CN113260044A - CSI fingerprint positioning method, device and equipment based on double-layer dictionary learning - Google Patents

Abstract

The embodiment of the invention provides a CSI fingerprint positioning method, a device and equipment based on double-layer dictionary learning, wherein when an instruction for positioning a target is obtained, channel state information of the target is collected and used as CSI data of the target channel state information; acquiring a first sparse code and a region label of the target CSI data based on a first dictionary learning model obtained through pre-training; acquiring a second sparse code of the first sparse code based on a second dictionary learning model obtained by pre-training and the region label, and acquiring a position label of the second sparse code as a position label of the target by using a classifier in the second dictionary learning model; the classifier is used for classifying according to the difference of the position labels of the sparse coding. The scheme can improve the positioning accuracy and reduce the storage pressure and the data processing complexity.

Description

Method, Apparatus and Equipment for CSI Fingerprint Positioning Based on Double-layer Dictionary Learning

technical field

The present invention relates to the technical field of fingerprint positioning, in particular to a CSI fingerprint positioning method, device and equipment based on double-layer dictionary learning.

Background technique

In wireless communication, the fingerprint positioning method is widely used because it can be positioned without changing the hardware of the device. In specific applications, the Channel State Information (CSI) fingerprint positioning method is based on the principle that the indoor environment is complex, and the signal reflection and refraction form different signal strength information at different positions. , RP), and the mapping relationship between the signal features indicated by the channel state information of the reference node, so that the position labels and signal features in the area are stored in a one-to-one correspondence to obtain a location fingerprint database. In this way, when locating a target such as a certain terminal, the channel state information of the target can be obtained, and the location label matching the channel state information of the target can be determined from the location fingerprint database as the location label of the target in this area, so as to realize the detection of the target. positioning.

However, in order to map richer scene information, the channel state information often contains relatively many channel features. Therefore, the channel state information is easily affected by noise, multipath, and personnel movement, etc., resulting in large fluctuations, resulting in significant differences between the channel state information in the location fingerprint database and the channel state information obtained during positioning, affecting the matching effect and causing The problem of reduced positioning accuracy. Moreover, the dimension of channel state information is relatively high. Therefore, the location fingerprint database constructed by the traditional method is large in scale, which leads to the problems of storage pressure and data processing complexity.

SUMMARY OF THE INVENTION

The purpose of the embodiments of the present invention is to provide a CSI fingerprint positioning method, device and device based on double-layer dictionary learning, so as to achieve the effects of improving positioning accuracy and reducing storage pressure and data processing complexity. The specific technical solutions are as follows:

In a first aspect, an embodiment of the present invention provides a CSI fingerprint positioning method based on double-layer dictionary learning, and the method includes:

When the instruction to locate the target is obtained, the channel state information of the target is collected as the target channel state information CSI data;

Based on the first dictionary learning model obtained by pre-training, the first sparse coding and the region label of the target CSI data are obtained; wherein, the first dictionary learning model is used to perform the analysis of the region to which the CSI data belongs based on the principle of minimizing the reconstruction error. Discriminate, and is a sparse coding model trained by using multiple sample CSI data and the position label of each sample CSI data;

Obtain the second sparse code of the first sparse code based on the pre-trained second dictionary learning model and the region label, and use the classifier in the second dictionary learning model to obtain the second sparse code The location label of , as the location label of the target;

Wherein, the second dictionary learning model is used to enhance the discrimination degree of the second sparse coding based on dictionary atomic local constraints, and uses the first sparse coding of multiple sample CSI data and the position label of each sample CSI data The sparse coding model obtained by training; the classifier is used to classify according to the difference of the sparsely coded position labels.

In a second aspect, an embodiment of the present invention provides a CSI fingerprint positioning device based on double-layer dictionary learning, the device comprising:

an information collection module, configured to collect the channel state information of the target as the target channel state information CSI data when the instruction to locate the target is obtained;

a first coding module, configured to obtain the first sparse coding and the region label of the target CSI data based on the first dictionary learning model obtained by pre-training; wherein, the first dictionary learning model is used to minimize the reconstruction error based on In principle, the region to which the CSI data belongs is discriminated, and it is a sparse coding model trained by using multiple sample CSI data and the position label of each sample CSI data;

The second encoding module is configured to obtain the second sparse encoding of the first sparse encoding based on the second dictionary learning model obtained by pre-training and the region label, and use the classifier in the second dictionary learning model, obtaining the position label of the second sparse encoding as the position label of the target;

Wherein, the second dictionary learning model is used to enhance the discrimination degree of the second sparse coding based on dictionary atomic local constraints, and uses the first sparse coding of multiple sample CSI data and the position label of each sample CSI data The sparse coding model obtained by training; the classifier is used to classify according to the difference of the sparsely coded position labels.

