CN115201750A - Ultra-wideband positioning system NLOS identification method - Google Patents

Abstract

The invention discloses an NLOS (non line of sight) identification method of an ultra-wideband positioning system, which comprises the steps of acquiring channel impulse response time sequence data of an anchor point or a mobile tag in the UWB system; performing wavelet transformation on the acquired channel impulse response data, and calculating to acquire a time-frequency information graph of the channel impulse response data; constructing a deep neural network, and generating a time-frequency information graph by using the acquired LOS and NLOS channel impulse response data to train the time-frequency information graph to obtain an NLOS identification model; and transplanting the constructed model to a mobile tag hardware platform of the UWB system, and performing NLOS identification and positioning performance verification in an indoor environment. According to the invention, the time-frequency information graph of the channel impulse response time sequence is processed by using the deep neural network, time and frequency information is effectively extracted, and the constructed deep neural network can effectively perform NLOS identification in the UWB indoor positioning system.

Description

Ultra-wideband positioning system NLOS identification method

Technical Field

The invention belongs to the technical field of indoor positioning, and particularly relates to an ultra-wideband positioning system NLOS identification method.

Background

In order to meet the increasing demand of people on high-precision positioning of personnel and assets, the radio positioning technology is rapidly developed in various indoor positioning scenes, and has wide market application prospects. With the increase of human indoor activities, various wireless positioning indoor services such as bluetooth, WIFI, zigBee, near Field Communication (NFC), and UWB are emerging. Among these indoor positioning technologies, UWB is considered to be one of the most promising wireless cooperative indoor positioning solutions. Currently, ultra-wideband is widely applied to indoor positioning systems due to its characteristics of low power consumption, high transmission rate and strong penetrability. Compared to other radio frequency technologies, UWB has a bandwidth greater than 500MHz, extremely short transmit pulses, high temporal and spatial resolution, and considerable multipath error rejection. It differs from existing narrowband communications in that the UWB spectrum is made up of multiple sub-bands, which enables wireless transmission data to propagate through obstacles at relatively high data rates. In addition, the characteristics of UWB narrow pulse and high time resolution are utilized, the distance measurement precision can be effectively improved, and the positioning precision is improved.

Currently, UWB positioning technology has proven to achieve centimeter-level accuracy in line-of-sight (LOS) propagation. However, various obstacles are encountered when using UWB in a practical indoor environment. These obstacles, such as the human body, concrete walls, glass windows, metal sheets and wooden doors, may block and reflect UWB wireless signals, introduce multipath interference and generate NLOS signals. The transmit signal arrives at the receiving end with different degrees of delay under NLOS conditions compared to LOS conditions. Furthermore, when the delay spread is relatively large, the received NLOS signal becomes attenuated and distorted. Therefore, in ultra-wideband positioning applications, it is important to reduce and eliminate errors caused by NLOS propagation.

Current LOS/NLOS identification methods can be classified into two different types based on Channel Impulse Response (CIR) and other signal metrics. For example, the statistical parameter differences of the estimated distance information may be used for NLOS and LOS classification. Distance noise under LOS conditions follows a gaussian distribution with a mean of zero, whereas distance noise under NLOS conditions can be modeled as a gaussian distribution with a non-zero mean. Therefore, hypothesis testing can be performed based on the distance variance and mean difference. The variance of the distance measurements is also considered and a binary hypothesis test is used to determine NLOS conditions. Jo et al combines ray tracing algorithms with statistical-based or map-based methods for NLOS identification. However, these techniques may not be applicable when the tag is moving in an actual positioning system, as it requires multiple measurements of a pair of positions. Furthermore, the detection threshold may vary in different sites and environments.

