degree as the positioning of a user to be positioned in the appointed area, further obtaining an accurate positioning result in a dynamically-changed complex environment, greatly reducing positioning errors and improving the positioning accuracy.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the disclosure.
In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the related art will be briefly described below, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
Fig. 1 schematically shows a flow chart of a method of processing fingerprint data according to an embodiment of the present disclosure;
FIG. 2 schematically illustrates a flow diagram for training a fingerprint optimizer in a method of processing fingerprint data according to an embodiment of the disclosure;
fig. 3 schematically shows a detailed flowchart of operation S204 according to an embodiment of the present disclosure;
FIG. 4 schematically illustrates a flowchart for training a fingerprint optimizer in a method of processing fingerprint data according to an embodiment of the present disclosure in an exemplary scenario;
FIG. 5 schematically illustrates a flow chart of a method of fingerprint location according to an embodiment of the present disclosure;
fig. 6 schematically shows a block diagram of an apparatus for processing fingerprint data according to an embodiment of the present disclosure;
FIG. 7 schematically illustrates a block diagram of an apparatus for fingerprint location according to an embodiment of the present disclosure; and
fig. 8 schematically shows a block diagram of an electronic device according to an embodiment of the present disclosure.
Detailed Description
With the emergence of 5G communication systems and the application of indoor positioning technologies in the technical fields of artificial intelligence, the internet of things, 5G and the like, the 5G-based fingerprint positioning has a good application prospect, wireless signals acquired by a user terminal accessed to a 5G base station are generally used as collected crowdsourcing fingerprint data for positioning, however, the wireless fingerprint signals are greatly changed due to the influence of factors such as indoor environment and the like in the crowdsourcing data collection process, and in order to reduce the influence, in the related art, a method for establishing a mapping relationship between fingerprint database data and crowdsourcing data is generally adopted, and combining PDR (Pedestrian Dead Reckoning) and crowd-sourced fingerprint positioning to carry out PDR according to the initial position of the user, and calibrating the position of the user by acquiring accurate motion sensors such as an accelerometer, a gyroscope, a geomagnetic meter and the like after obtaining the estimated PDR position.
In the process of implementing the invention, the inventor finds that at least the following technical problems exist in the related art: the existing method for fusing PDR and crowdsourcing fingerprint positioning has high requirement on the accuracy of a sensor used by mobile equipment participating in crowdsourcing, an accurate motion sensor is needed to calibrate the position of a user, errors of the PDR are large, and the positioning accuracy is difficult to guarantee.
To make the objects, technical solutions and advantages of the embodiments of the present disclosure more clear, the technical solutions of the embodiments of the present disclosure will be described clearly and completely with reference to the drawings in the embodiments of the present disclosure, and it is obvious that the described embodiments are some embodiments of the present disclosure, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments disclosed herein without making any creative effort, shall fall within the protection scope of the present disclosure.
In view of this, embodiments of the present disclosure provide a method, an apparatus, a device, and a medium for fingerprint positioning and processing, which reduce positioning errors caused by obvious disturbance and interference of fingerprint data due to influences of source factors such as environment during fingerprint data acquisition, so as to obtain accurate positioning results in a dynamically changing complex environment, reduce positioning errors, and improve positioning accuracy.
Fig. 1 schematically shows a flow chart of a method of processing fingerprint data according to an embodiment of the present disclosure.
Referring to fig. 1, a method for processing fingerprint data according to an embodiment of the present disclosure includes the following operations:
operation S101: collecting fingerprint data of a terminal, wherein the fingerprint data carries interference generated by a source of a fingerprint in a collecting process to the positioning of the terminal;
operation S102: inputting the fingerprint data into a trained fingerprint optimizer to perform feature extraction to obtain the positioning features of the fingerprint data; wherein the positioning feature is configured to position the terminal, and the fingerprint optimizer is configured to remove the interference from the fingerprint data.
In an actual application scenario, a terminal (user Equipment, UE) includes but is not limited to a smart phone, a tablet computer, a notebook computer, a smart watch, a smart bracelet, and the like, fingerprint data may be data in various forms such as signal strength and channel state information, taking a 5G application scenario as an example, the fingerprint data may be a wireless signal acquired by a terminal mobile device accessing a 5G base station, in an indoor scenario covered by the 5G base station in high density, a mobile device of a user connects to a network, and can continuously receive and transmit the wireless signal with the base station, and the wireless signal may be used as fingerprint data for uploading, fingerprint library construction, updating, user positioning, and the like.
For example, the fingerprint data acquired in operation S101 carries interference generated by a source of a fingerprint in an acquisition process to position the terminal, the fingerprint data may be acquired in real time in a complex indoor environment, and factors such as variability of the environment, instability of a wireless channel, and heterogeneity of acquisition equipment all seriously affect adaptability and robustness of a crowdsourced fingerprint, for example, opening or closing of a door and a window, movement of a terminal providing the fingerprint data and the like all cause great interference to propagation of a wireless signal, and noise is introduced into the wireless signal in the propagation process due to reflection, diffraction and the like, and these influences all significantly affect the fingerprint data and interfere with positioning the terminal.