In a third aspect, an embodiment of the present invention provides an electronic device, the electronic device includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; The processor is used for storing the computer program; when the processor is used for executing the program stored in the memory, the above-mentioned first aspect provides the steps of the CSI fingerprint positioning method based on double-layer dictionary learning.

Beneficial effects of the embodiment of the present invention:

In the solution provided by the embodiment of the present invention, the first sparse coding and the region label of the target CSI data are obtained based on the first dictionary learning model obtained by pre-training, and the first dictionary learning model and the region label obtained by pre-training are obtained. The second sparse coding of sparse coding. Therefore, the target CSI data can be sparsely encoded by the dictionary learning model, so as to realize the concise expression and deep feature extraction of high-latitude target CSI data, thereby reducing the noise of the target CSI data and reducing the storage pressure and data processing complexity. . Moreover, through the two-layer dictionary learning model: the first dictionary learning model and the second dictionary learning model, the region labels and the second sparse coding to enhance the discrimination are obtained respectively, so as to ensure that the sparse coding of different positioning regions has high discrimination and different fingerprints. The sparse coding of points, that is, labels of different positions, has a high degree of discrimination, thereby improving the accuracy of localization. And, based on the second sparse coding, the pre-trained classifier is used to obtain the position label of the second sparse coding as the position label of the target; wherein, the classifier is used to classify according to the difference of the sparsely coded position labels. Therefore, the matching process in fingerprint positioning can be simplified to classify the output of the second dictionary learning model, which can reduce the matching process, that is, the complexity of data processing. It can be seen that this solution can improve the positioning accuracy and reduce the storage pressure and data processing complexity.

Of course, not all of the advantages described above are required and achieved to implement any product or method of the present invention.

Description of drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. For those of ordinary skill in the art, other embodiments can also be obtained according to these drawings without creative efforts.

FIG. 1 is a schematic flowchart of a CSI fingerprint positioning method based on double-layer dictionary learning according to an embodiment of the present invention;

2 is an example diagram of an application scenario of a CSI fingerprint positioning method based on double-layer dictionary learning provided by an embodiment of the present invention;

3 is an example diagram of a training process of a dictionary learning model in a CSI fingerprint positioning method based on double-layer dictionary learning provided by an embodiment of the present invention;

4 is a schematic structural diagram of a CSI fingerprint positioning device based on double-layer dictionary learning provided by an embodiment of the present invention;

FIG. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art based on the present application fall within the protection scope of the present invention.

其中，

和

分别代表发射和接受信号向量，而向量

表示加性高斯白噪声，H矩阵就表示信道状态信息，它是每个子载波信道信息的集合，可以利用

和

估计得到。其中H＝[H₁,H₂,...,H_n]^T，N为子载波数，Hi以复数形式出现：H_i＝|H_i|exp{j∠H_i}。其中，|H_i|和∠H_i分别是子载波i的幅度相应和相位相应。由于衰落效应和频率偏移，CSI数据的相位与幅度相比略显嘈杂，因此，利用CSI数据的幅度信息可以降低噪声。Orthogonal Frequency Division Multiplexing (OFDM) is a bandwidth-limited digital multi-carrier modulation method in wireless communication, and has become the most widely used multi-carrier modulation technology. In Wi-Fi networks, OFDM divides the signal into multiple orthogonal sub-channels with different frequencies, where the channel state information reflects the characteristics of the communication link between the transmitter and receiver, including distance, scattering, fading, etc. influence of the signal. Channel status information can be represented as:

in,

and

represent the transmit and receive signal vectors, respectively, while the vector

Represents additive white Gaussian noise, and the H matrix represents the channel state information, which is a collection of channel information for each subcarrier, which can be used

and

estimated. _{Wherein H =} [ ^H ₁ , H _{2 ,} . Among them, |H _i | and ∠H _i are the amplitude response and phase response of subcarrier i, respectively. Due to the fading effect and frequency offset, the phase of the CSI data is slightly noisy compared to the amplitude, therefore, the noise can be reduced by utilizing the amplitude information of the CSI data.

Dictionary learning is also known as sparse representation. It learns a set of atoms and combines them to form a dictionary, so that a given signal can make good use of sparse representation and can be reconstructed by a linear combination of several atoms in the dictionary matrix. Sparse coding refers to the process of obtaining the sparse signal x from the original signal y based on a dictionary D, which is usually expressed as the following constraint optimization objective:

where the equality constraint y=Dx is too strict for the optimization problem, so a smaller threshold can be used to relax the optimization problem. Moreover, in order to make the algorithm more suitable for classification problems, the class label information or classifier parameters can be embedded in the objective function to make the sparse coding more discriminating.

In combination with the above content, an embodiment of the present invention provides a CSI fingerprint positioning method based on double-layer dictionary learning, and the method can be applied to an electronic device that provides a positioning service. In a specific application, the electronic device may specifically include: a desktop computer, a portable computer, a mobile terminal, a wearable device, an Internet TV, a server, and the like.

