CN114679356A - A method of channel full-dimensional parameter extraction independent of likelihood function - Google Patents

Abstract

The invention discloses a channel full-dimensional parameter extraction method that does not depend on the likelihood function. Signal data is received through multiple channels; the signal data is preprocessed to obtain the time delay peak value of the signal data; the time delay peak value is used as input information, The azimuth and elevation angles corresponding to each sub-path in the delay peak are calculated by the forward and backward spatial smoothing MUSIC algorithm; the complex amplitude of the sub-path is calculated according to the corresponding azimuth and elevation angles of each sub-path; the aggregate delay peak and each sub-path correspond to The azimuth angle and elevation angle, and the complex amplitude are obtained to obtain the channel full-dimensional parameter set; the present invention can reduce the influence of the Doppler frequency shift on the estimation of the complex amplitude of the sub-path by preprocessing the signal, so that the estimated complex amplitude is more accurate, Compared with SAGE, the parameter estimation algorithm based on likelihood function that introduces EM iteration, it does not need iteration, requires shorter time, and does not converge to the local optimal solution due to improper initial value setting, thereby estimating false paths. The problem.

Description

A method of channel full-dimensional parameter extraction independent of likelihood function

technical field

The invention belongs to the technical field of wireless channel parameter extraction, and in particular relates to a channel full-dimensional parameter extraction method that does not depend on a likelihood function.

Background technique

In order to meet the demands of future wireless communication networks (increase data rate, reduce delay, energy and cost), design and evaluate various advanced wireless communication technologies in different communication systems, it is necessary to have the ability to capture the characteristics exhibited by the above technologies on the corresponding channels. ability.

Quasi Deterministic Radio Channel Generator (QuaDRiGa), a geometry-based statistical ray tracing channel modeling method, has been generally recognized by the industry, but to adapt to specific application scenarios, it is necessary to input the various parameters of the scene for the model. These channel parameters can only be obtained from a large number of actual channel measurement data.

In terms of parameter extraction, to fully extract the parameters of each dimension of the channel, the most widely used is the Space Alternating Generalized Expectation-maximization (SAGE) algorithm. However, since SAGE is a likelihood function-based parameter estimation algorithm that introduces the iteration of Expectation-Maximization (EM) to reduce the complexity of the Maximum Likelihood (ML) algorithm, each iteration needs to search for subpaths The dimensional parameters of , make it satisfy the maximum likelihood condition, so it is very time-consuming when the number of sub-paths is large. And due to the characteristics of the EM algorithm, when there are some sub-paths with similar parameters or the initial parameter values are not set properly, it is easy to make the result converge to the local optimal solution and estimate the false path.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a channel full-dimensional parameter extraction method that does not depend on the likelihood function, so as to improve the accuracy of channel parameter extraction.

The present invention adopts the following technical scheme: a channel full-dimensional parameter extraction method that does not depend on the likelihood function, comprising the following steps:

Receive signal data through multiple channels; wherein the signal data is PN sequence or CFR data;

Preprocess the signal data to obtain the delay peak value of the signal data;

Taking the delay peak as input information, the forward and backward spatial domain smoothing MUSIC algorithm is used to calculate the azimuth and elevation angles corresponding to each subpath in the delay peak;

Calculate the complex amplitude of each subpath according to the corresponding azimuth and elevation angles of the subpath;

Collect the delay peak value, the azimuth angle and elevation angle corresponding to each subpath, and the complex amplitude to obtain the channel full-dimensional parameter set.

Preferably, when the signal data is a PN sequence, the preprocessing is:

Sliding correlation is performed on the PN sequence.

Preferably, before preprocessing the CFR data, it further includes:

Perform IFFT on the CFR data to obtain the CIR data corresponding to the CFR data.

Preferably, when the signal data is CFR data, the preprocessing is:

Use cyclic prefix length to divide CIR data into noise segment and valid signal segment;

Determine the first noise threshold according to the noise segment, and select the first delay position of the valid signal segment according to the first noise threshold;

Determine the union of the first delay positions of the CIR data of the same time slot received by different receiving channels, and obtain the first delay position set;

Taking the intersection of multiple first sets of different time slots to obtain a second set of delay positions;

Calculate the covariance matrix corresponding to each element in the second delay position set, and perform eigenvalue decomposition on the covariance matrix to obtain the maximum eigenvalue and the minimum eigenvalue;

Calculate the ratio of the maximum eigenvalue and the minimum eigenvalue, select elements whose ratio is greater than the second noise threshold in the second delay position set to form a third delay position set, and use the third delay position set as the delay peak value of the signal data .

Preferably, when the signal data is a PN sequence, after calculating the complex amplitude of the subpath according to the azimuth angle and the elevation angle corresponding to each subpath, the method further includes:

Calculate the phase difference between different time slots of each subpath;

Determine the Doppler frequency shift of the subpath according to the phase difference;

Construct local PN sequence based on Doppler frequency shift;

The local PN sequence is used to perform sliding correlation on the PN sequence, and the execution is continued until the complex amplitude of each subpath is obtained again.

Preferably, when the Doppler shift difference between each subpath is less than the difference threshold:

Calculate the average Doppler shift of each delay cluster by using the weighted average method;

A local PN sequence is generated based on the average Doppler shift.

Preferably, when the signal data is a PN sequence, the channel full-dimensional parameter set further includes:

Doppler shift for each subpath.

