A multi-antenna UWB NLOS identification method

Abstract

The application discloses a multi-antenna UWB NLOS identification method, compared with a common identification method based on a single-antenna propagation model, the method introduces a phase parameter of a UWB receiving signal as an identification feature, utilizes correlation of multi-path antenna receiving signals, and compares with a threshold value trained in advance by using a machine learning method to perform NLOS identification.The feature information extracted by the method is rich and has good distinguishability.

Description

Technical Field

[0001] This invention relates to the field of wireless positioning, and more particularly to a multi-antenna NLOS identification method. Background Technology

[0002] Wireless positioning refers to the process of estimating the geographical location of a mobile terminal in a wireless mobile communication network by measuring the characteristic parameters of received radio waves and using specific algorithms based on the measured wireless signal data. This provides accurate terminal location information and services. Currently, mainstream wireless positioning technologies include those based on Time of Arrival (TOA), Direction of Arrival (DOA), and Received Signal Strength Indication (RSSI). These technologies are highly sensitive to whether the distance between the transmitter and receiver is line-of-sight (LOS), especially in indoor environments where walls or other obstructions frequently cause non-line-of-sight (NLOS) errors, leading to significant degradation in positioning accuracy. Therefore, effectively identifying the presence of non-line-of-sight (NLOS) errors is of great importance.

[0003] Multi-antenna UWB positioning based on Phase Difference of Arrival (PDOA) is one of the DOA positioning technologies developed in recent years. Because it possesses simultaneous ranging and direction-finding capabilities, it can achieve two-dimensional positioning using only a single base station, offering significant advantages in terms of construction difficulty and deployment cost. The high temporal resolution of UWB signals gives TOA-based ranging better multipath resistance, resulting in a clear advantage in ranging accuracy compared to other wireless signals. However, PDOA-based direction finding is much more sensitive to the propagation environment. Its phase resolution range is only half a wavelength; even a slight obstruction or diffraction during UWB signal propagation can cause significant phase jitter in the received signal, thus interfering with the positioning effect.

[0004] Most existing UWB positioning and NLOS identification technologies are based on single-antenna ranging models, utilizing the statistical characteristics of propagation energy attenuation or flight delay parameters for NLOS identification. This results in limited available information and has significant limitations for UWB direction finding based on PDOA. Summary of the Invention

[0005] This invention discloses a multi-antenna UWB NLOS identification method. Compared with common identification methods based on single-antenna propagation models, this method introduces the phase parameters of the UWB received signal as identification features. It utilizes the correlation of the received signals from multiple antennas and compares them with a pre-trained threshold obtained through machine learning to identify NLOS. This method extracts rich feature information with good discriminative power.

[0006] This invention provides a method for NLOS identification in multi-antenna UWB, comprising:

[0007] 1) A multi-antenna UWB base station receives UWB signals from multiple channels in the same frame from the transmitting tag;

[0008] 2) Based on the I and Q signals of the IQ coherent demodulator of each channel, obtain the composite amplitude at the first path position and the phase difference values ​​from the N1th to the N2th points before the first path position;

[0009] 3) Calculate the composite amplitude at the first path position of all channels and the extreme values ​​of the phase differences from the N1th to the N2th phases before and after the first path position to form a feature vector. Compare this feature vector with a threshold parameter trained by machine learning to determine whether the environment is LOS or NLOS. Optionally, obtaining the composite amplitude at the first path position and the phase differences from the N1th to the N2th phases before and after the first path position specifically includes:

[0010] 1) Calculate the composite amplitude at the first arrival point using the following formula:

[0011]

[0012] Among them, id fp Indicates the location of the first route;

[0013] 2) Calculate the phase values ​​from N1 to N2 before the first arrival point using the following formula:

[0014] θ m (id fp +k)=tan2 -1 (I(id fp +k),Q(id fp +k)), k=-N1,…,-1,0,1,…,N2;

[0015] 3) Group the consecutive positions and calculate the phase difference between the first N1 and last N2 positions of the first arrival path position for each channel, and normalize it to (0, 2π).

[0016] Optionally, the step of calculating the composite amplitude at the first path position of all channels and the extreme values ​​of the phase differences from the N1th to the N2th phases before the first path position constitutes a feature vector, specifically including:

[0017] 1) Compare the magnitude of the composite amplitude at the first path position of all channels to obtain the maximum amplitude, minimum amplitude, and the ratio of the maximum amplitude to the minimum amplitude;

[0018] 2) Compare the phase differences between the first N1 and last N2 positions of all channels to obtain the maximum (N1+N2) phase differences.

[0019] 3) The eigenvector consists of the maximum amplitude, the minimum amplitude, the ratio of the maximum amplitude to the minimum amplitude, and (N1+N2) maximum phase differences.

[0020] Optionally, the step of comparing and determining the LOS or NLOS environment using threshold parameters trained by machine learning specifically includes:

[0021] In advance, multiple sets of feature vectors consisting of the maximum amplitude, minimum amplitude, ratio of maximum amplitude to minimum amplitude, and (N1+N2) maximum phase difference values ​​are collected in known LOS and NLOS environments respectively.

