NLOS identification method based on GMM-PSO-EM algorithm

Abstract

The invention discloses an NLOS identification method based on a GMM-PSO-EM algorithm, and the method comprises the steps: carrying out the preprocessing and noise reduction of time delay data, determining the distribution number through the combination of AIC and BIC criteria, iteratively updating GMM parameters through the PSO-SA-EM algorithm, building a GMM model, and determining an NLOS probability value in a scene through a region area determined by a PDF value. The GMM-PSO-EM algorithm-based NLOS identification method has excellent NLOS identification performance, has relatively high effectiveness and robustness when processing an NLOS problem, can solve the problems of relatively high NLOS identification complexity, need of a large amount of data marking work and relatively poor performance under the condition of relatively small data volume in a complex indoor environment, and is relatively simple to implement, relatively high in calculation speed and relatively high in accuracy. The method is especially suitable for the recognition of the NLOS signal in the positioning process of the UWB system in a complex indoor environment.

Description

Technical Field

[0001] The invention relates to a method for identifying wireless signals propagated in non-line-of-sight (NLOS), in particular to a NLOS identification method based on a GMM-PSO-EM algorithm, and belongs to the technical field of ultra-wideband. Background Art

[0002] Positioning technology is widely used in modern life. Although there are various positioning methods, in indoor environments, ultra-wideband (UWB) positioning technology has become a popular choice due to its advantages of high precision, low power consumption and anti-multipath interference. UWB positioning mainly relies on ranging technology. In complex environments with multiple scenes, due to factors such as building structures and obstacles, the signal will encounter strong multipath effects during propagation, that is, the signal will reach the receiving end after reflection, refraction and scattering through multiple different paths, resulting in many problems such as signal delay spread, fading and phase change. At the same time, the non-line of sight (NLOS) propagation phenomenon between the positioning base station and the target to be positioned is extremely significant and the problem is extremely prominent, which seriously interferes with the positioning system's accurate judgment of the target position. Therefore, in the UWB positioning system, it is very important to accurately identify the NLOS signal. Correct NLOS identification can not only significantly improve the positioning accuracy, but also enhance the robustness of the system, so as to adapt to various complex indoor multi-scene environments.

[0003] In the indoor NLOS identification problem based on UWB, firstly, due to the difference in delay and intensity between NLOS signals and LOS (Line Of Sight) signals, NLOS signals are easily submerged by background clutter and noise. Secondly, the delay and intensity changes of NLOS signals are much smaller than those of LOS signals. The characteristics of NLOS signals and LOS signals may have similar patterns and amplitudes, which makes it difficult to distinguish NLOS signals from LOS signals by conventional time domain or frequency domain filtering methods. Therefore, how to effectively separate NLOS signals is a key issue in the industry. Among the existing NLOS identification methods, algorithms based on channel impulse response require high computational complexity and need to be recalibrated or adjusted in different environments. Although machine learning-based algorithms have a high recognition rate, they require sample data collection in advance and are more complicated to apply on site. The judgment method based on geometric features identifies NLOS conditions by analyzing the geometric features of signal propagation, which is sensitive to environmental changes and may result in inaccurate identification in complex environments. Algorithms based on deep learning and transfer learning require a large amount of training data and may need to retrain the model in new environments to obtain better performance. The statistical analysis method identifies NLOS conditions by performing statistical analysis on the resulting range measurements in complex indoor environments. Since a variety of physical effects may distort the direct path LOS signal in different ways, multi-class identification becomes challenging. Summary of the invention

[0004] In view of the problems existing in the above-mentioned prior art, the present invention provides a NLOS identification method based on the GMM-PSO-EM algorithm, which can solve the problems of high complexity of NLOS identification in complex indoor environments, the need for a large amount of data labeling work, and poor performance when the data volume is small. The implementation is relatively simple and the calculation speed is faster. It is particularly suitable for the identification of NLOS signals in the UWB system positioning process in complex indoor environments.