As shown in FIG. 1, an embodiment of the present invention provides a CSI fingerprint positioning method based on double-layer dictionary learning. The method may include the following steps:

S101, when an instruction to locate the target is obtained, collect channel state information of the target as the target channel state information CSI data.

Specifically, there may be various manners for obtaining the instruction for locating the target. Exemplarily, the instruction for locating the target sent by the target may be received, or when it is detected that there is a target, the instruction for locating the target may be determined to be obtained.

S102, based on the pre-trained first dictionary learning model, obtain the first sparse coding and the region label of the target CSI data.

The first dictionary learning model is used to discriminate the region to which the CSI data belongs based on the principle of minimizing the reconstruction error, and is a sparse coding model trained by using a plurality of sample CSI data and the position label of each sample CSI data.

S103, based on the pre-trained second dictionary learning model and the region label, obtain the second sparse coding of the first sparse coding, and use the classifier in the second dictionary learning model to obtain the position label of the second sparse coding as the target location label.

The second dictionary learning model is used to enhance the discrimination of the second sparse coding based on the local constraints of dictionary atoms, and is a sparse coding obtained by using the first sparse coding of multiple sample CSI data and the position label of each sample CSI data Encoding model; the classifier is used to classify by differences in sparsely encoded location labels.

In the solution provided by the embodiment of the present invention, the target CSI data can be sparsely encoded through a dictionary learning model, so as to realize the concise expression and deep feature extraction of the high-latitude target CSI data, so as to reduce the noise reduction of the target CSI data and reduce the Storage pressure and data processing complexity. Moreover, through the two-layer dictionary learning model: the first dictionary learning model and the second dictionary learning model, the region labels and the second sparse coding to enhance the discrimination are obtained respectively, so as to ensure that the sparse coding of different positioning regions has high discrimination and different fingerprints. The sparse coding of points, that is, labels of different positions, has a high degree of discrimination, thereby improving the accuracy of localization. In addition, the matching process in fingerprint positioning can be simplified to classify the output of the second dictionary learning model, which can reduce the matching process, that is, the complexity of data processing. It can be seen that this solution can improve the positioning accuracy and reduce the storage pressure and data processing complexity.

In an optional embodiment, the above-mentioned first dictionary learning model is obtained by training the following steps:

Obtain sample data based on a plurality of sample CSI data and the position label of each sample CSI data;

The sample data, the specifications of the dictionary learning model, and the initial parameters of the dictionary learning model are input into the objective function of the first dictionary learning model, and the objective function of the first dictionary learning model is iteratively trained:

When the objective function of the first dictionary learning model converges, the trained objective function of the first dictionary learning model is used as the first dictionary learning model; the convergence includes: the number of iterations reaches the first iteration threshold, or the objective function of the current iteration is equal to Between the objective functions of the previous iteration, the difference of the output results is less than the first difference threshold;

When the objective function of the first dictionary learning model does not converge, the sample data is used to obtain an updated first sparse code, and the updated first sparse code is used to update the trained objective function of the first dictionary learning model.

In an optional embodiment, the above-mentioned second dictionary learning model is obtained by training the following steps:

Obtain sample data based on a plurality of sample CSI data and the position label of each sample CSI data;

The first sparse coding of the sample data, the specifications of the dictionary learning model, and the initial parameters of the dictionary learning model are input into the objective function of the first dictionary learning model, and the objective function of the first dictionary learning model is iteratively trained:

When the objective function of the second dictionary learning model converges, the trained objective function of the second dictionary learning model is used as the second dictionary learning model; the convergence includes: the number of iterations reaches the second iteration threshold, or the objective function of the current iteration is equal to Between the objective functions of the previous iteration, the difference of the output results is less than the second difference threshold;

When the objective function of the second dictionary learning model does not converge, use multiple sample CSI data and the graph Laplacian matrix of the position label of each sample CSI data to obtain the updated second sparse code, and use the updated first The second sparse coding updates the objective function of the trained second dictionary learning model.

Illustratively, as shown in FIG. 2 . The CSI fingerprint positioning method based on double-layer dictionary learning provided by the embodiment of the present invention can be regarded as including two stages: the first stage is an offline training stage, and the second stage is an online positioning stage. The above two optional embodiments are the offline training stage, and the embodiment of FIG. 1 of the present invention is the online positioning stage. In the offline training phase, the training data is Y, the dictionary size is K, and the parameters are α, τ, λ ₁ , λ ₂ , θ. so:

更新所训练的第一字典学习模型的目标函数。其中，X⁽¹⁾为更新后的第一稀疏编码，D⁽¹¹⁾为所训练的第一字典学习模型的目标函数，D⁽¹²⁾为更新后的所训练的第一字典学习模型的目标函数。For the first dictionary learning model: the sparse coding can be updated by X ⁽¹⁾ = (D ^(11)T D ⁽¹¹⁾ +τI+τΛ) ⁻¹ D ^(11)T Y; by

Update the objective function of the trained first dictionary learning model. Wherein, X ⁽¹⁾ is the updated first sparse coding, D ⁽¹¹⁾ is the objective function of the trained first dictionary learning model, D ⁽¹²⁾ is the updated target of the trained first dictionary learning model function.

