Method for locating the launch position in non-line-of-sight conditions

Abstract

The present invention discloses a method for locating a transmitting position in a non-line-of-sight situation. By clustering analysis of path loss data and detecting local outliers, the method can effectively identify and mark NLOS areas, improve positioning performance in a non-line-of-sight environment, maintain high positioning accuracy in various complex environments, and is unaffected by multipath effects and obstructions. The method is applicable not only to indoor environments but also to outdoor environments. The electromagnetic fingerprint positioning technology based on received signals does not require the deployment of additional hardware equipment, thereby reducing system costs. The method utilizes existing wireless signal infrastructure, thereby reducing system maintenance and operating costs.

Description

Technical Field

[0001] The present invention relates to the field of electromechanical equipment, and in particular to a method for locating a transmitting position in a non-line-of-sight situation. Background Art

[0002] Due to the openness of radio wave propagation, radio services are very susceptible to interference. Therefore, countries around the world have strict regulations on radio use and are responsible for monitoring and management by specialized functional departments. Traditional radio monitoring systems still have the following problems: (1) The "general practitioner" radio monitoring method based on spectrum engineering is difficult to resolve the contradictions between transmission processing delay and network bandwidth, between antenna bandwidth and antenna gain, and between radio service complexity and single receiver architecture. It lacks spectrum analysis and signal analysis capabilities for different radio services; (2) Focusing on frequency domain data acquisition, the receiver architecture is single and lacks multi-dimensional data acquisition capabilities and knowledge acquisition capabilities such as time domain, spatial domain, and transform domain. As a result, the radio monitoring system has massive data but lacks a complete knowledge system; (3) The direction finding and positioning function is basically ineffective under non-line of sight (NLOS) conditions; (4) When determining whether the emission parameters are compliant through radio administrative law enforcement and supervision and inspection, it is difficult to efficiently complete compliance inspection and compatibility management when faced with massive amounts of heterogeneous data. Therefore, the emission position positioning under non-line of sight conditions is the last difficulty in acquiring radio monitoring knowledge.

[0003] In summary, a method for locating a transmitting position in a non-line-of-sight situation is needed to address the deficiencies in the prior art. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the present invention provides a method for locating a transmitting position in a non-line-of-sight situation, aiming to solve the above-mentioned problems.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for locating a transmitting position in a non-line-of-sight situation, comprising the following steps:

[0006] Step S1: Data collection: select a specific scene where the transmission position needs to be located, select multiple detection points in the scene, and collect feature data at each detection point;

[0007] Step S2: Data preprocessing: converting the collected received signal strength into a relative path loss linear value and performing standardization on the converted data;

[0008] Step S3: Feature extraction, feature extraction is performed through distribution fingerprint, delay fingerprint and spatial spectrum fingerprint, and the corresponding model is established;

[0009] Step S4: Fingerprint generation: combining the distribution fingerprint, delay fingerprint, and spatial spectrum fingerprint to generate a complete emission position fingerprint, and storing the generated emission position fingerprint and its corresponding location information in the fingerprint library;

[0010] Step S5: Fingerprint matching: In the scene where positioning is required, the received signal characteristics are collected in real time, the real-time collected data is converted into the corresponding transmission position fingerprint, the real-time generated transmission position fingerprint is matched with the fingerprints in the fingerprint library, the most similar fingerprint is found, and the transmitter location is determined based on the matching results;

[0011] Step S6: Identify non-line-of-sight, calculate path loss, divide the path loss data into multiple detection zones, cluster the data in each detection zone, identify LOS and NLOS areas, calculate the inter-cluster distance of the clustering results, and determine whether NLOS exists based on a preset threshold;

[0012] Step S7: Output the result, output the final transmitter position, mark the NLOS area in the result, and provide detailed positioning information.

[0013] Optionally, the data collected in step S1 includes but is not limited to received signal strength, multipath structure, and round-trip time of signal propagation.

