CN111835365B

CN111835365B - Comb filtering Turbo equalization algorithm suitable for short wave communication

Info

Publication number: CN111835365B
Application number: CN202010749832.9A
Authority: CN
Inventors: 张凯
Original assignee: Shaanxi Fenghuo Communication Group Co Ltd
Current assignee: Shaanxi Fenghuo Communication Group Co Ltd
Priority date: 2020-07-30
Filing date: 2020-07-30
Publication date: 2024-07-26
Anticipated expiration: 2040-07-30
Also published as: CN111835365A

Abstract

The invention belongs to the field of short wave communication and discloses a comb filtering Turbo equalization algorithm suitable for short wave communication, which comprises the steps that firstly, a comb filter performs path-by-path separation on multipath signals to extract clear and hierarchical multipath user useful signalsEach signal can extract soft informationThereby playing a diversity effect; secondly, all soft information is combined to obtain L (x), and the L (x) is input into a decoder for decoding; then, an average information is constructed based on the soft information L ^out (x) output from the decoderAnd simulate the received signal in combination with known multipath channel parametersFinally, the comb filter uses the analog received signalAnd extracting the multipath signal again by the true received signal r; the algorithm is simple and feasible, does not need to solve the inverse characteristic of the channel, avoids complex matrix inversion operation, greatly reduces the calculation complexity, and has the advantages of simplicity, fast convergence, easiness in implementation and the like.

Description

Comb filtering Turbo equalization algorithm suitable for short wave communication

Technical Field

The invention relates to the field of short-wave communication, in particular to a comb filtering Turbo equalization algorithm suitable for short-wave communication.

Background

In mobile communication, the position of a receiver is constantly changed, and meanwhile, due to the diversity of communication environments, electromagnetic waves can generate reflection phenomena when encountering obstacles or ionosphere, and diffuse reflection phenomena can be generated when encountering obstacles with uneven surfaces. Therefore, in actual communication, the receiver receives the superposition of signals from different paths, and even if noise interference is small, the situation that normal demodulation cannot be performed can occur. The presence of multipath can cause Inter-symbol interference (Inter-Symbol Interference, ISI), degrading the performance of the communication system and presenting an erroneous floor. The front code element and the rear code element are overlapped together in the time domain, and the frequency selective fading is caused in the frequency domain, namely, the deep fading is generated for a certain frequency, and part of detail information is completely lost.

Equalization is an effective method for solving multipath effects, and conventional equalization can be classified into time domain equalization and frequency domain equalization. Along with the proposal of modern iterative decoding ideas of coding (such as Turbo/LDPC codes), people gradually recognize that an iterative mode can bring additional performance gain to a communication system, so that the idea of joint iterative decoding is proposed; the essence is a process in which information is continuously cycled between the equalizer and the decoder until decoding is successful or a maximum number of iterations is reached. Since the operation mechanism of this process is very similar to the decoding mechanism of Turbo codes, joint iterative decoding is also called Turbo equalization.

The publication No. CN102185617A, simplified Turbo equalization algorithm, proposes a method for reducing the complexity of Turbo equalization calculation. The variance of the transmitted symbols is basically the same in the first iteration, the coefficient of the equalizer is fixed at the moment, and when the iteration reaches a certain number of times, the variance of the transmitted modulation symbols is small, the relationship between the coefficient of the equalizer and the transmitted symbols is small, and the coefficient can be considered to be unchanged, so that matrix inversion operation is avoided when one symbol is calculated.

The main idea of the existing equalization technology is to obtain the inverse of the channel through some calculation, and to interact the received aliasing signal with the inverse of the channel so as to counteract the multipath effect of the channel. A transversal filter is typically employed in the time domain to approximate the inverse of the channel, and the more the order of the filter is, the closer to the true inverse channel, the better the equalization performance. However, good performance comes at the expense of computational complexity, so the prior art has focused largely on how to solve the inverse of the channel quickly, with low complexity, in noisy environments. The defects of the method mainly include the following three points:

1) Most Turbo equalization techniques employ frequency domain equalization, requiring a fast fourier transform (Fast Fourier Transform, FFT) FFT/inverse fast fourier transform (INVERSE FFT, IFFT), which increases the complexity of the system. Meanwhile, due to the introduction of FFT, the data length can only select specific values (such as 512 and 1024) conforming to FFT rules, and the design of the data frame length is greatly limited, so that the practical use is inflexible; 2) Filters of finite order are typically employed in engineering practice to approximate the inverse of the channel, so there is a truncation effect; a compromise between computational complexity and performance is required; 3) The unavoidable channel matrix inversion operation in the time domain Turbo equalization algorithm is performed, the operation amount of the inversion operation is increased in a cubic manner along with the increase of the order of the transverse filter, and engineering realization is almost impossible when the order is large.

