CN117222005B

CN117222005B - Fingerprint positioning method, fingerprint positioning device, electronic equipment and storage medium

Info

Publication number: CN117222005B
Application number: CN202311478044.0A
Authority: CN
Inventors: 王霄峻; 李大帅; 贺晨琳; 程科; 魏诗雨; 刘鑫; 许睿哲; 夏灵均
Original assignee: Network Communication and Security Zijinshan Laboratory
Current assignee: Network Communication and Security Zijinshan Laboratory
Priority date: 2023-11-08
Filing date: 2023-11-08
Publication date: 2024-04-02
Anticipated expiration: 2043-11-08
Also published as: CN117222005A

Abstract

The invention provides a fingerprint positioning method, a fingerprint positioning device, electronic equipment and a storage medium, and relates to the technical field of positioning, wherein the fingerprint positioning method comprises the following steps: determining an angle time delay domain channel amplitude matrix of the position of the terminal to be positioned; inputting the angle time delay domain channel amplitude matrix into a region determination model to obtain a target region corresponding to a terminal to be positioned, which is output by the region determination model, wherein the region determination model is obtained based on the angle time delay domain channel amplitude matrix sample and region sample training; and determining the position information of the terminal to be positioned according to the angle time delay domain channel amplitude matrix and at least one target angle time delay domain channel amplitude matrix sample in the target area. The area determination model is adopted to determine the area where the terminal to be positioned is located, and then the terminal to be positioned is subjected to positioning estimation, so that a positioning result with higher accuracy can be obtained, and meanwhile, the positioning time required by the terminal to be positioned in position estimation can be greatly shortened, so that the real-time requirement of the terminal to be positioned is met.

Description

Fingerprint positioning method, fingerprint positioning device, electronic equipment and storage medium

Technical Field

The present invention relates to the field of positioning technologies, and in particular, to a fingerprint positioning method, a fingerprint positioning device, an electronic device, and a storage medium.

Background

In the traditional wireless network positioning method, due to the existence of multipath effect, reflection, diffraction and other phenomena occur in the process that a wireless signal is transmitted and propagated to a base station by a terminal to be positioned, so that a larger transmission time delay exists between the wireless signal received by the base station and the wireless signal transmitted by the terminal, the error is larger when the terminal to be positioned is positioned according to the received wireless signal, the positioning result is greatly influenced, the accuracy of the obtained positioning result is lower, and the real-time requirement of positioning the terminal to be positioned cannot be met in the whole positioning process.

Disclosure of Invention

The invention provides a fingerprint positioning method, a device, electronic equipment and a storage medium, which are used for solving the defects that in the prior art, errors are large when a terminal to be positioned is positioned according to a received wireless signal, the positioning result is greatly influenced, the accuracy of the obtained positioning result is low, and the whole positioning process cannot meet the real-time requirement of positioning the terminal to be positioned.

The invention provides a fingerprint positioning method, which comprises the following steps:

determining an angle time delay domain channel amplitude matrix of the position of the terminal to be positioned;

inputting the angle delay domain channel amplitude matrix into a region determination model to obtain a target region corresponding to the terminal to be positioned, which is output by the region determination model, wherein the region determination model is obtained based on angle delay domain channel amplitude matrix samples and region samples through training;

and determining the position information of the terminal to be positioned according to the angle time delay domain channel amplitude matrix and at least one target angle time delay domain channel amplitude matrix sample in the target area.

According to the fingerprint positioning method provided by the invention, the area determination model is obtained based on training of the following steps: acquiring an angle time delay domain channel amplitude matrix sample of a position of a reference terminal and an area sample of the reference terminal; and training the original area determination model by taking the angle time delay domain channel amplitude matrix sample as training data and taking the area sample as a training label to obtain a trained area determination model.

According to the fingerprint positioning method provided by the invention, the determining of the angular delay domain channel amplitude matrix of the position of the terminal to be positioned comprises the following steps: acquiring a channel frequency response matrix of the position of the terminal to be positioned; and determining the angle delay domain channel amplitude matrix according to the channel frequency response matrix.

According to the fingerprint positioning method provided by the invention, the determining the position information of the terminal to be positioned according to the angle delay domain channel amplitude matrix and at least one target angle delay domain channel amplitude matrix sample in the target area comprises the following steps: determining similarity distances between the angle delay domain channel amplitude matrix and the target angle delay domain channel amplitude matrix samples according to the target angle delay domain channel amplitude matrix samples; and determining the position information of the terminal to be positioned according to at least one similarity distance.

According to the fingerprint positioning method provided by the invention, the training of the original area determination model by taking the angle delay domain channel amplitude matrix sample as training data and the area sample as a training label to obtain a trained area determination model comprises the following steps: an input layer in the model is determined by adopting the original region, and the angle delay domain channel amplitude matrix sample is normalized to obtain a first angle delay domain channel amplitude matrix sample; adopting at least one characteristic processing layer in the original area determination model to perform characteristic processing on the first angle time delay domain channel amplitude matrix sample to obtain a second angle time delay domain channel amplitude matrix sample; adopting the original region to determine a full connection layer in a model, performing feature integration on the second angle time delay domain channel amplitude matrix sample, and determining a prediction region corresponding to the integrated second angle time delay domain channel amplitude matrix sample; and updating model parameters of the original region determination model according to the integrated second angle time delay domain channel amplitude matrix sample and the prediction region to obtain the trained region determination model.

According to the fingerprint positioning method provided by the invention, each characteristic processing layer comprises a convolution layer and a pooling layer, at least one characteristic processing layer in the original area determination model is adopted to perform characteristic processing on the first angle delay domain channel amplitude matrix sample to obtain a second angle delay domain channel amplitude matrix sample, and the method comprises the following steps: for each characteristic processing layer, if the current characteristic processing layer is not the last characteristic processing layer, adopting a convolution layer in the current characteristic processing layer to perform characteristic processing on the first angle delay domain channel amplitude matrix sample to obtain a third angle delay domain channel amplitude matrix sample; compressing the third angle time delay domain channel amplitude matrix sample by adopting a pooling layer in the current feature processing layer to obtain a fourth angle time delay domain channel amplitude matrix sample, wherein the fourth angle time delay domain channel amplitude matrix sample is used for determining the second angle time delay domain channel amplitude matrix sample by the last feature processing layer, and the fourth angle time delay domain channel amplitude matrix sample is input data of the next adjacent feature processing layer of the current feature processing layer; if the current feature processing layer is the last feature processing layer, performing feature processing on the first angle delay domain channel amplitude matrix sample by adopting a convolution layer in the current feature processing layer to obtain a fifth angle delay domain channel amplitude matrix sample; and compressing the fifth angle time delay domain channel amplitude matrix sample by adopting a pooling layer in the current characteristic processing layer to obtain the second angle time delay domain channel amplitude matrix sample, wherein the second angle time delay domain channel amplitude matrix sample is input data of the full-connection layer.

