CN106059730A - Adaptive pilot frequency structure optimization design method based on sparse channel estimation - Google Patents

Abstract

The invention provides an adaptive pilot frequency structure optimization design method based on sparse channel estimation with the purpose of addressing the problem of real-time selecting optimal pilot frequency structure under a specific wireless channel. The method is different from traditional determinant pilot frequency structure design method in that the method considers variability of the wireless channel in actual transmission, combines the design of channel structure and channel estimation, such that the optimized pilot frequency structure has higher estimation precision and higher suitability. A simulation experiment result shows that the adaptive pilot frequency structure optimization design method, with proper increasing of system operation complexity as a cost, effectively increases the precision of sparse channel estimation, enables the optimized pilot frequency structure to be more representative and has better engineering application potential.

Description

An Adaptive Pilot Structure Optimization Design Method Based on Sparse Channel Estimation

technical field

The invention belongs to the communication field, in particular to an adaptive pilot structure optimization design method in the sparse channel estimation process.

Background technique

In Orthogonal Frequency Division Multiplexing (OFDM) wireless mobile communication systems, wireless channels usually exhibit sparseness. Since the traditional channel estimation methods such as the least squares (LS) method and the minimum mean square error (MMSE) method are only suitable for the estimation of dense channels, the performance of traditional channel estimation methods is not ideal when the channel is sparse. Therefore, channel estimation algorithms under sparse channels have become a new research hotspot in the field of communication.

Compressed Sensing (CS) is a new and promising theory in the field of applied mathematics in recent years. It has been widely studied and applied in many fields. According to the theory of compressed sensing, when a signal can be sparsely represented in a specific space, it can be sampled at a rate much lower than the Nyquist sampling rate and reconstructed with high probability by an optimized method. . It can be seen that when the wireless channel shows sparseness, the channel estimation process in OFDM system is similar to the reconstruction process of the original sparse signal in compressed sensing. Therefore, applying the CS theory to the channel estimation algorithm of the OFDM system will be different from the traditional channel estimation algorithm and greatly improve the channel estimation performance in sparse channels.

Although there have been many studies on sparse channel estimation algorithms based on compressive sensing theory, most of them focus on the improvement and innovation of reconstruction algorithms. Studies have shown that in the recovery process of the sparse channel, the structure of the pilot also plays a very important role in the final sparse channel estimation performance, so the present invention will mainly solve the problem of optimal design of the pilot structure in the sparse channel estimation process . Among the existing pilot structure design methods, most of the pilot structure design standards adopted in the orthogonal frequency division multiplexing (OFDM) system are based on reducing the cross-correlation of the measurement matrix Deterministic pilot structure design, this standard may also be referred to as MIP (mutual incoherence property). For example, a series of deterministic pilot design methods represented by the invention patent "Pilot Optimization Method, Device and Channel Estimation Method for Sparse Channels" are all based on the MIP standard. It can be seen that in this invention, the design of the entire pilot structure is completed before signal transmission and is independent from the subsequent channel estimation process. Although this type of deterministic pilot design method reduces the complexity of its engineering implementation, it fails to take into account the uncertainty of the wireless channel in the actual channel transmission process, so the pilot optimized by this type of method It is imprecise that the structure is used for the channel transmission process in any case.

Contents of the invention

Aiming at the deficiencies in current deterministic pilot structure design methods, the present invention proposes an adaptive pilot structure optimization design method based on sparse channel estimation. This method is different from the common deterministic pilot design method. It will optimize the design of the pilot structure for a specific wireless channel, and effectively improve the mean square error (Mean Square Error) of sparse channel estimation at the cost of appropriately increasing system complexity. Error, MSE) performance, and make the pilot structure more applicable.

In order to be distinguished from the deterministic pilot structure design method, the main features and steps of the present invention include the following aspects:

(1) The present invention starts with a series of random pilot structures, and utilizes the sparse channel estimation results after actual signal transmission to react on the reconstruction of the pilot structure. This method is different from the common deterministic pilot structure design, so that the pilot structure is continuously adjusted during continuous transmission until it converges.