In operation S102, the fingerprint data with interference is input into a trained fingerprint optimizer for feature extraction, so as to extract a location feature of the fingerprint data, the fingerprint optimizer is configured to remove the interference from the fingerprint data, and exclude other interference irrelevant to location, so as to ensure that the fingerprint feature extracted by the fingerprint optimizer only retains feature information effective for location, and remove disturbance features caused by environment, human interference, and other sources in the fingerprint acquisition process, so as to achieve an effect of optimizing the fingerprint data, so that the feature optimized by the fingerprint optimizer can accurately locate the terminal, and greatly improve the location accuracy of the fingerprint data.
Based on the above operations, the method for processing fingerprint data according to the embodiment of the disclosure obtains the positioning feature by inputting the acquired fingerprint data of the terminal into a trained fingerprint optimizer for feature extraction, where the fingerprint optimizer is configured to remove interference, which is carried in the fingerprint data, generated by a source of the fingerprint in an acquisition process to position the terminal, and the positioning feature is configured to position the terminal, so that a positioning error caused by obvious disturbance and interference of the fingerprint data due to influence of source factors such as an environment and the like in a scene in which an environment dynamically changes is not achieved.
Fig. 2 schematically shows a flow chart for training a fingerprint optimizer in a method of processing fingerprint data according to an embodiment of the present disclosure.
Referring to fig. 2, a fingerprint optimizer provided in the embodiment of the present disclosure may be trained by the following steps:
operation S201: acquiring a training sample, inputting the training sample into the fingerprint optimizer for feature extraction to obtain features of the training sample, wherein the training sample comprises a preset fingerprint sample and a crowdsourcing fingerprint sample, and the features of the training sample comprise actual position information of the training sample and actual source information representing that the training sample belongs to the preset fingerprint sample or the crowdsourcing fingerprint sample;
in operation S202: inputting the characteristics of the training samples into a preset state discriminator and a positioning decision maker, and respectively extracting the source characteristics and the positioning characteristics of the training samples;
operation S203: determining a loss function of the state discriminator and the positioning decision maker based on the source feature and the positioning feature;
in operation S204: and iteratively optimizing parameters of the state discriminator, the positioning decision device and the fingerprint optimizer based on the loss functions of the state discriminator and the positioning decision device, performing gradient inversion on the optimized loss function of the state discriminator, and updating the parameters of the fingerprint optimizer to obtain the trained fingerprint optimizer.
In an actual application scenario, before inputting the fingerprint data of the user into the trained fingerprint optimizer in S101, the fingerprint optimizer may be trained through operations S201 to S204.
Specifically, in operation S201, the training samples may be a preset fingerprint sample and a crowd-sourced fingerprint sample, where the preset fingerprint sample may be manually collected fingerprint data, actual location information of the preset fingerprint sample is known, and for example, indoor fingerprint data may be collected by dividing a certain divided area into a limited number of location points and by a manual collection method. The crowd-sourced fingerprint samples may be fingerprint samples collected by the terminal in real-time. The feature extraction may specifically adopt implementation manners such as feature space mapping, for example, low-dimensional vector space mapping.
The fingerprint optimizer is used for extracting features, so that the actual source information of the training samples can be obtained, namely, the sources of the training samples can be determined, the training samples belong to preset fingerprint samples or crowd-sourced fingerprint samples, and the actual position information of the training samples can also be determined.
In a possible implementation manner, a training network may be constructed to train the fingerprint optimizer and perform the training operation, the training network may at least include a preset fingerprint optimizer, a state discriminator and a positioning decision device, and the number of each part is not limited, and may be set according to a specific practical application scenario. Optionally, the state discriminator may also be a domain discriminator, and is configured to perform domain feature extraction, so as to classify the input data according to different domains to which the data stream belongs.
In operation S202, the features of the training samples are input into a state discriminator and a positioning decision device, the source features of the training samples are extracted by the state discriminator, and the positioning features of the training samples are extracted by the positioning decision device. For example, the feature data stream extracted by the fingerprint optimizer is input into the state discriminator, and the source features of the training sample are obtained through state feature extraction, and the training target can be continuous training, so that the state discriminator can achieve the effect of accurately identifying different sources of the training sample, that is, whether the sample is a preset fingerprint sample or a crowd-sourced fingerprint sample can be identified. For another example, the feature data stream extracted by the fingerprint optimizer is input into a positioning decision-making device, the positioning features of the training samples are extracted through feature fitting position information, and the training targets can be continuously trained, so that the positioning decision-making device can accurately identify the position information of the training samples, that is, output accurate positions.
In one possible implementation, operation S203 may include: calculating a loss function of the state discriminator according to the actual source information and the source characteristics of the training samples; and calculating a loss function of the positioning decision device according to the actual position information and the positioning characteristics of the training samples.
In one possible implementation, the loss function of the state arbiter and the loss function of the position decider may be cross-entropy loss functions (cross-entropy loss functions).
For example, the penalty function of the state discriminator and the penalty function of the location decider may be cross entropy penalty functions expressed by:
Figure BDA0003615087190000101
where y is the true probability distribution and y' is the predicted probability distribution.
Fig. 3 schematically shows a detailed flowchart of operation S204 according to an embodiment of the present disclosure;
referring to fig. 3, in one possible implementation, operation S204 may include the following operations:
operation S401: iteratively optimizing parameters of the state discriminator, the positioning decider and the fingerprint optimizer based on the loss functions of the state discriminator and the positioning decider;
operation S402: stopping the iterative optimization when the iterative optimization reaches the saddle point of the parameter, and performing gradient inversion on the loss function of the state discriminator;
operation S403: and updating the parameters of the fingerprint optimizer, and taking the fingerprint optimizer after the parameters are updated as a trained fingerprint optimizer.
