CN103873398B - Joint merges multiple-limb regulatable view window mouthful length equilibrium detection method and device - Google Patents

Abstract

The present invention relates to the technical field of cooperative communication in wireless communication, in particular to a self-adaptive system used in cooperative communication, which can effectively combine and process received signals from source nodes and multiple relay nodes, and significantly improve system performance. The joint combined multi-branch adjustable observation window length equalization detection method and device is characterized in that the following processing is performed on the signals of the M+1 branches from the source node and M (M≥1) relay nodes: each branch The output signal is first sampled by analog-to-digital conversion, the sampled signal is weighted by the corresponding tap coefficient, and then added as the output of the branch, and the output of each branch is added to obtain the final output. It is worth noting that the weight of each branch In the training phase, the training sequence is used for joint adjustment, and the length of the observation window of each branch can be adjusted to the optimal length of the observation window according to the specific channel envelope of each branch.

Description

Combined multi-branch adjustable observation window length equalization detection method and device

technical field

The present invention relates to the technical field of cooperative communication in wireless communication, in particular to a self-adaptive system used in cooperative communication, which can effectively combine and process received signals from source nodes and multiple relay nodes, and significantly improve system performance. The joint combined multi-branch adjustable observation window length equalization detection method and device.

Background technique

Cooperative wireless communication technology does not require each node to contain multiple antennas to achieve cooperative diversity. It relies on neighboring nodes to share each other's antennas and perform cooperative transmission, thereby forming a virtual antenna array similar to multi-antenna transmission. Cooperative communication Combining the advantages of diversity technology and relay transmission technology, the performance gain of multi-antenna and multi-hop transmission can be obtained. Depending on the strategy of relay cooperation, relay networks can generally be classified into decode-and-forward (DF) and amplify-and-forward (AF) networks. Among these two strategies, the AF strategy is more attractive in practical systems because it has less computational load on the relay side.

One of the advantages of cooperative communication is that it can improve the effective signal-to-noise ratio (SNR) at the destination node, so that better system performance can be obtained. For cooperative wireless communication, the information sent by the source node reaches the destination node through different transmission paths, one of which is that the source node directly transmits to the destination node, and the other reaches the destination node through a relay node. Therefore, the system performance partly depends on the combined processing technique of the signal from the source node and the relay node. There have been a lot of research on combining processing techniques, such as Maximum Ratio Combining (MRC), Selective Combining (SC) and Switching Diversity Combining (SDC). These schemes assume that at the destination node, the channel state information ( CSI) is known, and the channel is a frequency flat fading channel. However, in high-speed wireless communication applications, the transmission bandwidth is larger than the associated bandwidth of the channel, making the channel frequency selective. For high-speed communication applications in cooperative communication networks, existing techniques for frequency flat-fading channels need to be improved, or new techniques should be proposed to eliminate the impact of frequency-selective channels. A fractionally spaced 2-branch fixed observation window length (FOWL) equalization detector has been proposed that combines the input signals from two independent channels, one frequency selective (assumed to be from a relay) and the other is a Gaussian channel (assumed to come from the source node). The design of the detector is implemented based on the minimum mean square error (MMSE) standard, but the tap coefficient vector of the detector is derived using the orthogonality theorem on the premise that the CSI between all nodes is known (the adaptive algorithm is not given). But in fact, at the destination node, the CSI between all nodes is unknown, the channel from the source node may also be a frequency selective channel, and more nodes may participate in cooperative communication. Therefore, it is necessary to consider improving the two-branch FOWL equalization detector so that it can handle these situations. In addition, wireless channels have environmental dependencies (eg, the geometric layout of buildings), because the channel envelopes between different nodes are different because of the different environments. The observation window length (OWL) is an important parameter for the BER performance of the equalization detector. The optimal OWL required to obtain the optimal BER performance depends on the specific channel envelope, so for different channels envelope, the desired optimal OWL is different. Especially for the actual cooperative communication system, the channel envelope between the destination node and the node is unknown, so the optimal OWL cannot be obtained in advance. In order to achieve better bit error rate performance, the equalization detector needs to have the ability to adaptively adjust the OWL of each branch according to the specific channel envelope corresponding to each branch.

Contents of the invention

Aiming at the deficiencies in the prior art, the present invention proposes an adaptive joint combination that can effectively combine and process received signals from source nodes and multiple relay nodes and significantly improve system performance in cooperative communication Multi-branch adjustable observation window length equalization detection method and device.

The present invention can reach through the following measures:

A joint and combined multi-branch adjustable observation window length equalization detection method, characterized in that the following processing is performed on the signals of the M+1 branches from the source node and M (M≥1) relay nodes: the output of each branch The signal is first sampled by analog-to-digital conversion. The sampled signal is weighted by the corresponding tap coefficient, and then added as the output of the branch. The output of each branch is added to obtain the final output. The tap coefficient weight of each branch is in the training stage. Joint adjustment using training sequences.