Another technical solution of the present invention: a channel full-dimensional parameter extraction device that does not depend on a likelihood function, comprising:

The receiving module is used to receive signal data through multiple channels; the signal data is PN sequence or CFR data;

The preprocessing module is used to preprocess the signal data to obtain the delay peak value of the signal data;

The first calculation module is used for taking the delay peak value as input information, and calculating the azimuth angle and the elevation angle corresponding to each subpath in the delay peak value by adopting the forward and backward spatial smoothing MUSIC algorithm;

The second calculation module is used to calculate the complex amplitude of the subpath according to the azimuth angle and the elevation angle corresponding to each subpath;

The aggregation module is used to aggregate the delay peak value, the azimuth angle and elevation angle corresponding to each subpath, and the complex amplitude to obtain a channel full-dimensional parameter set.

Another technical solution of the present invention: a channel full-dimensional parameter extraction device that does not depend on a likelihood function, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, the processor executes the computer The above-mentioned method for channel full-dimensional parameter extraction that does not depend on the likelihood function is implemented in the program.

Another technical solution of the present invention: a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned channel full-dimensional parameter that does not depend on the likelihood function is realized Extraction Method.

The beneficial effects of the present invention are: the present invention can reduce the influence of Doppler frequency shift on the estimation of the complex amplitude of the sub-path by preprocessing the signal, so that the estimated complex amplitude is more accurate, compared with the SAGE and other methods that introduce EM iteration. The parameter estimation algorithm based on the likelihood function does not require iteration, requires shorter time, and does not have the problem of converging to the local optimal solution and estimating the false path caused by improper initial value setting.

Description of drawings

1 is a flowchart of a method for extracting full-dimensional parameters of a channel that does not depend on a likelihood function according to an embodiment of the present invention;

2 is a schematic diagram of a frame format of a transmitting end of a channel measurement system in an embodiment of the present invention;

3 is a schematic diagram of a frame format of a transmitting end of an NR system in an embodiment of the present invention;

Fig. 4 is the processing flow chart of input data in the channel parameter extraction method in the embodiment of the present invention;

5 is a schematic diagram of sub-array selection for forward and backward spatial smoothing processing in an embodiment of the present invention;

6 is a schematic diagram of time delay position selection based on eigenvalue assistance provided in an embodiment of the present invention;

Fig. 7 is the comparison diagram of method estimation result and SAGE running time in the embodiment of the present invention;

8 is an angular power spectrum diagram drawn by parameters obtained by processing the actual data collected by the base station with the method and the forward and backward spatial smoothing MUSIC algorithm in the embodiment of the present invention;

9 is an angular power spectrum diagram reconstructed after using the proposed method to obtain angular domain parameters for data collected by a base station in a single-user line-of-sight scenario according to an embodiment of the present invention;

10 is a comparison diagram of the received power of the beam and the received power of the original beam obtained by the method and the forward and backward spatial smoothing MUSIC algorithm processing the actual data collected by the base station in the embodiment of the present invention;

FIG. 11 is a schematic structural diagram of an apparatus for extracting full-dimensional parameters of a channel that does not depend on a likelihood function according to an embodiment of the present invention.

Detailed ways

The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

The current 5G New Radio (NR) system often uses beamforming to improve system capacity, so the acquisition of angular domain parameters is particularly important. The NR system uses the Sounding Reference Signal (SRS) to detect the channel in the uplink, but currently it is mainly used to obtain the channel frequency response (CFR), and obtain effective angle information from the existing CFR to configure the beam resources of the base station. Has practical significance.

In practical applications, the delay estimation often uses sliding correlation based on pseudo-noise (PN) or constant envelope zero auto-correlation (CAZAC) sequences, and its resolution depends on the chip width. With the continuous increase of the signal bandwidth that can be sent by the current wireless transmitting device, the chip width can be correspondingly reduced, so the time delay resolution based on this method is continuously improved.

In terms of estimating the angle, the multi-signal classification (Multiple Signal Classification, MUSIC) feature structure algorithm not only has a shorter estimation time and higher resolution in terms of angle estimation, but generally requires the number of incoming waves to be known and the incoming waves are Uncorrelated, and the number of paths in the channel is often unknown and since each path has the same probe signal, the received signals of different paths are coherent.

In terms of deploying beams at base stations, most NR systems currently only perform eigenvalue decomposition on the obtained multi-channel CFR data, use eigenvalue information to roughly deploy beams and allocate resources, or directly bring CFR into the angle estimation algorithm. molecular path and get its angle.

In fact, by performing Inverse Fast Fourier Transform (IFFT) on the frequency-domain CFR to obtain the time-domain Channel Impulse Response (CIR), the system time-domain resolution can be used to preliminarily separate the paths. The delay cluster is obtained, and the spatial domain is further separated to extract finer angle information. However, due to the existence of the leakage mechanism, the angle domain information of different paths will be leaked to different delay positions, so some preprocessing operations are required to ensure the extracted angle. quality and quantity of information.

The present invention discloses a channel full-dimensional parameter extraction method that does not depend on the likelihood function. As shown in FIG. 1, the method includes the following steps: Step S110, receiving signal data through multiple channels; wherein the signal data is a PN sequence or CFR data ( That is, the channel frequency response data); Step S120, preprocessing the signal data to obtain the delay peak value of the signal data; Step S130, using the delay peak value as the input information, using the forward and backward spatial domain smoothing MUSIC algorithm to calculate and obtain each delay peak value in the delay peak value. The azimuth and elevation angles corresponding to the sub-paths; step S140, calculating the complex amplitude of the sub-paths according to the azimuth and elevation angles corresponding to each sub-path; Amplitude, get the channel full-dimensional parameter set.