(2) In the process of continuously adjusting the pilot frequency structure, the present invention uses the minimum l ₁ norm model in the convex optimization algorithm and the genetic algorithm (genetic algorithm, GA) theory to screen and reconstruct the pilot frequency. Compared with a single pilot design method using the MIP criterion as a pilot structure design standard, it is more direct and comprehensive.

(3) In order to improve the convergence speed of the pilot structure and the estimation performance of the pilot structure, the present invention can be realized by changing the number of initial populations and the probability of gene crossover and mutation in the GA iteration process.

(4) After completing each reconstruction of the pilot structure, the receiving end needs to feed back a series of new pilot structure information to the sending end as the start of the next loop iteration, and repeat the above steps many times until the pilot structure convergence. ,

In summary, implementing the above-mentioned adaptive pilot optimization design method has the following beneficial effects:

(1) The present invention continuously modulates the structure design of the pilot in the transmission signal according to the actual transmission results, so that the finally designed pilot structure is more in line with the characteristics of the actual channel, that is, more representative.

(2) The present invention combines the respective characteristics of the greedy algorithm and the convex optimization algorithm, which not only ensures the lower implementation complexity of the system in the channel reconstruction process, but also ensures the accuracy in the channel recovery process, and achieves the estimation performance and calculation Good compromise between complexity.

(3) The present invention utilizes the GA theory to reconstruct and update the pilot information, ensuring that each reconstruction of the pilot structure changes towards the specified target direction, so that the cyclic system in the present invention can finally converge.

(4) The pilot structure optimized by the adaptive pilot optimization design method in the present invention can improve the estimation accuracy of the entire OFDM sparse channel estimation system.

Description of drawings

In order to illustrate the technical method of the present invention more clearly, the following will briefly introduce the accompanying drawings used in the description of the implementation manner.

Fig. 1 is a schematic flow chart of an adaptive pilot optimization design method of the present invention;

Fig. 2 is a schematic flow chart of a specific embodiment of the present invention;

Fig. 3 is a comparison of the convergence of the adaptive pilot optimization design method in the present invention under different initial population individual numbers.

Fig. 4 is a comparison result diagram of estimation performance between an embodiment of the present invention and several existing representative pilot structure design methods.

Fig. 5 is a comparison result of running time between an embodiment of the present invention and several existing representative pilot structure design methods.

detailed description

In order to describe the objectives, technical methods and method features of the present invention more clearly and completely, the present invention will be further described in detail below in conjunction with specific embodiments and drawings in the embodiments.

Before introducing the embodiments of the present invention, it is necessary to establish a basic OFDM signal input and output system model. Assume that the total number of subcarriers of the OFDM signal is N, and the number of pilots is p, which are respectively located in subcarriers k ₁ , k ₂ ,…,k _p (1≤k ₁ ,k ₂ ,…,k _p ≤N )superior. We define vector p=[k ₁ ,k ₂ ,...,k _p ] ^T as a pilot position vector (pilotposition vector, PPV). In addition, the transmission data of each OFDM signal on the i-th subcarrier at the transmitting end and the receiving end are defined as x(i)(i=1,2,...,N) and y(i)(i=1,2, ..., N). Thus, we can get the representation of the received signal in the frequency domain:

Y=XH+N ^* (1)

Among them, Y=[y(1),y(2),…,y(N)] ^T , X=diag[x(1),x(2),…,x(N)], H=F _{N × L} h. Among them, F _N×L is a part of the discrete Fourier transformation matrix (Discrete Fourier Transformation, DFT), which is composed of the first L columns of the N-dimensional DFT transformation matrix:

in, N ^* is an N-dimensional additive complex Gaussian white noise vector with variance ^σ2 .

In order to select p pilots from N subcarriers, here we define a p×N-dimensional pilot selection matrix (pilot selection matrix, PSM) for pilot selection: in, is a unit column vector of length N, in which only the elements on k _i are 1, and the rest of the elements are 0. Then, the PSM is multiplied to the left by both ends of (1), and the following equation can be obtained,

separate order

Therefore, the above formula can be further simplified as Y _P =X _p F _p h+N _p . Here, we set T=X _p F _p , and T is the measurement matrix in CS theory. So far, the input and output model of the pilot signal in the entire OFDM system has been constructed, and the model can finally be expressed as Y _p =Th+N _p .