In the present invention, the received signal from the source node is denoted as r _SD (k), and the received signal from the relay node is denoted as r _RiD (k), where 1≤i≤M, the received signals of M+1 branches pass through A After /D conversion, they are respectively recorded as r _{SD (n)} , r _R1D (n) ... r _{RiD (n)} , M+1 branch signals are respectively sent to M+1 sub-equalizers using FIR filter structure Equalization processing is carried out so that the A/D sampled signal is weighted by the corresponding tap coefficient, where the signal vectors of the received signals r _SD (k) and r _RiD (k) are written as:

Among them, m represents the observation window length of the sub-equalizer, T _f represents the bit symbol duration of the sub-equalizer, and T _s represents the sampling period of the A/D; the sampling in the bit symbol duration forms a processing unit, and l represents in The number of processing units in the observation window;

with u _SD(n) The tap coefficient vector corresponding to u _RiD(n) is defined as follows

Among them, the initial values of c _SD(n) and c _RiD(n) are c _i =[0, 0, ..., 0] ^T , and m represents the length of the observation window. Further, the (M+1)m-dimensional row vector is introduced:

with (M+1)m-dimensional column vectors:

(12),

Then the outputs of M+1 branches are respectively recorded as y _SD (n), y _RiD (n), where 1≤i≤M,

The final output signal y(n)=y _SD (n)+y _R1D (n)+...+y _RMD (n) after the combination, the tap coefficients of the M+1 sub-equalizers are adaptively adjusted and updated jointly, then The final output is governed by:

(13)

In the present invention, attention is paid to the estimation of the transmitted symbol, rather than the estimation of the tap coefficient vector, so the error is defined as

where d(n) represents the data symbol sent by the source node, represents an estimate of d(n),

Based on the principle of minimizing interference, then the ultimate design purpose of the present invention can be expressed as the following constrained optimal problem:

Minimize the Euclidean norm of tap coefficient vector increments

and subject to the following constraints on the joint merged multi-branch (JCMB) FOWL equalized detector output

in Represents the Euclidean norm, formula (19) is close to (13).

The meaning of the optimal criterion in the present invention is as follows: given the input signal vector u(n), the tap coefficient vector should be changed from c(n) to c(n+1) in the minimum (minimum mean square sense), so that The output u(n)c(n+1) filtered by the updated tap coefficient c(n+1) will be equal to d(n), which should satisfy the constraint condition (13), because the purpose of the proposed detector is to jointly combine processing from Based on the signals of the source node and multiple relay nodes, an estimate of the transmitted signal is obtained. In order to solve this constrained optimal problem, the cost function of the proposed detector is set as:

in, is the Lagrangian multiplier, Represents the square operation of the Euclidean norm, let the cost function (20) take the derivative of c(n+1), and let the result of the derivative be zero, we can get:

Substitute (21) into equation (19) to get the unknown multiplier ,have

In order to overcome the problem of gradient noise amplification, the parameter , combining the results of formulas (21) and (22), we can get the pair linear estimation based on MMSE standard:

The adaptive joint merge processing algorithm for :

Among them, both μ and δ are positive numbers, different from the single branch equalization detector used for point-to-point communication, the tap coefficient vector (c _SD(n) and ) and the merge process is based on the MMSE criteria for joint selection.

The present invention also includes the adjustment of the length of the adjustable observation window of the M+1 multi-channel branch, specifically: let the length of the steady-state observation window of the branch i be L _i , with represent the corresponding steady-state tap coefficient vector and input signal vector respectively, and n represents the discrete-time index, then the segmental steady-state error SSE of this branch is defined as:

where d(n) represents the desired signal, , with are the tap coefficient vectors and the input signal vector For the first M values of , the mean square SSE is defined as

In order to complete the adjustment of the length of the observation window, first set the cost function of OWL adjustment, and the cost function of each branch to search for the optimal OWL is defined as

in with It is a normal number with a small value, and it is preset according to the system needs.

with represent the steady-state mean square SSE when the OWL corresponding to the SD and R _i -D branches is L _SD and L _RiD , respectively, with represent the steady-state mean square SSE of the SD and R _i -D branches, respectively:

in

The different outputs of the various branches can be used to calculate the corresponding error signal:

The gradient method is used to obtain the corresponding OWL based on the standards (26) and (27), but the OWL must be an integer, which restricts the adaptive adjustment of the OWL. In order to solve this problem, the concept of pseudo-fractional OWL (PF-OWL) is used to enable Adjust the OWL instantaneously. At this time, the length of the observation window is no longer limited to integers. The real OWL does not change until the PF-OWL accumulates to a certain extent. Based on the cost functions (26) and (27), the JCMB- AOWL balance, the algorithm is as follows:

in Indicates the pseudo-score OWL of the corresponding branch, the parameters α and γ are normal numbers, α is the leakage factor, and it needs to satisfy α<<γ, L _k(n) represents the corresponding branch The true OWL at discrete time n, The initial value of ,when The cumulative amount of change exceeds a certain threshold, and L _k(n) is rounded and updated as follows:

in Express and To the nearest integer, η is a small integer threshold.

According to equations (36) and (37), the real OWL L _k(n) remains unchanged until accumulated to a certain extent , assuming that the OWL of all branches is changed, define

with Represents the OWL adjusted signal vector, with Indicates the corresponding tap coefficient vector after adjustment, and the adaptive adjustment process of the tap coefficient is as follows:

First, take the R _i -D branch as an example to illustrate the change of the signal vector and the tap coefficient vector,

If you increase p _a taps:

If p _r taps are reduced:

Among them, p _a and p _r are determined by formula (37).