Comb filters are devices commonly used in the field of digital signal processing and are composed of a number of passband or stopband filters arranged at equal intervals in order to allow only certain specific frequency ranges of signals to pass or not, the spectral characteristics of which are comb-like and are therefore called comb filters. The invention borrows the concept of comb filters in the field of digital signal processing, however the working principle is completely different.

Disclosure of Invention

Aiming at the problems existing in the prior art, the invention aims to provide a comb filtering Turbo equalization algorithm suitable for short wave communication, which aims to effectively resist various problems caused by propagation of multiple paths through a certain equalization algorithm, and on the premise of acquiring multi-channel parameters, the comb filter is used for separating signals which are aliased together and extracting soft information, so that multiple versions of useful signals which are subjected to different fading are obtained, the diversity effect is realized, the reliability and the effectiveness of a system are further improved, the algorithm is simple and easy to implement, the inverse characteristic of a solution channel is not required, complex matrix inversion operation is avoided, the calculation complexity is greatly reduced, and the advantages of simplicity, fast convergence, easiness in realization and the like are achieved.

The main idea of the invention is that: the comb filter separates received aliasing signals path by path based on soft information generated by the decoder, extracts a plurality of noise-added versions of the useful signals, achieves diversity effect, decodes the enhanced signals by the decoder, and feeds the decoded information back to the comb filter. The process is iterated repeatedly, and finally the purpose of improving the performance of the communication system is achieved.

In order to achieve the above purpose, the present invention is realized by the following technical scheme.

A comb filtering Turbo equalization algorithm suitable for short wave communication comprises the following steps:

Step 1, setting a short wave communication channel to share p+q+1 steps, wherein p steps are arranged in front of a main path, q steps are arranged behind the main path, and a multi-path parameter at a t-th moment is h (t) = (h _-p(t),…,h₀(t),…,h_q (t)), then a codeword of a user message u to be transmitted after channel coding is c= (c ₀,c₁,…,c_j,…,c_n-1),c_j epsilon {0,1}, and then high-order modulation is carried out to obtain a mapping symbol x= (x ₀,x₁,…,x_j,…,x_n-1), wherein x _j＝1-2c_j,(x_j = ±1) and n are codeword lengths;

The transmitting end transmits a data symbol x ^prefix, and the receiving end receives a signal r after the data symbol is transmitted through a multipath channel;

Step 2, the receiving end adopts comb filtering Turbo equalization algorithm to equalize and decode the received signal, and outputs estimated code word

Further, in step1, the high-order modulation is BPSK modulation, QPSK modulation, 8PSK modulation.

Further, in step 1, the data symbol x ^prefix is:

x^prefix＝(x_-l,…,x_-1,x₀,x₁,…,x_j,…,x_n-1,x_n,…,x_n+l-1)

wherein, (x _-l,…,x_-1) is a front guard band, (x _n,…,x_n+l-1) is a rear guard band, the front guard band and the rear guard band are respectively any sequences composed of + -1, and the lengths of the front guard band and the rear guard band are respectively greater than or equal to p+q;

The mathematical expression of the received value of the received signal at the t moment is as follows:

r_t＝h_-p(t)x_p+t+…+h_-1(t)x_1+t+h₀(t)x_t+h₁(t)x_-1+t+…+h_q(t)x_-q+t+w_t,(-p<＝t<n+q)

Wherein w _t is a sampling value of normal distribution with a mean value of 0 and a variance of w-t.

Further, step 2 comprises the sub-steps of:

Sub-step 2.1, initializing: let the current iteration number l=0, set the maximum iteration number L of joint decoding, the initial decoder output soft information vector is set to an all 0 vector, i.e. L ^out (x) = (0, …,0, …, 0);

step 2.2, the received aliasing signals are separated path by path based on a signal separation algorithm to obtain a path separation matrix D ^dispart, and a mean matrix M ^dispart and a variance matrix V ^dispart are constructed;

substep 2.3, calculating a log-likelihood ratio L (x _j,k) for each equivalent received value;

step 2.4, soft information self-adaptive fusion is carried out to obtain a log likelihood ratio vector L (x);

Step 2.5, the log-likelihood ratio vector L (x) is sent to a decoder for decoding, and soft information L ^out (x) is output;

Step 2.6, adding 1 to the iteration variable L, and if L is less than L, entering a step 2.2; otherwise, entering a substep 2.7;

Sub-step 2.7, the soft information L ^out (x) output by the decoder is judged to obtain the decoding estimated codeword

Further, substep 2.2 comprises the substeps of:

Substep 2.2.1, constructing average information based on the soft information vector L ^out (x) output by the decoder Sum of variance informationWherein,

Wherein, tan h represents hyperbolic tangent; The size of (3) reflects the decoder's estimate of symbol x _j;

Sub-step 2.2.2, let soft information bit sequence number j=0, if j < n, enter sub-step 2.2.2.1; otherwise, entering a substep 2.2.3;

substep 2.2.2.1, let Calculating the analog output vector at the j-th momentWherein,