By preprocessing the signal, the present invention can reduce the influence of the Doppler frequency shift on the estimation of the complex amplitude of the sub-path, so that the estimated complex amplitude is more accurate. It does not require iteration, requires shorter time, and does not have the problem of converging to the local optimal solution and estimating the false path caused by improper initial value setting.

Specifically, in the channel measurement system, the transmitter sends a BPSK-modulated PN sequence in a certain frame format, and the receiver completes multi-channel reception by an array with known arrangements and antenna patterns; the 5G-NR system is similar, the transmitter is controlled by the UE Acting as and modulating the SRS (ie, sounding reference signal) according to the protocol frame format, the receiving end acts as the BS and processes the received SRS to obtain channel CFR data of multiple receiving channels.

In the embodiment of the present invention, the obtained channel mathematical model of complete parameters is expressed as:

和f_l,p分别对应第l簇内第p条子路径的复幅度和多普勒频移。

为第l簇内第p条子路径对应的方向向量，其中θ_l,p和

分别为其仰角和方位角。Among them, the time domain response of h(τ, t), L and P _l are the number of delay clusters and the number of sub-paths included in delay cluster l, respectively. The delay resolution of the system can only distinguish each delay. The sub-paths within the cluster cannot be distinguished because of the cluster, and the delay of each delay cluster is expressed as τ _l . τ and t are different descriptions of different time scales. τ is the unit of PN chip width or the sampling interval of the NR system, while t is the unit of the frame interval of PN or the slot interval of the NR system, and δ(τ) is the unit shock function.

and f _l,p correspond to the complex amplitude and Doppler frequency shift of the pth subpath in the lth cluster, respectively.

is the direction vector corresponding to the p-th sub-path in the l-th cluster, where θ _{l, p} and

are the elevation and azimuth angles, respectively.

The baseband data processed by the receiver is the down-converted baseband data sampled by multiple RF channels, and includes the array arrangement and antenna pattern information in the specified reference frame. Taking the uniform plane array placed on the XOZ plane as an example, each The signal received by the receiving channel (that is, the received PN sequence) is expressed as:

Among them, N _m (τ, t) is the additive Gaussian noise on the mth receiving channel, M and N are the array elements of the uniform area array in the X-axis and Y-axis directions, respectively, a(τ) is the transmitted signal Baseband signal expression, where τ is the unit of the system sampling interval, which is a single chip width for the channel measurement system.

为从

方向入射来波的阵列导向矢量的第m元素，其包含阵列排布信息以及天线方向图信息，可展开为：For NR systems, it is the inverse of the system bandwidth.

for from

The m-th element of the array steering vector of the incoming wave in the direction, which contains the array arrangement information and the antenna pattern information, can be expanded as:

是第m阵元方向图在

上的复幅度，a_x(m)、a_z(m)是第m阵元在XOZ平面的坐标，

是载波在第m阵元相对参考原点的距离所引起的相移。in,

is the m-th array element pattern in

The complex amplitude on , a _x (m), a _z (m) are the coordinates of the mth array element in the XOZ plane,

is the phase shift caused by the distance of the carrier at the mth element relative to the reference origin.

More specifically, as shown in Figure 2, in the channel measurement system, the transmitting end uses the full bandwidth to cyclically transmit the BPSK-modulated PN sequence in a certain frame format with a period T _s . The number of PN sequence chips is K, and a single code The slice duration is _Tp . The receiving end completes multi-channel reception by an array with known arrangement and antenna pattern. As shown in Figure 3, the transmitter of the NR system is acted by the UE. The UE modulates the SRS according to the frame format specified in the protocol while continuously sending radio frames. The occupied time domain resource position is shown in Figure 2. In the frequency domain protocol As shown in Figure 3, it is sent in a 2-comb manner. The total bandwidth of the NR system includes 272 RBs, each RB includes 12 REs, and each RE corresponds to a subcarrier with a frequency interval of 30 kHz. The receiving end is acted by the BS and processes the received SRS, and obtains channel CFR data of multiple receiving channels after processing. Due to the limited processing capacity of the implementing base station, each collection task can only collect SRS of more than 40 slots in the time domain, and the SRS of each slot can only obtain 68 consecutive RBs in the frequency domain, and the base station needs to store the data. The CFR data is down-sampled by a quarter, so each acquisition task can only obtain channel CFR data of more than 40 slots on 68*12÷2÷4=102 REs.

Preferably, when the signal data is a PN sequence, the preprocessing is: performing sliding correlation on the PN sequence. Specifically, according to the description of the channel space-time characteristics in the QuaDRiGa channel model theory, the spatial scattering environment of the wireless channel between the transmitting end moving in a small range and the stationary receiving end hardly changes. Therefore, multiple snapshots (different snapshots) are used to receive PN sequence data or multiple slots (time slots) SRS data to complete channel exploration or angle domain information collection. The survey signal received in the channel measurement system uses the local PN sequence to perform sliding correlation in each snapshot, and obtains the delay peak value of each PN sequence. The channel CFR of each slot obtained by the BS processing in the NR system is converted into the time-domain channel impulse response CIR by using the inverse fast Fourier transform IFFT.