Learning the model for the second dictionary:

获得图拉普拉斯矩阵；通过Z⁽²¹⁾＝(D^(21)TD⁽²¹⁾+λ₁L)^-1D^(21)Tx获取更新的第二稀疏编码；通过D⁽²²⁾＝X⁽²¹⁾Z^(21)T(Z⁽²¹⁾Z^(21)T+Δ)^-1更新第二字典学习模型中的字典，以及通过

更新分类器参数。able to pass

Obtain the graph Laplacian matrix; obtain the updated second sparse code by Z ⁽²¹⁾ = (D ^(21)T D ⁽²¹⁾ +λ ₁ L) ^-1 D ^(21)T x; by D ⁽²²⁾ =X ⁽²¹⁾ Z ^(21)T (Z ⁽²¹⁾ Z ^(21)T +Δ) ^-1 to update the dictionary in the second dictionary learning model, and by

Update the classifier parameters.

为第i个类别为c类的样本CSI数据，b_c为支持向量机的第c类偏差。Among them, L is the graph Laplacian matrix, D ⁽²¹⁾ is the objective function of the trained second dictionary learning model, Δ is a diagonal matrix, and the elements on the diagonal are Lagrange multipliers, D ⁽²²⁾ is the dictionary in the updated second dictionary learning model, Z ⁽²¹⁾ is the updated second sparse coding, U', b' are the updated classifier parameters, n is the number of input sample CSI data, l is the loss function, u _c is the c- _th hyperplane parameter of the Support Vector Machine (SVM), zi is the second sparse coding of the i-th sample CSI data,

is the sample CSI data with the i-th category as the c-type, and b _c is the c-th type deviation of the support vector machine.

In an optional embodiment, the above-mentioned acquisition of sample data based on a plurality of sample CSI data and the position label of each sample CSI data may specifically include the following steps:

Image each sample CSI data to obtain multiple CSI images;

For each CSI image, obtain the original average amplitude of each pixel point of the CSI image, and perform dispersion normalization on the original average amplitude of each pixel point to obtain a first normalized CSI image;

Based on the first normalized CSI image, sample data is obtained.

Exemplarily, based on the first normalized CSI image, there may be various ways to obtain sample data. In an optional implementation manner, the first standardized CSI image may be directly used as sample data. In another optional real-time manner, the above-mentioned obtaining sample data based on the first standardized CSI image may specifically include the following steps:

For each first normalized CSI image, using the original average amplitude, dispersion normalization is performed on the corresponding pixel points in the first normalized CSI image to obtain sample data.

Using the CSI data trace to enhance the positioning effect, the CSI of each reference point (Reference Point, RP) is collected multiple times. For example, at the i-th RP, we group H CSI amplitude measurements from W subcarriers to construct a H×W matrix:

是在第i个RP的第h个时间戳中，来自第w个子载波的CSI幅度特征。由于字典学习在应对图像分类任务上具有出色性能，因此，可以将CSI数据图像化。首先应用一个标准化过程从CSI数据矩阵中得到更清晰的数据，也就是计算每个图象在每个位置点l_i处的原始平均振幅：in,

is the CSI amplitude feature from the wth subcarrier in the hth timestamp of the ith RP. Due to the excellent performance of dictionary learning on image classification tasks, CSI data can be visualized. First apply a normalization process to get clearer data from the CSI data matrix, that is to calculate the raw average amplitude of each image at each position _li :

其中，A_i为每个图象在每个位置点l_i处的原始平均振幅。

where A _i is the original average amplitude of each image at each location point _li .

与最小值

则第h行第w列的单个元素的标准化结果为

Use dispersion normalization on each row of the CSI image: for the CSI image at position l _i , record the maximum value of each row separately

with the minimum value

Then the normalized result of a single element in the hth row and the wth column is

Finally, the normalized CSI magnitude image is

And, in order to maintain the original output power of the image, the _CSI image of each location point li can be normalized again using the previously calculated average amplitude:

为位置点l_i的再一次标准化后的CSI图像，

为位置点l_i的标准化后的CSI图像，A_max为所有位置点中CSI数据的最大幅值。in,

is the normalized CSI image of the position point l _i again,

is the normalized CSI image of the position point _{li, and Amax is} the maximum amplitude of the CSI data in all the position points.