According to the fingerprint positioning method provided by the invention, the determining of the angular delay domain channel amplitude matrix according to the channel frequency response matrix comprises the following steps: according to the channel frequency response matrix, determining an angle time delay domain channel response matrix of the position of the terminal to be positioned; and determining the angle time delay domain channel amplitude matrix according to the angle time delay domain channel response matrix.

According to the fingerprint positioning method provided by the invention, the determining of the angular delay domain channel amplitude matrix according to the angular delay domain channel response matrix comprises the following steps: determining an angle time delay domain channel energy matrix according to the angle time delay domain channel response matrix; and determining the angle delay domain channel amplitude matrix according to the angle delay domain channel energy matrix.

According to the fingerprint positioning method provided by the invention, the determining the position information of the terminal to be positioned according to at least one similarity distance comprises the following steps: determining a target similarity distance meeting a preset distance range from the at least one similarity distance; determining position estimation coordinates according to the position information of the target similarity distance corresponding to the reference terminal; and determining the position estimation coordinates as the position information of the terminal to be positioned.

The invention also provides a fingerprint positioning device, which comprises:

the matrix determining module is used for determining an angle delay domain channel amplitude matrix of the position of the terminal to be positioned;

the area determining module is used for inputting the angle delay domain channel amplitude matrix into an area determining model to obtain a target area corresponding to the terminal to be positioned, which is output by the area determining model, wherein the area determining model is obtained based on angle delay domain channel amplitude matrix samples and area sample training;

and the position determining module is used for determining the position information of the terminal to be positioned according to the angle delay domain channel amplitude matrix and at least one target angle delay domain channel amplitude matrix sample in the target area.

The invention also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the fingerprint positioning method as described in any of the above when executing the program.

The invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a fingerprint positioning method as described in any of the above.

The invention also provides a computer program product comprising a computer program which when executed by a processor implements a fingerprint positioning method as described in any one of the above.

The fingerprint positioning method, the fingerprint positioning device, the electronic equipment and the storage medium provided by the invention are characterized in that the angular delay domain channel amplitude matrix of the position of the terminal to be positioned is determined; inputting the angle delay domain channel amplitude matrix into a region determination model to obtain a target region corresponding to the terminal to be positioned, which is output by the region determination model, wherein the region determination model is obtained based on angle delay domain channel amplitude matrix samples and region samples through training; and determining the position information of the terminal to be positioned according to the angle time delay domain channel amplitude matrix and at least one target angle time delay domain channel amplitude matrix sample in the target area. According to the method, the area of the terminal to be positioned is firstly determined by adopting an area determination model, then the terminal to be positioned is subjected to positioning estimation, the positioning error existing in the whole process is small, and then the positioning result of the terminal to be positioned is slightly influenced, so that the positioning result with higher accuracy can be obtained, and meanwhile, the positioning time required by the position estimation of the terminal to be positioned can be greatly shortened in the whole process, so that the real-time requirement of the terminal to be positioned is met.

Drawings

In order to more clearly illustrate the invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic flow chart of a fingerprint positioning method according to the present invention;

FIG. 2 is a schematic view of a scenario in which a target area to be located is divided according to the present invention;

FIG. 3a is a schematic diagram of the amplitude distribution of an ADCPM sample provided by the present invention;

FIG. 3b is a schematic diagram of the amplitude distribution of ADCAM samples according to the present invention;

FIG. 4 is a schematic view of a visual structure of the area determination model provided by the present invention;

FIG. 5 is a second flowchart of a fingerprint positioning method according to the present invention;

FIG. 6 is a schematic diagram of a fingerprint positioning device according to the present invention;

fig. 7 is a schematic structural diagram of an electronic device provided by the present invention.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

In order to better understand the embodiments of the present invention, a simple explanation can be made on the fingerprint positioning technology related to the embodiments of the present invention:

the fingerprint positioning technology refers to a mode of acquiring a sufficient number of signal fingerprints in a specific area as position references by utilizing the characteristic that multipath effects have different effects on fingerprint signals in different positions, so as to position terminals in the area. Compared with the traditional wireless network positioning technology, the fingerprint positioning technology has the greatest advantage that the positioning precision is not affected by multipath effect, but multipath information can be fully utilized, and the parameter characteristics of signal fingerprint are enhanced.

It should be noted that, the execution body according to the embodiment of the present invention may be a fingerprint positioning device, or may be an electronic device, and the electronic device may include: base station, computer, mobile terminal, wearable equipment, etc.

Wherein, the electronic device may be associated with at least one terminal, and optionally, the terminal may include: computer, mobile terminal, wearable device, etc.

Alternatively, the electronic device and the terminal may be connected by a wireless communication technology, where the wireless communication technology may include, but is not limited to, one of the following: fourth generation communication technology (the 4th Generation mobile communication technology,4G), fifth generation communication technology (the 5th Generation mobile communication technology,5G), wireless fidelity technology (Wireless Fidelity, wiFi), and the like.

The following further describes embodiments of the present invention by taking an electronic device as an example.

As shown in fig. 1, a flow chart of a fingerprint positioning method provided by the present invention may include:

101. and determining an angle delay domain channel amplitude matrix of the position of the terminal to be positioned.

The angular delay domain channel amplitude matrix (Angle Delay Channel Amplitude Matrix, ADCAM) refers to an amplitude matrix of the angular delay domain corresponding to the channel frequency response of the terminal to be positioned at the base station side in the scattering environment, namely a fingerprint matrix corresponding to the fingerprint signal.

The electronic equipment can firstly acquire the fingerprint signal when the fingerprint signal sent by the terminal to be positioned reaches the base station through channel estimation, and then determine the fingerprint matrix corresponding to the fingerprint signal, namely ADCAM, so as to provide data support for the subsequent positioning estimation of the terminal to be positioned.

The channel estimation refers to an amplitude matrix of channel frequency response corresponding to a terminal to be positioned at a base station side in an angle delay domain in a scattering environment.

