A noise vector direct estimation method in a receiver multiplexing single-target direction finding system

Abstract

The application discloses a noise vector direct estimation method for a receiver multiplexing single-target direction finding system of a BLE AoA / AoD type direction finding system, and belongs to the technical field of measurement and testing. The method estimates the autocorrelation coefficients of each antenna and a reference antenna by calculating the sum of the single-time acquisition data of each antenna in an antenna array and the single-time acquisition data of the reference antenna and then taking the average, and constructs an estimation matrix of a noise space feature vector from the estimated autocorrelation coefficients of each antenna and the reference antenna, without needing to perform matrix autocorrelation, eigenvalue decomposition, eigenvalue sorting and other operations. With the noise space feature vector provided by the application and a traditional direction finding algorithm, such as Root-MUSIC, high-precision direction finding effects can be obtained with lower computing resources, and the application is suitable for Internet of Things devices with limited computing resources.

Description

TECHNICAL FIELD

[0001] The present application relates to wireless high-precision direction finding technology, and particularly discloses a noise vector direct estimation method in a receiver multiplexing single-target direction finding system, and belongs to the technical field of measurement and testing. BACKGROUND

[0002] Array signal processing is an important branch of signal processing, which mainly sets multiple sensors at different positions in space to form a sensor array. Traditional array signal processing mainly receives and processes the spatial signal field in parallel, as shown in the formula (1), the purpose is to extract the characteristic parameters of the array received signal, and suppress the interference and noise or information not interested. Figure 1

[0003] Generally, considering N far-field narrow-band signals incident to a direction finding array composed of M antennas, the signals received by each array element are sampled in parallel by the receiver, so that the sampling signals satisfy the formula (2):

[0004] X(t)=AS(t)+N(t)

[0005] Wherein, X(t) is the M parallel channel data collected by the direction finding array in the formula (2), which can be further expressed as the formula (3): Figure 1

[0006] X(t)=[x1(t),…,x M (t)] T

[0007] S(t) is N far-field narrow-band signals, which can be further expressed as the formula (4):

[0008] S(t)=[s1(t),…,s N (t)] T

[0009] N(t) is the noise signal of M channels, which can be further expressed as the formula (5):

[0010] N(t)=[n1(t),…,n M (t)] T

[0011] A is a to-be-estimated steering vector matrix composed of N incident signals:

[0012] A=[a1,…,a N ] T

[0013] Wherein, the steering vector of the i-th incident signal to is:

[0014] a i =[exp(-jω0τ 1i ​​..., exp(-jω0τ Mi )] T

[0015] where ω0= 2πf0, f0is the center frequency of the direction finding signal, τ 1i is the delay of the i-th incident signal at the 1st array element relative to the reference array element, τ Mi is the delay of the i-th incident signal at the M-th array element relative to the reference array element.

[0016] If the observed array is a uniform linear array (ULA) with an antenna spacing of d and the first antenna as the reference point, the following is further obtained:

[0017]

[0018] where θ i is the angle between the i-th incident signal to be estimated and the normal line of the linear array.

[0019] The array signal processing algorithm is to estimate the incident direction of different signals through the known antenna array structure and the measured signal X(t).

[0020] The positioning technology based on array signal processing has been researched for nearly half a century, and there are many achievements in related super-resolution and anti-multipath direction finding, array error calibration, etc., among which the most representative is the method based on matrix characteristic space decomposition.

[0021] First, the covariance matrix of the data is obtained, and the covariance matrix of the data is:

[0022] R XX = E[XX H ] = AE[SS H ]A H + δ 2 I = AR SS A H + δ 2 I

[0023] Since the signal and the noise are independent of each other, they can be decomposed into two parts of signal and noise:

[0024] R XX = U s ∑ s U s H + U N ∑ N U N H

[0025] where U sis the subspace spanned by the eigenvectors corresponding to the large eigenvalues, that is, the signal space; U N is the subspace spanned by the eigenvectors corresponding to the small eigenvalues, that is, the noise space.

[0026] Ideally, the signal subspace and the noise subspace are orthogonal, so the steering vector of the signal subspace is also orthogonal to the noise subspace, that is:

[0027] a H (θ)U N =0

[0028] The classical MUSIC algorithm is based on the above formula. Considering that the length of the actually collected signal is limited, the maximum likelihood estimation of the covariance matrix is:

[0029]

[0030] where X(t) is sampled at L points to obtain X, and at this time, by the same calculation, can obtain The eigenvector estimation matrix of the noise subspace is obtained. At this time, a(θ) and are not necessarily completely orthogonal, so the search can be obtained by minimizing:

[0031]

[0032] Finally, the spectral peak search formula of the MUSIC method is obtained:

[0033]

[0034] Through the derivation of the above MUSIC method, we can obtain the main calculation process needed by this kind of algorithm (taking ULA as an example): calculate the autocorrelation matrix of M roots of antenna and M roots of antenna L sampling points Calculate the eigenvalues and eigenvectors of the M*M dimensional autocorrelation matrix ; sort the eigenvalues to obtain M-D noise eigenvalues and corresponding eigenvectors According to the traversal incident angle θ, obtain P MUSIC ; obtain D maximum values in P MUSIC , which correspond to the incident signal directions of D signals.

