CN-122027401-A - Adaptive-fast convergence sparse recovery algorithm for millimeter wave large-scale MIMO

CN122027401ACN 122027401 ACN122027401 ACN 122027401ACN-122027401-A

Abstract

The invention discloses a self-adaptive-fast convergence sparse recovery algorithm of millimeter wave large-scale MIMO, which is a synchronous orthogonal matching tracking algorithm of self-adaptive multidimensional joint correlation. The algorithm significantly accelerates the convergence process by selecting multiple atoms in parallel in a single iteration. And the adaptive termination criterion can be realized without prior information such as channel sparsity, noise power and the like. Simulation results show that the convergence rate of the algorithm is improved by about 50% compared with that of the traditional SOMP algorithm, and the equivalent normalized mean square error can be obtained under the condition of lacking channel sparsity and noise power priori information. Particularly, in the environment with the signal-to-noise ratio of 8-23dB, the sparsity estimation error can be effectively reduced even if the channel noise power information is not available, and the sparsity estimation error is superior to the traditional algorithm. This shows that the algorithm of the invention has strong adaptability and can effectively cope with the dynamic change of the channel.

Inventors

ZHENG ZHIAN
LI JIAJUN
WU XINQI
ZHU JUNJIE
YAN JIABAO

Assignees

中南林业科技大学

Dates

Publication Date: 20260512
Application Date: 20260214

Claims (7)