In the channel measurement system, the received signal on the NM receiving channel is slidingly correlated with the local PN sequence:

T_p为单个码片的时间宽度，K为PN码的码长，N'(τ,t)表示接收阵列上的噪声与PN信号滑动相关的结果，由于KP_b很大，因此在后面可忽略噪声N'(τ,t)的影响。时延簇l对应的时延峰值表示为：Among them, <y(τ,t), a(τ)> means that a(τ) is used to perform sliding correlation on each receiving channel data y(τ,t) of the t-th slot, ⊙ represents the convolution operation,

T _p is the time width of a single chip, K is the code length of the PN code, N'(τ, t) is the result of the correlation between the noise on the receiving array and the PN signal slip, since KP _b is very large, it can be ignored later The effect of noise N'(τ,t). The delay peak corresponding to delay cluster l is expressed as:

In the NR system, the base station receives and processes the SRS signals received by all channels according to the OFDM architecture. The operation corresponding to the sliding correlation is to perform IFFT on the acquired CFR data to obtain the baseband CIR data of each slot. The expression is as follows:

T_s为系统采样周期，当

不是整数时，对应时延为τ_l的时延簇的功率成分会泄露到所有时延位置上。N为系统中的子载波数目，也是单个slot内的采样点数。

为每条子路表达式中不依赖角度的成分。Φ_l,p(t)为时延簇l内子路径p在第t个slot由多普勒频移造成的额外相位差，由于NR系统的应用场景多为城市宏小区这类低速移动场景，因此推导中假设系统是准静态的，即Φ_l,p(t)其不随n变化。Among them, Y(k,t) is the frequency domain received data of the t-th slot,

T _s is the sampling period of the system, when

When it is not an integer, the power component of the delay cluster corresponding to the delay τ _l will leak to all delay positions. N is the number of subcarriers in the system, and is also the number of sampling points in a single slot.

is the angle-independent component of each subpath expression. Φ _l,p (t) is the extra phase difference caused by the Doppler frequency shift of the subpath p in the delay cluster l at the t-th slot. Since the application scenarios of the NR system are mostly low-speed mobile scenarios such as urban macro cells, so The derivation assumes that the system is quasi-static, that is, Φ _l,p (t) does not vary with n.

For the NR system, since the channel CIR obtained by the IFFT of the channel CFR data of the base station has obvious peak leakage phenomenon, firstly, the cyclic prefix (Cyclic Prefix, CP) length is used to divide it into noise segment and effective signal segment. The maximum value of the amplitude is used as the noise threshold η. Use the threshold η to select the delay position Ω of the effective signal segment, and take the union of the set of delay positions on multiple receiving channels to resist the spatial fading caused by multipath on the array:

再取交集，以此来降低噪声干扰。得到新的时延位置集合：different slot

Then take the intersection to reduce noise interference. Get a new set of delay positions:

处所有通道的CIR峰值计算所有slot下的协方差矩阵：Since the leaked delay power at some delay positions may be less or there may be power leaked from multiple paths, the angle information of a large number of sub-paths is mixed and difficult to be distinguished by MUSIC. In order to obtain correct and reliable angle domain information, choosing to discard the CIR peaks at these delay positions will only lose a small amount of power of the angle information of the subpaths, but reduce the adverse effect of noise on the ZF algorithm, which can provide information for subsequent amplitude estimation. Assure. Efficient time-delay location selection methods are eigenvalue-assisted, and for arbitrary time-delay locations

Calculate the covariance matrix under all slots at the CIR peaks of all channels:

接下对

进行特征值分解，得到其最大和最小特征值：in,

next pair

Perform eigenvalue decomposition to get its maximum and minimum eigenvalues:

到集合

否则抛弃该时延位置，在后续提取角度和幅值信息时，只使用

内时延位置对应的CIR峰值。When λ _max /λ _min >η ₂ , keep the delay position

to the collection

Otherwise, discard the delay position, and only use

The CIR peak value corresponding to the inner delay position.

In one embodiment, the MUSIC algorithm of forward and backward spatial domain smoothing is used to process the peaks of multiple snapshots/slots, to distinguish sub-paths within each delay cluster and obtain their azimuth and elevation angles. First, vectorize the peaks of multiple channels:

为：Then select the sub-array of the forward and backward spatial domain smoothing algorithm, wherein the array elements of M columns contain p _x isomorphic sub-arrays interlaced with each other, and the array elements of N rows contain p _{z iso} -morphic sub-arrays interlaced with each other. The two-dimensional planar array is divided into p _x xp _z sub-area arrays of size M _s =M+1-p _x columns and N _s =N+1-p _z rows. Using V _l (t) of T snapshots/slots to obtain the covariance matrix after smoothing in the forward and backward spatial domains

for:

表示Kronecker积，(·)^H表示取共轭转置，

为M_s×M_s的单位矩阵；

为M_s×M_s的置换矩阵，它的反对角线上元素为1，其余元素均为0。in,

represents the Kronecker product, ( ) ^H represents the conjugate transpose,

is the identity matrix of M _s ×M _s ;

M _s ×M _s permutation matrix, its antidiagonal elements are 1, and other elements are 0.