[0014] Optionally, the distribution fingerprint in step S3 is statistically analyzed in the following manner:

[0015] Step A1: Use the fitdist function to perform kernel density estimation on the standardized data to obtain the distribution function;

[0016] Step A2: Calculate the cumulative distribution function and detect whether the test data meets the distribution characteristics;

[0017] Step A3: Calculate the ratio of the number of data groups that pass the scene to the total number of measurement data groups to obtain the pass rate.

[0018] Optionally, the distribution function in step A2 includes but is not limited to Rayleigh distribution, Rice distribution, Nakagami and Weibull distribution functions;

[0019] The Rayleigh distribution function is equal to the signal intensity x divided by the square of the standard deviation σ, multiplied by the base of the natural logarithm e to the power of negative x squared divided by twice the square of the standard deviation σ;

[0020] The Rice distribution function is calculated as follows: first determine the peak value A and standard deviation σ of the main signal, then calculate the corresponding probability density based on the Rice distribution function formula and the signal strength x, evaluate the probability distribution of the received signal strength, and calculate the bit error rate.

[0021] The Nakagami distribution function is calculated as follows: first determine the shape parameter μ of the fading severity and the second-order moment ω of the distribution function, calculate the probability density of the signal strength x, and evaluate the probability distribution and bit error rate performance of the signal severity;

[0022] The Weibull distribution function is calculated as follows: first determine the shape parameter k and the scale parameter λ, fit the test data in combination with the signal intensity x, calculate the probability density of the signal intensity x, and evaluate the life distribution and reliability analysis.

[0023] Optionally, the model in step S3 is established in the following manner:

[0024] The distribution fingerprint captures the statistical distribution information of the received signal strength and signal-to-noise ratio, performs sliding window processing on the data collected over a period of time, calculates the mean and variance of each window, and establishes a distribution feature model based on the calculation results;

[0025] The delay fingerprint is based on the time difference of arrival of different paths, which characterizes the time characteristics of multipath propagation, detects the time delay corresponding to the peak, and constructs the power delay spectrum model;

[0026] The spatial spectrum fingerprint describes the angle information of the signal arriving from different directions, obtains the angle spectrum, extracts the spatial power spectrum distribution characteristics, detects the arrival angle of the main peak position signal, and establishes the angle spectrum model.

[0027] Optionally, the spatial spectrum fingerprint is calculated using a minimum variance distortion-free response algorithm:

[0028] Step C1: define the input signal vector X(t) and the weight vector W(θ), and calculate the output signal y(t);

[0029] Step C2: Calculate the output power spectrum of the beamformer.

[0030] Optionally, the non-line-of-sight identification in step S6 adopts a radial NLOS region clustering test algorithm and is tested in the following manner:

[0031] Step S61: Set a threshold value, place the base station at the origin, and calculate the path loss value of each path along the radial direction;

[0032] Step S62: Randomly increase the path loss value to simulate a non-line-of-sight scenario, convert the path loss value into a multiple value and normalize it;

[0033] Step S63: clustering the data using the K-means clustering algorithm;

[0034] Step S64: Calculate the inter-class distance and determine the difference between the inter-class distance and the threshold value. If the inter-class distance is greater than the threshold value, NLOS occurs; otherwise, NLOS does not occur.

[0035] Optionally, the non-line-of-sight identification in step S6 adopts a radial NLOS area LOF local outlier factor detection algorithm in the following manner:

[0036] Step D1: Set the preset threshold, input the data to be detected, and select the K value;

[0037] Step D2: Calculate the local outlier factor LOFk using the LOF detection algorithm;

[0038] Step D3: Determine whether it is an outlier based on the comparison of LOFk with the preset threshold. If the local outlier factor LOFk is greater than the preset threshold, it is determined to be an outlier; otherwise, it is determined to be a non-outlier.