A base station refers to a radio transceiver station that performs information transfer with a terminal through a mobile communication switching center in a certain radio coverage area.

102. And inputting the angle time delay domain channel amplitude matrix into the region determination model to obtain a target region corresponding to the terminal to be positioned, which is output by the region determination model.

The area determination model is an Action-Decision Networks (ADNet) network model, which is a convolutional neural network model and is obtained based on angle delay domain channel amplitude matrix samples and area samples, and the number of the angle delay domain channel amplitude matrix samples and the area samples is not limited.

In the online stage, the electronic equipment inputs the acquired ADCAM of the position of the terminal to be positioned into a trained region determination model, and the region determination model can determine the region sample corresponding to the maximum probability in all probabilities as a target region output by the region determination model based on the probabilities of the ADCAM in different region samples.

103. And determining the position information of the terminal to be positioned according to the angle time delay domain channel amplitude matrix and at least one target angle time delay domain channel amplitude matrix sample in the target area.

After determining the target area, the electronic device may determine at least one target ADCAM sample within the target area; and the electronic equipment carries out positioning estimation on the terminal to be positioned according to the ADCAM corresponding to the terminal to be positioned and the at least one target ADCAM sample acquired before, so as to obtain the position information of the terminal to be positioned with higher accuracy.

In the embodiment of the invention, an angle delay domain channel amplitude matrix of the position of the terminal to be positioned is determined; inputting the angle time delay domain channel amplitude matrix into a region determination model to obtain a target region corresponding to a terminal to be positioned, which is output by the region determination model, wherein the region determination model is obtained based on the angle time delay domain channel amplitude matrix sample and region sample training; and determining the position information of the terminal to be positioned according to the angle time delay domain channel amplitude matrix and at least one target angle time delay domain channel amplitude matrix sample in the target area. According to the method, the area of the terminal to be positioned is firstly determined by adopting an area determination model, then the terminal to be positioned is subjected to positioning estimation, the positioning error existing in the whole process is small, and then the positioning result of the terminal to be positioned is slightly influenced, so that the positioning result with higher accuracy can be obtained, and meanwhile, the positioning time required by the position estimation of the terminal to be positioned can be greatly shortened in the whole process, so that the real-time requirement of the terminal to be positioned is met.

For a further understanding of embodiments of the present invention, the following detailed description of embodiments of the present invention will be provided:

in some embodiments, the determining, by the electronic device, an angular-delay domain channel amplitude matrix of a location of the terminal to be located may include: the electronic equipment acquires a channel frequency response (Channel Frequency Response, CFR) matrix of the position of the terminal to be positioned; the electronic device determines an angular delay domain channel amplitude matrix from the channel frequency response matrix.

The CFR matrix refers to channel frequency response corresponding to the terminal to be positioned at the base station side in a scattering environment.

The electronic equipment can obtain a CFR matrix corresponding to the fingerprint signal sent by the terminal to be positioned when reaching the base station through channel estimation, and further process the CFR matrix to obtain the ADCAM.

In some embodiments, the electronic device determining the angular delay domain channel amplitude matrix from the channel frequency response matrix may include: the electronic equipment determines an Angle time-delay domain channel response matrix (ADCRM) of the position of the terminal to be positioned according to the channel frequency response matrix; the electronic device determines an angle delay domain channel amplitude matrix according to the angle delay domain channel response matrix.

The ADCRM refers to a matrix of an angle delay domain of a channel frequency response corresponding to a terminal to be positioned at a base station side in a scattering environment.

After acquiring the CFR matrix, the electronic device may perform discrete fourier transform (Discrete Fourier Transform, DFT) on the CFR matrix to obtain adcmr, and further process the adcmr to obtain ADCAM.

Optionally, the determining, by the electronic device, the angular-delay domain channel response matrix of the location of the terminal to be located according to the channel frequency response matrix may include: and the electronic equipment determines an angle delay domain channel response matrix of the position of the terminal to be positioned according to a first formula.

Wherein, the first formula is:；

represents ADCRM; />Representing a terminal to be positioned +.>The number of antennas; />Representation->A phase shift DFT matrix of dimensions; />Representing a terminal to be positioned +.>A corresponding CFR matrix; />Representing the number of subcarriers of an orthogonal frequency division multiplexing (Orthogonal Frequency Division Multiplexing, OFDM) system; />Representing the number of cyclic prefixes of the OFDM system; />Representation->DFT unitary matrix of dimension>Is a conjugate matrix of (a).

Wherein,，/>representing phase-shifted DFT matricesMiddle->Line->Elements of a column; />，/>Representing DFT unitary matrix->Conjugate matrix of>Middle->Line->Column elements.

And the electronic equipment performs DFT conversion on the CFR matrix according to the first formula to obtain the ADCRM with higher accuracy.

In some embodiments, the electronic device determines an angular-delay domain channel amplitude matrix from the angular-delay domain channel response matrix, which may include: the electronic equipment determines an Angle time-delay domain channel energy matrix (ADCPM) according to the Angle time-delay domain channel response matrix; the electronic device determines an angle delay domain channel amplitude matrix according to the angle delay domain channel energy matrix.

The ADCPM refers to a power matrix of a channel frequency response in an angle delay domain, which corresponds to a terminal to be positioned at a base station side in a scattering environment.

After the electronic device obtains the adcm, hadamard (Hadamard) inner product operation can be performed on the adcm and the conjugate matrix of the adcm to obtain adcm, and then the adcm is processed to obtain ADCAM.

Optionally, the electronic device determines an angle delay domain channel energy matrix according to the angle delay domain channel response matrix, which may include: and the electronic equipment determines an angle delay domain channel energy matrix according to a second formula.

Wherein, the second formula is:；

represents adcm; />Representing the conjugate matrix of ADCRM; />Representing an inner product operation; />Representing the mathematical expectation of the inner product operation of the conjugate matrix of ADCRM and ADCRM.

Wherein,，/>representing an angular-delay domain channel response matrix>Middle->Line->Elements of a column; />Representing an angular-delay domain channel response matrix>Middle->Line->Elements of a column; the absolute value is indicated.

The electronic device generates a second formula based on the ADCRM and the conjugate matrix of the ADCRMAfter Hadamard inner product operation is carried out, the corresponding mathematical expectation is obtained, so that ADCPM with higher accuracy is obtained.

