Fingerprint database positioning method based on CWGAN and neural network

Abstract

The invention provides a fingerprint database positioning method based on a CWGAN and a neural network, and the method is applied to a 5G indoor scene, and comprises the steps: S1, training to obtain a first positioning model: S11, obtaining and filtering an off-line fingerprint database, obtaining a filtered off-line fingerprint database, and carrying out the normalization of the filtered off-line fingerprint database, and obtaining a first off-line fingerprint database; step S12, processing the first off-line fingerprint database based on the CWGAN, and forming a second off-line fingerprint database by the generated new fingerprints; step S13, combining the first offline fingerprint database and the second offline fingerprint database to obtain a combined offline fingerprint database; step S14, a CNN-LSTM model is obtained, training of the CNN-LSTM model is completed, and a first positioning model is obtained; s2, collecting fingerprints of the test points and carrying out data processing according to the step S1; and S3, inputting the fingerprint after the test point processing into a first positioning model, and outputting a predicted coordinate value of the test point by the first positioning model. The fingerprint database positioning method based on the CWGAN and the neural network is small in positioning error and high in positioning precision.

Description

Technical Field

[0001] The invention belongs to the field of positioning and navigation, and in particular relates to a fingerprint library positioning method based on CWGAN and a neural network. Background Art

[0002] At present, the development of 5G enables it to support higher-speed and lower-latency positioning, but the high cost of data collection and fluctuations in indoor reception signals are still the main factors restricting 5G's efficient and high-precision positioning in indoor scenarios.

[0003] Due to the presence of obstacles and metal shielding indoors, walls, floors, ceilings and other obstacles will cause signal reflection and scattering, and signal propagation will have multipath effects, showing non-stationary characteristics in time and space dimensions, which will lead to a decrease in indoor positioning accuracy and jitter of positioning results over a period of time. For the above scenarios, fingerprint library matching algorithms using deep learning in indoor positioning scenarios have been widely used, but the training of deep learning models requires a large amount of data, and the manpower and time costs of collecting data are huge, which greatly affects the efficiency of 5G indoor scene positioning.

[0004] Based on the above, this application provides a technical solution to solve the above technical problems. Summary of the invention

[0005] Aiming at the low accuracy, large fluctuation and low efficiency of 5G indoor positioning in the prior art, the present invention provides a fingerprint library positioning method based on CWGAN and neural network, which is applied to 5G indoor scenarios, including: Step S1, training to obtain a first positioning model, including: Step S11, acquiring and filtering an offline fingerprint library to obtain a filtered offline fingerprint library, normalizing the filtered offline fingerprint library to obtain a first offline fingerprint library; the fingerprint constituting the offline fingerprint library is a synchronization signal reference signal received power value SS-RSRP of a primary cell and a neighboring cell of a 5G base station collected by a device in a positioning area; Step S12: Processing the first offline fingerprint library based on CWGAN, wherein the CWGAN learns the characteristic distribution of fingerprints in the first offline fingerprint library and generates new fingerprints similar to the characteristic distribution of fingerprints in the first offline fingerprint library, and the generated new fingerprints constitute a second offline fingerprint library; Step S13: merging the first offline fingerprint library and the second offline fingerprint library to obtain a merged offline fingerprint library; Step S14, obtaining a CNN-LSTM model, inputting the merged offline fingerprint library into the CNN-LSTM model, and when the parameters of the CNN-LSTM model meet the preset conditions, completing the training of the CNN-LSTM model to obtain a first positioning model; Step S2, collecting the fingerprint of the test point and performing data processing according to step S1; Step S3: input the processed fingerprint of the test point into the first positioning model, and the first positioning model outputs the predicted coordinate value of the test point.

[0006] Preferably, obtaining the offline fingerprint library in step S11 includes: Step S111: Divide the positioning area into N grids, the number of 5G base stations in the positioning area is M Set the intersection of each grid as a reference point, obtain the fingerprint of each reference point, and the fingerprint feature vectors of all reference points constitute the offline fingerprint library , ; in, By i A reference point at a certain time t The fingerprint of the SS-RSRP of all 5G base stations received within 1≤ i≤N .