[0039] Beneficial effects of the present invention:

[0040] 1. In the present invention, distribution fingerprinting can generate accurate distribution fingerprints by collecting and analyzing the received signal strength distribution at different locations, thereby improving positioning accuracy; delay fingerprinting can more accurately identify the signal propagation path and reduce positioning errors by utilizing the time delay information of multipath propagation; spatial spectrum fingerprinting can further improve positioning accuracy, especially in complex environments, by analyzing the arrival angle and angular spread of the signal;

[0041] 2. In this invention, cluster analysis of path loss data and local outlier factor (LOF) detection can effectively identify and mark NLOS areas, improve positioning performance in non-line-of-sight environments, and maintain high positioning accuracy in various complex environments without being affected by multipath effects and obstructions.

[0042] 3. The present invention is applicable not only to indoor environments but also to outdoor environments. The electromagnetic fingerprint positioning technology based on received signals does not require the deployment of additional hardware equipment, thus reducing the cost of the system and utilizing the existing wireless signal infrastructure to reduce the maintenance and operation costs of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a schematic diagram of a distributed fingerprint simulation of the present invention.

[0044] Figure 2 Schematic diagram of an MVDR algorithm of the present invention.

[0045] Figure 3 Schematic diagram of a radial NLOS region clustering detection algorithm of the present invention.

[0046] Figure 4 This is a flow chart of a radial NLOS region clustering detection algorithm of the present invention.

[0047] Figure 5This is a flow chart of a radial NLOS region LOF local outlier factor detection algorithm of the present invention.

[0048] Figure 6 This is a schematic diagram of a WiFi positioning simulation model of the present invention.

[0049] Figure 7 This is a field strength distribution diagram obtained by transmitting WiFi signals in a room according to the present invention. DETAILED DESCRIPTION

[0050] In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0051] A method for locating a transmitter position in a non-line-of-sight situation, comprising the following:

[0052] Step S1: Data collection: select a specific scene where the transmission position needs to be located, select multiple detection points in the scene, and collect feature data at each detection point;

[0053] Step S2: Data preprocessing: converting the collected received signal strength into a relative path loss linear value and performing standardization on the converted data;

[0054] Step S3: Feature extraction, feature extraction is performed through distribution fingerprint, delay fingerprint and spatial spectrum fingerprint, and the corresponding model is established;

[0055] Step S4: Fingerprint generation: combining the distribution fingerprint, delay fingerprint, and spatial spectrum fingerprint to generate a complete emission position fingerprint, and storing the generated emission position fingerprint and its corresponding location information in the fingerprint library;

[0056] Step S5: Fingerprint matching: In the scene where positioning is required, the received signal characteristics are collected in real time, the real-time collected data is converted into the corresponding transmission position fingerprint, the real-time generated transmission position fingerprint is matched with the fingerprints in the fingerprint library, the most similar fingerprint is found, and the transmitter location is determined based on the matching results;

[0057] Step S6: Identify non-line-of-sight, calculate path loss, divide the path loss data into multiple detection zones, cluster the data in each detection zone, identify LOS and NLOS areas, calculate the inter-cluster distance of the clustering results, and determine whether NLOS exists based on a preset threshold;

[0058] Step S7: Output the result, output the final transmitter position, mark the NLOS area in the result, and provide detailed positioning information.

[0059] Optionally, the data collected in step S1 includes but is not limited to received signal strength, multipath structure, and round-trip time of signal propagation.

[0060] The distribution fingerprint in step S3 is statistically analyzed in the following way:

[0061] Step A1: Use the fitdist function to perform kernel density estimation on the standardized data to obtain the distribution function;

[0062] Step A2: Calculate the cumulative distribution function and detect whether the test data meets the distribution characteristics;

[0063] Step A3: Calculate the ratio of the number of data groups that pass the scene to the total number of measurement data groups to obtain the pass rate.

[0064] The distribution function in step A2 includes but is not limited to Rayleigh distribution, Rice distribution, Nakagami and Weibull distribution functions;

[0065] The Rayleigh distribution function is equal to the signal intensity x divided by the square of the standard deviation σ, multiplied by the base of the natural logarithm e to the power of negative x squared divided by twice the square of the standard deviation σ;

[0066] The Rice distribution function is calculated as follows: first determine the peak value A and standard deviation σ of the main signal, then calculate the corresponding probability density based on the Rice distribution function formula and the signal strength x, evaluate the probability distribution of the received signal strength, and calculate the bit error rate.