进行特征值分解，根据最小特征值的模和接收通道的信噪比设置门限ξ_l，绝对值大于门限的P_l个特征值对应时延簇l内的P_l条子路径，用小于门限的特征值对应的特征向量v_i,i＝P_l+1,...NM构造正交空间

时延簇l内的P_l条子路径各自对应的阵列导向向量

均与B_⊥正交。为时延簇l构造的MUSIC伪谱

并对其进行二维搜索寻找峰值：next to

Perform eigenvalue decomposition, set a threshold ξ _l according to the modulo of the smallest eigenvalue and the signal-to-noise ratio of the receiving channel, the P _l eigenvalues whose absolute value is greater than the threshold correspond to P _l subpaths in the delay cluster l, and use the features smaller than the threshold The eigenvectors corresponding to the values v _i , i=P _l +1,...NM construct an orthogonal space

Array steering vectors corresponding to P _l subpaths in delay cluster l

Both are orthogonal to B _⊥ . MUSIC Pseudospectrum Constructed for Delay Cluster l

and do a 2D search on it looking for peaks:

并将

该位置周围半径Δ置为零，再搜索下一峰值直至得到P_l条子路径的仰角和方位角。Δ的取值决定了将多大角度范围内子路径的合成一条径，其值得选取决于子阵列尺寸、snapshot/slot数目以及噪声，在对角度估计精度要求不高的一般应用情况下可取1～5°。After each peak position is obtained, record its azimuth and elevation positions

and will

The radius Δ around the position is set to zero, and the next peak is searched until the elevation and azimuth angles of P _l subpaths are obtained. The value of Δ determines how many sub-paths in the range of angles are combined into one path. The value of the value depends on the size of the sub-array, the number of snapshots/slots, and the noise. In general applications where the accuracy of angle estimation is not high, it can be taken from 1 to 5. °.

In another embodiment, when the signal data is CFR data, the preprocessing is: using the cyclic prefix length to divide the CIR data into a noise segment and a valid signal segment; determining a first noise threshold according to the noise segment, and according to the first noise threshold Select the first delay position of the valid signal segment; determine the union of the first delay positions of the CIR data of the same time slot received by different receiving channels, and obtain the first delay position set; for multiple first sets of different time slots Take the intersection to obtain the second delay position set; calculate the covariance matrix corresponding to each element in the second delay position set, and perform eigenvalue decomposition on the covariance matrix to obtain the maximum eigenvalue and the minimum eigenvalue; calculate the maximum eigenvalue The ratio of the value to the minimum eigenvalue, selecting elements with a ratio greater than the second noise threshold in the second delay position set to form a third delay position set, and using the third delay position set as the delay peak value of the signal data.

Use multiple snapshot PN sequences to receive data or multiple slot SRS data to complete channel exploration or angle domain information collection. The survey signal received in the channel measurement system is slidingly correlated with the local PN sequence in each snapshot. The channel CFR of each slot obtained by the BS processing in the NR system is converted into the time domain CIR using IFFT.

The system delay resolution is used to distinguish delay clusters. The distinction of sub-paths in the cluster and the estimation of azimuth and elevation angles are accomplished by the MUSIC algorithm for smoothing forward and backward spatial domains for uniform plane arrays. The sub-array selection methods of uniform plane arrays are as follows: shown in Figure 5. Divide the original array of size N×M into p _x ×p _z subarrays of size N _S × _MS . Finally, the ZF algorithm is used to calculate the complex amplitude of each subpath, and the phase difference between the complex amplitudes of the subpaths obtained by different snapshots or slots is used to estimate the Doppler frequency shift of the subpath.

Using the known subpath angles and ZF algorithm to analyze the complex amplitude and Doppler frequency shift of P _l subpaths in delay cluster l, first convert the noise-free part of V _l (t) into matrix product form:

然后利用ZF算法得到

的最小二乘估计：It can be obtained by using the angle estimation results of subpaths in cluster l

Then use the ZF algorithm to get

The least squares estimate of :

计算簇l内子路径的多普勒频移估计：Use T snapshots/slots

Compute Doppler shift estimates for subpaths within cluster l:

Among them, κ(t) is a factor used to cancel the 2π periodicity of the phase, and its calculation expression is:

和

后，在收发已有定时同步的基础上可以很容易的计算

进而得到复幅度估计

此外，无论是信道测量或是NR系统，要估计子路多普勒频移时，slot之间的间隔或PN不同帧之间的帧间隔T_f均需要满足：Got

and

After that, it can be easily calculated on the basis of the existing timing synchronization of sending and receiving

and then get the complex magnitude estimate

In addition, whether it is a channel measurement or an NR system, when estimating the sub-channel Doppler frequency shift, the interval between slots or the frame interval T _f between different PN frames needs to satisfy:

Among them, f _d is the maximum Doppler frequency shift that exists in the channel in the application scenario.

The system delay resolution can only distinguish between delay clusters, and the sub-paths within the cluster all carry the same signal, so the received signals of each sub-path are coherent. It is necessary to complete the distinction and calculate the azimuth and elevation angles of the MUSIC algorithm for the forward and backward spatial smoothing of the uniform plane array. Then, the MUSIC algorithm for the forward and backward spatial domain smoothing of the uniform plane array is completed to distinguish and calculate its azimuth and elevation. The phase difference between the complex amplitudes of the subpaths obtained from different snapshots or slots is used to estimate the Doppler shift of the subpaths.

More specifically, in the face of channel measurement applications in high-speed mobile scenarios, since the Doppler frequency shift on the subpath cannot be ignored, only the time of a single PN chip will cause a relatively obvious phase change, making the PN sliding correlation. The autocorrelation peak value of , is reduced, and there will be a large error in calculating the complex amplitude using this correlation peak value.

Therefore, first adjust the channel model of each snapshot as follows:

并用它与接收信号做滑动相关，从而抵消多普勒频移的影响。After obtaining the subpath Doppler shift estimate, the result can be used to construct a local PN sequence with Doppler shift

And use it to do sliding correlation with the received signal, so as to cancel the influence of Doppler frequency shift.