1. The adaptive-fast convergence sparse recovery algorithm is characterized in that in a millimeter wave massive MIMO-OFDM system, a distributed compressed sensing based sparse recovery synchronous orthogonal matching pursuit (Simultaneous Orthogonal Matching Pursuit ‌, SOMP) algorithm framework is adopted, an observation signal vector set formed by an observation signal vector of each pilot subcarrier on a pilot subcarrier set and received by a base station from one User Equipment (UE) is used as input, a multidimensional joint correlation (Multi-Dimensional Joint Correlation) atomic support set construction algorithm is adopted on the basis of a sensing matrix, an atomic support set is gradually and accurately constructed through multiple iterations, and coefficients of atoms corresponding to each pilot subcarrier in the atomic support set are recovered, so that the purposes of channel reconstruction and sparse recovery on each pilot subcarrier are achieved.
2. The adaptive-fast convergence sparse recovery algorithm for mmwave massive MIMO according to claim 1, comprising the steps of: step 1, initializing an atom support set Is an empty set and sets the iteration counter to be ; The inputs to the system are: Represented in the pilot subcarrier set On the mth pilot frequency subcarrier, the User Equipment (UE) transmits pilot frequency signals in continuous Q OFDM symbol time slots, and the base station obtains observation signal vectors containing noise in the corresponding Q time slots: ; the output of the system is the channel reconstructed in the angle domain , , wherein, To be in pilot subcarrier set Inner subcarrier Up the reconstructed channel; In order to perceive the matrix of the device, Is a set of atomic supports which are arranged in a row, To support the collection by atoms Is indexed with The set of atoms on the top of the table, Is an atomic coefficient; Step2, effective atomic detection: In each iteration, based on the current residual signal, a Multi-dimensional joint correlation (Multi-Dimensional Joint Correlation) atom support set construction algorithm is adopted, and joint judgment is carried out on candidate atoms through joint sequencing correlation coefficients JRCC (Joint Ranking Correlation Coefficient, JRCC) and joint power threshold correlation coefficients JPTCC (Joint Power Threshold Correlation Coefficient, JPTCC), so that a candidate atom set of the current iteration is obtained And outputs the maximum ranking correlation coefficient of all candidate atoms ; Step 3, support set expansion: The valid atomic index detected in the current iteration is incorporated into the existing joint support set: ; and 4, orthographic projection: On the updated joint support set, the least square method is adopted to carry out orthogonal projection estimation on coefficients of each atom: pinv is a function of calculating the pseudo-inverse of a matrix; and 5, updating residual errors: using the current support set Atomic coefficient of Based on Reconstructing the signal and converting the reconstructed signal from the original signal Is removed to obtain an updated residual signal , Updated residual signal For the next iteration; if the termination criterion (namely the termination condition) is not met, returning to the step 2 to continue iteration, otherwise, terminating the iteration; termination criteria, the number of atoms meeting the joint ordering correlation coefficient in the current iteration is lower than a preset threshold, namely Or no new valid atoms are detected.
3. The adaptive-fast convergence sparse recovery algorithm for mmwave massive MIMO according to claim 1, wherein in the termination criterion, the threshold is , M is the number of subcarriers.
4. The adaptive-fast convergence sparse recovery algorithm for millimeter wave massive MIMO according to claim 1, wherein in the multi-dimensional joint correlation atomic support set construction algorithm in step 2, the input is (1) the first Residual signal of multiple iterations (2) Existing support sets ; In the first place In the iteration, the current residual signal and the existing support set are taken as inputs, and candidate atoms are gradually screened through a joint sequencing correlation coefficient (JRCC) and a Joint Power Threshold Correlation Coefficient (JPTCC), so that the effective atoms newly added in the iteration of the round are determined.
5. The adaptive-fast convergence sparse recovery algorithm of millimeter wave massive MIMO according to claim 4, wherein the flow of the multidimensional joint-related atomic support set construction algorithm is as follows: step 21, constructing joint candidate atoms: based on the residual signal, using JRCC parameters to calculate the sorting correlation coefficient of each atom Constructing a preliminary joint candidate atom set ; Step 22, expanding a candidate atom set: Combining the joint candidate atom set constructed in step 21 with the existing support set to form an extended candidate atom set for further decision: ; step 23, combining candidate atom selection (strict JPTCC parameters): for an extended candidate atom set Firstly, calculating power value of each candidate atom, then updating current maximum path power, and finally adopting strict JPTCC parameters based on the updated maximum path power Screening to obtain a preliminary support set of the round of iteration ; Step 24, loose parameter rollback mechanism: When the support set obtained by screening in step 23, the algorithm further determines whether the maximum joint ordering correlation coefficient of the current iteration is satisfied If the conditions are satisfied, the potential effective atoms are considered to exist, and JPTCC parameters are determined to be strict Switching to loose And re-executing the maximum path power threshold verification mechanism (formulas (12) - (14)) based on the updated parameters without changing the candidate atom set to update the effective atom-support set for the current iteration The loose power threshold back-off mechanism can effectively ensure that even atoms with lower power are not missed; step 25, adding an atom support set Output of the newly added atom support set 。
6. The adaptive-fast convergence sparse recovery algorithm of mmwave massive MIMO of claim 5, wherein in step 21, the set of pilot subcarriers is Wherein For each subcarrier in the set, for the number of subcarriers Calculating a matrix Is associated with the residual vector on the subcarrier Is ranked, and the top ranking is selected Atomic index of (a) as candidate vector Then, combining the candidate vectors of all sub-carriers to form an atomic candidate matrix : (6); Wherein the matrix Simultaneously contains real atoms and interference atoms, defines a joint candidate atom index vector Which is composed of matrix After removal of the repeat atoms, all atoms present in (a) consist of: (7) Wherein the method comprises the steps of The number of different atomic indexes after the union operation is represented, From the matrix Chinese statistical joint candidate atomic index vector Frequency of occurrence of each element (candidate atomic index) of (a) to obtain : (8a) (8b); Representing candidate atomic index vectors A vector of the number of the first q bits (joint ordering correlation coefficient JRCC) in M subcarriers, wherein the kth element thereof Representing candidate atoms Is not limited to the above-mentioned JRCC, Representation of Maximum JRCC of all candidate atoms in (if) Exceeding a preset threshold ( Representing a downward rounding), then the corresponding atomic index is added to the joint candidate set for the current iteration Wherein, the method comprises the following steps of, In order to set the parameters to be in the preset, As the number of subcarriers of the pilot subcarrier set, Representation pair And (3) with Is rounded down; (9)。
7. The adaptive-fast convergence sparse recovery algorithm for mmwave massive MIMO according to claim 5 or 6, characterized in that, In step 23, atom candidate set Each atom in (1) Power on subcarriers Obtained by orthogonal projection, the calculation formula is: (10) Wherein the method comprises the steps of Representing modular operation of vector elements, atom candidate set Wherein Representing the number of atoms in the collection; Representing a slave perception matrix The selected index belongs to Is of the column vector of (2) The matrix being formed, i.e. , Representing element-wise squares of vectors, correspondingly, when When the maximum path power on the first subcarrier is expressed as: (11) When (when) At the time of the first The maximum path power on the sub-carriers is equal to the maximum value of the square of the correlation value between each atom in the sensing matrix and the observation vector Wherein, the method comprises the steps of, Representing a perception matrix Middle (f) Conjugate transpose of individual atoms (i.e. column vectors), Ranging from 1 to the perception matrix The pilot frequency subcarrier set is Wherein For the number of sub-carriers, For the index of the pilot sub-carrier, ; To refine the candidate set, the path power of each atom is compared to a maximum value, and only atoms with sufficient path power are retained to be included in the support set by comparing the path power of each candidate atom to a threshold value Comparing to obtain the number of sub-carriers with the atomic path power exceeding the threshold value, and recording as The calculation mode is as follows: (12) Wherein, the An indicator variable (binary indicator variable) for two values, for indicating candidate atoms In the first place Whether the path power on a subcarrier exceeds a threshold for maximum power The calculation mode is as follows: (13) Wherein the method comprises the steps of And Respectively represent candidate atoms Path power of (c) and (d) Maximum path power on a subcarrier; Is defined as JPTCC when In which In order to set the parameters to be in the preset, And, the candidate atom Is regarded as effective, and the effective atomic support set obtained by the maximum path power threshold verification mechanism is recorded as The definition is: (14); And Respectively take out , 。