In an optional implementation manner, the above-mentioned first dictionary learning model is:

为字典原子约束项，f(D_l)为非相干促进项，用于增强属于不同区域的CSI数据之间的区分度。Wherein, D ⁽¹⁾ is the first dictionary in which the sub-dictionaries are aggregated, Z ⁽¹⁾ is the first sparse coding of the target CSI data, n is the number of regions corresponding to the first dictionary, and D _l is In the first dictionary, the sub-dictionary belongs to the region l, X _l is the sparse coding of the CSI data Y _l collected in the region l on the corresponding sub-dictionary D _l , α, τ are constant parameters,

is the dictionary atomic constraint term, and f(D _l ) is the incoherent promotion term, which is used to enhance the discrimination between CSI data belonging to different regions.

In an optional implementation manner, the above-mentioned second dictionary learning model is:

为字典原子局部约束项，D⁽²⁾为第二字典，

为支持向量判别项，用于区分属于不同位置标签的稀疏编码，u_c是与支持向量判别项所代表的支持向量的第c类超平面相关联的法向量，b_c是与支持向量对应的偏差。Among them, Z ⁽²⁾ is the second sparse coding of CSI data, U, b are both classifier parameters, λ ₁ and λ ₂ are two constant parameters,

is the dictionary atomic local constraint term, D ⁽²⁾ is the second dictionary,

is the support vector discriminant item, used to distinguish sparse coding belonging to different position labels, u _c is the normal vector associated with the c-th hyperplane of the support vector represented by the support vector discriminant item, b _c is the corresponding support vector deviation.

Illustratively, as shown in FIG. 3 . The objective function in the training process is similar to the above model, the difference is that the parameters of the objective function are continuously updated during the training process, until the objective function converges, the parameters of the objective function are the same as those of the above model. Therefore, both the first dictionary learning model and the second dictionary learning model described above are objective functions of the converged dictionary learning model. In order to reduce the burden on data processing brought by the location fingerprint database containing a large number of fingerprint points, in the first dictionary learning model, region-specific sub-dictionaries are trained. In order to make the sparse coding of CSI magnitudes of different partitions have a high degree of discrimination, the following incoherent promotion terms are first introduced:

表示X_l的互补矩阵，即在所有的训练数据中排除X_l本身。其中，X_l和分区l是对应的，也就是说训练数据Y_l可以很好地被X_l表示而不是X_j,(l≠j)。如下式所示，

在优化过程中要尽可能的小，从而确保D_lX_j和Y_l并不接近，而这一点将显著增加指纹点分区时的判别效果：Wherein, D1 corresponds to the sub-dictionary of the region ₁ , and _X1 is the sparse coding of the CSI training data Y1 collected in the region ₁ on the corresponding sub-dictionary D1 _.

Represents the complementary matrix of X _l , i.e. excludes X _l itself in all training data. Among them, X _l and partition l are corresponding, that is to say, the training data Y _l can be well represented by X _l instead of X _j , (l≠j). As shown in the following formula,

In the optimization process, it should be as small as possible to ensure that D _l X _j and Y _l are not close, which will significantly increase the discriminative effect of fingerprint point partitioning:

And, in order to ensure that X _l is as sparse as possible, in order to reduce the computational time-consuming caused by l ₀ -norm regularization or l ₁ -norm regularization for sparse coding in most existing models, in order to train an efficient region For the classification model, in the embodiment of the present invention, the l _{2,1 -norm} is used to ensure that the lines of the coding coefficients are sparse, and the calculation of the l _{2,1 -norm} is relatively easy. Finally, the objective function that defines the first dictionary learning model is:

为字典原子约束项，使第一字典学习模型的计算过程保持稳定。第一字典学习模型对输入的CSI数据进行区域判别，而不涉及更细节的指纹匹配，因此不额外训练分类器，而是以重构误差的最小化作为衡量标准来进行区域判别：

也就是说，y_new根据它的稀疏编码来分区的，具体方法是将其分配给能够得到最小化重构误差的对象类。综上，在双层学习模型中，第一字典学习模型可以通过引入一个非相干促进项来保证CSI数据在对应区域子字典上的稀疏表示能力。in,

For dictionary atomic constraints, the calculation process of the first dictionary learning model is kept stable. The first dictionary learning model performs regional discrimination on the input CSI data without involving more detailed fingerprint matching, so no additional training of the classifier is performed, but the minimization of the reconstruction error is used as the criterion for regional discrimination:

That is, y _new is partitioned according to its sparse encoding by assigning it to the object class that minimizes the reconstruction error. To sum up, in the two-layer learning model, the first dictionary learning model can ensure the sparse representation ability of CSI data on the corresponding region sub-dictionary by introducing an incoherent promotion term.

并且，考虑到CSI数据的敏感性及不稳定性，引入字典原子局部约束项，它可以进一步增强第二字典学习模型对CSI数据流形结构描述的真实性。最终，提出第二层的目标函数如下：In addition, in order to make sparse coding also have the ability to locate fingerprints, the first sparse coding of the first dictionary learning model can be used as the input of the second dictionary learning model, and a support vector discriminant term can be introduced to force the sparse coding from different fingerprint points. can be separated, which is specifically defined as:

Moreover, considering the sensitivity and instability of CSI data, the dictionary atomic local constraint term is introduced, which can further enhance the authenticity of the second dictionary learning model's description of the CSI data manifold structure. Finally, the objective function of the proposed second layer is as follows:

In this way, the sparse encoding of the first dictionary learning model is input into the second dictionary learning, and the location label is predicted through the classifier parameters U and b.