[0007] Preferably, filtering the offline fingerprint library in step S11 includes: Step S112, i The feature vector of the fingerprint of a single sampling of reference points is expressed as ,delete There are vectors with null values ​​in the , and filter out SS-RSRP values ​​higher than the threshold The sample is used as the fingerprint of the reference point ;

[0008] Step S113: Obtaining time t The extreme points of the fingerprint received within the envelope are interpolated by curves according to the extreme points to obtain the upper envelope. and lower envelope , and for the upper envelope and the lower envelope Take the average and get the average envelope , and calculate the remaining component : ; Step S114: Repeat step S113 as a new signal until the difference between the number of extreme points and zero-crossing points is no more than 1, and the average value of the maximum and minimum envelopes at any point is 0. At this time, the remaining component is the first-order eigenmode function component ; Step S115, calculate the first order residual: ; make Alternative , repeat step S113 to step S114 until the residual is monotonic or no longer oscillates; Step S116: Discard the previous p The remaining intrinsic mode function components are accumulated to obtain the fingerprint after filtering of the reference point. , n is the total number of eigenmode function components; Step S117: Merge all reference points Get the filtered offline fingerprint library .

[0009] Preferably, step S11 further includes: Step S118: filter the offline fingerprint database Each column of k The column-normalized SS-RSRP value is expressed as , Indicates k The minimum value of SS-RSRP, Indicates k The maximum value of SS-RSRP in each column is combined to form the normalized first offline fingerprint library. , normed Indicates normalization processing; Step S119: Use one-hot encoding to process the number of each reference point. The corresponding unique hot codes of all reference points: ,in, For the i The one-hot code corresponding to each sample is is the total number of samples.

[0010] Preferably, CWGAN includes a generator and a discriminator, the generator and the discriminator are respectively composed of a multi-layer perceptron MLP and a fully connected FC layer, and step S12 includes: Step S121: The unique hot code C of all reference points is used as a priori condition and compared with the random noise vector z The splicing is used as the input of the generator, and after passing through the MLP and FC layers, the enhanced fingerprint corresponding to the reference point is output. ; Step S122: Obtain discriminator input variables, wherein the discriminator input includes: True fingerprint and one-hot code for sampling C The concatenated vector, the enhanced fingerprint generated by the generator based on random noise With one-hot encoding C The vector composed of splicing; Step S123: After the discriminator input variables are input into the MLP and FC layers, the discriminator outputs a one-dimensional real number score value for the real fingerprint and the enhanced fingerprint.

[0011] Step S124: Calculate the loss function of the discriminator :

[0012] The loss function of the generator :

[0013] in, z It follows a standard normal distribution A random noise vector, represents the distribution of real fingerprints, For distribution The real samples sampled in ; Step S125, update the network parameters through the back propagation algorithm BP iteration so that the loss function and Minimize. When the loss function reaches the preset condition, end the CWGAN training and save the CWGAN model parameters.

[0014] Preferably, step S12 further includes: Step S126, concatenate the one-hot code and random noise vector of each reference point, and input them into CWGAN to obtain an enhanced fingerprint corresponding to the reference point; Step S127: Use the coordinates of the reference points as the coordinates of the corresponding enhanced fingerprints, merge the enhanced fingerprints of all reference points, and form a second offline fingerprint library.

[0015] Preferably, step S13 includes: Based on the collection time and generation order, the first offline fingerprint database And the second offline fingerprint library The length of fingerprint segmentation for each reference point is s The first offline fingerprint library And the second offline fingerprint library Merge to get the merged offline fingerprint library .