Another class of CIR-based NLOS identification methods has the main idea that the first path is significantly larger than the energy of the delay path. In previous studies, NLOS identification was performed based on traditional statistical features of the CIR of the received signal, such as kurtosis, peak advance delay, mean excess delay, and Root Mean Square (RMS), which appear to be different values under NLOS and LOS conditions. The discrimination threshold of NLOS varies in different locations and environments. Therefore, it is difficult to identify NLOS under various scenarios using standard statistical parameters. To address this problem, some non-parametric machine learning methods are used for NLOS and LOS classification. For example, NLOS recognition using an Input Vector Machine (IVM) has the advantage of few input vectors with minimal sparsity. However, in UWB positioning systems, as anchor and tag observations increase, their efficiency may be severely reduced. In addition, a Support Vector Machine (SVM) and a correlation vector machine (RVM) are applied to NLOS recognition, and decision boundaries are established according to input features by using the support vector and the correlation vector, respectively, thereby obtaining more accurate and robust NLOS recognition. However, manually selected vector features generated from UWB signal propagation path loss models may not be sufficient to meet the identification requirements in various positioning scenarios. With the development of machine learning, convolutional Neural Networks (CNNs) and long term memory (LSTM) have been demonstrated to exhibit superior performance in ultra-wideband LOS and NLOS time-series data classification. The CNN is used to automatically detect and extract the CIR characteristics of the received UWB signal, and then the CNN output is fed into the LSTM for classification. However, this method does not take into account the difference in spectral information between the received LOS and NLOS signals, and the serial structure design of CNN and LTSM is relatively complex. Simone et al correct under LOS or NLOS conditions by high semantic feature extraction of CIR signals, and realize effective ranging error mitigation at the edge by using deep learning and graph optimization techniques. However, the method only utilizes the time domain information of the received signal, and the NLOS positioning accuracy only reaches the decimeter level.

Disclosure of Invention

Aiming at the problem of NLOS identification of UWB systems by different obstacles such as human bodies, concrete walls, glass windows, metal plates, wooden doors and the like which possibly exist indoors, the invention provides a LOS and NLOS signal identification method based on a CIR time-frequency diagram and a deep neural network model. First, the neural network model takes the time-frequency information graph of the original channel impulse response CIR of the base station and the tag of the UWB positioning system as input. Meanwhile, the original CIR signal input by Continuous Wavelet Transform (CWT) is processed to generate a time-frequency information map, which can be regarded as an a × a pixel image. Then, a deep learning neural network is designed, the time-frequency information graph is used for training the deep learning neural network, and the trained model can be applied to NLOS and LOS signal identification received on a UWB system mobile tag.

The invention discloses an ultra-wideband positioning system NLOS identification method, which comprises the following steps:

acquiring CIR time sequence data of an anchor point or a mobile tag in a UWB system;

performing wavelet transformation on the obtained CIR data, and calculating a time-frequency information graph of the obtained CIR;

constructing a deep neural network, and training the deep neural network by using the acquired LOS and NLOS data to obtain an NLOS identification model;

and transplanting the constructed model to a mobile tag hardware platform of the UWB system to perform NLOS identification in an indoor environment.

Further, each anchor or mobile tag transmit symbol is modeled as:

where P is the amplitude of the emission of a pulse and s (τ) is the shape of a single Gaussian pulse with t _s The N pulses of a cycle constitute a specific frame;

the transmission channel pulses are as follows:

wherein gamma is _m And beta _m Respectively representing the fading coefficient and the time delay of the mth path, so that the received signal is an accumulation of attenuated and delayed transmission signals in several different transmission paths, described as:

where ψ (t) represents a variance of

Additive white gaussian noise.

Further, a time-frequency information map of the CIR is generated using a continuous wavelet transform, which is defined as:

where p (τ) is the received impulse response sequence signal, κ _b,μ (τ) is the scaling and shifting of the basic wavelet κ (τ) expressed as:

wherein b is a scale factor, theta is a time delay factor, a non-analytic Moret wavelet is used as a basic wavelet, and a frequency domain mathematical expression is defined as:

in the formula of omega ₀ The default value is 6.

And finally, drawing a time-frequency information graph according to the value of the continuous wavelet transform coefficient matrix of the input CIR signal.

Further, the deep neural network comprises 4 convolutional layers, 4 max pool layers and 1 flat layer;

firstly, processing a time-frequency information graph of an input CIR into a two-dimensional matrix through pixel mathematical transformation;

initializing through a random core, and performing convolution on an input time-frequency information graph by a convolution layer to extract edge characteristics of the input time-frequency information graph;

in order to reduce the computational complexity, the size of a matrix output by the convolutional layer is optimized by using the maximum pool layer;

after extracting and optimizing the time-frequency information graph characteristics of the CIR, putting the obtained characteristic matrix into a flat layer to be used as a one-dimensional vector for transformation;

to balance the information of the time-frequency information map and the CIR, the output of the flat layer is sent to the hidden layer to reduce the size of the subsequent combination.