Optionally, the electronic device determines an angle delay domain channel amplitude matrix according to the angle delay domain channel energy matrix, which may include: and the electronic equipment determines an angle delay domain channel amplitude matrix according to a third formula.

Wherein, the third formula is:；

indicating ADCAM.

Wherein,，/>channel amplitude matrix representing angle time delay domain>Middle (f)Line->Column elements.

And the electronic equipment performs quadratic root finding on each element in the ADCPM according to the third formula so as to obtain the ADCAM with higher accuracy.

At this time, the relative size relation between the elements in the ADCAM determined by the above process is maintained, and the element values of the elements can be effectively reduced, so that the subsequent calculation and storage overhead of the ADCAM is reduced.

Optionally, in determining the probability that the ADCAM is in a sample of a different area, since the output result (the score value corresponding to the ADCAM) of the area determination model has no range limitation, the electronic device may input the output result into a Softmax classifier of the area determination model, where the Softmax classifier normalizes the output result by using a probability distribution formula to obtain the probability distribution of the ADCAM in the sample of the different area.

Wherein, the probability distribution formula is:；

indicating ADCAM is +.>Sample of individual region->Probability of each region sample, and the sum of all probabilities is 1;representing a Softmax activation function; />Input data representing a Softmax activation function, i.e. the output result of the region determination model (ADCAM at +. >Fractional value of individual region samples).

Optionally, in the offline stage, the electronic device acquires samples of a plurality of areas as follows: the electronic equipment acquires a target area to be positioned; the electronic equipment determines the shape and the area of the target area to be positioned; the electronic equipment divides the target area to be positioned according to the need based on the shape and the area to obtain a plurality of area samples, so that the original area determination model can be trained conveniently.

In addition, the electronic device may further assign different numbered labels to the plurality of region samples in sequence.

Exemplary, as shown in fig. 2, a schematic view of a scenario in which a target area to be located is divided is provided in the present invention. As can be seen from fig. 2: after the electronic equipment acquires the target area to be positioned, the shape of the target area to be positioned can be determined to be rectangular and the area; the electronic device sets a region width range to 127 meters based on the shape and the area, and sets a base station at a coordinate origin point, the base station being 300 meters apart from the center of the target region to be located. The electronic equipment divides the target area to be positioned to obtain nine area samples. At this time, the electronic device may assign a number tag to the first area sample of the first row to be area1, assign a number tag to the second area sample of the first row to be area2, and sequentially assign the number tags "area1" to "area9" to the respective area notes from left to right and from top to bottom according to this rule.

Optionally, in the off-line stage, the electronic device acquires the ADCAM sample at the position in each area sample as follows: the electronic equipment uniformly divides the samples of each region according to preset intervals to obtain a plurality of reference points; the electronic equipment traverses each reference point, acquires ADCAM samples of the positions of the reference points corresponding to the reference terminals, and facilitates subsequent training of the original region determination model.

In addition, the electronic equipment can also determine the reference position information of the reference terminal corresponding to each reference point and the serial number label of the sample in the area.

Illustratively, in connection with FIG. 2 described above, it can be seen from FIG. 2 that: setting the preset interval to be 2 meters, and uniformly dividing the samples of each area according to the preset interval by the electronic equipment to obtain 4032 reference points; the electronics can obtain 4032 ADCAM samples based on these 4032 reference points, and thus 4032 ADCAM samples.

The same reference point may correspond to one reference terminal or may correspond to a plurality of reference terminals, which is not specifically limited herein.

Exemplary, as shown in fig. 3a, a schematic diagram of the amplitude distribution of the adcm sample provided by the present invention is shown. In fig. 3a, the number of antennas corresponding to the adcm samples Cyclic prefix number of 32, OFDM->48. Fig. 3a is a three-dimensional bar graph drawn by taking a column of adcm samples (TOA components corresponding to different paths) as an X-axis, a row of the adcm samples (AOA components corresponding to different paths) as a Y-axis, and the element values of each element in the adcm samples as a Z-axis. As shown in fig. 3b, the amplitude distribution of the ADCAM sample provided by the invention is shown. In fig. 3b, the number of antennas to which the ADCAM sample corresponds +.>Cyclic prefix number of 32, OFDM->48. Fig. 3b is a three-dimensional bar graph drawn with the column of ADCAM samples (TOA components corresponding to different paths) as the X-axis, the row of ADCAM samples (AOA components corresponding to different paths) as the Y-axis, and the element values of the elements in the ADCAM samples as the Z-axis. Because the elements in the ADCAM sample are obtained by squaring the elements in the adcm, the relative magnitude relationship between the elements in the ADCAM sample remains consistent, and the amplitude (i.e., the element value) near the 19 th AOA component in the angle domain is still the highest, and the other two sides are lower.

Note that, the ADCAM samples and adcm samplesThe difference is that: the ADCPM sample has huge element value of individual elements, which can reach 2.8X10 ⁶ The difference between the element values of the ADCAM sample and most elements is more than 3 orders of magnitude, the existence of the special values inevitably brings additional expenditure to the subsequent calculation process, and the difference between the elements in the ADCAM sample can be effectively reduced because the element values in the ADCPM sample are square roots, and the maximum element value in the ADCAM sample is 1687, so that the calculation expenditure can be reduced, and the subsequent calculation process is more convenient.

Optionally, the electronic device may further store the ADCAM sample, reference position information of the reference terminal corresponding to each reference point, and a serial number tag of the region sample in a database, so that the method is convenient for subsequent training of the original region determination model and use in estimating the position information of the terminal to be located in an online stage.

In some embodiments, the electronic device region determination model is trained based on the following steps: the method comprises the steps that electronic equipment obtains an angle time delay domain channel amplitude matrix sample of a position of a reference terminal and a region sample of the reference terminal; the electronic equipment trains an original area determination model by taking an angle time delay domain channel amplitude matrix sample as training data and taking an area sample as a training label to obtain a trained area determination model.

Optionally, the process of determining the angle delay domain channel amplitude matrix sample by the electronic device is similar to the process of determining the angle delay domain channel amplitude matrix by the electronic device, which is not described in detail herein.

Alternatively, the training label may be a numbered label corresponding to the region sample.