Corresponding to the above method embodiments, the embodiments of the present invention further provide a CSI fingerprint positioning device based on double-layer dictionary learning.

As shown in FIG. 4 , a structure of a CSI fingerprint positioning device based on double-layer dictionary learning provided by an embodiment of the present invention includes:

The information collection module 401 is configured to collect the channel state information of the target as the target channel state information CSI data when the instruction to locate the target is obtained;

The first coding module 402 is configured to obtain the first sparse coding and the region label of the target CSI data based on the first dictionary learning model obtained by pre-training; wherein, the first dictionary learning model is used for minimizing reconstruction based on The error principle is used to discriminate the region to which the CSI data belongs, and is a sparse coding model trained by using multiple sample CSI data and the position label of each sample CSI data;

The second encoding module 403 is configured to acquire the second sparse encoding of the first sparse encoding based on the second dictionary learning model obtained by pre-training and the region label, and use the classifier in the second dictionary learning model , obtain the position label of the second sparse encoding as the position label of the target;

Wherein, the second dictionary learning model is used to enhance the discrimination degree of the second sparse coding based on dictionary atomic local constraints, and uses the first sparse coding of multiple sample CSI data and the position label of each sample CSI data The sparse coding model obtained by training; the classifier is used to classify according to the difference of the sparsely coded position labels.

In the solution provided by the embodiment of the present invention, the target CSI data can be sparsely encoded through a dictionary learning model, so as to realize the concise expression and deep feature extraction of the high-latitude target CSI data, so as to reduce the noise reduction of the target CSI data and reduce the Storage pressure and data processing complexity. Moreover, through the two-layer dictionary learning model: the first dictionary learning model and the second dictionary learning model, the region labels and the second sparse coding to enhance the discrimination are obtained respectively, so as to ensure that the sparse coding of different positioning regions has high discrimination and different fingerprints. The sparse coding of points, that is, labels of different positions, has a high degree of discrimination, thereby improving the accuracy of localization. In addition, the matching process in fingerprint positioning can be simplified to classify the output of the second dictionary learning model, which can reduce the matching process, that is, the complexity of data processing. It can be seen that this solution can improve the positioning accuracy and reduce the storage pressure and data processing complexity.

Optionally, the first dictionary learning model is obtained by training the following steps:

obtaining sample data based on the plurality of sample CSI data and the position label of each sample CSI data;

The sample data, the specifications of the dictionary learning model, and the initial parameters of the dictionary learning model are input into the objective function of the first dictionary learning model, and the objective function of the first dictionary learning model is iteratively trained:

When the objective function of the first dictionary learning model converges, use the trained objective function of the first dictionary learning model as the first dictionary learning model; the convergence includes: the number of iterations reaches a first iteration threshold, or Between the objective function of the current iteration and the objective function of the previous iteration, the difference between the output results is less than the first difference threshold;

When the objective function of the first dictionary learning model does not converge, use the sample data to obtain an updated first sparse code, and use the updated first sparse code to set the target of the trained first dictionary learning model function to update.

An embodiment of the present invention further provides an electronic device, as shown in FIG. 5 , including a processor 501 , a communication interface 502 , a memory 503 and a communication bus 504 , wherein the processor 501 , the communication interface 502 , and the memory 503 pass through the communication bus 504 complete communication with each other,

a memory 503 for storing computer programs;

When the processor 501 is used to execute the program stored in the memory 503, the following steps are implemented:

When the instruction to locate the target is obtained, the channel state information of the target is collected as the target channel state information CSI data;

Based on the first dictionary learning model obtained by pre-training, the first sparse coding and the region label of the target CSI data are obtained; wherein, the first dictionary learning model is used to perform the analysis of the region to which the CSI data belongs based on the principle of minimizing the reconstruction error. Discriminate, and is a sparse coding model trained by using multiple sample CSI data and the position label of each sample CSI data;

Obtain the second sparse code of the first sparse code based on the pre-trained second dictionary learning model and the region label, and use the classifier in the second dictionary learning model to obtain the second sparse code The location label of , as the location label of the target;

Wherein, the second dictionary learning model is used to enhance the discrimination degree of the second sparse coding based on dictionary atomic local constraints, and uses the first sparse coding of multiple sample CSI data and the position label of each sample CSI data The sparse coding model obtained by training; the classifier is used to classify according to the difference of the sparsely coded position labels.