Further, the connection between the hidden layers is realized by the following modes:

r _i ＝q _i *e _j +u _i

wherein q is _i And u _i Respectively representing the weight coefficient and the bias coefficient between hidden layers, e _j Is a neuron vector, the neuron vector of the hidden layer is defined by an active function of the rectifying linear unit ReLU as follows:

further, for the final NLOS and LOS binary classification, the LOSs function is calculated as:

where D is the number of samples, f is a label with a value of 0 or 1, and p (f) represents the predicted likelihood; in back propagation, automatically optimizing the weight coefficient and the deviation coefficient of each layer to ensure that the prediction gradually approaches a target; the weight coefficient and the deviation coefficient are updated by:

where x is the learning rate, u ^new And u ^old New and old deviation coefficients, q, respectively ^new And q is ^old Respectively a new weighting factor and an old weighting factor.

The invention has the following beneficial effects:

compared with the existing NLOS classification method only using CIR, the method effectively extracts time and frequency information, efficiently converts the signal type identification problem into the image classification problem, has more advantages for extracting and classifying image features, and can effectively carry out NLOS identification in the UWB indoor positioning system.

Compared with a real indoor advancing route, the indoor UWB positioning method has the advantages that the indoor UWB positioning test is carried out by utilizing the comparison results of different NLOS identification methods, the ranging curve is fitted from the identified measurement value and the low positioning average error, and the NLOS identification task under the complex positioning environment can be completed through further verification, so that the aim of effectively improving the precision of the UWB positioning system is fulfilled.

Drawings

FIG. 1 is an overall flow diagram of the present invention;

FIG. 2 is a partial architecture diagram of the identification method of the present invention;

FIG. 3 is another architectural diagram of a portion of the identification method of the present invention;

FIG. 4 the difference between CIR of LOS and NLOS;

FIG. 5 CIR time-frequency information diagram under LOS;

FIG. 6 CIR time-frequency information diagram under NLOS;

FIG. 7 UWB positioning system results based on least squares LS positioning and weighted minimum WLS two-times positioning.

Detailed Description

The present invention is further described with reference to the drawings, but the present invention is not limited thereto in any way, and any modifications or alterations based on the teaching of the present invention shall fall within the scope of the present invention.

The working flow of NLOS and LOS identification based on CIR time-frequency diagram and deep neural network structure is shown in figures 1-3 and comprises the following steps:

step 1: CIR time series data from anchor or mobile tags in a UWB system is received. UWB-based positioning systems, where each anchor or tag transmitted symbol can be modeled as:

where P is the amplitude of the emission of a pulse and s (τ) is the shape of a single Gaussian pulse with t _s The N pulses of a cycle constitute a particular frame. The transmission channel pulses may be as follows:

in the formula of gamma _m And beta _m Respectively representing the fading coefficient and the time delay of the mth path. Thus, the received signal is an accumulation of attenuated and delayed transmission signals in several different transmission paths, which can be described as:

where ψ (t) represents a variance of

Additive White Gaussian Noise (AWGN).

For ultra-wideband based positioning systems, the location of the tag is unknown, while the locations of at least three base stations are known in advance. In order to measure the distance between the tag and the base station in the positioning work, a time of flight (TOF) method has emerged, which has the advantage that there is no error due to clock synchronization deviation in UWB-based positioning systems. The ranging error is determined by the detection of the first occurrence of the received CIR signal. CIR signals are typically affected by external interference and physical phenomena such as obstacle attenuation and multipath attenuation. Fig. 2 shows the difference between CIRs of LOS and NLOS. By comparing the first path position and the peak path position in the accumulator, the first path and peak path for LOS is much closer than in the case of NLOS. Further, the normalized magnitude of LOS is greater than the normalized magnitude of NLOS. Therefore, LOS and NLOSCIR can be considered as one-dimensional time series having respective characteristics due to the difference in transmission paths.

Step 2: and constructing a CIR signal time-frequency diagram for NLOS and LOS identification. Since the LOS and CIR signals under NLOS conditions are different in nature, it is predicted that the signal of the first path and the signal of the peak path should exhibit different frequencies in the frequency domain. In order to highlight the frequency domain characteristics of the CIR signal and retain the time-varying information of the CIR signal, the invention constructs a CIR signal time-frequency diagram for NLOS and LOS identification.