[0035] Through the analysis of the above classical direction finding system and direction finding algorithm, it can be known that the classical direction finding hardware architecture needs multiple radio frequency receivers to sample in parallel, and the hardware cost and running power consumption are high; at the same time, the above algorithm steps are more, and the calculation amount is complex, which is not suitable for low-power Internet of Things devices to direction finding.

[0036] To solve the problems of hardware complexity and power consumption, a single receiver multiplexing direction finding system can be used. The system switches through a radio frequency switch to make a single radio frequency receiver obtain the antenna array signals in turn, as shown in Figure 2 The latest Bluetooth Low Energy (BLE) direction finding uses the above hardware architecture. To compensate for the signal loss caused by antenna switching, the direction finding signal uses a periodic sine signal. In addition, to simplify the direction finding algorithm and obtain a unique identifier of the direction finding target in each collected signal, the single collected direction finding signal only contains a single direction finding target. The above simplifications are to enable low-cost and low-power Internet of Things systems such as BLE to complete high-precision direction finding. Although the single receiver multiplexing direction finding system simplifies the hardware architecture, the single collected measurement signal is still positioned using the classic direction finding algorithm, which has the defects of large calculation amount and high power consumption, so the precision of the Internet of Things device with limited computing resources is poor when used for single target direction finding.

[0037] In summary, the present application proposes a noise vector direct estimation method for a single receiver switch multiplexing direction finding system for single target direction finding to overcome the above-mentioned defects. SUMMARY

[0038] The present application aims to overcome the deficiencies of the above background art and provides a noise vector direct estimation method for a single receiver multiplexing direction finding system to directly estimate the noise spatial feature vector to eliminate the complex operation in the traditional direction finding process, achieve the purpose of high-precision single target direction finding on Internet of Things devices, and solve the technical problem that Internet of Things devices with limited computing resources are not suitable for precise direction finding in a single receiver multiplexing direction finding system.

[0039] The present application uses the following technical solutions to achieve the above-mentioned purposes:

[0040] To achieve high-precision single target direction finding on a single receiver multiplexing Internet of Things node with limited computing resources, the present application proposes a noise spatial feature vector direct estimation method. The incident signal direction is obtained by subsequently using the Root-MUSIC algorithm.

[0041] In the single target case, the signal collected by the array antenna can be simplified as:

[0042] X(t)=a1s(t)+N(t)

[0043] where s(t) is the direction finding signal incident from a single target, and for a ULA with M antennas, the steering vector of the single target is:​

[0044]

[0045] where

[0046] The covariance matrix can be expressed as:

[0047]

[0048] For R xx , we can get M eigenvalues, which are:

[0049] λ1= M + δ 2

[0050] λ2= λ3=... = λ M = δ 2

[0051] Obviously, λ2,..., λ M are smaller eigenvalues, and their corresponding eigenvectors are:

[0052]

[0053] Therefore, U noise is the eigenvector of the noise space corresponding to the ULA in the single-target case. By observation, we can find that although each vector of U noise is orthogonal to the vector of the signal space, the eigenvectors of the noise space are not orthogonal. Therefore, we use the Schmidt orthogonalization technique to orthogonalize the vectors in U noise , and directly obtain a closed-form expression of the orthogonal eigenvectors of the noise space in the single-target direction-finding case:

[0054]

[0055] By comparing the above eigenvectors with R xx , we can find that U N can be directly expressed by the elements in R xx :

[0056]

[0057] Considering that the length of the actual collected signal is limited, we use to estimate U N :

[0058]

[0059] where:

[0060]

[0061] wherein: x1(i), 0≤i≤N-1 is N sampling points from the reference antenna, the starting antenna can be used as the reference antenna for the general ULA array; x m x1(i), 0≤i≤N-1, 1<m≤M is N sampling points from the mth antenna other than the reference antenna; * represents the conjugate of the data.

[0062] The noise spatial orthogonal eigenvector proposed by observation The expression form, we can find that at this time, only the average result of the conjugate multiplication of the ULA direction finding array reference antenna sampling signal and the M-1 antenna sampling signal is obtained, the direction finding algorithm complexity can be greatly simplified. The method proposed by the application omits the mathematical operations such as M-antenna autocorrelation operation (conjugate multiplication average between M-antennas), eigenvalue decomposition and eigenvalue sorting, and greatly simplifies the complexity of the direction finding algorithm.