[0067] The Nakagami distribution function is calculated as follows: first determine the shape parameter μ of the fading severity and the second-order moment ω of the distribution function, calculate the probability density of the signal strength x, and evaluate the probability distribution and bit error rate performance of the signal severity;

[0068] The Weibull distribution function is calculated as follows: first determine the shape parameter k and scale parameter λ, fit the test data in combination with the signal intensity x, calculate the probability density of the signal intensity x, and evaluate the life distribution and reliability analysis;

[0069] The Rayleigh distribution function is:

[0070] , where σ is the variance;

[0071] The Rice distribution function is:

[0072] , where A is the peak value of the main signal, σ 2is the total power of the multipath signal components, I0(.) is the first kind zero-order Bessel function;

[0073] The Nakagami distribution function is:

[0074] , where μ is the shape parameter representing the severity of fading, ω is the second moment of the distribution function, and Γ(.) is the gamma function;

[0075] The Weibull distribution function is:

[0076] , where r≥0, λ is the scale parameter and λ>0, k is the shape parameter and k>0, representing the severity of the fading.

[0077] The model in step S3 is established in the following way:

[0078] The distribution fingerprint captures the statistical distribution information of the received signal strength and signal-to-noise ratio, performs sliding window processing on the data collected over a period of time, calculates the mean and variance of each window, and establishes a distribution feature model based on the calculation results;

[0079] The delay fingerprint is based on the time difference of arrival of different paths, which characterizes the time characteristics of multipath propagation, detects the time delay corresponding to the peak, and constructs the power delay spectrum model;

[0080] The spatial spectrum fingerprint describes the angle information of the signal arriving from different directions, obtains the angle spectrum, extracts the spatial power spectrum distribution characteristics, detects the arrival angle of the main peak position signal, and establishes the angle spectrum model.

[0081] The delay fingerprint is related to multipath propagation, and the multipath channel model is:

[0082] , where represents the complex gain of the k-th multipath signal, is the baseband representation of the transmitted signal, is the time delay of the signal along the kth path;

[0083] The second-order matrix of the multipath channel model is:

[0084] , where is a second-order matrix, is the mathematical expectation, for exist The complex conjugate of the moment.

[0085] The power spectral density of the delay fingerprint is:

[0086] ,

[0087] ,

[0088] Where S 21 is the equivalent scattering parameter of the network composed of transmitting and receiving antennas, and S(t) is the time domain representation of the transmitted signal.

[0089] The spatial spectrum fingerprint is calculated using the minimum variance distortion-free response algorithm:

[0090] Step C1: Define the input signal vector X(t) and the weight vector W(θ), and calculate the output signal y(t):

[0091] ;

[0092] Step C2: Calculate the output power spectrum of the beamformer,

[0093] ,

[0094] Where R is the covariance matrix of the array output x(t), ;

[0095] In step S6, the non-line-of-sight identification adopts the radial NLOS region clustering test algorithm and is tested in the following way:

[0096] Step S61: Set a threshold value, place the base station at the origin, and calculate the path loss value of each path along the radial direction;

[0097] Step S62: Randomly increase the path loss value to simulate a non-line-of-sight scenario, convert the path loss value into a multiple value and normalize it;

[0098] Step S63: clustering the data using the K-means clustering algorithm;

[0099] Step S64: Calculate the inter-class distance and determine the difference between the inter-class distance and the threshold value. If the inter-class distance is greater than the threshold value, NLOS occurs; otherwise, NLOS does not occur.

[0100] In step S6, non-line-of-sight identification uses the radial NLOS region LOF local outlier factor detection algorithm in the following way:

[0101] Step D1: Set the preset threshold, input the data to be detected, and select the K value;

[0102] Step D2: Calculate the local outlier factor LOFk using the LOF detection algorithm;

[0103] Step D3: Determine whether it is an outlier based on the comparison of LOFk with the preset threshold. If the local outlier factor LOFk is greater than the preset threshold, it is determined to be an outlier; otherwise, it is determined to be a non-outlier.