Continuous Wavelet Transform (CWT) is capable of detecting certain features in the CIR Signal and image, and therefore is chosen to generate a time-frequency Information map of the CIR (see non-patent paper C. Peng, et al, A Noise-Robust Modulation Signal Classification Method Based on circuits Wavelet Transform,5th IEEE Information Technology and mechanics Engineering conference, ITOEC 2020, june 12,2020-June 14,2020, institute of Electrical and Electronics Engineers Inc., chongq, china,2020, pp.745-750.). The continuous wavelet transform is the basis of wavelet analysis theory and can be defined as:

where p (τ) is the received impulse response sequence signal, κ _b,μ (τ) is the scaling and shifting of the basic wavelet κ (τ). Can be expressed as:

wherein b is a scale factor, theta is a time delay factor, a non-analytic Moret wavelet is used as a basic wavelet, and a frequency domain mathematical expression is defined as:

in the formula of omega ₀ The default value is 6. And finally, drawing a time-frequency information graph according to the value of the CWT coefficient matrix of the input CIR signal.

CIR time-frequency information diagrams of LOS and NLOS are shown in FIGS. 4 and 5. Obviously, the time-frequency information diagrams of NLOS and LOS are different in terms of energy propagation at the frequency and time of the same sub-wave. Furthermore, LOS energy is much more concentrated than NLOS energy in the time-frequency information plot due to CIR signal delay and multipath fading. Unlike the features in the CIR, the features in the time-frequency information map are two-dimensional. The invention combines the time and frequency characteristics of the time-frequency information diagram to increase the information provided for LOS and NLOS identification.

And 3, step 3: in order to ensure high feature extraction efficiency, the time-frequency information two-dimensional image feature extraction module of the CIR comprises 4 convolution layers, 4 maximum pool layers and 1 flat layer. Thus, the input time-frequency information graph is processed into a two-dimensional matrix through pixel mathematical transformation. And initializing through a random core, and performing convolution on the input time-frequency information graph by the convolution layer to extract the edge characteristics of the input time-frequency information graph. To reduce computational complexity, the matrix size of the convolutional layer output is optimized using the maximum pool layer. After the time-frequency information graph characteristics of the CIR are extracted and optimized, the obtained characteristic matrix is put into a flat layer and is used as a one-dimensional vector to be transformed. To balance the information of the time-frequency information map and the CIR, the output of the flat layer is sent to the hidden layer to reduce the size of the subsequent combination.

The connection between the hidden layers is realized by the following modes:

r _i ＝q _i *e _j +u _i (7)

wherein q is _i And u _i Respectively representing the weight coefficient and the bias coefficient between the hidden layers. e.g. of the type _j Is a neuron vector, the neuron vector of the hidden layer is defined by an active function of a rectifying linear unit (ReLU) as:

at the output layer, the Sigmoid activation function is defined as:

for the final NLOS and LOS binary classification, the LOSs function can be calculated as:

where D is the number of samples, f is a label with a value of 0 or 1, and p (f) represents the predicted likelihood. In back-propagation, the weighting and bias coefficients for each layer may be automatically optimized to ensure that the prediction approaches the target gradually. The weight coefficient and the bias coefficient may be updated by:

where x is the learning rate. At each training phase, these equations are used to update the hyperparameter and reduce the loss function until its value becomes steady state. Finally, the proposed model-driven prediction is closer to the predetermined target.

In the invention, 4 layers of convolution CNN are designed for constructing a network model for NLOS identification. FIG. 3 depicts the structure of a deep neural network model. The detailed parameters of the neural network are as follows:

TABLE 1 deep neural network model Structure

It is noted that for 4 convolutional layers, the filling method is "same", and the size of the input matrix can be maintained to simplify the operation. The number of convolution kernels is 32, 64, 128 and 128, and the size of the convolution kernel is 3 × 3. To solve the gradient vanishing problem and the short model training time, reLU is used as the activation function for each convolutional layer. A maxporoling layer is placed after each convolutional layer and the sliding window size is set to 2 x 2. Furthermore, the Dropout regularization coefficient for each layer is set to 0.2 to handle the overfitting. The output of the fourth pooling layer is sent to the underlying flattening layer for converting the multidimensional vectors into one-dimensional vector data, which is then fully connected with the hidden layer of 64 cells to reduce the size. The 64 units output by the hidden layer are combined together, and then the output layer connected with the 1 unit classifies the NLOS signal and the LOS signal.

The present invention uses 70000 NLOS and 70000 LOS measurements from different locations in typical indoor obstruction environments (including wooden doors, concrete walls, sheet metal, human body, glass windows, etc.) to construct an overall data set. 105000 samples are randomly selected from the data set for NLOS recognition model training and testing of the deep neural network to reduce the likelihood of overfitting the recognition model at certain specific locations. The selected test data and training data samples were randomly mixed, with the number of training and test data sets being 87500 and 17500. The designed network was trained in 20 rounds. The final overall NLOS recognition rate is 87.45%, and the fact that the NLOS and the LOS can be effectively judged on the received CIR time sequence in the UWB positioning system under the indoor obstacle environment is proved.