In the off-line stage, the electronic equipment can build a proper convolutional neural network model, namely training the original area determination model. In the training process, firstly acquiring an ADCAM sample at the position of a reference terminal and a region sample at the position of the reference terminal; the electronic equipment trains the original area determination model by taking the ADCAM sample as training data and taking the area sample as a training label until the loss function value corresponding to the original area determination model is reduced to an expected range, and stores the node parameter weight and each bias corresponding to the final original area determination model to obtain the trained area determination model for on-line stage prediction.

The expected range may be set before the electronic device leaves the factory, or may be user-defined according to the actual situation, which is not specifically limited herein.

Alternatively, the loss function may employ a cross entropy loss function.

In some embodiments, the electronic device trains the original area determination model with the angle delay domain channel amplitude matrix sample as training data and with the area sample as a training tag, to obtain a trained area determination model, which may include: the electronic equipment adopts an original area to determine an input layer in the model, and performs normalization processing on the angular time delay domain channel amplitude matrix samples to obtain first angular time delay domain channel amplitude matrix samples; the electronic equipment adopts at least one characteristic processing layer in an original area determination model to perform characteristic processing on the first angle time delay domain channel amplitude matrix sample to obtain a second angle time delay domain channel amplitude matrix sample; the electronic equipment adopts an original area to determine a full connection layer in a model, performs characteristic integration on a second angle time delay domain channel amplitude matrix sample, and determines a prediction area corresponding to the integrated second angle time delay domain channel amplitude matrix sample; the electronic equipment updates model parameters of the original region determination model according to the integrated second angle time delay domain channel amplitude matrix sample and the prediction region to obtain a trained region determination model.

It should be noted that the original region determination model may include an input layer, a feature processing layer, and a full connection layer.

Alternatively, the feature processing layer may include: a convolution layer and a pooling layer.

Wherein, each characteristic processing layer in at least one characteristic processing layer is connected in series. The convolution layers are an indispensable core component in the original region determination model, are key different from other neural networks, and the number of the convolution layers is not limited.

The pooling layer performs the specific operation of selectively compression screening the feature map to reduce the number of parameters in the original region determination model.

The preset times can be set before the electronic equipment leaves the factory, can be customized by a user according to actual conditions, and are not particularly limited. For example, the preset number of times is set to 2.

In the process of training the original region determination model, the electronic device can input the ADCAM sample as input data into an input layer, and the input layer can normalize the ADCAM sample to obtain output data which is easy to extract features, namely a first ADCAM sample. Specifically, the number of antennas isCyclic prefix number of 128, OFDM->For example, the dimension of the ADCAM sample obtained at 96 is 128×96, and since the input data of the conventional convolutional neural network model is usually 3-channel image feature data, the dimension is +. >Therefore, in order to better exploit the feature extraction capability of the original region determination model, before the ADCAM samples are input to the convolution layer, the ADCAM samples are rearranged into a dimension of 64×64×3, that is, the ADCAM samples are subjected to normalization processing, so as to obtain first ADCAM samples, where features of the first ADCAM samples are easier to extract.

The electronic device inputs the first ADCAM sample as input data into at least one feature processing layer, and the at least one feature processing layer performs feature processing on the first ADCAM sample to obtain a second ADCAM sample with better quality. In the feature processing process, the electronic device may input the first ADCAM sample to the first feature processing layer to obtain a first feature result output by the first feature processing layer, and input the first feature result to a feature processing layer adjacent to the first feature processing layer to obtain a second feature result output by the adjacent feature processing layer until a feature result is obtained, where the feature result is the second ADCAM sample.

Then, the electronic device adopts the full connection layer to integrate the characteristics of the finally output second ADCAM sample, determines a prediction area corresponding to the integrated second ADCAM sample, and maps the prediction area to a sample space. For example, assuming that the dimension of the final output second ADCAM sample is 16×16×8, in order to effectively map the features proposed in the second ADCAM sample to the sample space, a flattening (flat) operation is performed on the second ADCAM sample, that is, the dimension of the second ADCAM sample is changed to 2048×1, and then a full connection layer with a length of 128 is used to integrate the second ADCAM sample with the dimension of 2048×1.

Finally, the electronic equipment stores the node parameter weights and various biases corresponding to the final original area determination model according to the integrated second ADCAM sample and the prediction area until the loss function value corresponding to the original area determination model is reduced to an expected range, and a trained area determination model is obtained.

It should be noted that, in order to prevent the overfitting of the original region determination model, a Dropout operation with a probability of 0.5 may be added; the length of the full connection layer of the last layer is the number of divided region samples, and in connection with fig. 2, for nine region sample classification application scenarios, the length of the full connection layer may be set to 9.

In some embodiments, each feature processing layer of the electronic device includes a convolution layer and a pooling layer, and the electronic device performs feature processing on the first angle delay domain channel amplitude matrix sample by using at least one feature processing layer in the original region determination model to obtain a second angle delay domain channel amplitude matrix sample, including at least one implementation manner of the following:

implementation 1: for each feature processing layer, if the current feature processing layer is not the last feature processing layer, adopting a convolution layer in the current feature processing layer to perform feature processing on the first angle delay domain channel amplitude matrix sample to obtain a third angle delay domain channel amplitude matrix sample; and compressing the third angle time delay domain channel amplitude matrix sample by adopting a pooling layer in the current feature processing layer to obtain a fourth angle time delay domain channel amplitude matrix sample, wherein the fourth angle time delay domain channel amplitude matrix sample is used for determining a second angle time delay domain channel amplitude matrix sample by the last feature processing layer, and the fourth angle time delay domain channel amplitude matrix sample is input data of the next adjacent feature processing layer of the current feature processing layer.

Judging whether the current feature processing layer is the last feature processing layer in all feature processing layers in the process of carrying out feature processing on the first ADCAM sample by at least one feature processing layer, if not, carrying out feature processing on the first ADCAM sample by adopting a convolution layer in the current feature processing layer to obtain a third ADCAM sample; and compressing the third ADCAM sample by adopting a pooling layer in the current characteristic processing layer to obtain a fourth ADCAM sample, so as to determine a second ADCAM sample by the last characteristic processing layer.

Implementation 2: for each feature processing layer, if the current feature processing layer is the last feature processing layer, adopting a convolution layer in the current feature processing layer to perform feature processing on the first angle time delay domain channel amplitude matrix sample to obtain a fifth angle time delay domain channel amplitude matrix sample; and compressing the fifth angle time delay domain channel amplitude matrix sample by adopting a pooling layer in the current characteristic processing layer to obtain a second angle time delay domain channel amplitude matrix sample, wherein the second angle time delay domain channel amplitude matrix sample is input data of the full-connection layer.