In the solution provided by the embodiment of the present invention, the target CSI data can be sparsely encoded through a dictionary learning model, so as to realize the concise expression and deep feature extraction of the high-latitude target CSI data, so as to reduce the noise reduction of the target CSI data and reduce the Storage pressure and data processing complexity. Moreover, through the two-layer dictionary learning model: the first dictionary learning model and the second dictionary learning model, the region labels and the second sparse coding to enhance the discrimination are obtained respectively, so as to ensure that the sparse coding of different positioning regions has high discrimination and different fingerprints. The sparse coding of points, that is, labels of different locations, has a high degree of discrimination, thereby improving the accuracy of localization. In addition, the matching process in fingerprint positioning can be simplified to classify the output of the second dictionary learning model, which can reduce the matching process, that is, the complexity of data processing. It can be seen that this solution can improve the positioning accuracy and reduce the storage pressure and data processing complexity.

The communication bus mentioned in the above electronic device may be a peripheral component interconnect standard (Peripheral Component Interconnect, PCI) bus or an Extended Industry Standard Architecture (Extended Industry Standard Architecture, EISA) bus or the like. The communication bus can be divided into an address bus, a data bus, a control bus, and the like. For ease of presentation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

The communication interface is used for communication between the above electronic device and other devices.

The memory may include random access memory (Random Access Memory, RAM), and may also include non-volatile memory (Non-Volatile Memory, NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

The above-mentioned processor can be a general-purpose processor, including a central processing unit (Central Processor Unit, CPU), a network processor (Network Processor, NP), etc.; it can also be a digital signal processor (Digital SignalProcessor, DSP), application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), Field-Programmable Gate Array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

In another embodiment provided by the present invention, a computer-readable storage medium is also provided, and a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, any of the above-mentioned two-layer-based Steps of a dictionary-learned CSI fingerprinting method.

In yet another embodiment provided by the present invention, there is also provided a computer program product including instructions, which, when running on a computer, enables the computer to execute any of the two-layer dictionary learning-based CSI fingerprint positioning methods in the above-mentioned embodiments .

In the above-mentioned embodiments, it may be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented in software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer may be a general purpose computer, special purpose computer, computer network, or other programmable device. The computer instructions may be stored in or transmitted from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions may be downloaded from a website site, computer, server or data center Transmission to another website site, computer, server, or data center is by wire (eg, coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (eg, infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that includes an integration of one or more available media. The usable media may be magnetic media (eg, floppy disks, hard disks, magnetic tapes), optical media (eg, DVD), or semiconductor media (eg, Solid State Disk (SSD)), among others.

It should be noted that, in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any relationship between these entities or operations. any such actual relationship or sequence exists. Moreover, the terms "comprising", "comprising" or any other variation thereof are intended to encompass a non-exclusive inclusion such that a process, method, article or device that includes a list of elements includes not only those elements, but also includes not explicitly listed or other elements inherent to such a process, method, article or apparatus. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in a process, method, article or apparatus that includes the element.

Each embodiment in this specification is described in a related manner, and the same and similar parts between the various embodiments may be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the apparatus and device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and reference may be made to some descriptions of the method embodiments for related parts.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims (10)

1. a CSI fingerprint positioning method based on double-layer dictionary learning, is characterized in that, described method comprises: When the instruction to locate the target is obtained, the channel state information of the target is collected as the target channel state information CSI data; Based on the first dictionary learning model obtained by pre-training, the first sparse coding and the region label of the target CSI data are obtained; wherein, the first dictionary learning model is used to perform the analysis of the region to which the CSI data belongs based on the principle of minimizing the reconstruction error. Discriminate, and is a sparse coding model trained by using multiple sample CSI data and the position label of each sample CSI data; Obtain the second sparse code of the first sparse code based on the pre-trained second dictionary learning model and the region label, and use the classifier in the second dictionary learning model to obtain the second sparse code The location label of , as the location label of the target; Wherein, the second dictionary learning model is used to enhance the discrimination degree of the second sparse coding based on dictionary atomic local constraints, and uses the first sparse coding of multiple sample CSI data and the position label of each sample CSI data The sparse coding model obtained by training; the classifier is used to classify according to the difference of the sparsely coded position labels. 2. method according to claim 1, is characterized in that, described first dictionary learning model, adopts following steps to train to obtain: obtaining sample data based on the plurality of sample CSI data and the position label of each sample CSI data; The sample data, the specifications of the dictionary learning model, and the initial parameters of the dictionary learning model are input into the objective function of the first dictionary learning model, and the objective function of the first dictionary learning model is iteratively trained: When the objective function of the first dictionary learning model converges, use the trained objective function of the first dictionary learning model as the first dictionary learning model; the convergence includes: the number of iterations reaches a first iteration threshold, or Between the objective function of the current iteration and the objective function of the previous iteration, the difference between the output results is less than the first difference threshold; When the objective function of the first dictionary learning model does not converge, use the sample data to obtain an updated first sparse code, and use the updated first sparse code to set the target of the trained first dictionary learning model function to update. 3. method according to claim 1, is characterized in that, described second dictionary learning model, adopts following steps to train to obtain: obtaining sample data based on the plurality of sample CSI data and the position label of each sample CSI data; The first sparse coding of the sample data, the specifications of the dictionary learning model, and the initial parameters of the dictionary learning model are input into the objective function of the first dictionary learning model, and the objective function of the first dictionary learning model is iteratively trained: When the objective function of the second dictionary learning model converges, the objective function of the trained second dictionary learning model is used as the second dictionary learning model; the convergence includes: the number of iterations reaches the second iteration threshold, or Between the objective function of the current iteration and the objective function of the previous iteration, the difference between the output results is less than the second difference threshold; When the objective function of the second dictionary learning model does not converge, use the plurality of sample CSI data and the graph Laplacian matrix of the position label of each sample CSI data to obtain an updated second sparse code, and The objective function of the trained second dictionary learning model is updated using the updated second sparse coding. 4. The method according to any one of claims 2 to 3, wherein the acquiring sample data based on the plurality of sample CSI data and the position label of each sample CSI data comprises: Image each sample CSI data to obtain multiple CSI images; For each CSI image, obtain the original average amplitude of each pixel point of the CSI image, and perform dispersion normalization on the original average amplitude of each pixel point to obtain a first normalized CSI image; The sample data is obtained based on the first normalized CSI image. 5. The method according to claim 4, wherein the obtaining the sample data based on the first standardized CSI image comprises: For each first normalized CSI image, using the original average amplitude, perform dispersion normalization on the corresponding pixel points in the first normalized CSI image to obtain the sample data. 6. The method according to claim 1, wherein the first dictionary learning model is:

Wherein, D ⁽¹⁾ is the first dictionary in which the sub-dictionaries are aggregated, Z ⁽¹⁾ is the first sparse coding of the target CSI data, n is the number of regions corresponding to the first dictionary, and D _l is In the first dictionary, the sub-dictionary belongs to the region l, X _l is the sparse coding of the CSI data Y _l collected in the region l on the corresponding sub-dictionary D _l , α, τ are constant parameters,

is the dictionary atomic constraint term, and f(D _l ) is the incoherent promotion term, which is used to enhance the discrimination between CSI data belonging to different regions. 7. The method according to claim 1, wherein the second dictionary learning model is:

Among them, Z ⁽²⁾ is the second sparse coding of CSI data, U, b are both classifier parameters, λ ₁ and λ ₂ are two constant parameters,

is the dictionary atomic local constraint term, D ⁽²⁾ is the second dictionary,

is the support vector discriminant term, used to distinguish sparse coding belonging to different position labels, u _c is the normal vector associated with the c-th hyperplane of the support vector represented by the support vector discriminant term, b _c is the Bias corresponding to the support vector. 8. A CSI fingerprint positioning device based on double-layer dictionary learning, wherein the device comprises: an information collection module, configured to collect the channel state information of the target as the target channel state information CSI data when the instruction to locate the target is obtained; a first coding module, configured to obtain the first sparse coding and the region label of the target CSI data based on the first dictionary learning model obtained by pre-training; wherein, the first dictionary learning model is used to minimize the reconstruction error based on In principle, the region to which the CSI data belongs is discriminated, and it is a sparse coding model trained by using multiple sample CSI data and the position label of each sample CSI data; The second encoding module is configured to obtain the second sparse encoding of the first sparse encoding based on the second dictionary learning model obtained by pre-training and the region label, and use the classifier in the second dictionary learning model, obtaining the position label of the second sparse encoding as the position label of the target; Wherein, the second dictionary learning model is used to enhance the discrimination degree of the second sparse coding based on dictionary atomic local constraints, and uses the first sparse coding of multiple sample CSI data and the position label of each sample CSI data The sparse coding model obtained by training; the classifier is used to classify according to the difference of the sparsely coded position labels. 9. The device according to claim 8, wherein the first dictionary learning model is obtained by training the following steps: obtaining sample data based on the plurality of sample CSI data and the position label of each sample CSI data; The sample data, the specifications of the dictionary learning model, and the initial parameters of the dictionary learning model are input into the objective function of the first dictionary learning model, and the objective function of the first dictionary learning model is iteratively trained: When the objective function of the first dictionary learning model converges, use the trained objective function of the first dictionary learning model as the first dictionary learning model; the convergence includes: the number of iterations reaches a first iteration threshold, or Between the objective function of the current iteration and the objective function of the previous iteration, the difference between the output results is less than the first difference threshold; When the objective function of the first dictionary learning model does not converge, use the sample data to obtain an updated first sparse code, and use the updated first sparse code to set the target of the trained first dictionary learning model function to update. 10. An electronic device, comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; memory for storing computer programs; The processor is configured to implement the method steps of any one of claims 1-7 when executing the program stored in the memory.

Images