[0016] Preferably, the step S14 includes: Step S141: Sample the merged offline fingerprint library to obtain an input sequence , the input sequence is input into the CNN model and the LSTM model respectively, the CNN model outputs a one-dimensional vector, and the LSTM model updates the cell state and hidden state ; Step S142: concatenate the output of the last time step of the LSTM model and the one-dimensional vector, and output the two-dimensional coordinates based on the multi-layer perceptron. ; Step S143: Based on the time back propagation BPTT gradient calculation method, the loss function Solve for the gradient of network parameters; Step S144: Update the model weight matrix and bias parameters based on the Adam optimization algorithm.

[0017] Preferably, step S14 further includes: The loss function is the root mean square error of the positioning coordinates, expressed as: , in, is the batch size, For the i The predicted coordinates of the points , For the i The actual coordinates of the points ; When the loss function When the value does not decrease continuously for more than a preset number of times, it is determined that the CNN-LSTM model parameters meet the preset conditions, and the CNN-LSTM model training is completed.

[0018] The present invention can bring at least one of the following beneficial effects: The present invention proposes a fingerprint library positioning method based on CWGAN and neural network, which is applied to 5G indoor scenarios. The high-frequency components of samples in the fingerprint library are filtered to overcome signal fluctuations, and then the original fingerprint library is processed based on the CWGAN model to automatically expand the number of fingerprints that meet the requirements, generate enhanced fingerprints with similar feature distribution to the original fingerprints, and obtain the enhanced fingerprint library. After merging the filtered fingerprint library with the enhanced fingerprint library, the merged fingerprint library is input into the trained first positioning model, and the first positioning model learns the implicit relationship between fingerprint features and in time sequence and performs positioning. The fingerprint information collected in multiple steps is integrated for positioning, which effectively improves the fluctuation of the positioning result, reduces the positioning error, greatly improves the positioning accuracy, and at the same time reduces the overhead and cost of collecting fingerprint data, thereby improving the positioning efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The preferred implementation scheme will be described below in a clear and understandable manner with reference to the accompanying drawings to further illustrate the above-mentioned characteristics, technical features, advantages and their implementation methods.

[0020] Figure 1A schematic diagram of the steps of a fingerprint library positioning method based on CWGAN and neural network provided in an embodiment of the present invention; Figure 2 A top view of an indoor positioning scene; Figure 3 A schematic diagram of the CWGAN model structure in an embodiment of the present invention; Figure 4 This is a schematic diagram of the CNN-LSTM model structure in an embodiment of the present invention; Figure 5 A schematic diagram comparing the cumulative distribution function of positioning errors of a fingerprint library positioning method based on CWGAN and neural network provided by the present invention and other methods. DETAILED DESCRIPTION

[0021] Various aspects of the present invention are described in further detail below.

[0022] Unless otherwise defined or indicated, all professional and scientific terms used herein have the same meanings as those familiar to those skilled in the art. In addition, any methods and materials similar or equivalent to those described herein can be applied to the present invention.

[0023] The following is an explanation of the terms.

[0024] Unless otherwise clearly specified and limited, the "or" mentioned in the present invention includes the relationship of "and". The "and" is equivalent to the Boolean logic operator "AND", the "or" is equivalent to the Boolean logic operator "OR", and "AND" is a subset of "OR".

[0025] It is to be understood that although the terms "first", "second", etc. may be used herein to describe different elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element. Therefore, the first element may be referred to as the second element without departing from the teachings of the present invention.

[0026] In the present invention, the terms "comprising", "including" or "comprising" indicate that various components can be applied together in the mixture or composition of the present invention. Therefore, the term "consisting mainly of..." is included in the terms "comprising", "including" or "comprising".

[0027] Unless otherwise clearly specified and limited, the terms "connected", "connected" and "connection" in the present invention should be understood in a broad sense, for example, it can be a fixed connection, or it can be connected through an intermediary medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0028] For example, if an element (or component) is referred to as being on, coupled to, or connected to another element, the element may be directly formed on, coupled to, or connected to the other element, or there may be one or more intervening elements between them. In contrast, if the expressions "directly on," "directly coupled to," and "directly connected to" are used herein, then no intervening elements are indicated. Other words used to describe the relationship between elements should be similarly interpreted, such as "between" and "directly between," "attached" and "directly attached," "adjacent" and "directly adjacent," etc.