[0063] The method is particularly suitable for a receiver multiplexing direction finding system, which needs to be switched by a radio frequency switch to sequentially collect the direction finding data of all antennas, for example, a BLE AoA / AoD receiver multiplexing direction finding system.

[0064] The noise vector direct estimation method of the receiver multiplexing single target direction finding system provided by the application comprises the following steps:

[0065] Step 1: for the receiver multiplexing direction finding system, complete the switching of M antennas according to the designed antenna switching mode, and record the N-point sampling data obtained by each antenna as a single round of collected signal;

[0066] Step 2: for the receiver multiplexing direction finding system, the frequency offset compensation algorithm is needed to eliminate the influence of the carrier frequency offset caused by the switching of the switch, and the specific method is to perform steps 3-5, and then perform step 4 frequency offset compensation on the result calculated in step 3 to construct the estimation matrix of the noise spatial feature vector; for the parallel collection system, step 2 can be ignored, and steps 3 and 5 are executed;

[0067] Step 3: obtain the average result of the sum of the conjugate multiplication results of the M-1 antenna collection data and the reference antenna sampling data by formula (2), and the specific method is to perform conjugate calculation on the same incident far-field narrowband signal collected by the current antenna and the reference antenna in the direction finding system antenna array, and take the average value of the sum of the conjugate calculation results of each incident far-field narrowband signal collected by the current antenna and the reference antenna, the average value is the autocorrelation coefficient of the current antenna and the reference antenna in the direction finding system antenna array.

[0068] Step 4: When the carrier frequency offset of the direction finding system exceeds the threshold value, the autocorrelation coefficients of each antenna in the antenna array of the direction finding system and the reference antenna are compensated for frequency offset, and then step 5 is entered; when the carrier frequency offset of the direction finding system does not exceed the threshold value, step 5 is directly entered;

[0069] Step 5: According to the calculation result of step 3 or the calculation result after frequency offset compensation of step 4, M-1 characteristic vectors orthogonal in the noise space are directly calculated to obtain a matrix as shown in formula (1), and a noise space characteristic vector estimate is obtained

[0070] A direction finding method of a receiver multiplexing single-target direction finding system, which adopts a spectrum estimation type direction finding algorithm to estimate the direction of an incident far-field narrowband signal. Preferably, the noise space characteristic vector estimate obtained by the above noise vector direct estimation method is substituted into the Root-MUSIC algorithm to obtain the direction of the single-target incident signal.

[0071] The technical scheme of the present application has the following beneficial effects: the method proposed in the present application only needs to obtain the conjugate multiplication results of the signals received by each array element of the antenna and the reference antenna, and can directly estimate the noise space characteristic vector in the single-target case The complex operations such as matrix autocorrelation, eigenvalue decomposition and sorting in the MUSIC algorithm are eliminated. The noise space characteristic vector estimated by the method proposed in the present application can be combined with traditional direction finding algorithms such as Root-MUSIC to obtain high-precision direction finding results with lower computing resources, and is suitable for Internet of Things devices with limited computing resources. BRIEF DESCRIPTION OF DRAWINGS

[0072] Figure 1 It is a linear array antenna parallel receiver direction finding model.

[0073] Figure 2 It is a low-cost receiver multiplexing direction finding system.

[0074] Figure 3 BLE CTE direction finding signal format.

[0075] Figure 4 It is a flowchart of the noise space characteristic vector direct estimation method proposed in the present application. DETAILED DESCRIPTION

[0076] In the following, taking a four-antenna ULA linear array as an example, combining the flowchart of the noise space characteristic vector direct estimation method shown in FIG. Figure 4 The specific implementation of the method proposed in the present application applied to the BLE AoA / AoD direction finding algorithm is given.

[0077] For Figure 2 ​The receiver multiplexes single-target measurement system, BLE AoA / AoD direction finding method, includes the following six steps, wherein steps one to five are the noise space characteristic vector direct estimation method proposed by the application.

[0078] The first step: according to the BLE AoA / AoD direction finding specification, for the ULA containing 4 antennas, through the RF switch, each antenna is polled and switched according to the antenna switching mode to collect the incident N narrowband signals once, and 4 parallel channel CTE signals are obtained, and the format of the CTE signal is as shown in the figure. Figure 3

[0079] The second step: using the existing algorithm, according to the ReferencePeriod in the collected 4 parallel channel CTE signals, the frequency of the CTE signal of this round of single collection is estimated, and the deviation of the CTE signal frequency of this round of single collection from 250kHz is calculated, denoted as Δf.

[0080] The third step: using the formula (2) in the application, the

[0081] The fourth step: if Δf exceeds 1kHz, the in the third step needs to be corrected, and the corrected values are respectively: t slot is the time interval of the RF switch switching.