[0104] like Figure 1 As shown in the figure, regarding the distributed fingerprint, a house with two bedrooms, one living room, and one bathroom is used as an example. The house dimensions are 14.5m × 9.7m. The master bedroom is 7.2m × 5.5m, the second bedroom is 5.6m × 5.5m, and the living room is 14.5m × 4.2m. The wall height is 3m and the door height is 2m. Dipole antennas are used in the master bedroom, second bedroom, and living room, respectively, operating at 2.4GHz and 5.8GHz for transmission. N detection areas are set up in each of these areas for data collection and statistical analysis.

[0105] Twelve, nine, and 16 sets of field strength data were collected in the master bedroom, second bedroom, and living room, respectively. This data was converted into relative path loss linear values ​​(PL_line). The statistical characteristics of the field strength data collected in each of the three rooms were analyzed using the following Matlab process:

[0106] Step 1: Assume that each set of field strength data obtained in each scenario satisfies a specific distribution, and use the fitdist function to perform kernel density estimation on the data to obtain the distribution function. Taking the Rayleigh distribution as an example, pd_ray = fitdist(PL_line,'Rayleigh');

[0107] Step 2: Calculate its cumulative distribution function cdfv = cdf(pd_ray, PL_line) and use KS, x2, and AD to determine whether the test data meets the hypothesized distribution characteristics. Taking KS as an example, Hks = kstest(PL_line, [PL_line, cdfv], alpha), where alpha = 0.05, that is, the confidence interval is 0.95, and the output Hks = 0 and 1 respectively indicate that the hypothesis is passed and failed;

[0108] Step 3: Calculate the ratio of the number of data groups that pass in each scenario to the total number of measured data groups to obtain the pass rate.

[0109] Table 1 shows KS, x 2 Small-scale fading pass rate under three detection methods including AD

[0110]

[0111] From the above table, we can see that in the three simulation scenarios, KS, x 2 The conclusions drawn by the three detection methods, AD, and , are that the probability of field strength data satisfying Weibull distribution is relatively high. Among them, at 2.4GHz, the master bedroom, second bedroom, and living room adopt x 2The Weibull distribution pass rates for detection were 83.33%, 77.78%, and 87.5%, respectively. The Rayleigh distribution pass rate was lower. For example, at 2.4 GHz, the field strength distribution in the living room had a Rayleigh distribution pass rate of 0 when using all three detection methods. For the same room, the average pass rates of the three detection methods generally increased as the frequency band increased.

[0112] like Figure 2 As shown in the figure, the transmission position fingerprint includes distribution fingerprint, delay fingerprint and angle fingerprint. The angle fingerprint is represented by the angle spectrum, which mainly includes the number of wave clusters, arrival angle, angle power spectrum, angle spread and its distribution. It is the key to studying the statistical characteristics of spatial fading of wireless channels, and the minimum variance distortionless response (MVDR) algorithm is the basis for angular power spectrum estimation. The basic idea of ​​the MVDR algorithm is to adjust the network weight coefficient wi(θ) so that the weight vector W (that is, W is equal to a(θk)) is orthogonal to the incoming wave direction vector a(θk), thereby achieving the maximum output y(t)=aH(θk)a(θk)=M.

[0113] ,in,

[0114] ,

[0115] ,

[0116] but, ,

[0117] Calculate the output power spectrum of the beamformer,

[0118] ,

[0119] Where R is the covariance matrix of the array output x(t), ;

[0120] Assume that in the far field there is a desired signal d(t) with a direction of arrival of θd and J interference signals ij(t) (j=1,2,...,J) with directions of arrival of θij. For an M-element equidistant linear array in a plane, the received signal at the mth element can be expressed as:

[0121] ,

[0122] The received signal on each array element is expressed in matrix form, and the following formula is obtained:

[0123] ,

[0124] Where A is the response matrix, the vector composed of J+1 signals is s(t), the noise vector is n(t), the direction vector of the desired signal is a(θd), and the direction vector corresponding to each interference signal is a(θij);

[0125] When the weight vector is w=[w1,w2,...,wM]T, the beamforming output is expressed as:

[0126] ,

[0127] The average power of y(t) is:

[0128] ,

[0129] Among them, R=E[x(t)xH(t)] is the covariance matrix of x(t).