In a possible implementation, the operation S401 may further include: minimizing a loss function of the state discriminator with respect to features of the preset fingerprint samples and the crowd-sourced fingerprint samples input to the state discriminator based on the loss function of the state discriminator; based on the loss function of the location decider, minimizing the loss function of the location decider for the characteristics of the preset fingerprint sample input to the location decider.
Specifically, in the training process, iterative optimization may be performed by minimizing the loss function, and the parameter is continuously updated until the stopping condition is met, for example, a predetermined number of iterations may be set, so that after the preset number of iterations is completed, the updating of the parameter is stopped, or for example, a threshold required to be reached by the loss function may be set, and when the loss function converges to meet the threshold requirement, the iterative computation is stopped.
In a possible implementation, the operation S401 may further include: constructing a loss function of a network based on the state discriminator and the loss function of the positioning decider, wherein the network comprises the state discriminator, the positioning decider and the fingerprint optimizer; and iteratively optimizing the parameters of the state discriminator, the positioning decision maker and the fingerprint optimizer by iteratively optimizing the loss function of the network.
Fig. 4 schematically illustrates a flowchart for training a fingerprint optimizer in a method of processing fingerprint data according to an embodiment of the present disclosure in an exemplary scenario.
The above-described operation will be explained below in conjunction with the exemplary scenario illustrated in fig. 4.
Illustratively, a training network may be constructed to train the fingerprint optimizer and perform the training operation, the training network may at least include a preset fingerprint optimizer, a state discriminator and a positioning decision device, and the number of each part is not limited, and may be set according to a specific practical application scenario.
As shown in fig. 4, the training network may include a fingerprint optimizer, a state discriminator and a positioning decision device, where the state discriminator may be a domain discriminator for domain feature extraction, so as to classify input data according to different domains to which data streams belong. The training network may architecturally include: fingerprint optimizers, e.g. denoted GfThe parameter set may be θf(ii) a The location decider, for example, may be denoted as GyWith a parameter set of thetayThe positioning decision device can be used for position output; the state discriminator can be represented, for example, by GdWith parameter set thetadThe state discriminator may be used to classify the fingerprint source.
The training samples may include crowd-sourced fingerprints and manually-gathered fingerprints, which are first introduced into the fingerprint optimizer, and through feature extraction, features of the extracted training samples, i.e., a feature data stream, may be output, where the features of the training samples include actual location information of the training samples and actual source information that characterizes the training samples as belonging to the manually-gathered fingerprints or the crowd-sourced fingerprints. For example, the training sample may be a data set comprising manually acquired fingerprint data
Figure BDA0003615087190000111
And crowd-sourced fingerprint data
Figure BDA0003615087190000112
And a plurality of samples are waited.
Then, inputting the features into a state discriminator and a positioning decision maker, and extracting source features of the training samples through the state discriminator, wherein the source features can comprise fingerprint source labels, for example; the localization features of the training samples are extracted by an input localization decider, which may include, for example, fingerprint location tags.
For example, with LyLoss function representing the location decider, in LdRepresenting state-discriminator penalty functions, discriminating based on stateThe loss functions of the positioning decision device and the positioning decision device iteratively optimize parameters of the training network, in other words, the state arbiter and the positioning decision device can jointly form a standard feedforward system structure, and the parameters of the network can be updated according to the loss functions of the positioning decision device and the loss functions of the state arbiter.
Optionally, the penalty function of the state discriminator and the penalty function of the position decision device may be cross entropy penalty functions, for example, cross entropy penalty functions expressed by the following formula:
Figure BDA0003615087190000121
where y is the true probability distribution and y' is the predicted probability distribution.
Optionally, the loss function of the network may be determined by:
Figure BDA0003615087190000122
wherein, thetafFor the above-mentioned parameters of the fingerprint optimizer, θdAs a parameter of the above-mentioned state discriminator, θyAs a parameter of the location decider, LyIs a loss function of the location decider, LdFor the loss function of the state discriminator, λ is a scaling coefficient, i represents a serial number of a training sample, the 1 st to nth training samples are preset fingerprint samples, the N +1 st to nth training samples are crowdsourcing fingerprint samples, and N are positive integers.
In the process of iterative optimization, that is, in the process of loop training, for the loss function of the network, the parameters of the positioning decision-maker and the fingerprint optimizer can be updated by minimizing the function, and the parameters of the state arbiter can be updated by maximizing the loss function, that is, the following parameter saddle points are used as the target for solving:
Figure BDA0003615087190000123
Figure BDA0003615087190000124
in the process of iteratively optimizing the parameters of the state discriminator, the location decider and the fingerprint optimizer by iteratively optimizing the loss function of the network, the loss function of the network may be iteratively optimized by, for example, a gradient descent method, as follows:
Figure BDA0003615087190000131
Figure BDA0003615087190000132
Figure BDA0003615087190000133
wherein μ is the learning rate, and the above is the specific process of selecting the gradient descent method to update the parameters, so that the model reaches the parameter saddle point. It can also be understood that the training goal may be, as training continues, so that the state discriminator can achieve the effect of accurately identifying different sources of the training samples, i.e., whether the samples are preset fingerprint samples or crowd-sourced fingerprint samples can be identified, so that the positioning decision maker can accurately identify the position information of the training samples, i.e., output an accurate position.