[0005] To achieve the above purpose, the NLOS identification method based on the GMM-PSO-EM algorithm specifically includes the following steps:

[0006] Step 1, determine the number of Gaussian distributions in GMM based on AIC and BIC;

[0007] Step 2, use the PSO-SA-EM algorithm to determine the parameters of the distribution of each component of GMM and obtain the GMM model;

[0008] Step 3, collect the delay data of LOS and NLOS scenarios;

[0009] Step 4, fitting the time delay GMM model of LOS and NLOS;

[0010] Step 5, calculate the PDF value of each time-delay data, use the geometric method to determine the probability that the current propagation is NLOS, and perform NLOS identification;

[0011] The NLOS probability value calculation method is as follows:

[0012]

[0013] Where: PDF NLOS and PDF LOS are the probability density function values ​​of given delay values ​​under NLOS model and LOS model respectively; Area NLOS and Area LOS are the geometric area obtained by square of edge probability values ​​based on NLOS and LOS models; TotalArea is the sum of non-overlapping area of ​​NLOS and LOS; Probability NLOS is the NLOS probability value;

[0014] If PDF NLOS and PDF LOS If the values ​​are equal, it is considered that there is no overlapping area and the output NLOS probability value is 50%; otherwise, it is considered that there is an overlapping area and the TotalArea is calculated; the NLOS probability value Probability is calculated by dividing the NLOS non-overlapping area by the TotalArea NLOS , the probability value greater than or equal to 50% is identified as NLOS, otherwise it is LOS.

[0015] Furthermore, Step 1 is as follows:

[0016] Step 1-1, introduce the histogram to visualize the peak information and determine a rough distribution range;

[0017] Step 1-2, use AIC and BIC to make judgments, draw the curves of AIC and BIC within the distribution range, and smooth the curves;

[0018] Step 1-3, select the value in the curve that minimizes the AIC or BIC value as the distribution quantity.

[0019] Furthermore, the PSO-SA-EM algorithm in Step 2 is as follows:

[0020] Step 2-1, initialization:

[0021] ① Initialize the particle swarm size N and randomly assign each particle’s position pBest(0) and velocity v i (0);

[0022] ② Randomly initialize the individual optimal pBest(0) and the global optimal gBest(0) for each particle;

[0023] ③ Set the initial temperature T(0) and cooling rate α of the SA algorithm;

[0024] ④Determine the convergence conditions and number of iterations of the EM algorithm;

[0025] Step 2-2, determine the annular domain: assign a neighborhood to each particle i, and the leading particle x lead Index: (i+1) mod N, lagging particle x lag Index of: (i-1+n) mod N;

[0026] Step 2-3 Determine the optimal neighborhood: Compare the individual optimal solutions of the leading particle and the lagging particle, and determine the optimal neighborhood solution nBest;

[0027] Step 2-4, PSO-SA iteration, perform the following steps for each particle:

[0028] ① Use the PSO algorithm to update the position and velocity of the particles, as shown below:

[0029] v i (t+1)=w·v i (t)+c1·r1·(pBest i -x i (t))+c2·r2·(gBest-x i (t))

[0030] x i (t+1)=x i (t)+v i (t+1)

[0031] Where: v i (t) is the velocity of the ith particle at time t; x i (t) is the position of the ith particle at time t; w is the inertia weight used to balance the exploration and development capabilities of the particle; c1 is the cognitive coefficient that affects the movement of the particle to the individual optimal position; c2 is the social coefficient that affects the movement of the particle to the global optimal position; r1 and r2 are random numbers between 0 and 1;

[0032] ②Neighborhood optimal guidance: Calculate the current particle position x i The difference Δx between (t+1) and the best neighbor nBest is expressed as follows:

[0033] Δx=nBest-x i (t+1)

[0034] Update the particle position x according to the difference Δx and step size λ i (t+1)', expressed as follows:

[0035] x i (t+1)'=x i (t+1)+λΔx

[0036] Where: x i (t+1) is the particle position before updating;

[0037] ③ Use the probability acceptance criterion of the SA algorithm to decide whether to accept the new solution: calculate the fitness of the new position and compare it with the individual optimality of the current particle, and then accept the new position with probability according to the SA criterion to decide whether to accept the new solution. The calculation formula of the SA criterion is as follows:

[0038]

[0039] Where: ΔE is the difference between the fitness of the new position and the optimal fitness of the current individual; T(t) is the current temperature;

[0040] ④ Update individual optimality and global optimality;

[0041] ⑤ Update the temperature in simulated annealing, as shown below:

[0042] T(t+1)=T(t) * α

[0043] Where: α is the cooling rate;

[0044] Step 2-5, initialize the EM algorithm and use the optimal particle position found by the PSO-SA algorithm as the initial parameters of the EM algorithm;

[0045] Step 2-6, execute the E step and M step of the EM algorithm until convergence;

[0046] Step 2-7, output the final GMM parameters.