After being adjusted by the OWL of each branch, the signal vector and tap coefficient vector of the proposed equalizer detector can be expressed as

in with represent the adjusted signal vector and the tap coefficient vector of the corresponding branch respectively, and then, the tap coefficient ( and ) and channel merging are jointly processed based on the MMSE standard, and the update algorithm is as follows:

In the present invention, the tap coefficients of all branches and the combined processing are jointly implemented, so the errors calculated by the (30)-(33) formulas are segmental steady-state errors. This is because the tap coefficient vector shown in equation (43) is given by with constitute, while with Both are segmented tap coefficient vectors shown in formula (43).

The present invention also proposes an equalization detection device adopting the above-mentioned combined multi-branch adjustable observation window length equalization detection method, which is characterized in that it consists of one source node S, one destination node D and two or more relay nodes R _i , i=1,2,3,..., the source node, the destination node and the relay node all contain only one transmitting antenna and one receiving antenna and are in a cooperative communication system in half-duplex mode, with M+1 The sub-equalizers are respectively used to process the signal r _SD(k) received by the destination node from the source node and the signal r _RiD(k) received by the M source node from the relay node, and there are also M+1 sub-equalizers for adaptive adjustment The adaptive adjuster of the tap coefficient of the device, the sub-equalizer adopts the FIR filter structure, and the output of the M+1 sub-equalizer is added by the adder to obtain a joint combined signal.

In the present invention, there are also M+1 length adjusters that are respectively connected to the M+1 path equalizers in one-to-one correspondence to adjust the length of the observation window. The PF-OWL adjustment unit connected to the output end of the steady-state error calculation unit and the OWL adjustment unit connected to the output end of the PF-OWL adjustment unit, and the output end of the OWL adjustment unit is correspondingly connected to the equalizer.

Compared with the prior art, the present invention can jointly process signals from multiple independent frequency selective channels (source node-destination node and multiple source nodes-relay node-destination node), and the tap coefficients of all branches are used for each The specific channel corresponding to a branch is jointly adjusted using an adaptive algorithm, and at the destination node, there is no need to assume that the CSI between all nodes is known. In addition, the OWL of each branch can be adaptively adjusted according to the corresponding specific channel envelope. Applicable Wide range and high system performance.

Description of drawings:

Accompanying drawing 1 is a system model diagram of single source node, single destination node multi-relay cooperative network.

Accompanying drawing 2 is a kind of structural representation of the present invention.

Accompanying drawing 3 is another structural representation of the present invention.

Accompanying drawing 4 is the primary channel realization based on formula (1) in the present invention.

Accompanying drawing 5 is the BER performance (SNR=10dB) curve chart of equalization detector under different OWL in the present invention.

Accompanying drawing 6 is the OWL evolution curve (SNR=10dB) of the AOWL equalization detector in the present invention.

Accompanying drawing 7 is the bit error rate performance comparison of the combined 2-branch FOWL equalization detector proposed by the present invention and the prior art.

Figure 8 shows the BER performance of the combined (SC) MB-FOWL and JCMB-FOWL equalization detection devices of the present invention when different numbers of cooperative nodes are used.

Accompanying drawing 9 is the BER performance comparison graph of SCMB-AOWL, JCMB-FOWL and JCMB-AOWL equalization detector in the present invention when adopting different cooperative relay numbers (M=1 and M=2).

detailed description:

Below in conjunction with accompanying drawing, the present invention will be further described:

As shown in Figure 1, taking the following system as an example, the present invention proposes a method and device for joint and merged multi-branch adjustable observation window length equalization detection. The system consists of a source node (S) and a destination node (D ) and several relay nodes ( )constitute. The source, destination and relay nodes all contain only one transmitting antenna and one receiving antenna, and are in half-duplex mode, and cannot receive and transmit signals at the same time. Assume that relays work in amplify-and-forward (AF) mode and all relays have the same average power limit. A transmission process includes two stages: in the first stage, the source node broadcasts information to the destination and relay nodes; in the second stage, the relay amplifies and forwards the noisy signal it received in the first stage to the destination node.

S→D, S→R _i , and R _i →D( is the channel impulse response between the i-th relay) respectively denoted as ,

with , where L _SD , L _SR and L _RiD denote the corresponding channel memory lengths, respectively. The channel between all nodes is assumed to be a frequency-selective channel, and the discrete-time channel impulse response is modeled by a zero-mean Gaussian random variable subject to an exponential power delay envelope

in , Describe the delay spread characteristics of the channel, Represents the Dirac δ function, _PR represents the average power of the multipath component.

In the first stage, the signal received by the i-th relay and destination node is given by

where s(k) represents the signal sent by the source node, with Represents additive white Gaussian noise with zero mean and variance with .

In the second stage, the signal received by the destination node from the i-th relay R _i is

where n _RiD(k) denotes the corresponding additive white Gaussian noise with zero mean and variance .

β _i is the relay gain, in order to satisfy its power limit, set the relay The gain is

where P is the average energy of each symbol transmitted by each node, are the source node and the relay The k-th channel coefficient of the inter-channel, N _{0, i} represents the variance of the zero-mean white Gaussian noise in the signal received from the i-th relay at the destination node.