This is divided into two specific cases. When the Doppler frequency shift between sub-paths in the cluster is quite different (special scenario), the Doppler frequency shift of each sub-path can be used to construct a corresponding local sequence to extract the sub-paths that do not contain The complex magnitude of the Doppler shift effect. For example, to extract the pth subpath in the lth cluster, first perform sliding correlation to obtain the peak value of the corresponding delay cluster l:

的相关结果，A为时延簇l内其他子路径对p子路径滑动相关峰值的干扰，当子路径之间的多普勒频移较大时，该干扰随着PN序列的长度增大而减小。接下来由已估计出的子路径角度和时延l的峰值向量计算消除多普勒频移后的复幅度：Among them, N”(τ, t) is the noise and

The correlation result, A is the interference of other subpaths in the delay cluster l to the sliding correlation peak of the p subpath. When the Doppler frequency shift between the subpaths is large, the interference increases with the length of the PN sequence. decrease. Next, calculate the complex amplitude after eliminating the Doppler shift from the estimated subpath angle and the peak vector of the delay l:

为第l簇内第p子路径对应的导向矢量估计。in,

Estimate the steering vector corresponding to the p-th subpath in the l-th cluster.

When the difference in Doppler frequency shift between sub-paths is small, such as a general scenario such as an urban macro cell, the influence of factor A cannot be ignored, so it is impossible to eliminate the influence of Doppler frequency shift for each path individually, but The influence of Doppler frequency shift on subpaths in each delay cluster is cancelled as much as possible. In order to ensure the elimination effect of the path Doppler frequency shift with large amplitude, the average Doppler frequency shift on the delay cluster l is determined by means of weighted average.

对接收序列做滑动相关得到

按照之前的复幅度估计处理流程，用已经求得的子路径角度和峰值向量完成多普勒频移部分消除的子路径复幅度估计。Next use

Perform sliding correlation on the received sequence to get

According to the previous complex amplitude estimation processing flow, the sub-path complex amplitude estimation for partial elimination of Doppler frequency shift is completed using the obtained sub-path angles and peak vectors.

As shown in Fig. 4, for the channel sounding application based on the PN sequence, the Doppler frequency shift estimation result of the subpath is used to eliminate the influence of the Doppler frequency shift on the estimation of the complex amplitude of the subpath. For the NR system, since the channel data estimated by LS has a power leakage phenomenon in the time domain as shown in Figure 6, the eigenvalue-assisted time-delay location selection method is used for preprocessing to determine the high signal-to-noise ratio and angle information. Easy-to-extract time-delay locations, thereby reducing the adverse effects of CIR leakage.

Specifically, when the signal data is a PN sequence, after calculating the complex amplitude of the subpath according to the azimuth angle and the elevation angle corresponding to each subpath, the method further includes: calculating the phase difference of different time slots of each subpath; determining the phase difference of the subpath according to the phase difference. Doppler frequency shift; construct a local PN sequence based on the Doppler frequency shift; use the local PN sequence to perform sliding correlation on the PN sequence, and continue to execute until the complex amplitude of each subpath is obtained again.

In one embodiment, when the Doppler frequency shift difference between each subpath is less than the difference threshold: the weighted average method is used to calculate the average Doppler frequency shift of each delay cluster; based on the average Doppler frequency Shift to generate a local PN sequence.

Finally, according to the channel parameter estimation results obtained from multiple snapshots/slots, calculate the channel parameters of large and small scales and their probability and statistical distribution, and draw the spectrum diagrams such as Power Azimuth Spectrum (PAS) in Figure 8 and Figure 9. When the signal data is a PN sequence, the channel full-dimensional parameter set further includes: the Doppler frequency shift of each subpath.

To sum up, the method proposed in the present invention uses the space-time separable characteristics of the channel to distinguish each subpath in the channel and further extracts parameters. In the delay domain, PN sliding correlation or IFFT is used to distinguish the paths and obtain the delay information of the paths. In the airspace, the MUSIC algorithm based on forward and backward airspace smoothing is introduced to distinguish the sub-paths within the path and obtain their azimuth and elevation angles. Then use the angle information and the correlation peaks or CIR peaks of multiple receiving channels to analyze the subpath complex amplitude, and finally estimate the corresponding Doppler according to the phase difference of the subpath complex amplitude in different snapshots/slots frequency shift. In addition, the introduction of Doppler frequency shift cancellation in the complex amplitude estimation module can obtain more accurate complex amplitudes in high-speed mobile scenarios. In the delay estimation module, an eigenvalue-assisted delay location selection preprocessing method is introduced, so that the proposed method can more effectively extract the angle domain information in the channel data of the NR system. Moreover, the simulation verifies the efficiency of the method, and the processing results of the measured data demonstrate the effectiveness of the angle domain information extracted by the method.

The following content is the simulation verification of the method of the present invention. The channel parameter table shown in Table 1 is set to simulate the parameters of each subpath in the wireless multipath channel, and the channel parameters are extracted by SAGE and the proposed method respectively. Both methods use a PN sequence of length K=511, and the width of each chip is dt=3.69*10 ^-6 s. The azimuth and elevation angles of the subpaths are randomly selected between 1° and 180°. According to the description of QuaDRiGa, the Doppler frequency shifts and angles of sub-paths in the same cluster are mostly close, so this is taken into account when setting co-delay paths. In addition, the default number of SAGE paths is known, and the proposed method defaults to appropriate settings for various thresholds, and other related configurations are shown in Table 2.

Table 1

Table 2

Add paths in turn and record the running results of each proposed algorithm and SAGE. When the number of paths is different, the running time of the two algorithms and the number of iterations of SAGE are shown in Figure 7. The specific estimation is recorded in Table 3, and it is considered that When the estimated delay deviation of each path is less than dt, the angle deviation is less than 1°, the Doppler frequency shift is less than 1Hz, and the amplitude estimation error is less than 0.01, the estimation is correct.