[0029] It should also be noted that the words "front", "rear", "left", "right", "upper" and "lower" used in the following description refer to directions in the drawings. The words "inner" and "outer" used refer to directions toward or away from the geometric center of a particular component, respectively. It can be understood that these terms are used here to describe the relationship of one element, layer or region relative to another element, layer or region as shown in the drawings. In addition to the orientations described in the drawings, these terms should also include other orientations of the device.

[0030] Other aspects of the present invention will be apparent to those skilled in the art in view of the disclosure herein.

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the specific implementation methods of the present invention will be described below with reference to the accompanying drawings. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings and other implementation methods can be obtained based on these drawings without creative work.

[0032] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present application. The drawings only show the components related to the present application and are not drawn according to the number, shape and size of the components in actual implementation. The type, quantity and proportion of each component in actual implementation may be changed arbitrarily, and the component layout may also be more complicated. For example, the thickness of the components in the drawings may be exaggerated for the sake of clarity.

[0033] Example

[0034] In view of the low accuracy, large fluctuation and low efficiency of 5G indoor positioning in existing technologies, such as Figure 1 As shown, the present invention provides a deep learning fingerprint library positioning method based on CWGAN, which is applied to 5G indoor scenarios, including: Step S11, acquiring and filtering an offline fingerprint library to obtain a filtered offline fingerprint library, normalizing the filtered offline fingerprint library to obtain a first offline fingerprint library; the fingerprint constituting the offline fingerprint library is a synchronization signal reference signal received power value SS-RSRP of a primary cell and a neighboring cell of a 5G base station collected by a device in a positioning area; Step S12: Processing the first offline fingerprint library based on CWGAN, wherein the CWGAN learns the characteristic distribution of fingerprints in the first offline fingerprint library and generates new fingerprints similar to the characteristic distribution of fingerprints in the first offline fingerprint library, and the generated new fingerprints constitute a second offline fingerprint library; Step S13: merging the first offline fingerprint library and the second offline fingerprint library to obtain a merged offline fingerprint library; Step S14, obtaining a CNN-LSTM model, inputting the merged offline fingerprint library into the CNN-LSTM model, and when the parameters of the CNN-LSTM model meet the preset conditions, completing the training of the CNN-LSTM model to obtain a first positioning model; Step S2, collecting the fingerprint of the test point and performing data processing according to step S1; Step S3: input the processed fingerprint of the test point into the first positioning model, and the first positioning model outputs the predicted coordinate value of the test point.

[0035] In a specific embodiment of the present invention, obtaining the offline fingerprint library in step S11 includes: Step S111: Divide the positioning area into N grids, the number of 5G base stations in the positioning area is M Set the intersection of each grid as a reference point, obtain the fingerprint of each reference point, and the fingerprint feature vectors of all reference points constitute the offline fingerprint library , ; in, By i A reference point at a certain time t The fingerprint of the SS-RSRP of all 5G base stations received within 1≤ i≤N .

[0036] In this embodiment, the indoor scene is as follows Figure 2As shown in the figure, the size of the positioning room is 14m×9m, and 6 5G base stations AP1, AP2, AP3, AP4, AP5, and AP6 are deployed. The room is gridded at 1m intervals, and 5G devices are fixed at the intersections of each grid. The RSRP value within the collection time T is used as the fingerprint of each grid intersection. In step S1, filtering the data can effectively overcome the fluctuation of the received signal and improve the positioning accuracy; in step S2, CWGAN means using the conditional Wasserstein generative adversarial network to learn the fingerprint feature distribution of different grids, and generate the corresponding enhanced fingerprint based on the original limited number of data sets, thereby expanding the data set and reducing the manpower consumption and collection cost of data collection.