Further, when the above objectives are achieved, that is, it is ensured that the features extracted by the location decider can satisfy the location task, and the state arbiter can also distinguish the fingerprint data sources, the gradient inversion can be performed on the loss function of the optimized state arbiter and the parameters of the fingerprint optimizer can be updated, so as to obtain the trained fingerprint optimizer. Referring to fig. 4, when the iterative optimization reaches the saddle point of the parameter, the iterative optimization is stopped, the loss function of the state discriminator is gradient-inverted, and the output of the state discriminator may be gradient-inverted, for example, by providing a gradient inversion layer. The lambda can be a scaling coefficient after gradient inversion, can be a hyper-parameter, and gradient inversion (namely gradient inversion) is realized by multiplying the gradient by the lambda, and the lambda can be a constant, so that unsupervised fingerprint feature optimization is realized.
Therefore, the acquired crowdsourcing data fingerprint is similar to the manually acquired feature distribution, and the stable feature of the crowdsourcing fingerprint data under the conditions of environmental change and manual disturbance can be obtained. It can also be understood that above-mentioned operation can make the unable accurate fingerprint data that distinguish of state arbiter derive from manual collection or crowdsourcing data to guarantee that the fingerprint characteristic that fingerprint optimizer drawed has only kept and has fixed a position effectual characteristic information, got rid of the disturbance characteristic that crowdsourcing fingerprint collection in-process environment or artificial interference brought, in order to reach the effect of optimizing crowdsourcing fingerprint.
After the gradient inversion operation, the parameters of the fingerprint optimizer may be updated, and the fingerprint optimizer with the updated parameters is used as a trained fingerprint optimizer, so as to complete a training process, so far, the trained fingerprint optimizer may be configured to input the fingerprint data from the terminal into the trained fingerprint optimizer for feature extraction in operation S102, so as to obtain the location features of the fingerprint data; wherein the positioning feature is configured to position the terminal, and the fingerprint optimizer is configured to remove the interference from the fingerprint data.
Fig. 5 schematically shows a flow chart of a method of fingerprint localization according to an embodiment of the present disclosure.
As shown in fig. 5, a method for locating a fingerprint provided by an embodiment of the present disclosure may include the following operations:
s501: acquiring fingerprint data of a user to be positioned in a specified area and a pre-configured fingerprint database, wherein the fingerprint database comprises real positioning and fingerprint characteristics of different position points in the specified area;
s502: inputting the fingerprint data into a trained fingerprint optimizer to perform feature extraction so as to obtain the positioning features of the fingerprint data;
s503: calculating the matching degree between the fingerprint characteristics of the different position points and the positioning characteristics of the fingerprint data;
s504: and determining the real location of the position point corresponding to the fingerprint feature with the highest matching degree as the location of the user to be located in the specified area.
Specifically, the preconfigured fingerprint database may be constructed by dividing the designated area into a limited number of location points and collecting the indoor fingerprint data by a manual collection method. Because the trained fingerprint optimizer can remove the interference of the environment in the acquisition process of the fingerprint data on the positioning fingerprint data terminal, so that the stable characteristics of the crowdsourcing fingerprint data under the conditions of environment change and artificial disturbance are obtained, the fingerprint data is input into the trained fingerprint optimizer for characteristic extraction, and the positioning characteristics of the fingerprint data can be obtained, namely, the fingerprint characteristics of crowdsourcing data fingerprints, which are subjected to optimization processing and eliminate the interference of different fingerprint sources, can be understood as the fingerprint characteristics with the domain invariant characteristic. And then, matching and positioning the optimized crowdsourcing fingerprint information with a fingerprint database to obtain an accurate positioning result in a complex environment.
Based on the above operations, the fingerprint positioning method according to the embodiment of the present disclosure applies the above method for processing fingerprint data, performs feature extraction on fingerprint data of a user to be positioned to obtain a positioning feature, that is, a positioning feature from which interference generated by a fingerprint source on a positioning terminal in an acquisition process is removed, calculates a matching degree between fingerprint features of different location points in a fingerprint database and the positioning feature of the fingerprint data of the user by combining a pre-configured fingerprint database including real positioning and fingerprint features of different location points in an assigned area, and determines real positioning of a location point corresponding to a fingerprint feature with the highest matching degree as positioning of the user to be positioned in the assigned area, thereby obtaining an accurate positioning result in a dynamically changing complex environment, greatly reducing positioning errors, and improving positioning accuracy.
In a possible embodiment, the fingerprint optimizer is further trained by: acquiring a training sample, inputting the training sample into the fingerprint optimizer for feature extraction to obtain features of the training sample, wherein the training sample comprises a preset fingerprint sample and a crowdsourcing fingerprint sample, and the features of the training sample comprise actual position information of the training sample and actual source information representing that the training sample belongs to the preset fingerprint sample or the crowdsourcing fingerprint sample; inputting the characteristics of the training samples into a preset state discriminator and a positioning decision maker, and respectively extracting the source characteristics and the positioning characteristics of the training samples; determining a loss function of the state discriminator and the location decider based on the source feature and the location feature; and iteratively optimizing parameters of the state discriminator, the positioning decision-maker and the fingerprint optimizer based on the loss functions of the state discriminator and the positioning decision-maker, performing gradient inversion on the optimized loss function of the state discriminator, and updating the parameters of the fingerprint optimizer to obtain the trained fingerprint optimizer.