It can be seen from Figure 7 that the proposed algorithm and SAGE estimation have similar performance but shorter running time when the number of paths is small. The convergence of SAGE depends on the composition of its sub-paths. The number of iterations required for convergence cannot be determined in advance. When there is a simultaneous arrival path, the estimation effect is not good, and it is easy to converge to the local optimal solution, which does not conform to the actual situation. It is estimated that it fails, and the method proposed in the present invention deals with this problem more in line with the actual situation. The initial running time of the SAGE algorithm increases with the increase of the number of paths, while the running time of the proposed method mainly increases with the increase of the number of clusters, and the iteration time of the SAGE algorithm is mainly related to the number of iterations and the number of paths.

table 3

As shown in Figure 9, for the data collected by the base station in the single-user Line of Sight (LOS) scenario, the angular power spectrum reconstructed after obtaining the angular domain parameters using the proposed method, the position of the circle with a label represents the base station side 32 The directions of the 32 beams formed by each receiving unit using the Discrete Fourier Transform (DFT) codebook. As shown in Figure 8, the angle spectrum is reconstructed after extracting the angle domain parameters by directly bringing the CFR data of each channel of the base station into the forward and backward spatial domain smoothing MUSIC algorithm. It can be seen that the main power region of the angular spectrum constructed by the method of the present invention is the same as that of the latter, but the angular power spectrum constructed by the method of the present invention is more refined. In order to verify its effectiveness, the power ratio of the received CFR on the 32 beams was reconstructed with the respective extracted angle domain parameters and compared with the power ratio on the original 32 beams. The results are shown in Figure 10. The restored power ratio of the inventive method is closer to the original beam power ratio. In addition, it can be seen that in the main power region of the angular spectrum, the beam power reconstruction performance with strong received power is better, which verifies its effectiveness.

Compared with the parameter estimation algorithm based on likelihood function which introduces EM iteration like SAGE, the method of the present invention does not need iteration, the time required is shorter, and there is no convergence to the local optimal solution caused by improper initial value setting to estimate The problem of false trails. Compared with various estimation algorithms about channel parameters, it can effectively distinguish the subpaths in the delay cluster and obtain the parameters such as azimuth angle, elevation angle, complex amplitude and Doppler frequency shift. The present invention not only utilizes the characteristics of 5G large bandwidth, and uses the high resolution capability of the time delay domain to distinguish the delay clusters in the channel, but also utilizes the characteristics of the 5G large-scale antenna, and uses the front and rear of the coherent signal source with high angular resolution capability. The MUSIC algorithm is smoothed to the spatial domain to distinguish sub-paths within the cluster, so a large number of sub-paths can be extracted from the temporal and spatial domains and their parameters can be used to accurately characterize the channel. In addition, for 5G channel bathymetric applications based on PN sequences, a Doppler frequency cancellation method is provided, which reduces its impact on the estimation of subpath complex amplitudes. For the NR system, a corresponding preprocessing method is provided to extract the angle domain parameters from the channel CIR data by using this algorithm.

The present invention also discloses a channel full-dimensional parameter extraction device that does not depend on the likelihood function, as shown in FIG. 11 , comprising: a receiving module 210 for receiving signal data through multiple channels; wherein the signal data is a PN sequence or a CFR data; the preprocessing module 220 is used to preprocess the signal data to obtain the time delay peak value of the signal data; the first calculation module 230 is used to take the time delay peak value as the input information, and use the forward and backward spatial domain smoothing MUSIC algorithm to calculate the time delay the azimuth angle and the elevation angle corresponding to each subpath in the extension peak; the second calculation module 240 is used to calculate the complex amplitude of the subpath according to the azimuth angle and the elevation angle corresponding to each subpath; the aggregation module 250 is used to aggregate the time delay peak, The azimuth and elevation angles corresponding to each subpath, as well as the complex amplitude, obtain the channel full-dimensional parameter set.

It should be noted that the information exchange, execution process and other contents between the modules of the above-mentioned device are based on the same concept as the method embodiments of the present application, and the specific functions and technical effects brought by them can refer to the method embodiments section for details. It is not repeated here.

Those skilled in the art can clearly understand that, for the convenience and brevity of the description, only the division of the above-mentioned functional modules is used as an example for illustration. The internal structure of the device is divided into different functional modules to complete all or part of the functions described above. Each functional module in the embodiment may be integrated in one processing unit, or each unit may exist physically alone, or two or more units may be integrated in one unit, and the above-mentioned integrated units may be implemented in the form of hardware. , can also be implemented in the form of software functional units. In addition, the specific names of the functional modules are only for the convenience of distinguishing from each other, and are not used to limit the protection scope of the present application. For the specific working processes of the units and modules in the above-mentioned system, reference may be made to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

The invention also discloses a channel full-dimensional parameter extraction device that does not depend on the likelihood function, comprising a memory, a processor, and a computer program stored in the memory and running on the processor. A method for channel full-dimensional parameter extraction that does not depend on the likelihood function.

The device may be a computing device such as a desktop minicomputer, a notebook, a palmtop computer, and a cloud server. The apparatus may include, but is not limited to, a processor, a memory. Those skilled in the art can understand that the apparatus may include more or less components, or combine certain components, or different components, for example, may also include input and output devices, network access devices, and the like.