[0037] In a specific embodiment of the present invention, filtering the offline fingerprint library in step S11 includes: Step S112, i The feature vector of the fingerprint of a single sampling of reference points is expressed as ,delete There are vectors with null values ​​in the , and filter out SS-RSRP values ​​higher than the threshold The sample is used as the fingerprint of the reference point ; Step S113: Obtaining time t The extreme points of the fingerprint received within the envelope are interpolated by curves according to the extreme points to obtain the upper envelope. and lower envelope , and the upper envelope and the lower envelope Take the average and get the average envelope , and calculate the remaining component :

[0038] Step S114: Repeat step S113 as a new signal until the difference between the number of extreme points and zero-crossing points is no more than 1, and the average value of the maximum and minimum envelopes at any point is 0. At this time, the remaining component is the first-order eigenmode function component ; Step S115, calculate the first order residual: ; make Alternative , repeat step S113 to step S114 until the residual is monotonic or no longer oscillates; Step S116: Discard the previous p The remaining intrinsic mode function components are accumulated to obtain the fingerprint after filtering of the reference point. , n is the total number of eigenmode function components; Step S117: Merge all reference points Get the filtered offline fingerprint library .

[0039] In a specific embodiment of the present invention, step S11 further includes: Step S118: filter the offline fingerprint database Each column of k The column-normalized SS-RSRP value is expressed as , Indicates k The minimum value of SS-RSRP, Indicates k The maximum value of SS-RSRP in each column is combined to form the normalized first offline fingerprint library. , normed Indicates normalization processing; Step S119: Use one-hot encoding to process the number of each reference point. The corresponding unique hot codes of all reference points: ,in, For the i The one-hot code corresponding to each sample is is the total number of samples.

[0040] In a specific embodiment, p =2, that is, the first and second order modal components are filtered out.

[0041] Based on the traditional CWGAN network, this network adds the prior condition information of the reference point unique-hot code, so that the network can generate enhanced fingerprints of specified reference points without training a model for each reference point separately, significantly reducing the model training time and enhancing the universality of the model.

[0042] In a specific embodiment of the present invention, Figure 3 As shown, CWGAN includes a generator and a discriminator, the generator and the discriminator are respectively composed of a multi-layer perceptron MLP and a fully connected FC layer, and step S12 includes: Step S121: The grid unique hot code C of all grids is used as a priori condition and compared with the random noise vector z The splicing is used as the input of the generator, and after passing through the MLP and FC layers, the enhanced fingerprint corresponding to the grid is output. ; Step S122: Obtain discriminator input variables, wherein the discriminator input includes: The vector composed of the real fingerprint sampled in the grid and the grid unique hot code C, and the enhanced fingerprint generated by the generator based on random noise The vector formed by concatenating with the grid one-hot code C; Step S123: After the discriminator input variables are input into the MLP and FC layers, the discriminator outputs a one-dimensional real number score value for the real fingerprint and the enhanced fingerprint. or ; Step S124: Calculate the loss function of the discriminator : , The loss function of the generator : , in, z Follows standard normal distribution , represents the distribution of real fingerprints, For distribution The real samples sampled in ; Step S125, update the network parameters through the back propagation algorithm BP iteration so that the loss function and Minimize. When the loss function reaches the third preset condition, end the CWGAN training and save the CWGAN model parameters. In a specific embodiment of the present invention, step S12 further includes: Step S126: concatenate the one-hot encoding and random noise vector of each grid and input them into CWGAN to obtain an enhanced fingerprint of each grid; Step S127: Use the grid point coordinates as the coordinates of the enhanced fingerprint of the corresponding grid, merge the enhanced fingerprints of all grids, and form a second fingerprint library. .

[0043] In a specific application example, the generator network consists of 5 units connected in series, each of which consists of two hidden layers with (30, 20) neurons, the activation function between the hidden layers is LeakyRelu, and the learning rate is 0.001; the discriminator's MLP consists of 2 hidden layers with 50 neurons, the activation function between the hidden layers is LeakyRelu, and the learning rate is 0.0005.