The implementation process of each operation in the training process is specifically described in the implementation process of training the fingerprint optimizer in the method for processing fingerprint data, and is not described again here.
In one possible embodiment, the calculating the degree of matching between the fingerprint features of the different location points and the positioning features of the fingerprint data may further include calculating a confidence of a matching relationship between the different location points in the designated area and the fingerprint data by the following formula:
Z＝softmax(ReLU(XW0)W1)
Figure BDA0003615087190000151
wherein the content of the first and second substances,
Figure BDA0003615087190000152
between 0 and 1, Z represents the distance between different points in the designated area and the fingerprint dataX represents a location feature of the fingerprint data, W0And W1Represents a learnable weight matrix, where xiIs the ith characteristic value.
Specifically, after the fingerprint optimizer is trained, stable features of the fingerprint data can be crowdsourced under the conditions of environmental change, artificial disturbance and the like, so that the position point matching problem is converted into a fingerprint feature matching problem, which can be described by the above formula, and the fingerprint point with the highest confidence score is the current position estimation point.
In a possible implementation, the operation S504 may include: and updating and optimizing the confidence coefficient of the matching relationship between different position points in the designated area and the fingerprint data by adopting a gradient descent method, wherein the loss function is a cross entropy loss function represented by the following formula:
Figure BDA0003615087190000153
wherein N represents the number of location points in the fingerprint database, YifTaking 1 when the real position of the sample i is f, and taking 0 and Z if the real position of the sample i is not fifProbability that the fingerprint data i belongs to the position point f; and determining the real positioning of the position point with the highest probability as the positioning of the user to be positioned in the specified area.
Based on the above operations, the fingerprint positioning method according to the embodiment of the present disclosure applies the above method for processing fingerprint data, performs feature extraction on fingerprint data of a user to be positioned to obtain a positioning feature, that is, a positioning feature from which interference generated by a fingerprint source on a positioning terminal in an acquisition process is removed, calculates a matching degree between fingerprint features of different location points in a fingerprint database and the positioning feature of the fingerprint data of the user by combining a pre-configured fingerprint database including real positioning and fingerprint features of different location points in an assigned area, and determines real positioning of a location point corresponding to a fingerprint feature with the highest matching degree as positioning of the user to be positioned in the assigned area, thereby obtaining an accurate positioning result in a dynamically changing complex environment, greatly reducing positioning errors, and improving positioning accuracy. The problem that when the wireless signals have strong disturbance due to the fact that the crowdsourcing data are influenced by environment and human factors in the acquisition process of the crowdsourcing data in the complex environment is further solved, the influence of obvious change on matching positioning precision of fingerprint information is large, the influence of environment change and human disturbance in the crowdsourcing fingerprint data acquisition process is reduced, the robustness of crowdsourcing fingerprint positioning is greatly improved, and therefore the positioning accuracy of indoor positioning through crowdsourcing data in the complex environment is improved by the aid of the method for extracting the stable features in the multi-domain space.
Fig. 6 schematically shows a block diagram of an apparatus for processing fingerprint data according to an embodiment of the present disclosure.
As shown in fig. 6, an apparatus 600 for processing fingerprint data provided by an embodiment of the present disclosure may include:
a fingerprint acquisition module 601, configured to acquire fingerprint data of a terminal, where the fingerprint data carries interference generated by a source of a fingerprint in an acquisition process to positioning the terminal;
a feature extraction module 602, configured to input the fingerprint data into a trained fingerprint optimizer for feature extraction, so as to obtain a location feature of the fingerprint data; wherein the location feature is configured to locate the terminal, and the fingerprint optimizer is configured to remove the interference from the fingerprint data.
The implementation process of the functions and actions of each module in the above device is detailed in the implementation process of the corresponding step in the above method for processing fingerprint data, and is not described again here.
In a possible implementation, the apparatus for processing fingerprint data may further include: the training module is used for training the fingerprint optimizer in advance, and the training module comprises: the first sub-module is used for acquiring a training sample, inputting the training sample into the fingerprint optimizer and performing feature extraction to obtain features of the training sample, wherein the training sample comprises a preset fingerprint sample and a crowdsourcing fingerprint sample, and the features of the training sample comprise actual position information of the training sample and actual source information representing that the training sample belongs to the preset fingerprint sample or the crowdsourcing fingerprint sample; the second sub-module is used for inputting the characteristics of the training sample into a preset state discriminator and a positioning decision-making device and respectively extracting the source characteristics and the positioning characteristics of the training sample; a third sub-module, configured to determine a loss function of the state discriminator and the location decider based on the source feature and the location feature; and the fourth sub-module is used for iteratively optimizing the parameters of the state discriminator, the positioning decision device and the fingerprint optimizer based on the loss functions of the state discriminator and the positioning decision device, performing gradient inversion on the optimized loss function of the state discriminator and updating the parameters of the fingerprint optimizer to obtain the trained fingerprint optimizer.
Fig. 7 schematically shows a block diagram of the structure of the fingerprint positioning device according to the embodiment of the present disclosure.