The processor may be a central processing unit (Central Processing Unit, CPU), and the processor may also be other general-purpose processors, digital signal processors (Digital Signal Processors, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), Off-the-shelf programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, and the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory may in some embodiments be an internal storage unit of the device, such as a hard disk or memory of the device. In other embodiments, the memory may also be an external storage device of the device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a Secure Digital (SD, SD card) equipped on the device. ) card, flash card (Flash Card) and so on. Further, the memory may also include both an internal storage unit of the apparatus and an external storage device. The memory is used to store an operating system, an application program, a boot loader (Boot Loader), data, and other programs, such as program codes of the computer program, and the like. The memory may also be used to temporarily store data that has been output or is to be output.

The invention also discloses a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, realizes the above-mentioned method for extracting full-dimensional parameters of a channel that does not depend on a likelihood function.

The computer-readable medium may include at least: any entity or device capable of carrying the computer program code to the photographing device/terminal device, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM) , Random Access Memory), electrical carrier signals, telecommunication signals, and software distribution media. For example, U disk, mobile hard disk, disk or CD, etc. In some jurisdictions, under legislation and patent practice, computer readable media may not be electrical carrier signals and telecommunications signals.

Claims (10)

1. a channel full-dimensional parameter extraction method that does not depend on likelihood function, is characterized in that, comprises the following steps: Receive signal data through multiple channels; wherein the signal data is PN sequence or CFR data; Preprocessing the signal data to obtain a time delay peak value of the signal data; Taking the time delay peak value as input information, the azimuth angle and the elevation angle corresponding to each subpath in the time delay peak value are calculated by adopting the forward and backward spatial smoothing MUSIC algorithm; Calculate the complex amplitude of the sub-path according to the corresponding azimuth and elevation angles of each of the sub-paths; The time delay peak value, the azimuth angle and elevation angle corresponding to each subpath, and the complex amplitude are collected to obtain a channel full-dimensional parameter set. 2. a kind of channel full-dimensional parameter extraction method that does not depend on likelihood function as claimed in claim 1 is characterized in that, when described signal data is PN sequence, described preprocessing is: A sliding correlation is performed on the PN sequence. 3. a kind of channel full-dimensional parameter extraction method that does not depend on likelihood function as claimed in claim 1 or 2, is characterized in that, before the described CFR data is preprocessed, also comprises: Perform IFFT on the CFR data to obtain CIR data corresponding to the CFR data. 4. a kind of channel full-dimensional parameter extraction method that does not depend on likelihood function as claimed in claim 3 is characterized in that, when described signal data is CFR data, described preprocessing is: dividing the CIR data into a noise segment and a valid signal segment using a cyclic prefix length; Determine a first noise threshold according to the noise segment, and select a first delay position of the valid signal segment according to the first noise threshold; Determine the union of the first delay positions of the CIR data in the same time slot received by different receiving channels, and obtain the first delay position set; Taking the intersection of multiple first sets of different time slots to obtain a second set of delay positions; Calculate the covariance matrix corresponding to each element in the second delay position set, and perform eigenvalue decomposition on the covariance matrix to obtain a maximum eigenvalue and a minimum eigenvalue; Calculate the ratio of the maximum eigenvalue and the minimum eigenvalue, select elements whose ratio is greater than the second noise threshold in the second delay position set to form a third delay position set, and use the third delay position set as Delay peak value of the signal data. 5. a kind of channel full-dimensional parameter extraction method that does not depend on likelihood function as claimed in claim 2 is characterized in that, when described signal data is PN sequence, according to the azimuth angle corresponding to each described subpath and the elevation angle after calculating the complex magnitude of the subpath, it also includes: calculating the phase difference between different time slots of each of the subpaths; determining the Doppler shift of the subpath according to the phase difference; Construct a local PN sequence based on the Doppler shift; The sliding correlation is performed on the PN sequence by using the local PN sequence, and the execution is continued until the complex amplitude of each of the sub-paths is obtained again. 6. The method for extracting full-dimensional parameters of a channel that does not depend on a likelihood function according to claim 5, wherein when the Doppler shift difference between each of the sub-paths is less than a difference threshold : Calculate the average Doppler shift of each delay cluster by using the weighted average method; A local PN sequence is generated based on the average Doppler shift. 7. a kind of channel full-dimensional parameter extraction method that does not depend on likelihood function as claimed in claim 1 is characterized in that, when described signal data is PN sequence, described channel full-dimensional parameter set also comprises: Doppler shift for each of the subpaths. 8. A channel full-dimensional parameter extraction device that does not depend on a likelihood function, is characterized in that, comprising: a receiving module for receiving signal data through multiple channels; wherein the signal data is PN sequence or CFR data; a preprocessing module, configured to preprocess the signal data to obtain a delay peak value of the signal data; a first calculation module, configured to use the time delay peak as input information, and calculate the azimuth angle and elevation angle corresponding to each subpath in the time delay peak by adopting the forward and backward spatial smoothing MUSIC algorithm; The second calculation module is used to calculate the complex amplitude of the subpath according to the azimuth angle and the elevation angle corresponding to each of the subpaths; The aggregation module is configured to aggregate the delay peak value, the azimuth angle and elevation angle corresponding to each subpath, and the complex amplitude to obtain a channel full-dimensional parameter set. 9. A channel full-dimensional parameter extraction device that does not depend on a likelihood function, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processing When the computer executes the computer program, a method for extracting full-dimensional parameters of a channel that does not depend on a likelihood function according to any one of claims 1-7 is implemented. 10. A computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, a method according to any one of claims 1-7 is implemented A method for channel full-dimensional parameter extraction that does not depend on likelihood functions.

Images