[0047] Furthermore, Step 4 is as follows:

[0048] For a given data point x, the PDF calculation formula is as follows:

[0049]

[0050] Where: k is the weight of the kth component; is the Gaussian distribution of the kth component; μ k and σ k are the mean and variance of each Gaussian distribution, respectively;

[0051] According to the above formula, the distribution quantity and parameters in the NLOS and LOS cases are obtained, and the GMM model in the NLOS scenario and the GMM model in the LOS scenario are obtained.

[0052] Compared with the existing technology, the NLOS identification method based on the GMM-PSO-EM algorithm first pre-processes and de-noises the delay data, and determines the distribution quantity by combining the AIC and BIC criteria, and then iteratively updates the GMM parameters through the PSO-SA-EM algorithm. After the GMM model is established, the NLOS probability value in the scene is determined by combining the geometric method. The experimental results show that the NLOS identification method based on the GMM-PSO-EM algorithm has excellent NLOS identification performance, has strong effectiveness and robustness in dealing with NLOS problems, and is relatively simple to implement and has a faster calculation speed. It is particularly suitable for the identification of NLOS signals during the positioning process of UWB systems in complex indoor environments. In addition, by combining the distribution models of NLOS and LOS with the geometric method, this method can output the current NLOS probability value, which can provide a new perspective and idea for improving positioning accuracy in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a flow chart of the present invention;

[0054] Figure 2 is a flow chart of the NLOS identification method of the present invention;

[0055] Figure 3 It is the EM algorithm fitting graph;

[0056] Figure 4 It is the fitting diagram of the PSO-SA-EM algorithm of the present invention;

[0057] Figure 5 It is a comparison diagram of NLOS and LOS scene fitting of the present invention. DETAILED DESCRIPTION

[0058] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0059] like Figure 1 As shown, the NLOS identification method based on the GMM-PSO-EM algorithm specifically includes the following steps:

[0060] Step 1, determine the number of Gaussian distributions in the Gaussian Mixture Module (GMM) according to the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC).

[0061] Step 1-1, first, in order to determine the range of Gaussian mixture distribution, the histogram visualization peak information is introduced to determine a rough range of distribution quantity.

[0062] Step 1-2, then use AIC and BIC to make judgments. AIC and BIC are both criteria for statistical model selection, which are used to evaluate and compare the goodness of fit of different statistical models. Draw the curves of AIC and BIC within the distribution range, and then smooth the curves to reduce the impact of original data noise.

[0063] Step 1-3, finally select the value in the curve that minimizes the AIC or BIC value as the distribution quantity.

[0064] Step 2, use particle swarm optimization (PSO) and simulated annealing (SA) to improve the expectation maximization algorithm (EM), determine the parameters of the distribution of each component of GMM, and obtain the GMM model.

[0065] Since the EM algorithm is prone to fall into local optimality, the convergence speed may be slow when the initial parameters are far away from the optimal solution. Therefore, the PSO-SA algorithm is introduced to provide better initial parameters for the EM algorithm, thereby accelerating the convergence speed and improving the quality of the final solution. In addition, the probabilistic acceptance mechanism of SA allows the algorithm to accept suboptimal solutions with a certain probability during the search process, which helps to escape the local optimal trap, thereby improving the GMM model's ability to update parameters and fitting effect.

[0066] The EM algorithm after integrating PSO-SA is referred to as PSO-SA-EM. The algorithm steps are described as follows:

[0067] Step 2-1, initialization:

[0068] ① Initialize the particle swarm size N and randomly assign each particle’s position pBest(0) and velocity v i (0).

[0069] ② Randomly initialize the individual optimal pBest(0) and the global optimal gBest(0) for each particle.

[0070] ③ Set the initial temperature T(0) and cooling rate α of the SA algorithm.

[0071] ④Determine the convergence conditions and number of iterations of the EM algorithm.

[0072] Step 2-2, determine the annular domain: assign a neighborhood to each particle i, and the leading particle xlead Index: (i+1) mod N, lagging particle x lag The index of: (i-1+n)mod N, which determines that the particle group forms a closed loop, in which the lagging particle of the first particle is the last particle, and the leading particle of the last particle is the first particle.

[0073] Step 2-3 Determine the neighborhood optimal solution: Compare the individual optimal solutions of the leading particle and the lagging particle, and determine the neighborhood optimal solution nBest.