Indoor positioning test results and analysis

In order to verify that the positioning error of the UWB system is reduced through NLOS identification, an indoor positioning experiment is carried out, and Least Square (LS) method and Weighted Least Square (WLS) method are compared to solve the moving position of the UWB system label. In the indoor experiment, 6 base stations were arranged and 1 mobile tag was moved within a one-room-one-hall suite of about 50 square meters. The position coordinates of the 6 anchors are: a1 (0,4.93), A2 (1.42,1.62), A3 (4.57,6.22), A4 (6.02,2.72), A5 (4.57,0) and A6 (1.70,0.62). The tag moves along a predefined path as shown in fig. 6 (black straight track). In a UWB-based system, a two-way ranging method is employed to calculate the distance between an anchor and a tag in the UWB system. The received CIRs for these 6 anchors are input to a trained neural network model to classify as NLOS or LOS cases. If there are z base stations classified as NLOS propagation, w ₁ Minimum distance j in tag and base station representing NLOS propagation _{NLOS_min} Corresponding weights, then the remaining z-1 label weights can be determined as:

in the formula j _i Representing the distance between the ith base station and the tag. Similarly, in rIn LOS-propagating base stations, v ₁ Maximum distance j in tag and base station representing LOS propagation _{LOS_max} Corresponding weights, then the remaining r-1 label weights may be determined as:

in the formula j _t Representing the distance between the tth base station and the tag. In the present invention, z + r =6,w is obtained from the number of base stations used and the environment setting empirical value ₁ ＝0.1，v ₁ And =1. Thus, a weight matrix can be obtained:

W＝diag(w ₁ ,w ₂ ,...,w _n ,v ₁ ,v ₂ ,...,v _m ) (15)

fig. 6 shows CIR position estimate data collected for a moving trajectory, where the point set is estimated using WLS, the dashed-dotted trajectory is the curve using WLS and fitted, and the solid-black trajectory is the true moving path. In contrast, using WLS has a higher overall position estimation accuracy because its weights are used for NLOS measurements. The 9 th-order polynomial fitting curve of the position data set obtained by WLS estimation is smooth and almost consistent with an actual path, and the fact that the deep learning neural network can effectively reduce the influence of NLOS on the accurate positioning of the UWB system is proved.

The invention has the following beneficial effects:

compared with the existing NLOS classification method only using CIR, the method effectively extracts time and frequency information, efficiently converts the signal type identification problem into the image classification problem, has more advantages for extracting and classifying image features, and can effectively carry out NLOS identification in the UWB indoor positioning system.

Compared with a real indoor advancing route, the indoor UWB positioning method has the advantages that the indoor UWB positioning test is carried out by utilizing the comparison results of different NLOS identification methods, the ranging curve is fitted from the identified measurement value and the low positioning average error, and the NLOS identification task under the complex positioning environment can be completed through further verification, so that the aim of effectively improving the precision of the UWB positioning system is fulfilled.

The word "preferred" is used herein to mean serving as an example, instance, or illustration. Any aspect or design described herein as "preferred" is not necessarily to be construed as advantageous over other aspects or designs. Rather, use of the word "preferred" is intended to present concepts in a concrete fashion. The term "or" as used in this application is intended to mean an inclusive "or" rather than an exclusive "or". That is, unless specified otherwise or clear from context, "X employs A or B" is intended to include either of the permutations as a matter of course. That is, if X employs A; b is used as X; or X employs both A and B, then "X employs A or B" is satisfied under any of the foregoing instances.

Also, although the disclosure has been shown and described with respect to one or an implementation, equivalent alterations and modifications will occur to others skilled in the art based upon a reading and understanding of this specification and the annexed drawings. The present disclosure includes all such modifications and alterations, and is limited only by the scope of the appended claims. In particular regard to the various functions performed by the above described components (e.g., elements, etc.), the terms used to describe such components are intended to correspond, unless otherwise indicated, to any component which performs the specified function of the described component (e.g., that is functionally equivalent), even though not structurally equivalent to the disclosed structure which performs the function in the herein illustrated exemplary implementations of the disclosure. In addition, while a particular feature of the disclosure may have been disclosed with respect to only one of several implementations, such feature may be combined with one or other features of the other implementations as may be desired and advantageous for a given or particular application. Furthermore, to the extent that the terms "includes," has, "" contains, "or variants thereof are used in either the detailed description or the claims, such terms are intended to be inclusive in a manner similar to the term" comprising.