Example 1:

As shown in Figure 2, the present invention firstly proposes a method and device for joint and merged multi-branch fixed observation window length equalization detection, specifically a joint combined multi-branch fixed observation window length equalization detection device (referred to as JCMB- FOWL equalization detection device), the device adopts a fractionally spaced linear filter structure, which can jointly realize the functions of matched filtering and symbol space equalization, in order to effectively combine and process signals transmitted through multiple independent frequency selective channels, the present embodiment 1 A joint combined (JC) multi-branch (MB) fixed observation window length (FOWL) equalized detector is proposed. Its structure is shown in Fig. Received signals corresponding to multiple channels are jointly processed. The proposed equalization detector will jointly implement matched filtering, combining processing and adaptive equalization;

The equalization detector will directly process the sampling signal after the analog-to-digital conversion (A/D), and for SD and SR _i -D receiving signals, the input signal of the equalization detector is set as with . In each bit period, the output of the equalization detector produces a decision, and the tap coefficients of each branch are jointly adjusted by the adaptive filtering algorithm. The adaptation works in the training phase, using the training sequence to adapt to the channel, and then in the direct decision mode, Use hard decisions as the output of the equalization detector.

In the JCMB-FOWL equalization detector, corresponding to the received signal with The signal vector for is given by:

Among them, m represents the length of the observation window, T _f represents the bit symbol duration, and T _s represents the sampling period of A/D. It is assumed that the sampling within the duration of the bit symbol forms a processing unit, and l represents the number of processing units in the observation window.

corresponds to with The tap coefficient vector for is defined as follows:

in with The initial value of , m represents the observation window length, which is defined in formula (8);

Introduce (M+1)m-dimensional row vectors:

and (M+1)m-dimensional column vector

In the proposed detector, the outputs of multiple channels are jointly processed to estimate the transmitted data symbols. Therefore, the output of the JCMB-FOWL equalization detector is restricted by the following formula:

The detector focuses on the estimation of the transmitted symbols, rather than the estimation of the vector of tap coefficients, defining the error as:

where d(n) represents the data symbol sent by the source node, Represents an estimate of d(n):

in

Based on the principle of minimizing interference, the design criteria of the JCMB-FOWL equalized detector can be expressed as the following constrained optimal problem.

Minimize the Euclidean norm of tap coefficient vector increments

and subject to the following constraints on the output of the JCMB-FOWL equalized detector

in Represents the Euclidean norm, formula (19) is close to (13).

The meaning of the optimal standard is as follows: given the input signal vector u(n), the tap coefficient vector should change from c(n) to c(n+1) in the smallest (minimum mean square sense) way, so that the updated tap Coefficient c(n+1) filtered output will be equal to d(n). Constraint (13) should be satisfied, because the purpose of the proposed detector is to combine and process signals from the source node and multiple relay nodes to obtain an estimate of the transmitted signal.

In order to solve this constrained optimal problem, the Lagrangian multiplier method is used. According to equations (18) and (19), the cost function of the proposed detector is

Among them, ξ is the Lagrangian multiplier, Represents the square operation of the Euclidean norm. Let the cost function (20) take the derivative of c(n+1), and let the result of the derivative be zero, we can get

To solve the unknown multiplier ξ, substituting (21) into equation (19), we have

In order to overcome the problem of gradient noise amplification, the parameter δ is introduced. Combining the results of formulas (21) and (22), we can get the pair linear estimation based on MMSE standard Adaptive algorithm for:

Among them, μ and δ are normal numbers. Different from the single-tap equalized detector for point-to-point communication, in the JCMB-FOWL equalized detector, the tap coefficient vectors (cSD(n ₎ and _cRiD(n) ) and the combining process are jointly selected based on the MMSE criterion. In fact, as shown in FIG. 2, the output of each branch is firstly weighted by the corresponding tap coefficient, and then summed as the output of the branch. The output of the proposed detector is the sum of the outputs of each branch. The weights of each branch are jointly adjusted using the training sequence during the training phase. The size of the weight determines the importance of each branch to the detection performance.

Example 2:

The present invention also proposes a JCMB-AOWL equalization detection device based on the MMSE standard, and the structure of the detection device is as shown in Figure 3:

Among them, the definitions of segmental steady-state error (SSE) and mean square SSE are as follows: let the length of the steady-state observation window of branch i be L _i , with represent the corresponding steady-state tap coefficient vector and input signal vector respectively, n represents the discrete time index, then the branch SSE is defined as:

where d(n) represents the desired signal, , with are the tap coefficient vectors and the input signal vector For the first M values of , the mean square SSE can be defined as

In order to design and implement the JCMB-AOWL balanced detector, first set the cost function of OWL adjustment, and the cost function of each branch to search for the optimal OWL is defined as

in with It is a normal number with a small value, and it is preset according to the system needs. with Respectively represent the OWL corresponding to the SD and R _i -D branches as with The steady-state mean square SSE at . with Represent the steady-state mean square SSE of the SD and R _i -D branches, respectively:

in

The different outputs of each branch of the equalized detector can be used to calculate the corresponding error signal

The corresponding OWL can be solved based on the standard (26) and (27) using the gradient method. However, the OWL must be an integer, which restricts the adaptive adjustment of the OWL. In order to solve this problem, the concept of pseudo-fractional OWL (PF-OWL) is adopted to enable instantaneous adjustment of OWL. At this time, the length of the observation window is no longer limited to integers. The real OWL does not change until PF-OWL accumulates to a certain extent. Based on the cost functions (26) and (27), the JCMB-AOWL equalization based on the MMSE standard can be obtained, and the algorithm is as follows

in Indicates the pseudo-score OWL of the corresponding branch. The parameters α and γ are both normal numbers, and γ is the leakage factor, which must satisfy α<<γ. L _k(n) represents the corresponding branch The true OWL at discrete time n. The initial value of . when The cumulative amount of change exceeds a certain threshold, and L _k(n) is rounded and updated as follows:

in Express and to the nearest integer, is a small integer threshold.