[0074] Step 2-4, PSO-SA iteration, perform the following steps for each particle:

[0075] ① Use the PSO algorithm to update the position and velocity of the particles, as shown below:

[0076] v i (t+1)=w·v i (t)+c1·r1·(pBest i -x i (t))+c2·r2·(gBest-x i (t))

[0077] x i (t+1)=x i (t)+v i (t+1)

[0078] Where: v i (t) is the velocity of the ith particle at time t; x i (t) is the position of the ith particle at time t; w is the inertia weight, which is used to balance the exploration and development capabilities of the particle; c1 is the cognitive coefficient, which affects the movement of the particle to the individual optimal position; c2 is the social coefficient, which affects the movement of the particle to the global optimal position; r1 and r2 are random numbers between 0 and 1, which are used to simulate randomness.

[0079] ②Neighborhood optimal guidance: Calculate the current particle position x i The difference Δx between (t+1) and the best neighbor nBest is expressed as follows:

[0080] Δx=nBest-x i (t+1)

[0081] Update the particle position x according to the difference Δx and step size λ i (t+1)', expressed as follows:

[0082] x i (t+1)'=x i (t+1)+λΔx

[0083] Where: x i (t+1) is the particle position before updating.

[0084] ③ Use the probability acceptance criterion of the SA algorithm to decide whether to accept the new solution: calculate the fitness of the new position and compare it with the individual optimality of the current particle, and then accept the new position with probability according to the SA criterion to decide whether to accept the new solution. The calculation formula of the SA criterion is as follows:

[0085]

[0086] Where: ΔE is the difference between the fitness of the new position and the optimal fitness of the current individual; T(t) is the current temperature.

[0087] ④Update individual optimum and global optimum.

[0088] ⑤ Update the temperature in simulated annealing, as shown below:

[0089] T(t+1)=T(t)*α

[0090] Where: α is the cooling rate.

[0091] Step 2-5, initialize the EM algorithm and use the optimal particle position found by the PSO-SA algorithm as the initial parameters of the EM algorithm.

[0092] Step 2-6, execute the E step (calculate expectation) and M step (maximization) of the EM algorithm until convergence.

[0093] Step 2-7, output the final GMM parameters.

[0094] Step 3: Collect delay data for line-of-sight and non-line-of-sight scenarios.

[0095] In order to collect indoor time delay and other special environmental parameter data, the embodiment adds a circuit board that can complete sensor data collection and Wifi communication on the basis of Haoru Technology's LD150-I model UWB tag. The overall data collection and communication joint circuit board includes a power module, an MCU module, a DWM1000 and an IMU positioning module, a sensor acquisition module, a Wifi communication module, and an over-limit alarm module. The MCU uses the STM32F103C8T6 model microcontroller, the gas sensor module model is MQ-4, the carbon monoxide sensor module model is MQ-7, the temperature sensor model is DS18B20, and the Wifi module model is ESP8266. The functions realized are to collect harmful gas concentration, temperature, time delay, and positioning data. The data storage formats are TEMP, ppm_mq7, ppm_mq4, POSI_clock, POSI_distance(0, 1, 2...), POSI_accelX, POSI_accelY, POSI_accelZ, POSI_gyroX, POSI_gyroY, POSI_gyroZ, POSI_tof(0, 1, 2...), POSI_position_(x, y, z), which respectively represent temperature, carbon monoxide concentration, gas concentration, clock, distance from each base station to the tag, three-axis speed, three-axis acceleration, time delay from each base station to the tag, and position coordinates solved by the tag.

[0096] Step 4: Fit the Gaussian mixture distribution model of LOS and NLOS delays.

[0097] The probability density function of GMM can be expressed as the weighted sum of the probability density functions of each component. Therefore, for a given data point x, the probability density function (PDF) is calculated as follows:

[0098]

[0099] Where: k is the weight of the kth component; is the Gaussian distribution of the kth component; μ k and σ k are the mean and variance of each Gaussian distribution, respectively.

[0100] In this way, the distribution quantity and parameters in the two cases of NLOS and LOS can be obtained, thereby obtaining the GMM model in the NLOS scenario and the GMM model in the LOS scenario.

[0101] Step 5: Calculate the PDF value of each time-delayed data according to the GMM formula, use the geometric method to determine the probability that the current propagation is NLOS, and perform NLOS identification, as follows:

[0102] Step 5-1, after determining the number of GMM distributions and updating the GMM parameters, construct the GMM models for the NLOS and LOS scenarios. First, calculate the PDF value of each delay data according to the GMM model formula.