In the process of judging whether the current feature processing layer is the last feature processing layer in all feature processing layers, if so, carrying out feature processing on the first ADCAM sample by adopting a convolution layer in the current feature processing layer to obtain a fifth ADCAM sample; and compressing the fifth ADCAM sample by adopting the pooling layer in the current characteristic processing layer to obtain a second ADCAM sample for later use as input data of the full connection layer.

For each feature processing layer, in the process of performing feature processing on the first ADCAM sample, the convolution layer may perform convolution operation on the first ADCAM sample, that is, perform feature extraction on the first ADCAM sample, to obtain a next ADCAM sample with better quality. The number of convolution layers is illustratively two, with a total of 8 convolution kernels of size 3 x 3. The convolution layer uses a ReLU function as an activation function, and the convolution kernel performs convolution operation with the first ADCAM sample, so that an operation result (i.e., a next ADCAM sample) is a feature extracted from the first ADCAM sample by the convolution layer.

The pooling layer generally performs region division on the next ADCAM sample in a non-overlapping manner, and then performs a downsampling operation in each divided region, so as to compress all the eigenvalues of each region into one value, thereby obtaining a new ADCAM sample. Illustratively, the pooling layer compresses the next ADCAM sample by using maximum pooling, the pooling window is 2×2, the dimension of the next ADCAM sample input to the pooling layer is 64×64×8, and the dimension of the output data (i.e., the new ADCAM sample) after maximum pooling becomes 32×32×8.

Alternatively, the downsampling operation may include a maximum pooling operation or an average pooling operation.

Exemplary, as shown in fig. 4, a visual structure diagram of the area determination model provided by the present invention is shown. The region determination model may include one input layer, two convolution layers, two pooling layers, one flattening layer, one Dropout layer, two activation function layers (ReLU function and Softmax activation function, respectively).

In some embodiments, the determining, by the electronic device, the location information of the terminal to be located according to the angle delay domain channel amplitude matrix and at least one target angle delay domain channel amplitude matrix sample in the target area may include: aiming at each target angle time delay domain channel amplitude matrix sample, the electronic equipment determines the similarity distance between the angle time delay domain channel amplitude matrix and the target angle time delay domain channel amplitude matrix sample; the electronic equipment determines the position information of the terminal to be positioned according to at least one similarity distance.

The electronic equipment can determine respective similarity distances between the ADCAM corresponding to the terminal to be positioned and at least one target ADCAM sample, namely, how many target ADCAM samples exist, and the electronic equipment can acquire how many similarity distances; and the electronic equipment performs positioning estimation on the terminal to be positioned based on all the similarity distances to obtain the position information of the terminal to be positioned.

Optionally, the determining, by the electronic device, a similarity distance between the angle-delay domain channel amplitude matrix and the target angle-delay domain channel amplitude matrix sample may include: and the electrical equipment obtains the similarity distance between the angle time delay domain channel amplitude matrix and the target angle time delay domain channel amplitude matrix sample according to the distance formula.

Wherein, the distance formula is:；

wherein, representRepresent ADCAM; />Representing the +.f in at least one target ADCAM sample within the target area>The euclidean distance, i.e., similarity distance, between the individual target ADCAM samples.

The larger the Euclidean distance is, the description of ADCAM and the first region in the target regionThe farther the distance between the individual target ADCAM samples, the lower the similarity between the two; on the contrary, ADCAM and the first +.in the target area are described>The closer the distance between the individual target ADCAM samples, the higher the similarity between the two.

In some embodiments, the determining, by the electronic device, location information of the terminal to be located according to the at least one similarity distance may include: the electronic equipment determines a target similarity distance meeting a preset distance range from at least one similarity distance; the electronic equipment determines position estimation coordinates according to the position information of the reference terminal corresponding to the target similarity distance; the electronic device determines the position estimation coordinates as position information of the terminal to be positioned.

The preset distance range may be set before the electronic device leaves the factory, or may be user-defined, which is not specifically limited herein.

In the process of determining the position information of the terminal to be positioned according to at least one similarity distance, the electronic equipment can adopt a weighted nearest neighbor rule classification (Weighted K Nearest Neighbors, WKNN) algorithm to perform positioning estimation on the terminal to be positioned, at this time, the electronic equipment can determine the target similarity distance meeting a preset distance range from the at least one similarity distance, namely determine K similarity distances with the closest similarity distances, determine the K similarity distances as K target similarity distances, and further perform positioning estimation on the terminal to be positioned to obtain position estimation coordinates; the electronic equipment determines the position estimation coordinates as the position information of the terminal to be positioned.

Optionally, the determining, by the electronic device, the position estimation coordinate according to the position information of the target similarity distance corresponding to the reference terminal may include: and the electronic equipment obtains position estimation coordinates according to a position estimation formula in the WKNN algorithm.

The position estimation formula is as follows:，/>；

indicate->The object similarity distance corresponds to the position estimation coordinates of the reference terminal, >；Indicate->The object similarity distance corresponds to the weight of the reference terminal and satisfies +.>；/>Indicate->The target similarity distance corresponds to the similarity distance between the target ADCAM sample of the reference terminal and the ADCAM corresponding to the terminal to be positioned, and since the similarity distance may be small, a small positive number ∈0 is added to prevent the case that the denominator is approximately 0>；/>Indicate->The target similarity distances correspond to the location information of the reference terminal. />

Because of the introduction of the weighting concept, the larger the weight assigned to the reference terminal corresponding to the reference point with the closer similarity distance to the terminal to be positioned is, the larger the influence on the position estimation result is; conversely, the smaller the weight assigned to the reference terminal corresponding to the reference point at which the similarity distance of the terminal to be located is closer, the smaller the influence on the location estimation result.

Exemplary, as shown in fig. 5, a flowchart of a fingerprint positioning method provided by the present invention is shown. As can be seen from fig. 5: in the off-line stage, the electronic equipment can divide the target area to be positioned to obtain a plurality of area samples, and respectively endow the area samples with numbered labels; determining a plurality of reference points from a target area to be positioned, and constructing a fingerprint library based on the angle delay domain channel amplitude matrix samples of the reference terminals corresponding to the plurality of reference points; then, an initial area determining model is built, and training is carried out on the initial area determining model based on all the numbered labels and the fingerprint database, so that a trained area determining model is obtained; in an online stage, the electronic equipment acquires an angle time delay domain channel amplitude matrix of the position of the terminal to be positioned, and then inputs the angle time delay domain channel amplitude matrix into a region determination model to obtain a target region output by the region determination model; and then, according to the angle time delay domain channel amplitude matrix and at least one target angle time delay domain channel amplitude matrix sample in the target area, carrying out positioning estimation on the terminal to be positioned by adopting a WKNN algorithm, and outputting the position information of the terminal to be positioned.