According to equations (36) and (37), the true OWLL _k(n) remains unchanged until accumulated to a certain extent . Assuming that the OWL of all branches is changed, define with Represents the OWL adjusted signal vector, with Represents the corresponding tap coefficient vector after adjustment. The adaptive adjustment process of the tap coefficient is as follows.

First, with The branch is taken as an example to illustrate the change of the signal vector and the tap coefficient vector.

If you increase p _a taps:

If p _r taps are reduced:

Among them, p _a and p _r are determined by formula (37).

After being adjusted by the OWL of each branch, the signal vector and tap coefficient vector of the proposed equalizer detector can be expressed as

in with Respectively represent the adjusted signal vector and the tap coefficient vector of the corresponding branch.

Then, the tap coefficient ( and ) and channel merging are jointly processed based on the MMSE standard, and the update algorithm is as follows

In the proposed equalized detector, the tap coefficients of all branches and the combined processing are implemented jointly, so it is worth noting that the errors calculated by equations (30)-(33) are the segmental steady-state errors of the detector. This is because the tap coefficient vector shown in equation (43) is given by with constitute, while with Both are segmented tap coefficient vectors shown in formula (43).

Below in conjunction with simulation result, performance of the present invention is further analyzed:

In order to evaluate the performance of the JCMB equalization detection device proposed in the present invention, the modulation method adopts binary phase modulation (BPSK), assuming that all nodes work at the same power, in the simulation, consider a cooperative with a single sending and receiving pair and multiple relays In the communication network, the channels between nodes are semi-stable frequency selective channels. The channel impulse response coefficient is modeled by formula (1), and the parameters are set as , , , , , . The parameters μ and δ in formulas (23) and (46) are both set to 0.5. For a FOWL equalization detector, OWL is set to a 3-bit symbol duration, that is, in equations (6) and (7) . For the AOWL equalization detector, the initial OWL is also set to , the leakage factor α is 0.005, and the parameter η is set to 1.

(1) Verification of adaptive adjustment of observation window length:

Firstly, the adjustment ability of the proposed JCMB-AOWL equalization detector to OWL is verified. To facilitate the verification of adaptive OWL adjustments, the peer-to-peer communication case (S-D) is considered where no relay participates in cooperative transmission. Emulation is for the same channel. Figure 4 shows the impulse response of this channel produced by the channel model (1).

For the channel shown in Fig. 4, the optimal OWL is difficult to obtain directly due to the nonlinear relationship between OWL and steady-state MSE. In order to verify that the AOWL equalization detector can adaptively adjust OWL according to the specific channel envelope. First, the optimal OWL required for the channel envelope in Figure 4 will be obtained by simulation. Figure 5 presents the BER performance of the equalization detector when using different OWLs. For this channel, 100 data packets are sent, each data packet contains 10000 symbols, 500 of which are used as training sequences. Each point on the BER curve is obtained by averaging the BER over 100 packets. It can be seen from the figure that when the OWL is 45-105, the BER performance is very close. However, the increase of OWL will increase the complexity of the detector. Considering the complexity problem, the optimal OWL is defined as the minimum OWL that can make the balanced detector obtain the performance close to the optimal BER. According to this definition, it can be seen from Figure 5 that the optimal OWL is about 45.

Next, the ability of the AOWL equalization detector to adjust to OWL is evaluated. In the simulation, the parameters Set to 5. For this same channel, 10 packets are sent, each containing 15000 symbols. The OWL evolution curve is obtained by averaging the OWL evolution curve under each data packet. Figure 6 presents the OWL evolution curve of the AOWL equalization detector. For the specific channel envelope shown in Figure 4, it can be seen from Figure 6 that the proposed equalization detector can adaptively adjust the OWL to about 45. The obtained OWL is close to the optimal OWL in Fig. 5.

It can be seen from the above simulation results that the ability of the AOWL equalization detector to adaptively adjust OWL according to the specific channel envelope has been verified.

(2) Bit error rate performance comparison:

In order to compare the BER performance of different equalization detection schemes, a Monte Carlo simulation is established based on the channel model (1). In the simulation, it is assumed that at the destination node, the signal-to-noise ratio of the received signal from the source node and multiple relay nodes is the same, and the parameters of the SD and R _i -D branches Set to 6 and 8 respectively. For each channel implementation, a packet of 10,000 symbols is sent, of which 500 symbols are used as training sequences. The BER performance and OWL evolution curve are obtained from the bit error rate and OWL evolution curve achieved by averaging 500 random channels, respectively.