[0044] In a preferred embodiment of the present invention, step S13 includes: Based on the collection time and generation order, the first fingerprint library And the second fingerprint library The fingerprint segmentation length of each grid is sThe first fingerprint library And the second fingerprint library Merge to get the merged fingerprint library .

[0045] Furthermore, the fingerprint database is merged at a ratio of 8:2. Divide into training set and validation set, train the generator and discriminator mentioned above. Expand the data set without manual repeated collection, save manpower and time costs, and improve the prediction accuracy of the model.

[0046] In a specific embodiment of the present invention, the CNN-LSTM model is composed of a one-dimensional convolutional layer (CNN1d) and a long short-term memory network (LSTM), the CNN1d layer is composed of multiple convolution kernels, and the LSTM is composed of multiple LSTM units, each unit includes a cell state (Cell State), a forget gate (Forget Gate), an input gate (Input Gate) and an output gate (OutputGate), and the step S14 includes: Step S141: Sample the merged offline fingerprint library to obtain an input sequence , the input sequence is input into the CNN model and the LSTM model respectively, the CNN model outputs a one-dimensional vector, and the LSTM model updates the cell state and hidden state ; Step S142: concatenate the output of the last time step of the LSTM model and the one-dimensional vector, and output the two-dimensional coordinates based on the multi-layer perceptron. ; Step S143: Based on the time back propagation BPTT gradient calculation method, the loss function Solve for the gradient of network parameters; Step S144: Update the model weight matrix and bias parameters based on the Adam optimization algorithm.

[0047] Specifically, the CNN model outputs a one-dimensional vector including: The convolution kernel performs a one-dimensional convolution operation along the feature dimension, and the output of the convolution layer is expressed as:

[0048] in, is the nth element of the output sequence, x is the input fingerprint sequence, is the kth element of the convolution kernel, and K is the length of the convolution kernel. The output of the convolution layer is passed through the maximum pooling layer to further extract the local features of the receptive field, and then flattened through the flattening layer to obtain a one-dimensional vector.

[0049] Specifically, LSTM updates the cell state and hidden state Includes: In each time step calculation, the forget gate, input gate and output gate are used to update the cell state and hidden state .

[0050] In a specific embodiment of the present invention, step S14 further includes: The loss function is the root mean square error of the positioning coordinates, expressed as: , in, is the batch size, For the i The predicted coordinates of the points , is the real coordinate of the i-th point ; When the loss function When the value is not higher than the preset threshold, the CNN-LSTM model training ends.

[0051] Specifically, in a practical application example, the convolution layer is set to 2 layers, the fingerprint feature dimension is 6, the number of convolution layer channels is 10 and 32 respectively, the sequence length is s=10, and the batch size of the data batching is =128, learning rate =0.001, and the number of neurons in the MLP hidden layer is (128, 256, 128).

[0052] Figure 5 The figure shows a comparison of the cumulative distribution function of positioning errors of the fingerprint library positioning method based on CWGAN and neural network provided by the present invention (yellow) and those of the methods using only multi-layer perceptron MLP (red), only CNN-LSTM model (blue), and only empirical mode decomposition EMD and CNN-LSTM model (green) in a 5G indoor scenario, where the vertical axis is the cumulative distribution of errors and the horizontal axis represents the positioning error in meters.

[0053] In summary, the present invention has achieved the following effects: The present invention proposes a fingerprint library positioning method based on CWGAN and neural network, which is applied to 5G indoor scenarios. The high-frequency components of samples in the fingerprint library are filtered to overcome signal fluctuations, and then the original fingerprint library is processed based on the CWGAN model to automatically expand the number of fingerprints that meet the requirements, generate enhanced fingerprints similar to the original fingerprint feature distribution, obtain an enhanced fingerprint library, merge the first fingerprint library and the second fingerprint library, and input the merged fingerprint library into the trained first positioning model. The first positioning model learns the implicit relationship between fingerprint features and in time sequence and performs positioning. The fingerprint information collected in multiple steps is integrated for positioning, which effectively improves the fluctuation of positioning results, reduces positioning errors, greatly improves positioning accuracy, and reduces the overhead and cost of collecting fingerprint data, thereby improving positioning efficiency.