The whole positioning process can greatly reduce the positioning time length required by the online stage while causing little influence on positioning errors, and meets the real-time requirement of the terminal to be positioned. On the one hand, although the selection range of the reference point is limited by using the WKNN algorithm to carry out position estimation, the reference point with very close similarity distance is possibly excluded, so that the average positioning error is increased, if the area determination model is properly trained, the area where the terminal to be positioned is positioned can be correctly predicted, the positioning error is not greatly increased, and the amplification is within the acceptable range; on the other hand, the time required for positioning estimation of the terminal to be positioned is greatly reduced, the time required for 1000 times of online positioning in the nine-region sample classification scene is 18.25 seconds, and the time required for 1000 times of online positioning in the unclassified scene is only 11.88 percent of the time required in the unclassified scene, and the method has great significance for practical application scenes because the selection range of the reference point is limited, so that the calculation process of multiple similarity distances is saved.

The fingerprint positioning device provided by the invention is described below, and the fingerprint positioning device described below and the fingerprint positioning method described above can be referred to correspondingly.

Fig. 6 is a schematic structural diagram of a fingerprint positioning device according to the present invention, which may include:

the matrix determining module 601 is configured to determine an angular delay domain channel amplitude matrix at a location of a terminal to be located;

the area determining module 602 is configured to input the angular delay domain channel amplitude matrix into an area determining model, obtain a target area corresponding to the terminal to be positioned output by the area determining model, where the area determining model is obtained based on the angular delay domain channel amplitude matrix sample and the area sample training;

the location determining module 603 is configured to determine location information of the terminal to be located according to the angle delay domain channel amplitude matrix and at least one target angle delay domain channel amplitude matrix sample in the target area.

Optionally, the area determining module 602 is specifically configured to obtain an angle delay domain channel amplitude matrix sample at a location of the reference terminal, and an area sample at which the reference terminal is located; and training the original area determination model by taking the angle time delay domain channel amplitude matrix sample as training data and taking the area sample as a training label to obtain a trained area determination model.

Optionally, the matrix determining module 601 is specifically configured to obtain a channel frequency response matrix of a location where the terminal to be located is located; and determining the angular delay domain channel amplitude matrix according to the channel frequency response matrix.

Optionally, the position determining module 603 is specifically configured to determine, for each target angle delay domain channel amplitude matrix sample, a similarity distance between the angle delay domain channel amplitude matrix and the target angle delay domain channel amplitude matrix sample; and determining the position information of the terminal to be positioned according to at least one similarity distance.

Optionally, the area determining module 602 is specifically configured to normalize the angle delay domain channel amplitude matrix sample by using an input layer in the original area determining model to obtain a first angle delay domain channel amplitude matrix sample; at least one characteristic processing layer in the original area determination model is adopted to perform characteristic processing on the first angle time delay domain channel amplitude matrix sample, so as to obtain a second angle time delay domain channel amplitude matrix sample; determining a full connection layer in a model by adopting the original region, performing feature integration on the second angle time delay domain channel amplitude matrix sample, and determining a prediction region corresponding to the integrated second angle time delay domain channel amplitude matrix sample; and updating the model parameters of the original region determination model according to the integrated second angle time delay domain channel amplitude matrix sample and the prediction region to obtain the trained region determination model.

Optionally, the area determining module 602 is specifically configured to perform, for each of the feature processing layers including a convolution layer and a pooling layer, feature processing on the first angular delay domain channel amplitude matrix sample by using the convolution layer in the current feature processing layer if the current feature processing layer is not the last feature processing layer, to obtain a third angular delay domain channel amplitude matrix sample; compressing the third angle time delay domain channel amplitude matrix sample by adopting a pooling layer in the current feature processing layer to obtain a fourth angle time delay domain channel amplitude matrix sample, wherein the fourth angle time delay domain channel amplitude matrix sample is used for determining the second angle time delay domain channel amplitude matrix sample by the last feature processing layer, and the fourth angle time delay domain channel amplitude matrix sample is input data of the next adjacent feature processing layer of the current feature processing layer; if the current feature processing layer is the last feature processing layer, adopting a convolution layer in the current feature processing layer to perform feature processing on the first angle delay domain channel amplitude matrix sample to obtain a fifth angle delay domain channel amplitude matrix sample; and compressing the fifth angle time delay domain channel amplitude matrix sample by adopting a pooling layer in the current characteristic processing layer to obtain the second angle time delay domain channel amplitude matrix sample, wherein the second angle time delay domain channel amplitude matrix sample is input data of the full-connection layer.

Optionally, the matrix determining module 601 is specifically configured to determine an angular delay domain channel response matrix of the location of the terminal to be located according to the channel frequency response matrix; and determining the angular delay domain channel amplitude matrix according to the angular delay domain channel response matrix.

Optionally, the matrix determining module 601 is specifically configured to determine an angle delay domain channel energy matrix according to the angle delay domain channel response matrix; and determining the angle delay domain channel amplitude matrix according to the angle delay domain channel energy matrix.

Optionally, the location determining module 603 is specifically configured to determine, from the at least one similarity distance, a target similarity distance that meets a preset distance range; determining position estimation coordinates according to the position information of the target similarity distance corresponding to the reference terminal; and determining the position estimation coordinates as the position information of the terminal to be positioned.

As shown in fig. 7, a schematic structural diagram of an electronic device provided by the present invention may include: processor 710, communication interface (Communications Interface) 720, memory 730, and communication bus 740, wherein processor 710, communication interface 720, memory 730 communicate with each other via communication bus 740. Processor 710 may invoke logic instructions in memory 730 to perform a fingerprint positioning method comprising: determining an angle time delay domain channel amplitude matrix of the position of the terminal to be positioned; inputting the angle delay domain channel amplitude matrix into a region determination model to obtain a target region corresponding to the terminal to be positioned, which is output by the region determination model, wherein the region determination model is obtained based on angle delay domain channel amplitude matrix samples and region samples through training; and determining the position information of the terminal to be positioned according to the angle time delay domain channel amplitude matrix and at least one target angle time delay domain channel amplitude matrix sample in the target area.