First, the proposed JCMB-FOWL equalization detector is compared with the literature [S. Wei, DL Goeckel, and M.C. Valenti, “Asychronous cooperative diversity,” IEEE Transactions on Wireless Communications, vol. 5, no. 6, pp. 1547 -1557, Jun. 2006.] The BER performance of the scheme in the literature [S. Wei, DL Goeckel, and MC Valenti, "Asychronous cooperative diversity," IEEE Transactions on Wireless Communications, vol. 5, no. 6, pp.1547- 1557, Jun. 2006.], assuming that the CSI between nodes is known, the tap coefficient vector of the detector is derived based on the MMSE standard and the orthogonal principle (only two branches are considered, and one branch corresponds to frequency selectivity channel, the other branch corresponds to the Gaussian channel). By comparing the performance, the correctness of the proposed adaptive equalization detector can be verified. Fig. 7 shows the BER performance of the 2-branch FOWL equalization detector and the scheme in [21]. For the scheme in the literature [S. Wei, DLGoeckel, and MC Valenti, “Asychronous cooperative diversity,” IEEETransactions on Wireless Communications, vol. 5, no. 6, pp. 1547-1557, Jun.2006.], ,in , , the correlation matrix in the simulation Obtained by time averaging method , used to estimate The number of symbols N is 2000. It can be seen from the figure that the BER performance of the JCMB-FOWL equalization detector is close to that of the literature [S. Wei, DL Goeckel, and MC Valenti, "Asychronous cooperative diversity," IEEE Transactions on Wireless Communications, vol. 5, no. 6, pp. 1547-1557, Jun. 2006.]. The simulation results verify the correctness of the proposed scheme. In fact, for the actual cooperative communication system, at the destination node, the CSI between nodes is unknown, and JCMB-FOWL can adaptively adjust the tap coefficient according to the specific channel, so the proposed scheme is more suitable for the actual cooperative communication system.

Next, the effect of the number of cooperative nodes on the BER performance of the JCMB-FOWL balanced detector is evaluated. Fig. 8 presents the BER performance of combining (SC)MB-FOWL and the proposed JCMB-FOWL equalization detector respectively when using different numbers of cooperating nodes. The SC scheme means that the tap coefficients of each branch in the MB equalization detector are obtained independently similar to the point-to-point communication, and then the output of each branch is added with equal gain as the combination method of the detector output. The JC and SC schemes in the simulation use the same simulation parameters. It can be seen from the figure that with the increase of the number of cooperative nodes, the BER performance of the JCMB-FOWL equalized detector is significantly improved (when BER= ^10-5 , compared with the use of 1 About 8dB performance gain is obtained, and 5 relays are used to obtain about 12dB performance gain). This is because the number of cooperative nodes is an important factor affecting BER performance, and the diversity gain of single-relay cooperation is limited, and the more the number of cooperative nodes, the greater the diversity gain obtained. In addition, it can also be seen from the figure that the BER performance of the proposed JC scheme is better than that of the SC scheme. Especially, when BER=10 ^-3 , when 1, 3 and 5 cooperative relays are used, JCMB-FOWL can obtain 4dB, 4dB and 3.8dB performance gains compared with SCMB-FOWL equalization detector respectively. This is because in the JC scheme, the tap coefficients of all branches are jointly obtained based on the MMSE standard, and the weight of each branch is jointly adjusted using the training sequence in the training phase. The size of the weight determines the importance of each branch to the detection performance, while SC The scheme uses only equal gain combining.

Finally, the performance of the proposed JCMB-AOWL equalized detector is studied and compared with the performance of the JCMB-FOWL equalized detector. Figure 9 illustrates that the JCMB-AOWL equalized detector outperforms the SCMB-AOWL and JCMB-FOWL equalized detectors. The same simulation parameters are used for the three detectors in the simulation, and the step size parameter γ is set to 3 for the JC and SC MB-AOWL detectors. The SC scheme means that the OWL and tap coefficients of each branch in the MB equalization detector are obtained independently similar to the point-to-point communication, and then the outputs of each branch are added with equal gains as the combination method of the detector output. It can be seen from the figure that the BER performance obtained by the JCMB-AOWL equalization detector is better than that of SCMB-AOWL (when BER=10 ^-5 , the performance gain is about 2.8dB when using 1 relay, and the performance gain is about 2.8dB when using 2 relays. achieve 2.3dB performance gain). This is because in the JCMB-AOWL equalized detector, the tap coefficients of all branches are obtained jointly. Furthermore, it is worth noting that JCMB-AOWL achieves significant performance gain over JCMB-FOWL. Especially, when BER=10 ^-5 , when using 1 and 2 cooperative relays, JCMB-AOWL can obtain about 5dB and 4dB performance gain than JCMB-FOWL equalization detector respectively. This is because OWL is an important parameter for BER performance, and the JCMB-AOWL equalization detector can adaptively adjust the OWL of each branch according to the specific channel envelope corresponding to each branch.

The invention proposes a Joint Merge Multi-Branch (JCMB) equalization detector scheme for the cooperative communication network under the frequency selective channel. In order to evaluate the performance of the proposed scheme, a Monte Carlo simulation is built based on the frequency-selective channel model (1). The simulation results show that the JCMB-AOWL equalization detector can adaptively adjust the observation window length according to the specific channel envelope. Compared with the existing theoretically derived schemes that assume known CSI between nodes, the proposed JCMB-FOWL equalized detector does not need to assume known CSI, and can achieve very close BER performance. Under the same simulation parameter setting, the performance achieved by JC is better than that of SC equalized detector. Simulation results also show that JCMB-AOWL can achieve better BER performance than JCMB-FOWL equalization detector.