[0054] Based on this application, it should be understood by those skilled in the art that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number and aspect described herein can be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this device and / or practice this method.

[0055] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.

[0056] It should be noted that the above embodiments can be freely combined as needed. The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered as the protection scope of the present invention.

[0057] All documents mentioned in the present invention are cited as references in this application, just as each document is cited as reference separately. In addition, it should be understood that after reading the above content of the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the claims attached to this application.

Claims

1. A fingerprint library positioning method based on CWGAN and neural network, applied to 5G indoor scenarios, characterized in that: include: Step S1: training to obtain a first positioning model, the step S1 comprises: Step S11, acquiring and filtering an offline fingerprint library to obtain a filtered offline fingerprint library, normalizing the filtered offline fingerprint library to obtain a first offline fingerprint library; the fingerprint constituting the offline fingerprint library is a synchronization signal reference signal received power value SS-RSRP of a primary cell and a neighboring cell of a 5G base station collected by a device in a positioning area; Step S12: Processing the first offline fingerprint library based on CWGAN, wherein the CWGAN learns the characteristic distribution of fingerprints in the first offline fingerprint library and generates new fingerprints similar to the characteristic distribution of fingerprints in the first offline fingerprint library, and the generated new fingerprints constitute a second offline fingerprint library; Step S13: merging the first offline fingerprint library and the second offline fingerprint library to obtain a merged offline fingerprint library; Step S14, obtaining a CNN-LSTM model, inputting the merged offline fingerprint library into the CNN-LSTM model, and when the parameters of the CNN-LSTM model meet the preset conditions, completing the training of the CNN-LSTM model to obtain a first positioning model; Step S2, collecting the fingerprint of the test point and performing data processing according to step S1; Step S3: input the processed fingerprint of the test point into the first positioning model, and the first positioning model outputs the predicted coordinate value of the test point.

2. The fingerprint library positioning method based on CWGAN and neural network according to claim 1, characterized in that: The step S11 of obtaining an offline fingerprint database includes: Step S111: Divide the positioning area into N grids, the number of 5G base stations in the positioning area is M Set the intersection of each grid as a reference point, obtain the fingerprint of each reference point, and the fingerprint feature vectors of all reference points constitute the offline fingerprint library , ; in, By i A reference point at a certain time t The fingerprint of the SS-RSRP of all 5G base stations received within 1≤ i≤N .

3. The fingerprint library positioning method based on CWGAN and neural network according to claim 2 is characterized in that: The filtering of the offline fingerprint library in step S11 includes: Step S112, i The feature vector of the fingerprint of a single sampling of reference points is expressed as ,delete There are vectors with null values ​​in the , and filter out SS-RSRP values ​​higher than the threshold The sample is used as the fingerprint of the reference point ; Step S113: Obtaining time t The extreme points of the fingerprint received within the envelope are interpolated by curves according to the extreme points to obtain the upper envelope. and lower envelope , and for the upper envelope and the lower envelope Take the average and get the average envelope , and calculate the remaining component : ； Step S114: Repeat step S113 as a new signal until the difference between the number of extreme points and zero-crossing points is no more than 1, and the average value of the maximum and minimum envelopes at any point is 0. At this time, the remaining component is the first-order eigenmode function component ; Step S115, calculate the first order residual: ; make Alternative , repeat step S113 to step S114 until the residual is monotonic or no longer oscillates; Step S116: Discard the previous p The remaining intrinsic mode function components are accumulated to obtain the fingerprint after filtering of the reference point. , n is the total number of eigenmode function components; Step S117: Merge all reference points Get the filtered offline fingerprint library .