[0022] Using LOS and NLOS as labels, and the feature vector as training samples, a Bayesian classification algorithm is used to determine the threshold parameters corresponding to each component of the feature vector; or...

[0023] Using LOS and NLOS as labels, the feature vectors are divided into training and test sets according to cross-validation, and the threshold parameters corresponding to the attributes of each node of the decision tree are determined by the decision tree classification algorithm.

[0024] The above-described technical solution of the present invention has at least the following beneficial effects:

[0025] The method of this invention can improve existing NLOS identification methods in the following three aspects: 1) By introducing the phase information of the UWB received signal, this method can help determine whether the obtained first path location is correct; 2) The spacing between UWB multi-antennas is only a few centimeters, so the correlation between the received signals is high, which can be used as an effective means of identifying NLOS; 3) By using Bayesian classification or decision tree classification to train the judgment thresholds required for each feature parameter, the physical meaning of each judgment criterion is effectively preserved while ensuring the universality of the algorithm. Attached Figure Description

[0026] Figure 1 This is a flowchart of an NLOS identification method for multi-antenna UWB provided in Embodiment 1 of the present invention;

[0027] Figure 2 This is a flowchart illustrating how a Bayesian algorithm is used to determine the threshold parameters corresponding to each component of the feature vector, as provided in Embodiment 2 of the present invention. Detailed Implementation

[0028] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0029] Figure 1 This is a flowchart illustrating a multi-antenna UWB NLOS identification method provided in Embodiment 1 of the present invention. Figure 1 As shown, it includes the following steps:

[0030] S101: A multi-antenna UWB base station receives multiple channels of UWB signals from the same frame of the transmitting tag;

[0031] S102: Based on the I and Q signals of the IQ coherent demodulator of each channel, obtain the composite amplitude at the first path position and the phase difference values ​​from the N1th to the N2th steps before the first path position.

[0032] S103: Calculate the composite amplitude at the first path of arrival (LTA) position of all channels and the extreme values ​​of the phase differences from the N1th to the N2th LTA positions to form a feature vector. Use machine learning-trained threshold parameters to compare and determine whether the environment is LOS or NLOS. In this embodiment, the multi-antenna UWB base station is a wireless transceiver with at least two UWB antennas and corresponding UWB processing channels. With only two UWB antennas, it can achieve a 0-180° half-cycle direction finding function on a certain plane; with three or more UWB antennas, it can achieve a 0-360° full-cycle direction finding function on a certain plane. When the multi-antenna UWB base station acts as a receiver, it executes the above identification steps. Depending on the UWB TOA scheme, such as single-sided bidirectional ranging or double-sided bidirectional ranging, the multi-antenna UWB base station and the transmitting tag may perform one or more NLOS identification processes during a single TOA ranging operation.

[0033] In this embodiment, the IQ demodulator is used to recover information-containing low-frequency signals from high-frequency carrier signals. The receiver simultaneously generates two local reference signals and a cyclic correlation estimation channel (CIR) for the received signal. The I signal represents the correlation accumulation sequence of the RF carrier sine wave mixing, and the Q signal represents the correlation accumulation sequence of the same RF carrier sine wave mixed with a 90-degree phase shift. Based on the orthogonality of trigonometric functions, the I and Q signals can jointly represent the amplitude and phase changes of the sine wave.

[0034] In this embodiment, the first path position can be determined by finding the position corresponding to the first peak or slope peak in the I and Q signals that is greater than a set threshold.

[0035] Optionally, obtaining the composite amplitude at the first path position and the phase difference values ​​from N1 to N2 before and after the first path position specifically includes:

[0036] 1) Calculate the composite amplitude at the first arrival point using the following formula:

[0037]

[0038] Among them, id fp Indicates the location of the first route;

[0039] 2) Calculate the phase values ​​from N1 to N2 before the first arrival point using the following formula:

[0040] θ m (id fp +k)=tan2 -1 (I(id fp +k),Q(id fp +k)), k=-N1,…,-1,0,1,…,N2;

[0041] 3) Grouping consecutive positions, calculate the phase difference between the N1th and N2th positions before the first arrival point of each channel, and normalize it to (0, 2π):

[0042] θ d (id fp +k)=mod(θ m (id fp +k)-θ m (id fp +k-1),2π),k=-N1+1,…,0,1,…,N2;

[0043] Optionally, the step of calculating the composite amplitude at the first path position of all channels and the extreme values ​​of the phase differences from the N1th to the N2th phases before the first path position constitutes a feature vector, specifically including:

[0044] 1) Compare the magnitude of the composite amplitude at the first path position of all channels to obtain the maximum amplitude, minimum amplitude, and the ratio of the maximum amplitude to the minimum amplitude;

[0045] 2) Compare the phase differences between the first N1 and last N2 positions of all channels to obtain the maximum (N1+N2) phase differences.

[0046] 3) The eigenvector consists of the maximum amplitude, the minimum amplitude, the ratio of the maximum amplitude to the minimum amplitude, and (N1+N2) maximum phase differences.