Description

Adaptive-fast convergence sparse recovery algorithm for millimeter wave large-scale MIMO Technical Field The invention relates to a self-adaptive-fast convergence sparse recovery algorithm of millimeter wave massive MIMO. Background The fusion of millimeter wave massive MIMO with orthogonal frequency division multiplexing (Orthogonal Frequency Division Multiplexing, OFDM) is one of the key support technologies for fifth generation (5G) and future sixth generation (6G) mobile communication systems, capable of providing ultra wideband transmission and supporting extremely high data rates. In practical deployment, combining massive MIMO with beamforming techniques is critical to improving system coverage and overall throughput performance. Efficient beamforming typically relies on accurate spatial precoding design, and the implementation of spatial precoding is highly dependent on accurate Channel State Information (CSI), which is typically obtained by channel estimation. To balance system performance, hardware complexity, and power consumption, hybrid analog-to-digital precoding architectures are widely employed. However, since the number of radio frequency links is generally limited relative to the number of antennas, it is difficult for the system to make a complete observation of the high-dimensional spatial channel, and thus a large pilot overhead is inevitably introduced in channel estimation. Compressed sensing (Compressed Sensing, CS) theory utilizes sparse scattering characteristics of channels, effectively reduces pilot overhead required for channel estimation, and has been widely used in correlation research. Since greedy algorithms have low computational complexity, methods such as OMP and SOMP have been widely used for channel estimation for millimeter wave massive MIMO. Such methods typically transform the physical domain channel into an angular domain representation by a spatial fourier transform. Common support (common support) assumptions assume that all subcarriers share the same non-zero support set in the angular domain, which allows the multi-subcarrier channel estimation to be modeled as a joint sparse recovery problem. However, these methods often rely on known channel prior information (e.g., sparsity or noise power), which may not be available in a practical scenario. Meanwhile, compressed sensing is also applied to a Single-Input Single-Output (SISO) OFDM system, which estimates by using sparseness of a time domain channel. There is a literature that proposes a parallel atomic selection method that does not require prior information, but still easily interferes with atomic misselection under high signal-to-noise conditions. Therefore, it is necessary to design a new adaptive-fast convergence sparse recovery algorithm for millimeter wave massive MIMO. Disclosure of Invention The invention aims to provide a self-Adaptive-fast convergence sparse recovery algorithm of millimeter wave massive MIMO, which is also called as a synchronous orthogonal matching pursuit algorithm (Adaptive Multi-Dimensional Joint Correlation-based SOMP, AMDJC-SOMP) of self-Adaptive multidimensional joint correlation, so as to enhance the self-adaptability of the algorithm and effectively cope with the dynamic change of a channel. In a millimeter wave large-scale MIMO-OFDM system, a distributed compressed sensing based sparse recovery synchronization orthogonal matching pursuit (Simultaneous Orthogonal Matching Pursuit ‌, SOMP) algorithm framework is adopted, an observation signal vector set which is received by a base station and is formed by an observation signal vector of each pilot subcarrier on a pilot subcarrier set is taken as input, a multidimensional joint correlation (Multi-Dimensional Joint Correlation) atomic support set construction algorithm is adopted on the basis of a sensing matrix, an atomic support set is gradually and accurately constructed through multiple iterations, and coefficients of atoms corresponding to each pilot subcarrier in the atomic support set are recovered, so that the purposes of channel reconstruction and sparse recovery on each pilot subcarrier are achieved. The output of the present invention includes an atom support set, coefficients for each atom on each pilot subcarrier in the atom support set, and channels on each pilot subcarrier reconstructed from the support set and the coefficients for each atom. Based on the idea of distributed compressed sensing, participating in the channel estimation process is a set of pilot subcarriers consisting of a plurality of pilot subcarriers, each subcarrier on the set having a received observation vector. In the invention, the self-adaption and fast convergence characteristics are mainly embodied in the construction of an atomic support set. Unlike traditional SOMP algorithm, which selects only a single atom and relies on prior information such as sparseness or noise power as termination criterion, the invention selects multiple atoms i