As shown in fig. 7, an apparatus 700 for fingerprint location provided by an embodiment of the present disclosure may include:
an obtaining module 701, configured to obtain fingerprint data of a user to be located in a specified area and a preconfigured fingerprint database, where the fingerprint database includes real locations and fingerprint features of different location points in the specified area;
an optimization module 702, configured to input the fingerprint data into a trained fingerprint optimizer for feature extraction, so as to obtain a location feature of the fingerprint data;
a matching module 703, configured to calculate a matching degree between the fingerprint features of the different location points and the positioning features of the fingerprint data;
a determining module 704, configured to determine a true location of the location point corresponding to the fingerprint feature with the highest matching degree as a location of the user to be located in the designated area.
The implementation process of the functions and actions of each module in the above device is specifically detailed in the implementation process of the corresponding step in the above fingerprint positioning method, and is not described again here.
For the device embodiment, since it basically corresponds to the method embodiment, reference may be made to the partial description of the method embodiment for relevant points. The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of the invention. One of ordinary skill in the art can understand and implement it without inventive effort.
In the above embodiment, any multiple of the fingerprint acquisition module 601, the feature extraction module 602, the acquisition module 701, the optimization module 702, the matching module 703 and the determination module 704 may be combined and implemented in one module, or any one of the modules may be split into multiple modules. Alternatively, at least part of the functionality of one or more of these modules may be combined with at least part of the functionality of other modules and implemented in one module. At least one of the fingerprint acquisition module 601, the feature extraction module 602, the acquisition module 701, the optimization module 702, the matching module 703 and the determination module 704 may be implemented at least partially as a hardware circuit, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system on a chip, a system on a substrate, a system on a package, an Application Specific Integrated Circuit (ASIC), or may be implemented by hardware or firmware in any other reasonable manner of integrating or packaging a circuit, or implemented in any one of three implementations of software, hardware and firmware, or in a suitable combination of any of them. Alternatively, at least one of the fingerprint acquisition module 601, the feature extraction module 602, the acquisition module 701, the optimization module 702, the matching module 703 and the determination module 704 may be at least partly implemented as a computer program module which, when executed, may perform a corresponding function.
Fig. 8 schematically shows a block diagram of an electronic device according to an embodiment of the present disclosure.
Referring to fig. 8, an electronic device provided in an embodiment of the present disclosure includes a processor 810, a communication interface 820, a memory 830, and a communication bus 840, where the processor 810, the communication interface 820, and the memory 830 complete communication with each other through the communication bus 840;
a memory 830 for storing a computer program;
the processor 810, when executing the program stored in the memory 830, implements the fingerprint locating method and the fingerprint data processing method described in any of the previous embodiments.
The communication bus 840 may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The communication bus 840 may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown, but this does not mean that there is only one bus or one type of bus.
The communication interface 820 is used for communication between the electronic device and other devices.
The Memory 830 may include a Random Access Memory (RAM) or a non-volatile Memory (non-volatile Memory), such as at least one disk Memory. Alternatively, the memory 830 may be at least one storage device located remotely from the processor 810.
The Processor 810 may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; the Integrated Circuit may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, or a discrete hardware component.
Embodiments of the present disclosure also provide a computer-readable storage medium. The computer readable storage medium has a computer program stored thereon, and the computer program is used for realizing the fingerprint positioning method and the fingerprint data processing method in any one of the above embodiments when being executed by a processor.
The computer-readable storage medium may be contained in the apparatus/device described in the above embodiments; or may be present alone without being assembled into the device/apparatus. The computer readable storage medium carries one or more programs which, when executed, implement a method for fingerprint location and a method for processing fingerprint data according to any embodiment of the present disclosure.
According to embodiments of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, which may include, for example but is not limited to: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present disclosure, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
It is noted that, in this document, relational terms such as "first" and "second," and the like, may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The previous description is only for the purpose of describing particular embodiments of the present disclosure, so as to enable those skilled in the art to understand or implement the present disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the disclosure. Thus, the present disclosure is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (14)

1. A method of processing fingerprint data, comprising:
acquiring fingerprint data of a terminal, wherein the fingerprint data carries interference generated by a fingerprint source in an acquisition process to the positioning of the terminal;
inputting the fingerprint data into a trained fingerprint optimizer for feature extraction to obtain the positioning features of the fingerprint data; wherein the location feature is used to locate the terminal, and the fingerprint optimizer is used to remove the interference in the fingerprint data.
2. The method of claim 1, wherein the fingerprint optimizer trains by:
acquiring a training sample, inputting the training sample into the fingerprint optimizer for feature extraction to obtain features of the training sample, wherein the training sample comprises a preset fingerprint sample and a crowdsourcing fingerprint sample, and the features of the training sample comprise actual position information of the training sample and actual source information representing that the training sample belongs to the preset fingerprint sample or the crowdsourcing fingerprint sample;
inputting the characteristics of the training samples into a preset state discriminator and a positioning decision maker, and respectively extracting the source characteristics and the positioning characteristics of the training samples;
determining a loss function for the state arbiter and the location decider based on the source features and the location features;
and iteratively optimizing parameters of the state discriminator, the positioning decision-maker and the fingerprint optimizer based on the loss functions of the state discriminator and the positioning decision-maker, performing gradient inversion on the optimized loss function of the state discriminator and updating the parameters of the fingerprint optimizer to obtain the trained fingerprint optimizer.