Claims (4)

1. A combined multi-branch adjustable observation window length equalization detection method is characterized in that signals of M +1 branches from a source node and M (M is more than or equal to 1) relay nodes are processed as follows: the output signal of each branch is sampled by analog-to-digital conversion, the sampled signals are weighted by corresponding tap coefficients and then added to be used as the output of the branch, the outputs of the branches are added to obtain the final output, and the weight of each branch is jointly adjusted by using a training sequence in a training stage;

the received signal from the source node is denoted as r_SD(k) The received signal from the relay node is denoted asWherein i is more than or equal to 1 and less than or equal to M, and the received signals of M +1 branch are respectively marked asThe M +1 branch signals are respectively sent to M +1 sub-equalizers adopting FIR filtering structures for equalization processing, so that the signals after A/D sampling are weighted by corresponding tap coefficients, wherein the received signals r_SD(k) Andthe signal vector of (d) is noted as:

u_SD(n)＝[r_SD(n-1)，r_SD(n-2)，…，r_SD(n-m)]

$u_{R_{i}D}\left(n\right)=\&lsqb;r_{R_{i}D}(n-1),r_{R_{i}D}(n-2),\mathrm{...},r_{R_{i}D}(n-m)\&rsqb;$

m＝lT_f/T_swhere m denotes the observation window length of the sub-equalizer, T_fBit symbol duration, T, representing sub-equalizer_sRepresenting the sampling period of the A/D; the samples set within the bit symbol duration form a processing unit, l represents the number of processing units within the observation window;

and u_SD(n) andthe corresponding tap coefficient vector is defined as follows

c_SD(n)＝[c_SD，n1，c_SD，n2，…，c_SD，nm]^T

$c_{R_{i}D}\left(n\right)={\&lsqb;c_{R_{i}D,n1},c_{R_{i}D,n2},\mathrm{...},c_{R_{i}D,nm}\&rsqb;}^T$

Wherein c is_SD(n) andhas an initial value of c_i＝[0，0，…，0]^TM denotes the observation window length, and further, introduces an (M +1) M-dimensional row vector:

$u\left(n\right)=\&lsqb;u_{SD}\left(n\right),u_{R_{1}D}\left(n\right),u_{R_{2}D}\left(n\right),...,u_{R_{M}D}\left(n\right)\&rsqb;$

and (M +1) M-dimensional column vector:

the outputs of M +1 branches are respectively marked as y_SD(n)，Wherein i is more than or equal to 1 and less than or equal to M, y_SD(n)＝u_SD(n)c_SD(n)

$y_{R_{i}D}\left(n\right)=u_{R_{i}D}\left(n\right)c_{R_{i}D}\left(n\right)$

Combined and finally output signalIn the process, the adaptive regulator respectively updates tap coefficients for the outputs of the M +1 paths, so that the final output is restricted by the following formula:

$d\left(n\right)=u\left(n\right)c\left(n\right)=u_{SD}\left(n\right)c_{SD}\left(n\right)+\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}{u}_{R_{i}D}\left(n\right)c_{R_{i}D}\left(n\right).$

2. a method as claimed in claim 1, wherein the estimation of the transmitted symbols is considered instead of the estimation of the tap coefficient vector, so that the error defined during the adaptive adjustment process is:

$e\left(n\right)=d\left(n\right)-\hat{d}\left(n\right)$

where d (n) denotes a data symbol transmitted by the source node,denotes the estimation of d (n),

based on the interference minimization principle, minimizing the Euclidean norm of the tap coefficient vector increment:

||Δc(n)||＝||c(n+1)-c(n)||²

and subject to the following constraints:

$d\left(n\right)=u\left(n\right)c(n+1)=u_{SD}\left(n\right)c_{SD}(n+1)+\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}{u}_{R_{i}D}\left(n\right)c_{R_{i}D}(n+1)$

where | · | | represents the euclidean norm.

3. The method for detecting length equalization of a combined and merged multi-branch adjustable observation window according to claim 2, further comprising adjusting the length of the M +1 multi-branch adjustable observation window, specifically: let the steady state observation window length of branch i be L_i，Andrepresenting the corresponding steady state tap coefficient vector and input signal vector, respectively, and n represents the discrete time index, the branch segment steady state error SSE is defined as:

$e_M^{L_i}\left(n\right)=d\left(n\right)-c_{L_i,1:M}^T\left(n\right)u_{L_i,1:M}\left(n\right)$

where d (n) represents the desired signal, 1. ltoreq. M. ltoreq.L_i，Andare respectively a tap coefficient vectorAnd input signal vectorIs defined as the mean square SSE of

${\mathrm{\&xi;}}_M^{L_i}=E\left\{{\left(e_M^{L_i}\right(n\left)\right)}^2\right\}$

In order to complete the adjustment of the length of the observation window, a cost function for adjusting the length OWL of the observation window is firstly set, and the cost function for searching the optimal length OWL of the observation window in each branch is defined as

${\mathrm{\&xi;}}_{L_{SD}-\mathrm{\&Delta;}}^{L_{SD}}-{\mathrm{\&xi;}}_{L_{SD}}^{L_{SD}}\&le;{\mathrm{\&epsiv;}}_{SD}$