[0103] Step 5-2, estimate the probability that the current transmission is NLOS. Based on a large amount of experimental data, it is assumed that the probability of a random variable taking a value in a specific interval is proportional to the length of the interval. The core is to approximate the probability value through the area determined by the PDF value. The specific NLOS identification process is as follows: Figure 2 shown.

[0104] The NLOS probability value calculation method is as follows:

[0105]

[0106] Where: PDF NLOS and PDF LOS are the probability density function values ​​of given delay values ​​under NLOS model and LOS model respectively; Area NLOS and Area LOS are the geometric area obtained by square of edge probability values ​​based on NLOS and LOS models; TotalArea is the sum of non-overlapping area of ​​NLOS and LOS; Probability NLOS is the NLOS probability value.

[0107] If PDF NLOS and PDF LOS If the values ​​are equal, it is considered that there is no overlapping area, and the output NLOS probability value is 50%; otherwise, it is considered that there is an overlapping area, and the TotalArea (total area) is calculated, which is the sum of the non-overlapping areas of NLOS and LOS; the NLOS non-overlapping area divided by the total area is the calculated NLOS probability value Probability NLOS , the probability value greater than or equal to 50% is identified as NLOS, otherwise it is LOS.

[0108] The EM algorithm fitting effect when the number of GMM components is 4 is as follows Figure 3 As shown in Figure 1, the average residual of the PDF value of the fitting delay is 0.01245. The fitting effect diagram of the NLOS identification method based on the GMM-PSO-EM algorithm is shown in Figure 1. Figure 4As shown in Figure 2, the average residual of the PDF value of the fitted delay is only 0.01056, which is about 15.16% better than the EM algorithm. The GMM modeling visualization images of the NLOS and LOS scenarios are shown in Figure 2. Figure 5 As shown in the figure, it can be clearly seen that the delay has different distribution curves in the NLOS and LOS scenarios. In the LOS scenario, the path is direct and stable, and the distribution of delay is more concentrated, so the probability density distribution curve of delay is narrower. On the contrary, the probability density distribution curve of delay in the NLOS scenario is wider, indicating that the delay value is distributed in a larger range, reflecting the impact of multiple propagation paths.

[0109] In order to verify the recognition accuracy of the NLOS recognition method based on the GMM-PSO-EM algorithm, the NLOS recognition method based on the GMM-PSO-EM algorithm is compared with the unimproved GMM algorithm, CNN algorithm and SVM algorithm in terms of recall rate, precision and F1 score. The comparison results are shown in Table 1 below.

[0110] Table 1 Performance comparison of different methods

[0111]

[0112]

[0113] It can be seen from Table 1 that the recall rate, accuracy and F1 score of the NLOS recognition method based on the GMM-PSO-EM algorithm are higher than those of other algorithms. The average recall rate reached 92.4%, the average accuracy reached 93.7%, and the F1 score reached 93.2%. Compared with the other three methods, the F1 scores were improved by 3.78%, 2.3% and 5.2%, respectively, indicating that the NLOS recognition method based on the GMM-PSO-EM algorithm has better NLOS recognition accuracy due to better fitting of LOS and NLOS features.

[0114] Regarding the accurate identification of NLOS signals during the positioning process of the UWB system in indoor multi-scenario complex environments, the above-mentioned experimental simulation has verified that the NLOS identification method based on the GMM-PSO-EM algorithm can solve the problems of high complexity of NLOS identification in complex indoor environments, the need for a large amount of data labeling, and poor performance when the amount of data is small, thereby achieving higher NLOS identification accuracy, thereby effectively improving the UWB positioning accuracy in indoor multi-scenario complex environments.