[0047] Optionally, the step of comparing and determining the LOS or NLOS environment using threshold parameters trained by machine learning specifically includes:

[0048] In advance, multiple sets of feature vectors consisting of the maximum amplitude, minimum amplitude, ratio of maximum amplitude to minimum amplitude, and (N1+N2) maximum phase difference values ​​are collected in known LOS and NLOS environments respectively.

[0049] Using LOS and NLOS as labels, and the feature vector as training samples, a Bayesian algorithm is used to determine the threshold parameters corresponding to each component of the feature vector; alternatively, using LOS and NLOS as labels, the feature vector is divided into training and test sets according to cross-validation, and a decision tree classification algorithm is used to determine the threshold parameters corresponding to each node attribute of the decision tree.

[0050] Figure 2 This is a flowchart illustrating how a Bayesian algorithm is used to determine the threshold parameters corresponding to each component of the feature vector, as provided in Embodiment 2 of the present invention. Figure 2 As shown, taking the maximum amplitude value as an example, the process includes the following steps:

[0051] S201: Collect multiple sets of the maximum amplitude values ​​under known LOS and NLOS environments;

[0052] S202: Perform Gaussian fitting on the distribution of the maximum amplitude under both LOS and NLOS environments, assuming the fitted distribution under the LOS environment is as follows. The fitted distribution under NLOS environment is

[0053]

[0054] S203: Calculate the average cost function for identifying LOS / NLOS environments. Where c 12 This represents the cost factor for identifying an NLOS environment as a LOS environment, c 21 This indicates the cost factor for identifying a LOS environment as an NLOS environment;

[0055] S204: Obtain the threshold parameter corresponding to the maximum amplitude value according to the minimum cost Bayes criterion: In this embodiment, due to the blocking or diffraction effect of the NLOS environment, the channel attenuation of the UWB signal is more severe than that of the LOS environment, thus satisfying: μ1 > μ2. Based on the threshold parameter corresponding to the maximum amplitude value, the NLOS judgment criterion based on the maximum amplitude value is given as follows: if the maximum amplitude value is greater than η0, it is judged that the counter for the LOS environment is incremented by 1; otherwise, it is judged that the counter for the NLOS environment is incremented by 1. Similarly, by comparing each of the other components of the feature vector with the corresponding threshold parameter obtained using the method of this embodiment, the final counter values ​​for the LOS environment and the NLOS environment are obtained, and the larger value is taken as the final judgment result.

[0056] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for NLOS path identification in multi-antenna UWB, comprising: 1) A multi-antenna UWB base station receives UWB signals from multiple channels in the same frame from the transmitting tag; 2) Based on the I and Q signals of the IQ coherent demodulator of each channel, obtain the composite amplitude at the first path position and the phase difference values ​​from the N1th to the N2th points before the first path position; 3) Calculate the composite amplitude at the first path position of all channels and the extreme values ​​of the phase difference from the first path position to the last N2 positions to form a feature vector. Use the threshold parameters trained by machine learning to compare and determine the LOS or NLOS environment. The acquisition of the composite amplitude at the first arrival point and the phase difference values ​​from the N1th to the N2th phases before and after the first arrival point specifically includes: 1) Calculate the composite amplitude at the first arrival point using the following formula: Among them, id fp Indicates the location of the first route; 2) Calculate the phase values ​​from N1 to N2 before the first arrival point using the following formula: θ m (id fp +k)=tan2 -1 (I(id fp +k),Q(id fp +k)),k=-N1,…,-1,0,1,…,N2; 3) Grouping consecutive positions, calculate the phase difference between the first N1 and last N2 positions of the first arrival path position for each channel, and normalize it to (0, 2π). The calculation of the composite amplitude at the first arrival position of all channels and the extreme values ​​of the phase differences from the N1th to the N2th phases before and after the first arrival position constitutes a feature vector, specifically including: 1) Compare the magnitude of the composite amplitude at the first path position of all channels to obtain the maximum amplitude, minimum amplitude, and the ratio of the maximum amplitude to the minimum amplitude; 2) Compare the phase differences between the first N1 and last N2 positions of all channels to obtain the maximum (N1+N2) phase differences. 3) The eigenvector consists of the maximum amplitude, the minimum amplitude, the ratio of the maximum amplitude to the minimum amplitude, and (N1+N2) maximum phase differences.

2. The method as described in claim 1, characterized in that, Determining LOS or NLOS environments by comparing threshold parameters trained through machine learning, specifically including: In advance, multiple sets of feature vectors consisting of the maximum amplitude, minimum amplitude, ratio of maximum amplitude to minimum amplitude, and (N1+N2) maximum phase difference values ​​are collected in known LOS and NLOS environments respectively. Using LOS and NLOS as labels, and the feature vector as training samples, a Bayesian algorithm is used to determine the threshold parameters corresponding to each component of the feature vector; or... Using LOS and NLOS as labels, the feature vectors are divided into training and test sets according to cross-validation, and the threshold parameters corresponding to the attributes of each node of the decision tree are determined by the decision tree algorithm.

Images