${\mathrm{\&xi;}}_{L_{R_{i}D}-\mathrm{\&Delta;}}^{L_{R_{i}D}}-{\mathrm{\&xi;}}_{L_{R_{i}D}}^{L_{R_{i}D}}\&le;{\mathrm{\&epsiv;}}_{R_{i}D}$

Wherein_SDAndaccording to the preset requirements of the system,andrespectively represent a corresponding S-D and R_iThe observation window length OWL of the D branch is L_SDAnd L_RiDThe steady state mean square SSE of the time,andrespectively represent S-D and R_iSteady state mean square SSE of the D branch:

${\mathrm{\&xi;}}_{L_{SD}-\mathrm{\&Delta;}}^{L_{SD}}=E\left\{{\left(e_{L_{SD}-\mathrm{\&Delta;}}^{L_{SD}}\right(n\left)\right)}^2\right\}$

${\mathrm{\&xi;}}_{L_{R_{i}D}-\mathrm{\&Delta;}}^{L_{R_{i}D}}=E\left\{{\left(e_{L_{R_{i}D}-\mathrm{\&Delta;}}^{L_{R_{i}D}}\right(n\left)\right)}^2\right\}$

wherein

$e_{L_{SD}-\mathrm{\&Delta;}}^{L_{SD}}\left(n\right)=d\left(n\right)-c_{L_{SD},1:L_{SD}-\mathrm{\&Delta;}}^T\left(n\right)u_{L_{SD},1:L_{SD}-\mathrm{\&Delta;}}\left(n\right)$

$e_{L_{R_{i}D}-\mathrm{\&Delta;}}^{L_{R_{i}D}}\left(n\right)=d\left(n\right)-c_{L_{R_{i}D},1:L_{R_{i}D}-\mathrm{\&Delta;}}^T\left(n\right)u_{L_{R_{i}D},1:L_{R_{i}D}-\mathrm{\&Delta;}}\left(n\right)$

The different outputs of the various branches may be used to calculate respective error signals:

$e_{L_{SD}}^{L_{SD}}\left(n\right)=d\left(n\right)-y_{SD}\left(n\right)=d\left(n\right)-c_{L_{SD}}\left(n\right)u_{L_{SD}}\left(n\right)$

$e_{L_{R_{i}D}}^{L_{R_{i}D}}\left(n\right)=d\left(n\right)-y_{R_{i}D}\left(n\right)=d\left(n\right)-c_{L_{R_{i}D}}\left(n\right)u_{L_{R_{i}D}}\left(n\right),i=1,...,M$

based on a standard by means of a gradient methodAndthe corresponding observation window length OWL is obtained, but the observation window length OWL must be an integer, which constrains the observation windowIn order to solve the problem, a pseudo-fractional OWL (PF-OWL) concept is adopted to instantaneously adjust the observation window length OWL, and then the observation window length is not limited to an integer, and the real observation window length OWL is not changed until the PF-OWL is accumulated to a certain extent, so that JCMB-AOWL equalization based on MMSE standard can be obtained, and the algorithm is as follows:

$l_{f,SD}(n+1)=\left(l_{f,SD}\right(n)-\mathrm{\&alpha;})-\mathrm{\&gamma;}\&lsqb;{\left(e_{L_{SD}\left(n\right)}^{L_{SD}\left(n\right)}\right)}^2-{\left(e_{L_{SD}\left(n\right)-\mathrm{\&Delta;}}^{L_{SD}\left(n\right)}\right)}^2\&rsqb;$

$l_{f,R_{i}D}(n+1)=\left(l_{f,R_{i}D}\right(n)-\mathrm{\&alpha;})-\mathrm{\&gamma;}\&lsqb;{\left(e_{L_{R_{i}D}\left(n\right)}^{L_{R_{i}D}\left(n\right)}\right)}^2-{\left(e_{L_{R_{i}D}\left(n\right)-\mathrm{\&Delta;}}^{L_{R_{i}D}\left(n\right)}\right)}^2\&rsqb;$

wherein l_f，k(n) the pseudo-fraction OWL of the corresponding branch, the parameters α and gamma are normal numbers, α is a leakage factor, and the requirements of α < gamma and L are met_k(n)Represents the corresponding branch l_f，k(n) true OWL, l at discrete time n_f，kThe initial value of (n) is l_f，k(0)＝L_k(0) When l is_f，k(n) ofCumulative amount of change exceeds a certain threshold, L_k(n)Rounding update is performed as follows:

whereinRepresents taking and l_f，k(n) the nearest integer, η is an integer threshold.

4. An equalization detection apparatus using the method of equalizing detection of jointly merging multiple-branch adjustable observation window lengths according to any one of claims 1 to 3, characterized by comprising 1 source node S, 1 destination node D and more than two relay nodes R_iIn a cooperative communication system in which i is 1,2,3, … …, and the source node, the destination node, and the relay node all only include 1 transmitting antenna and 1 receiving antenna and are in a half-duplex mode, there are M +1 signals r respectively used for processing the signal received by the destination node from the source node_SD(k)And M paths of signals r received by the source node from the relay node_RiD(k)The sub-equalizer is also provided with a self-adaptive adjuster used for self-adaptively adjusting tap coefficients of the M +1 sub-equalizers, the sub-equalizers adopt an FIR filter structure, and the outputs of the M +1 sub-equalizers are added by the superimposer to obtain a combined signal.