3. The method of claim 2, wherein determining a loss function for the state arbiter and the location decider based on the source signature and the location signature comprises:
calculating a loss function of the state discriminator according to the actual source information and the source characteristics of the training samples;
and calculating a loss function of the positioning decision maker according to the actual position information and the positioning characteristics of the training samples.
4. The method of claim 3, wherein iteratively optimizing parameters of the state discriminator, the position decider and the fingerprint optimizer based on loss functions of the state discriminator and the position decider, performing gradient inversion on the loss functions of the state discriminator after iterative optimization and updating parameters of the fingerprint optimizer to obtain a trained fingerprint optimizer, comprising:
iteratively optimizing parameters of the state discriminator, the positioning decider and the fingerprint optimizer based on the loss functions of the state discriminator and the positioning decider;
stopping the iterative optimization when the iterative optimization reaches a saddle point of the parameter, and performing gradient inversion on a loss function of the state discriminator;
and updating the parameters of the fingerprint optimizer, and taking the fingerprint optimizer after the parameters are updated as a trained fingerprint optimizer.
5. The method of claim 4, wherein iteratively optimizing the parameters of the state arbiter, the location decider, and the fingerprint optimizer based on the loss functions of the state arbiter and the location decider comprises:
constructing a loss function of a network based on the loss functions of the state discriminator and the positioning decider, the network comprising the state discriminator, the positioning decider and the fingerprint optimizer;
and iteratively optimizing the parameters of the state discriminator, the positioning decision maker and the fingerprint optimizer by iteratively optimizing a loss function of the network.
6. The method of claim 5, wherein the loss function of the network is determined by:
Figure FDA0003615087180000021
wherein, thetafFor the parameters of the fingerprint optimizer, θdIs a parameter of the state discriminator, thetayIs a parameter of the location decider, LyIs a loss function of the location decider, LdAnd the loss function of the state discriminator is represented by lambda, i represents the serial number of the training sample, the 1 st to N training samples are preset fingerprint samples, the N +1 st to N training samples are crowdsourcing fingerprint samples, and N and N are positive integers.
7. The method of claim 5, wherein iteratively optimizing the parameters of the state arbiter, the location decider, and the fingerprint optimizer by iteratively optimizing a loss function of the network comprises: and performing iterative optimization on the loss function of the network by a gradient descent method.
8. The method of any of claims 2 to 6, wherein the penalty function of the state arbiter and the penalty function of the position decision maker are cross entropy penalty functions.
9. A method of fingerprint location, comprising:
acquiring fingerprint data of a user to be positioned in a specified area and a preconfigured fingerprint database, wherein the fingerprint database comprises real positioning and fingerprint characteristics of different position points in the specified area;
inputting the fingerprint data into a trained fingerprint optimizer to perform feature extraction so as to obtain the positioning features of the fingerprint data;
calculating the matching degree between the fingerprint features of the different position points and the positioning features of the fingerprint data;
and determining the real location of the position point corresponding to the fingerprint feature with the highest matching degree as the location of the user to be located in the specified area.
10. The method of claim 9, wherein the fingerprint optimizer trains by:
acquiring a training sample, inputting the training sample into the fingerprint optimizer for feature extraction to obtain features of the training sample, wherein the training sample comprises a preset fingerprint sample and a crowdsourcing fingerprint sample, and the features of the training sample comprise actual position information of the training sample and actual source information representing that the training sample belongs to the preset fingerprint sample or the crowdsourcing fingerprint sample;
inputting the characteristics of the training samples into a preset state discriminator and a positioning decision maker, and respectively extracting the source characteristics and the positioning characteristics of the training samples;
determining a loss function for the state arbiter and the location decider based on the source signature and the location signature;
iteratively optimizing parameters of the state discriminator, the positioning decision device and the fingerprint optimizer based on the loss functions of the state discriminator and the positioning decision device, performing gradient inversion on the optimized loss function of the state discriminator and updating the parameters of the fingerprint optimizer to obtain the trained fingerprint optimizer.
11. An apparatus for processing fingerprint data, comprising:
the terminal comprises a fingerprint acquisition module, a fingerprint acquisition module and a fingerprint processing module, wherein the fingerprint acquisition module is used for acquiring fingerprint data of the terminal, and the fingerprint data carries interference generated by a fingerprint source in an acquisition process on positioning of the terminal;
the characteristic extraction module is used for inputting the fingerprint data into a trained fingerprint optimizer for characteristic extraction to obtain the positioning characteristics of the fingerprint data; wherein the location feature is used to locate the terminal, and the fingerprint optimizer is used to remove the interference in the fingerprint data.
12. An apparatus for locating a fingerprint, comprising:
the system comprises an acquisition module, a positioning module and a positioning module, wherein the acquisition module is used for acquiring fingerprint data of a user to be positioned in a specified area and a pre-configured fingerprint database, and the fingerprint database comprises real positioning and fingerprint characteristics of different position points in the specified area;
the optimization module is used for inputting the fingerprint data into a trained fingerprint optimizer for feature extraction so as to obtain the positioning features of the fingerprint data;
the matching module is used for calculating the matching degree between the fingerprint characteristics of the different position points and the positioning characteristics of the fingerprint data;
and the determining module is used for determining the real positioning of the position point corresponding to the fingerprint feature with the highest matching degree as the positioning of the user to be positioned in the specified area.