[0130] From the perspective of improving the signal-to-interference ratio and suppressing interference and noise, the optimal weight vector can be calculated as:

[0131] ,

[0132] Table 2 shows the angle expansion of each wave cluster at the transmitter under different frequencies in indoor scenes.

[0133]

[0134] As can be seen from the table above, under the studied scenarios and simulation conditions, the effect of frequency on angular spread is very small.

[0135] like Figure 3 As shown in the figure, the radial NLOS area clustering test algorithm is used. The base station is located at the origin. In the r direction, a ray is calculated every 1° from the base station, with a distance from 1 to rmax kilometers, for a total of 360 rays. The maximum distance rmax is set to 20 km, with a step of 0.01. Assuming f is 100 MHz, hte is 50 m, and hre is 1.5 m, the Okumura model is used to calculate the path loss:

[0136] PL=69.55+26.16lg(f)−13.82lg(hte)−a(hre)+(44.9−6.55lg(hte))lg(d)(4-24)

[0137] Here, a(hre)=8.29(lg(1.54hre))²−1.1. A total of 1991 discrete path loss values ​​were obtained for each path. During the calculation, path loss values ​​were randomly added to simulate random non-line-of-sight (NLOS) scenarios along the radio wave propagation path. For example, assuming that non-line-of-sight scenarios occur at one or two random locations along a path, the simulated path loss along the propagation path is divided into detection zones every 0.25 km. Within each detection zone, the K-means algorithm is used to detect non-line-of-sight (NLOS).

[0138] like Figure 4 As shown in Figure 2, the process of the radial NLOS area clustering test algorithm is as follows:

[0139] Step 1: Calculate the path loss PL(1,2,3,…M) along the radial direction with the transmitter as the center. Divide the M discrete path loss values ​​into N equal parts, with n values ​​in each part.

[0140] Step 2: Convert the path loss PL into a multiple value, perform normalization, and cluster each set of data using the K-means clustering algorithm with a clustering factor of 2;

[0141] Step 3: Because each set of data can only have two situations, LOS and NLOS, the class distance δ is calculated for each set of data that has two categories in the cluster. If the class distance δ is less than the threshold value ξ, it is corrected to one category (i.e., no NLOS occurs); if the class distance δ is greater than the threshold value ξ, it is judged that NLOS occurs.

[0142] like Figure 5 As shown in Figure 1, the radial NLOS region LOF local outlier factor detection algorithm compares each sample point with its neighboring density (assigning each sample point an outlier factor (LOF) that depends on its neighborhood density) to quantify its degree of anomaly. For a sample set D with no identical points, assuming a total of n detection samples, the distance between any two data points can be calculated using Euclidean, Mahalanobis, Hamming, Chebyshev, or spherical distances.

[0143] Among them, the Minkowski distance calculation formula is: , parameter p=2 is Euclidean distance, p=1 is Manhattan distance;

[0144] The Mahalanobis distance calculation formula is: ,

[0145] According to the above distance formula, the distance d(O,P) between the k cluster points P of point O and point O can be calculated. The reachable distance between the k cluster points P of point O and point O is calculated by the following formula:

[0146] , dk(O) is the distance between O and its kth nearest neighbor.

[0147] The local reachability density Pk(O) in the outlier detection algorithm refers to the average reachable distance from all points P in the neighborhood of point O to point O, which is defined as:

[0148] ,

[0149] Nk(O) is the kth distance neighborhood of point O, satisfying: Nk(O)={P∈D / {O}|d(O,P)≤dk(O)}. The smaller the sum of the reachable distances between point O and its neighboring points in the same cluster, the larger Pk(O), and vice versa. Whether point O is an outlier is directly determined by LOF:

[0150] ,

[0151] It can be seen from the above formula that if LOF approaches 1, it means that point O and its neighboring points belong to the same cluster; if LOF is less than 1, it indicates that point O is a dense point; if LOF is greater than 1, point O can be judged as an outlier.