After the construction of the input and output model of the OFDM system is completed, Fig. 1 shows the general flow of the method for optimal design of adaptive pilot structure based on sparse channel estimation proposed by the present invention. The steps of the whole method can be mainly divided into the following five parts: generation of initial population, preliminary sparse channel estimation, screening of pilot structure, reconstruction of pilot structure, and cyclic iteration.

Fig. 2 is a schematic flow chart of an embodiment of an adaptive pilot structure optimization design method based on sparse channel estimation proposed by the present invention, and the specific steps of the method are:

S100: This method will generate z groups of random pilot structures to form an initial population Z ⁽⁰⁾ , and each individual in the population can be represented by a set of PPVs.

S200: Use z groups of individuals in the population Z ^(j) (j=0,1,...,C) as the pilot structure to transmit the OFDM signal in turn, and directly use the OMP algorithm to perform the OFDM signal received at the receiving end Preliminary channel estimation. Since the transmitting end sends z groups of OFDM signals composed of different pilot structures, the receiving end can obtain z groups of different channel estimation values h _guess (i), (i=1,2,...,z).

S300: Then, use the GA principle to screen and reconstruct excellent pilot structures. Here, this method utilizes the minimum l ₁ norm model in the convex optimization algorithm: st||Y-Th|| ₂ ≤σ _η , which gives the fitness criterion in GA iteration: V(i)=||Y _p -T _p ·h _guess (i)|| ₂ , i=1 ,2,...,z. Among them, ||x|| ₂ represents the two-norm of the vector x, and i represents the i-th individual in the population.

S400: After the fitness judgment standard is given, bring the h _guess (i) value of group z into the fitness judgment standard, and select the PPV corresponding to the first z/2 groups with the smallest h _guess (i) value as Excellent individuals replace the remaining z/2 group individuals.

S500: Simulate the gene crossover and mutation process in GA theory, so as to form z group of new individuals, and also form a new population Z ^(j+1) .

S600: Finally, feed back the new population information to the sending end, and repeat the steps from S200 to S600 until the number of cycles reaches C.

S700: After the number of cycles reaches C, select the pilot structure corresponding to the minimum fitness value V(i) in the population Z ^(C) at this time, and the pilot structure is optimized through the adaptive pilot in the present invention The design method selects the optimal pilot structure.

According to GA theory, although the genetic iterative process can continuously update and screen out pilot structures with better estimation performance, it is obviously unrealistic to exhaust all situations. This means that the pilot structure selected by the adaptive pilot optimization design method is not theoretically optimal, but only locally optimal. But despite this, we can still improve the estimation performance of the whole pilot structure by appropriately increasing the system complexity, so that it can meet the satisfactory solution in engineering.

In the simulation experiment, the embodiment of the present invention adopts an OFDM signal with 512 subcarriers and modulates it with 16-QAM. In addition, the number of cyclic prefixes is 128, the number of pilots is 30, the channel length is 60, and the sparsity is 6. For the GA iterative process, the gene crossover probability in this embodiment is 0.2, and the gene mutation probability is 0.02.

Fig. 3 is a schematic diagram of the convergence of the adaptive pilot optimization design method in the present invention under different initial population individual numbers. It gives the method of determining the number of cycles C in steps S200 and S600. It can be seen from Figure 3 that in the case of different initial population individuals, the number of cycles C required to make the minimum fitness of the population converge is different. In addition, when the number of iterations reaches C, the minimum fitness V _min ^(j) =min{V(i)} of the population will no longer change. This also means that the gene combination of the optimal individual in the population has tended to be stable. At this time, even if the population is iteratively updated, the individual genes in the population will not change. This is determined by the nature of the GA theory itself. Therefore, when the population information is stable, the optimal individual in the population is the local optimal pilot structure under a given channel.