13. An electronic device is characterized by comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus;
a memory for storing a computer program;
a processor for implementing the method of processing fingerprint data according to any one of claims 1 to 8 or the method of fingerprint location according to any one of claims 9 to 10 when executing a program stored in a memory.
14. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the method of processing fingerprint data according to any one of claims 1 to 8 or the method of fingerprint localization according to any one of claims 9 to 10.
CN202210443922.4A 2022-04-25 2022-04-25 Method, apparatus, device and medium for fingerprint positioning and processing fingerprint data Pending CN114783004A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210443922.4A CN114783004A (en) 2022-04-25 2022-04-25 Method, apparatus, device and medium for fingerprint positioning and processing fingerprint data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210443922.4A CN114783004A (en) 2022-04-25 2022-04-25 Method, apparatus, device and medium for fingerprint positioning and processing fingerprint data

Publications (1)

Publication Number Publication Date
CN114783004A true CN114783004A (en) 2022-07-22

Family

ID=82433592

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210443922.4A Pending CN114783004A (en) 2022-04-25 2022-04-25 Method, apparatus, device and medium for fingerprint positioning and processing fingerprint data

Country Status (1)

Country Link
CN (1) CN114783004A (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110296699A (en) * 2019-06-06 2019-10-01 深圳数位传媒科技有限公司 A kind of indoor orientation method and device based on earth's magnetic field
CN110958575A (en) * 2019-12-02 2020-04-03 重庆邮电大学 Positioning method and system based on WiFi fusion prediction
CN113596757A (en) * 2021-07-29 2021-11-02 天津大学 Rapid high-precision indoor fingerprint positioning method based on integrated width learning
CN113916231A (en) * 2021-09-28 2022-01-11 上海浦东发展银行股份有限公司 Mobile equipment positioning navigation method and device, mobile equipment and storage medium
CN114269014A (en) * 2021-12-22 2022-04-01 东南大学 Large-scale MIMO dynamic environment fingerprint positioning method based on domain adaptive network
CN114374931A (en) * 2020-10-19 2022-04-19 中国移动通信集团重庆有限公司 Fingerprint positioning method based on measurement learning of neighbor component analysis

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110296699A (en) * 2019-06-06 2019-10-01 深圳数位传媒科技有限公司 A kind of indoor orientation method and device based on earth's magnetic field
CN110958575A (en) * 2019-12-02 2020-04-03 重庆邮电大学 Positioning method and system based on WiFi fusion prediction
CN114374931A (en) * 2020-10-19 2022-04-19 中国移动通信集团重庆有限公司 Fingerprint positioning method based on measurement learning of neighbor component analysis
CN113596757A (en) * 2021-07-29 2021-11-02 天津大学 Rapid high-precision indoor fingerprint positioning method based on integrated width learning
CN113916231A (en) * 2021-09-28 2022-01-11 上海浦东发展银行股份有限公司 Mobile equipment positioning navigation method and device, mobile equipment and storage medium
CN114269014A (en) * 2021-12-22 2022-04-01 东南大学 Large-scale MIMO dynamic environment fingerprint positioning method based on domain adaptive network

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
YAROSLAV GANIN，ETAL: "《Unsupervised Domain Adaptation by Backpropagation》", 《ARXIV:1409.7495V2》 *
赵增华等: "《基于域自适应的Wi-Fi指纹设备无关室内定位模型》", 《通信学报》 *

Similar Documents

Publication Publication Date Title
CN110375753B (en) Map matching method, device, server and storage medium
CN110856112B (en) Crowd-sourcing perception multi-source information fusion indoor positioning method and system
CN108834077B (en) Tracking area division method and device based on user movement characteristics and electronic equipment
WO2018145611A1 (en) Effective indoor localization using geo-magnetic field
CN113342905B (en) Method and device for determining stop point
CN104394588B (en) Indoor orientation method based on Wi Fi fingerprints and Multidimensional Scaling
CN109165309B (en) Negative example training sample acquisition method and device and model training method and device
CN106471338A (en) Determine position in geographic area for the mobile device
CN110471055B (en) Method and device for predicting flight path, readable storage medium and terminal equipment
CN105517019A (en) Method for detecting LTE (Long Term Evolution) network performance by using integrated regression system
CN111383246B (en) Scroll detection method, device and equipment
CN112668238B (en) Rainfall processing method, rainfall processing device, rainfall processing equipment and storage medium
CN109902641B (en) Semantic alignment-based face key point detection method, system and device
US10677881B2 (en) Map assisted inertial navigation
CN108462939B (en) Indoor positioning method for geomagnetic time sequence analysis
CN114677575A (en) Scene migration method and device and electronic equipment
CN109889981B (en) Positioning method and system based on binary classification technology
CN114783004A (en) Method, apparatus, device and medium for fingerprint positioning and processing fingerprint data
CN108834053B (en) Positioning method, device and equipment
CN110910337A (en) Ship track fusion method and device and terminal equipment
CN111949840A (en) Topological graph structure construction method and device based on Internet of things data
JP6927161B2 (en) Learning devices, predictors, methods, and programs
KR20220159746A (en) Method and appratus for determining the path of a typhoon
CN111612816A (en) Method, device and equipment for tracking moving target and computer storage medium
CN111368624A (en) Loop detection method and device based on generation of 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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20220722