Claims

1. A NLOS identification method based on GMM-PSO-EM algorithm, characterized in that: The specific steps include: Step 1, determine the number of Gaussian distributions in GMM based on AIC and BIC; Step 2, use the PSO-SA-EM algorithm to determine the parameters of the distribution of each component of GMM and obtain the GMM model; Step 3, collect the delay data of LOS and NLOS scenarios; Step 4, fitting the time delay GMM model of LOS and NLOS; Step 5, calculate the PDF value of each time-delay data, use the geometric method to determine the probability that the current propagation is NLOS, and perform NLOS identification; The NLOS probability value calculation method is as follows: Where: PDF NLOS and PDF LOS are the probability density function values ​​of given delay values ​​under NLOS model and LOS model respectively; Area NLOS and Area LOS are the geometric area obtained by square of edge probability values ​​based on NLOS and LOS models; TotalArea is the sum of non-overlapping area of ​​NLOS and LOS; Probability NLOS is the NLOS probability value; If PDF NLOS and PDF LOS If the values ​​are equal, it is considered that there is no overlapping area and the output NLOS probability value is 50%; otherwise, it is considered that there is an overlapping area and the TotalArea is calculated; the NLOS probability value Probability is calculated by dividing the NLOS non-overlapping area by the TotalArea NLOS , the probability value greater than or equal to 50% is identified as NLOS, otherwise it is LOS.

2. The NLOS identification method based on the GMM-PSO-EM algorithm according to claim 1, characterized in that: Step 1 is as follows: Step 1-1, introduce the histogram to visualize the peak information and determine a rough distribution range; Step 1-2, use AIC and BIC to make judgments, draw the curves of AIC and BIC within the distribution range, and smooth the curves; Step 1-3, select the value in the curve that minimizes the AIC or BIC value as the distribution quantity.

3. The NLOS identification method based on the GMM-PSO-EM algorithm according to claim 1, characterized in that: The PSO-SA-EM algorithm in Step 2 is as follows: Step 2-1, initialization: ① Initialize the particle swarm size N and randomly assign each particle’s position pBest(0) and velocity v i (0); ② Randomly initialize the individual optimal pBest(0) and the global optimal gBest(0) for each particle; ③ Set the initial temperature T(0) and cooling rate α of the SA algorithm; ④Determine the convergence conditions and number of iterations of the EM algorithm; Step 2-2, determine the annular domain: assign a neighborhood to each particle i, and the leading particle x lead Index: (i+1) mod N, lagging particle x lag Index of: (i-1+n) mod N; Step 2-3 Determine the optimal neighborhood: Compare the individual optimal solutions of the leading particle and the lagging particle, and determine the optimal neighborhood solution nBest; Step 2-4, PSO-SA iteration, perform the following steps for each particle: ① Use the PSO algorithm to update the position and velocity of the particles, as shown below: v i (t+1)=w·v i (t)+c1·r1·(pBest i -x i (t))+c2·r2·(gBest-x i (t)) x i (t+1)=x i (t)+v i (t+1) Where: v i (t) is the velocity of the ith particle at time t; x i (t) is the position of the ith particle at time t; w is the inertia weight used to balance the exploration and development capabilities of the particle; c1 is the cognitive coefficient that affects the movement of the particle to the individual optimal position; c2 is the social coefficient that affects the movement of the particle to the global optimal position; r1 and r2 are random numbers between 0 and 1; ②Neighborhood optimal guidance: Calculate the current particle position x i The difference Δx between (t+1) and the best neighbor nBest is expressed as follows: Δx=nBest-x i (t+1) Update the particle position x according to the difference Δx and step size λ i (t+1)', expressed as follows: x i (t+1)'=x i (t+1)+λΔx Where: x i (t+1) is the particle position before updating; ③ Use the probability acceptance criterion of the SA algorithm to decide whether to accept the new solution: calculate the fitness of the new position and compare it with the individual optimality of the current particle, and then accept the new position with probability according to the SA criterion to decide whether to accept the new solution. The calculation formula of the SA criterion is as follows: Where: ΔE is the difference between the fitness of the new position and the optimal fitness of the current individual; T(t) is the current temperature; ④ Update individual optimality and global optimality; ⑤ Update the temperature in simulated annealing, as shown below: T(t+1)=T(t)*α Where: α is the cooling rate; Step 2-5, initialize the EM algorithm and use the optimal particle position found by the PSO-SA algorithm as the initial parameters of the EM algorithm; Step 2-6, execute the E step and M step of the EM algorithm until convergence; Step 2-7, output the final GMM parameters.

4. The NLOS identification method based on the GMM-PSO-EM algorithm according to claim 1, characterized in that: Step 4 is as follows: For a given data point x, the PDF calculation formula is as follows: Where: k is the weight of the kth component; is the Gaussian distribution of the kth component; μ k and σ k are the mean and variance of each Gaussian distribution, respectively; According to the above formula, the distribution quantity and parameters in the NLOS and LOS cases are obtained, and the GMM model in the NLOS scenario and the GMM model in the LOS scenario are obtained.

Images