[0152] WiFi positioning:

[0153] like Figure 1 、 6 As shown in Figure 7, it is assumed that WiFi signals are transmitted in sequence at different locations in each room for a total of 5000 times; the monitoring (receiving) antenna is located outdoors, assuming there are 12 different locations in total, then 5000×12 data are collected in each room. The small dots outside the rectangle in the figure represent monitoring locations, numbered 1-12; the boxes where Room1, Room2, Room3, and Room4 are located represent different rooms; the small towers in each room represent transmitting antennas, and a total of 36 different locations are indicated in the figure. It can be seen that the strongest signal area is located in the room where the transmitting source belongs, and the monitoring points close to the room receive the second strongest field strength.

[0154] like Figure 7 As shown, each room's antenna transmits once, and the received power of the surrounding monitoring (receiving) antenna is collected once. When an antenna transmits, the surrounding area receives a total of 12 different received field strengths. Therefore, when the antenna in the kth room transmits, the field strength data generated is:

[0155] ,

[0156] Where Pk represents the receiving field strength matrix of the kth room; pi,j represents the field strength value received by the jth antenna when the WiFi signal is transmitted for the i-th time in the room, where i = 1, 2, ... 5000 and j = 1, 2, ... 12.

[0157] The present invention's distribution fingerprint generates precise distribution fingerprints by collecting and analyzing the received signal strength distribution at different locations, thereby improving positioning accuracy. The delay fingerprint utilizes the time delay information of multipath propagation to more accurately identify the signal propagation path and reduce positioning errors. The spatial spectrum fingerprint further improves positioning accuracy, especially in complex environments, by analyzing the signal's angle of arrival and angular spread.

[0158] Through cluster analysis of path loss data and local outlier factor (LOF) detection, NLOS areas can be effectively identified and marked, improving positioning performance in non-line-of-sight environments. It can maintain high positioning accuracy in various complex environments and is not affected by multipath effects and obstructions.

[0159] Applicable not only to indoor environments but also to outdoor environments, the electromagnetic fingerprint positioning technology based on received signals does not require the deployment of additional hardware equipment, thus reducing the cost of the system and utilizing the existing wireless signal infrastructure, thereby reducing the maintenance and operating costs of the system.

[0160] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions or improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for locating a transmitter position in a non-line-of-sight situation, characterized in that: The following steps are involved: Step S1: Data collection: select a specific scene where the transmission position needs to be located, select multiple detection points in the scene, and collect feature data at each detection point; Step S2: Data preprocessing: converting the collected received signal strength into a relative path loss linear value and performing standardization on the converted data; Step S3: Feature extraction, feature extraction is performed through distribution fingerprint, delay fingerprint and spatial spectrum fingerprint, and the corresponding model is established; Step S4: Fingerprint generation: combining the distribution fingerprint, delay fingerprint, and spatial spectrum fingerprint to generate a complete emission position fingerprint, and storing the generated emission position fingerprint and its corresponding location information in the fingerprint library; Step S5: Fingerprint matching: In the scene where positioning is required, the received signal characteristics are collected in real time, the real-time collected data is converted into the corresponding transmission position fingerprint, the real-time generated transmission position fingerprint is matched with the fingerprints in the fingerprint library, the most similar fingerprint is found, and the transmitter location is determined based on the matching results; Step S6: Identify non-line-of-sight, calculate path loss, divide the path loss data into multiple detection zones, cluster the data in each detection zone, identify LOS and NLOS areas, calculate the inter-cluster distance of the clustering results, and determine whether NLOS exists based on a preset threshold; Step S7: Output the result, output the final transmitter position, mark the NLOS area in the result, and provide detailed positioning information.

2. The method for locating a transmitting position in a non-line-of-sight situation according to claim 1, characterized in that: The data collected in step S1 include but are not limited to received signal strength, multipath structure and round-trip time of signal propagation.