Figure 4 shows the channel estimation performance of the random pilot structure (Random-OMP), the equidistant pilot structure (Equalinterval-OMP) and the deterministic pilot structure based on the MIP criterion (MIP-OMP) under the OMP reconstruction algorithm. , and the estimated performance of the deterministic pilot structure designed based on the MIP criterion under the convex optimization reconstruction algorithm (MIP-Cov) is given. In addition, in order to better reflect the characteristics of the adaptive pilot structure optimization design method based on sparse channel estimation in the present invention, Fig. 4 respectively shows the channel estimation of the method when the number of individuals in the initial population is 200, 600, and 1000 performance.

It can be seen that when the number of individuals in the initial population is 200, the estimated performance of this method is almost the same as that of the MIP-OMP method. And when the number of individuals in the initial population increases to 1000, the estimation accuracy has surpassed the MIP-OMP method and is very close to the MIP-Cov method. It should be noted that due to the high complexity of the convex optimization algorithm in engineering implementation, it is usually replaced by the OMP algorithm in engineering. This also means that the high-precision channel estimation performance shown by the MIP-Cov method in simulation experiments is often difficult to achieve in actual engineering.

Fig. 5 is a comparison result of running time between the present invention and several existing representative pilot structure design methods. Considering that the adaptive pilot optimization method proposed in this paper combines pilot structure design and channel estimation process, for the sake of fairness, this paper defines the running time as the sum of the total running time of pilot structure design and channel estimation process.

According to the running result of the simulation software, the data in Fig. 5 can be obtained. It can be seen from Figure 5 that although the MIP-Cov method has the best channel estimation performance, its total running time is the longest, which is one of the reasons why it is difficult to implement in practical engineering. While the total running time of the MIP-OMP method is the shortest, its price is to reduce the accuracy of channel estimation. For the adaptive pilot optimization design method proposed in the present invention, although the total running time increases with the increase of the number of individuals in the initial population, and is higher than the running time of the MIP-OMP method, compared to the channel estimation accuracy In terms of almost the same MIP-Cov method, the computational complexity of this method has been effectively controlled.

The above description is only a specific embodiment of the present invention and a series of simulation results under this embodiment, and should not limit the scope of rights of the present invention. Therefore, any equivalent or similar changes to any feature disclosed in this specification still fall within the scope of the present invention.

Claims (5)

1. a kind of adaptive pilot structure optimization design method based on sparse channel estimation, is characterized in that, comprises following main steps: Step 1: Generate multiple groups of random pilot structure vectors and merge them into an initial population. Step 2: Apply the pilot structure in the initial population to different OFDM signals for sequential transmission. Step 3: The greedy algorithm is used to estimate the sparse channel of the received signal, and the pilot structure is further screened by using the estimated sparse channel estimation value. Step 4: Combine the GA theory to reconstruct the pilot structure after screening. Step 5: Feedback the reconstructed pilot structure information to the sending end for cyclic iteration until the pilot structure converges. 2. The method according to claim 1, characterized in that, the whole adaptive pilot structure optimization design method combines the design of the pilot structure with the channel estimation process, which is different from the traditional deterministic pilot structure The design method will select a more representative pilot structure for different wireless channels. 3. the method as claimed in claim 1 or 2, is characterized in that, in described step 3, has utilized the minimum _l1 norm model in the convex optimization algorithm as the fitness judgment standard to the more excellent pilot structure of estimation performance to filter. It makes up for the lack of recovery accuracy of the greedy algorithm in sparse channel reconstruction. 4. The method according to claim 1 or 2, characterized in that in step 4, the selected pilot structure is reconstructed in combination with GA theory to generate multiple sets of new pilot structure information. On the basis of ensuring the estimation accuracy, the huge complexity brought by the direct use of convex optimization algorithm is avoided. 5. The method according to claim 1 or 2, wherein in said step 5, the reconstructed pilot structure information is fed back to the transmitting end, and steps 2-5 in claim 1 are repeated until the pilot structure information is guided The frequency structure converges.