4. The fingerprint library positioning method based on CWGAN and neural network according to claim 3 is characterized in that: Step S11 also includes: Step S118: filter the offline fingerprint database Each column of k The column-normalized SS-RSRP value is expressed as , Indicates k The minimum value of SS-RSRP, Indicates k The maximum value of SS-RSRP in each column is combined to form the normalized first offline fingerprint library. , normed Indicates normalization processing; Step S119: Use one-hot encoding to process the number of each reference point. The corresponding unique hot codes of all reference points: ,in, For the i The one-hot code corresponding to each sample is is the total number of samples.

5. The fingerprint library positioning method based on CWGAN and neural network according to claim 4 is characterized in that: CWGAN includes a generator and a discriminator, the generator and the discriminator are respectively composed of a multi-layer perceptron MLP and a fully connected FC layer. Step S12 includes: Step S121: The unique hot code C of all reference points is used as a priori condition and compared with the random noise vector z The splicing is used as the input of the generator, and after passing through the MLP and FC layers, the enhanced fingerprint corresponding to the reference point is output. : Step S122: Obtain discriminator input variables, wherein the discriminator input includes: True fingerprint and one-hot code for sampling C The concatenated vector, the enhanced fingerprint generated by the generator based on random noise With one-hot encoding C The vector composed of splicing; Step S123: After the discriminator input variables are input into the MLP and FC layers, the discriminator outputs a one-dimensional real number score value for the real fingerprint and the enhanced fingerprint. or ; Step S124: Calculate the loss function of the discriminator : ， The loss function of the generator : ， in, z It follows a standard normal distribution A random noise vector, represents the distribution of real fingerprints, For distribution The real samples sampled in ; Step S125, update the network parameters through the back propagation algorithm BP iteration so that the loss function and Minimize. When the loss function reaches the preset condition, end the CWGAN training and save the CWGAN model parameters.

6. The fingerprint library positioning method based on CWGAN and neural network according to claim 5 is characterized in that: Step S12 also includes: Step S126, concatenate the one-hot code and random noise vector of each reference point, and input them into CWGAN to obtain an enhanced fingerprint corresponding to the reference point; Step S127: Use the coordinates of the reference points as the coordinates of the corresponding enhanced fingerprints, merge the enhanced fingerprints of all reference points, and form a second offline fingerprint library. .

7. The fingerprint library positioning method based on CWGAN and neural network according to claim 6 is characterized in that: Step S13 includes: Based on the collection time and generation order, the first offline fingerprint database And the second offline fingerprint library The length of fingerprint segmentation for each reference point is s The first offline fingerprint library And the second offline fingerprint library Merge to get the merged offline fingerprint library .

8. The fingerprint library positioning method based on CWGAN and neural network according to claim 7 is characterized in that: The step S14 comprises: Step S141: Sample the merged offline fingerprint library to obtain an input sequence , the input sequence is input into the CNN model and the LSTM model respectively, the CNN model outputs a one-dimensional vector, and the LSTM model updates the cell state and hidden state ; Step S142: concatenate the output of the last time step of the LSTM model and the one-dimensional vector, and output the two-dimensional coordinates based on the multi-layer perceptron. ; Step S143: Based on the time back propagation BPTT gradient calculation method, the loss function Solve for the gradient of network parameters; Step S144: Update the model weight matrix and bias parameters based on the Adam optimization algorithm.

9. The fingerprint library positioning method based on CWGAN and neural network according to claim 8, characterized in that: Step S14 also includes: The loss function is the root mean square error of the positioning coordinates, expressed as: ， in, is the batch size, For the i The predicted coordinates of the points , For the i The actual coordinates of the points ; When the loss function When the value does not decrease continuously for more than a preset number of times, it is determined that the CNN-LSTM model parameters meet the preset conditions, and the CNN-LSTM model training is completed.

Images