[0082] The fifth step: the corrected is substituted into the formula (1) in the application to obtain three noise space characteristic vectors, and a matrix

[0083] The sixth step: the is used to the spectrum estimation type direction finding algorithm, such as Root-MUSIC, to obtain the incident signal direction.

[0084] The above embodiments are only exemplary descriptions of the application and do not limit the protection scope thereof, and any form of equivalent replacement within the scope of the application can also be changed by those skilled in the art, and any form of equivalent replacement within the scope of the application falls within the protection scope of the application.​

Claims

1. A method for noise vector direct estimation in a receiver multiplexed single-target direction finding system, characterized in that, The method comprises the following steps: Step 1, for a single-target direction finding system with receiver multiplexing, polling each antenna in the antenna array to perform single-time acquisition to obtain a round of single-time acquisition CTE signals; Step 2, calculating a carrier frequency offset of the direction finding system according to the frequency of the round of single-time acquisition CTE signals; Step 3, estimating the autocorrelation coefficients of each antenna in the antenna array of the direction finding system and a reference antenna according to the round of single-time acquisition CTE signals; Step 4, when the carrier frequency offset of the direction finding system exceeds a threshold, performing frequency offset compensation on the autocorrelation coefficients of each antenna in the antenna array of the direction finding system and the reference antenna to enter Step 5, and when the carrier frequency offset of the direction finding system does not exceed the threshold, directly entering Step 5; Step 5, constructing an estimation matrix of noise space feature vectors.

2. The method of claim 1, wherein, The method for estimating the autocorrelation coefficients of each antenna in the antenna array of the direction finding system and the reference antenna in Step 3 is: performing conjugate calculation on the same incident far-field narrowband signal acquired by the current antenna in the antenna array of the direction finding system and the reference antenna, averaging the sum of the conjugate calculation results of each incident far-field narrowband signal acquired by the current antenna and the reference antenna, and obtaining the autocorrelation coefficients of the current antenna in the antenna array of the direction finding system and the reference antenna.

3. The method of claim 2, wherein the noise vector is estimated directly from the received signal vector. The expression for estimating the autocorrelation coefficient of each antenna in the antenna array of the direction finding system and the reference antenna in step 3 is: wherein, is the estimated value of the autocorrelation coefficient of the mth antenna in the antenna array of the direction finding system and the reference antenna, is the i-th incident far-field narrowband signal of a single acquisition of the reference antenna in the antenna array of the direction finding system, is the i-th incident far-field narrowband signal of a single acquisition of the mth antenna in the antenna array of the direction finding system, N is the number of incident far-field narrowband signals, is the number of antennas in the antenna array of the direction finding system, represents taking the conjugate of the data.

4. The method of claim 3, wherein the noise vector is estimated by The expression for compensating the self-correlation coefficient of each antenna in the antenna array of the direction finding system and the reference antenna in step 4 is: Wherein, is the modified estimation value of the self-correlation coefficient of the mth antenna in the antenna array of the direction finding system and the reference antenna, is the carrier frequency offset of the direction finding system calculated in step 2, is the time interval of the radio frequency switch in the direction finding system.

5. The method of claim 3, wherein the noise vector is estimated by The step 4 carries out frequency offset compensation on the autocorrelation coefficients of each antenna and the reference antenna in the antenna array of the direction finding system, and the step 5 constructs an estimation matrix of the noise spatial feature vector as follows: When the step 4 directly enters the step 5 when the carrier frequency offset of the direction finding system does not exceed the threshold value, the estimation matrix of the noise spatial feature vector constructed is as follows: Wherein, is the estimation matrix of the noise spatial feature vector, is the estimation value of the autocorrelation coefficient of the second antenna and the reference antenna in the antenna array of the direction finding system, is the estimation value of the autocorrelation coefficient of the third antenna and the reference antenna in the antenna array of the direction finding system, is the estimation value of the autocorrelation coefficient of the Mth antenna and the reference antenna in the antenna array of the direction finding system, is the estimation value of the autocorrelation coefficient of the second antenna and the reference antenna in the antenna array of the direction finding system, is the estimation value of the autocorrelation coefficient of the third antenna and the reference antenna in the antenna array of the direction finding system, is the estimation value of the autocorrelation coefficient of the Mth antenna and the reference antenna in the antenna array of the direction finding system.

6. A direction finding method for a receiver multiplexed single-target direction finding system employing a spectral estimation type direction finding algorithm to estimate the direction of an incident far field narrow band signal, characterised by, The noise space feature vectors estimated according to the method of any one of claims 1 to 5 are substituted into a spectral peak search formula.

7. The method of claim 6, wherein the receiver multiplexes the single target direction finding system. The spectral estimation algorithm is Root-MUSIC.

Images