Further, the logic instructions in the memory 730 described above may be implemented in the form of software functional units and may be stored in a computer readable storage medium when sold or used as a stand alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

In another aspect, the present invention also provides a computer program product comprising a computer program, the computer program being storable on a non-transitory computer readable storage medium, the computer program, when executed by a processor, being capable of performing the fingerprint positioning method provided by the above methods, the method comprising: determining an angle time delay domain channel amplitude matrix of the position of the terminal to be positioned; inputting the angle delay domain channel amplitude matrix into a region determination model to obtain a target region corresponding to the terminal to be positioned, which is output by the region determination model, wherein the region determination model is obtained based on angle delay domain channel amplitude matrix samples and region samples through training; and determining the position information of the terminal to be positioned according to the angle time delay domain channel amplitude matrix and at least one target angle time delay domain channel amplitude matrix sample in the target area.

In yet another aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform the fingerprint positioning method provided by the above methods, the method comprising: determining an angle time delay domain channel amplitude matrix of the position of the terminal to be positioned; inputting the angle delay domain channel amplitude matrix into a region determination model to obtain a target region corresponding to the terminal to be positioned, which is output by the region determination model, wherein the region determination model is obtained based on angle delay domain channel amplitude matrix samples and region samples through training; and determining the position information of the terminal to be positioned according to the angle time delay domain channel amplitude matrix and at least one target angle time delay domain channel amplitude matrix sample in the target area.

The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.

From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.

Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fingerprint positioning method, comprising:

inputting the angle time delay domain channel amplitude matrix into a region determination model to obtain a target region corresponding to the terminal to be positioned, which is output by the region determination model;

determining the position information of the terminal to be positioned according to the angle time delay domain channel amplitude matrix and at least one target angle time delay domain channel amplitude matrix sample in the target area;

wherein the region determination model is trained based on the following steps:

acquiring an angle time delay domain channel amplitude matrix sample of a position of a reference terminal and an area sample of the reference terminal;

an input layer in the model is determined by adopting an original area, and the angle time delay domain channel amplitude matrix sample is normalized to obtain a first angle time delay domain channel amplitude matrix sample;

determining at least one characteristic processing layer in a model by adopting the original region, wherein each characteristic processing layer comprises a convolution layer and a pooling layer, and aiming at each characteristic processing layer, if the current characteristic processing layer is not the last characteristic processing layer, carrying out characteristic processing on the first angle time delay domain channel amplitude matrix sample by adopting the convolution layer in the current characteristic processing layer to obtain a third angle time delay domain channel amplitude matrix sample; compressing the third angle time delay domain channel amplitude matrix sample by adopting a pooling layer in the current feature processing layer to obtain a fourth angle time delay domain channel amplitude matrix sample, wherein the fourth angle time delay domain channel amplitude matrix sample is used for determining a second angle time delay domain channel amplitude matrix sample by the last feature processing layer, and the fourth angle time delay domain channel amplitude matrix sample is input data of a next adjacent feature processing layer of the current feature processing layer; if the current feature processing layer is the last feature processing layer, performing feature processing on the first angle delay domain channel amplitude matrix sample by adopting a convolution layer in the current feature processing layer to obtain a fifth angle delay domain channel amplitude matrix sample; compressing the fifth angle time delay domain channel amplitude matrix sample by adopting a pooling layer in the current characteristic processing layer to obtain the second angle time delay domain channel amplitude matrix sample, wherein the second angle time delay domain channel amplitude matrix sample is input data of the full-connection layer;

Adopting the original region to determine a full connection layer in a model, performing feature integration on the second angle time delay domain channel amplitude matrix sample, and determining a prediction region corresponding to the integrated second angle time delay domain channel amplitude matrix sample;

and updating the model parameters of the original region determination model according to the integrated second angle time delay domain channel amplitude matrix sample and the prediction region to obtain a trained region determination model.

2. The method of claim 1, wherein the determining the angular-delay-domain channel amplitude matrix for the location of the terminal to be located comprises:

acquiring a channel frequency response matrix of the position of the terminal to be positioned;

and determining the angle delay domain channel amplitude matrix according to the channel frequency response matrix.

3. The method of claim 1, wherein the determining the location information of the terminal to be located based on the angle-delay domain channel amplitude matrix and at least one target angle-delay domain channel amplitude matrix sample within the target area comprises:

determining similarity distances between the angle delay domain channel amplitude matrix and the target angle delay domain channel amplitude matrix samples according to the target angle delay domain channel amplitude matrix samples;

And determining the position information of the terminal to be positioned according to at least one similarity distance.

4. The method of claim 2, wherein said determining said angular-delay domain channel magnitude matrix from said channel frequency response matrix comprises:

according to the channel frequency response matrix, determining an angle time delay domain channel response matrix of the position of the terminal to be positioned;

and determining the angle time delay domain channel amplitude matrix according to the angle time delay domain channel response matrix.

5. The method of claim 4, wherein said determining said angular-delay domain channel magnitude matrix from said angular-delay domain channel response matrix comprises:

determining an angle time delay domain channel energy matrix according to the angle time delay domain channel response matrix;

and determining the angle delay domain channel amplitude matrix according to the angle delay domain channel energy matrix.

6. A method according to claim 3, wherein said determining location information of the terminal to be located based on at least one similarity distance comprises:

determining a target similarity distance meeting a preset distance range from the at least one similarity distance;

Determining position estimation coordinates according to the position information of the target similarity distance corresponding to the reference terminal;

and determining the position estimation coordinates as the position information of the terminal to be positioned.

7. A fingerprint positioning device, comprising:

the area determining module is used for inputting the channel amplitude matrix of the angle time delay domain into an area determining model to obtain a target area corresponding to the terminal to be positioned, which is output by the area determining model;

the position determining module is used for determining the position information of the terminal to be positioned according to the angle delay domain channel amplitude matrix and at least one target angle delay domain channel amplitude matrix sample in the target area;

the area determining module is further used for obtaining an angle delay domain channel amplitude matrix sample of the position of the reference terminal and an area sample of the reference terminal;

8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the fingerprint positioning method according to any one of claims 1 to 6 when executing the program.

9. A non-transitory computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when executed by a processor, implements the fingerprint positioning method according to any of claims 1 to 6.