3. The method for locating a transmitting position in a non-line-of-sight situation according to claim 1, characterized in that: The distribution fingerprint in step S3 is statistically analyzed in the following manner: Step A1: Use the fitdist function to perform kernel density estimation on the standardized data to obtain the distribution function; Step A2: Calculate the cumulative distribution function and detect whether the test data meets the distribution characteristics; Step A3: Calculate the ratio of the number of data groups that pass the scene to the total number of measurement data groups to obtain the pass rate.

4. The method for locating a transmitting position in a non-line-of-sight situation according to claim 3, characterized in that: The distribution function in step A2 includes but is not limited to Rayleigh distribution, Rice distribution, Nakagami and Weibull distribution functions; The Rayleigh distribution function is equal to the signal intensity x divided by the square of the standard deviation σ, multiplied by the base of the natural logarithm e to the power of negative x squared divided by twice the square of the standard deviation σ; The Rice distribution function is calculated as follows: first determine the peak value A and standard deviation σ of the main signal, then calculate the corresponding probability density based on the Rice distribution function formula and the signal strength x, evaluate the probability distribution of the received signal strength, and calculate the bit error rate. The Nakagami distribution function is calculated as follows: first determine the shape parameter μ of the fading severity and the second-order moment ω of the distribution function, calculate the probability density of the signal strength x, and evaluate the probability distribution and bit error rate performance of the signal severity; The Weibull distribution function is calculated as follows: first determine the shape parameter k and the scale parameter λ, fit the test data in combination with the signal intensity x, calculate the probability density of the signal intensity x, and evaluate the life distribution and reliability analysis.

5. The method for locating a transmitting position in a non-line-of-sight situation according to claim 1, characterized in that: The model in step S3 is established in the following way: The distribution fingerprint captures the statistical distribution information of the received signal strength and signal-to-noise ratio, performs sliding window processing on the data collected over a period of time, calculates the mean and variance of each window, and establishes a distribution feature model based on the calculation results; The delay fingerprint is based on the time difference of arrival of different paths, which characterizes the time characteristics of multipath propagation, detects the time delay corresponding to the peak, and constructs the power delay spectrum model; The spatial spectrum fingerprint describes the angle information of the signal arriving from different directions, obtains the angle spectrum, extracts the spatial power spectrum distribution characteristics, detects the arrival angle of the main peak position signal, and establishes the angle spectrum model.

6. The method for locating a transmitting position in a non-line-of-sight situation according to claim 1, characterized in that: The spatial spectrum fingerprint is calculated using the minimum variance distortion-free response algorithm: Step C1: define the input signal vector X(t) and the weight vector W(θ), and calculate the output signal y(t); Step C2: Calculate the output power spectrum of the beamformer.

7. The method for locating a transmitting position in a non-line-of-sight situation according to claim 6, characterized in that: In step S6, the non-line-of-sight identification adopts the radial NLOS region clustering test algorithm, and is tested in the following manner: Step S61: Set a threshold value, place the base station at the origin, and calculate the path loss value of each path along the radial direction; Step S62: Randomly increase the path loss value to simulate a non-line-of-sight scenario, convert the path loss value into a multiple value and normalize it; Step S63: clustering the data using the K-means clustering algorithm; Step S64: Calculate the inter-class distance and determine the difference between the inter-class distance and the threshold value. If the inter-class distance is greater than the threshold value, NLOS occurs; otherwise, NLOS does not occur.

8. The method for locating a transmitting position in a non-line-of-sight situation according to claim 7, characterized in that: In step S6, the non-line-of-sight identification adopts the radial NLOS area LOF local outlier factor detection algorithm in the following way: Step D1: Set the preset threshold, input the data to be detected, and select the K value; Step D2: Calculate the local outlier factor LOFk using the LOF detection algorithm; Step D3: Determine whether it is an outlier based on the comparison of LOFk with the preset threshold. If the local outlier factor LOFk is greater than the preset threshold, it is determined to be an outlier; otherwise, it is determined to be a non-outlier.

Images