Each functional unit in the embodiments of the present invention may be integrated into one processing module, or each unit may exist alone physically, or a plurality of or more than one unit are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may also be stored in a computer readable storage medium. The storage medium mentioned above may be a read-only memory, a magnetic or optical disk, etc. Each apparatus or system described above may execute the storage method in the corresponding method embodiment.

In summary, the above-mentioned embodiment is an implementation manner of the present invention, but the implementation manner of the present invention is not limited by the above-mentioned embodiment, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be regarded as equivalent replacements within the protection scope of the present invention.

Claims (6)

1. An identification method of an ultra-wideband positioning system NLOS is characterized by comprising the following steps:

acquiring channel impulse response time sequence data of an anchor point or a mobile tag in a UWB system;

performing wavelet transformation on the acquired channel impulse response data, and calculating to acquire a time-frequency information graph of a channel impulse response sequence;

constructing a deep neural network, and training the deep neural network by using the acquired LOS and NLOS time-frequency information graph data to obtain an NLOS identification model;

and transplanting the constructed model to a mobile tag hardware platform of the UWB system to perform NLOS identification in an indoor environment.

2. The ultra-wideband positioning system NLOS identification method of claim 1, wherein each anchor or mobile tag transmission symbol is modeled as:

wherein P is oneThe amplitude of emission of each pulse, s (τ) being a single Gaussian pulse shape with t _s The N pulses of the cycle form a specific frame;

the transmission channel pulses are as follows:

wherein gamma is _m And beta _m Respectively representing the fading coefficient and the time delay of the mth path, M being the total number of paths, and therefore the received signal is the accumulation of the attenuated and delayed transmission signals in several different transmission paths, described as:

where ψ (t) represents a variance of

Additive white gaussian noise.

3. The method of claim 1, wherein the graph of the acquired impulse shock response time-frequency information is generated using a continuous wavelet transform, the continuous wavelet transform being defined as:

where p (τ) is the received impulse response sequence signal, κ _b,μ (τ) is the scaling and shifting of the basic wavelet κ (τ) expressed as:

wherein b is a scale factor, mu is a shift factor, theta is a time delay factor, a non-analytic Moret wavelet is used as a basic wavelet, and a frequency domain mathematical expression is defined as:

in the formula of omega ₀ The default value is a value of 6 which,

and finally, drawing a time-frequency information graph according to the value of the continuous wavelet transform coefficient matrix of the input CIR signal.

4. The ultra-wideband positioning system NLOS identification method of claim 1, wherein the deep neural network comprises 4 convolutional layers, 4 max-pool layers, and 1 flat layer;

firstly, processing an input time-frequency information graph into a two-dimensional matrix through pixel mathematical transformation;

initializing through a random core, and performing convolution processing on an input time-frequency information graph by a convolution layer to extract edge characteristics of the time-frequency information graph;

in order to reduce the computational complexity, the size of a matrix output by the convolutional layer is optimized by using the maximum pool layer;

after extracting and optimizing the characteristics of the time-frequency information graph, putting the obtained characteristic matrix into a flat layer to be used as a one-dimensional vector for transformation;

to balance the information of the time-frequency information map and the CIR, the output of the flat layer is sent to the hidden layer to reduce the size of the subsequent combination.

5. The method for identifying the NLOS of the ultra-wideband positioning system according to claim 1, wherein the connection between the hidden layers is implemented by:

r _i ＝q _i *e _j +u _i

wherein q is _i And u _i Respectively representing the weight coefficient and the bias coefficient between hidden layers, e _j Is a neuron vector, the neuron vector of the hidden layer is defined by an active function of the rectifying linear unit ReLU as follows:

6. the ultra-wideband positioning system (NLOS) identification method of claim 5, wherein for the final NLOS and LOS binary classification, the cross entropy function using the LOSs function is calculated as:

where D is the number of samples, f is a label with a value of 0 or 1, and p (f) represents the predicted likelihood; in back propagation, automatically optimizing the weight coefficient and the deviation coefficient of each layer to ensure that the prediction is gradually close to a target; the weight coefficient and the deviation coefficient are updated by:

where x is the learning rate, u ^new And u ^old New and old deviation coefficients, q, respectively ^new And q is ^old Respectively a new weighting factor and an old weighting factor.

Images