Sub-step 2.2.2.2, constructing a received vector r _j＝(r_j-p,…,r_j,…,r_j+q) at the j-th time and a difference vector d _j＝(d_j,-p,…,d_j,0,…,d_j,q), wherein,

Substep 2.2.2.3 calculating the mean value of d _j,k And variance of

Sub-step 2.2.2.4, add 1 to variable j, jump to sub-step 2.2.2;

Sub-step 2.2.3, constructing a path separation matrix D ^dispart:

constructing a mean matrix M ^dispart and a variance matrix V ^dispart:

Further, in substep 2.3, the log-likelihood ratio L (x _j,k) of each equivalent received value is:

Further, in the substep 2.4, the soft information is adaptively fused to obtain a log likelihood ratio vector L (x) as follows:

L (x) = (L (x ₀),…,L(x_j),…,L(x_n-1)), wherein,

Further, in sub-step 2.5, the output soft information L ^out (x) is:

L^out(x)＝(L^out(x₀),…,L^out(x_j),…,L^out(x_n-1))

Wherein L ^out(x_j) represents the log-likelihood ratio of the decoder output with respect to symbol x _j, which is defined as R _j is the received value at the j-th time; if L ^out(x_j) is not less than 0, judging the j bit as 0; on the contrary, it is judged as 1.

Further, substep 2.7, estimating the codewordWherein the criterion of the decision is

Compared with the prior art, the invention has the beneficial effects that:

The comb filtering Turbo equalization algorithm suitable for short wave communication adopts the comb filter to separate multipath aliasing signals from a time domain and extract soft information, so as to obtain multiple versions of user useful signals subjected to different fading, and realize the diversity effect; the decoder decodes the enhanced signal and feeds decoding information back to the comb filter; this process iterates repeatedly. The comb filtering Turbo equalization algorithm does not need inversion operation of a matrix, greatly reduces the calculation complexity, and has the advantages of simplicity, fast convergence, easiness in implementation and the like.

Drawings

The invention will now be described in further detail with reference to the drawings and to specific examples.

FIG. 1 is a block diagram of comb filter Turbo equalization logic of the present invention;

FIG. 2 is a graph showing the performance of the comb filter Turbo equalization algorithm and MMSE Turbo equalization algorithm of the present invention under different channels; wherein, the graph (a) is a performance graph under a time-invariant channel; fig. (b) is a performance diagram under a time-varying channel;

FIG. 3 is a graph showing the effect of different iteration numbers on performance of the comb filtering Turbo equalization algorithm of the present invention;

fig. 4 is a graph of the effect of the comb filtering Turbo equalization algorithm of the present invention on the performance of different paths.

Detailed Description

Embodiments of the present invention will be described in detail below with reference to examples, but it will be understood by those skilled in the art that the following examples are only for illustrating the present invention and should not be construed as limiting the scope of the present invention.

Referring to the logic block diagram of comb filter Turbo equalization of fig. 1, equalization is performed in an iterative manner. Firstly, a comb filter performs path-by-path separation on multipath signals to extract clear and hierarchical multipath user useful signalsEach signal can extract soft informationThereby playing a diversity effect; secondly, all soft information is combined to obtain L (x), and the L (x) is input into a decoder for decoding; then, an average information is constructed based on the soft information L ^out (x) output from the decoderAnd simulate the received signal in combination with known multipath channel parametersFinally, the comb filter uses the analog received signalAnd the true received signal r again extracts the multipath signal.

It should be noted that, in fig. 1, the multipath channel parameters are generally obtained by a "channel estimation" module, and the accuracy of the estimation directly affects the performance of the Turbo equalization system. Because the invention focuses on the comb filtering Turbo equalization algorithm, the receiving end is assumed to completely acquire the channel parameters for simplicity, and meanwhile, a Binary phase shift keying (Binary PHASE SHIFT KEY, BPSK) modulation mode is adopted to describe the technical scheme.

Specifically, the comb filtering Turbo equalization algorithm suitable for short wave communication of the invention comprises the following steps:

Step 1, setting a short wave communication channel to share p+q+1 steps, wherein p steps are arranged in front of a main path, q steps are arranged behind the main path, and a multi-path parameter at a t-th moment is h (t) = (h _-p(t),…,h₀(t),…,h_q (t)), then a codeword of a user message u to be transmitted after channel coding is c= (c ₀,c₁,…,c_j,…,c_n-1),c_j epsilon {0,1}, and then BPSK modulation is carried out to obtain a mapping symbol x= (x ₀,x₁,…,x_j,…,x_n-1), wherein x _j＝1-2c_j,(x_j = + -1) and n is a codeword length;

The transmitting end transmits the data symbol x ^prefix, and the receiving end receives the signal r _t after the data symbol is transmitted through a multipath channel.

Specifically, an information transmission model is built

The channel is assumed to have p+q+1 steps, p steps are arranged in front of the main path, q steps are arranged behind the main path, and the multipath parameter at the t-th moment is h (t) = (h _-p(t),…,h₀(t),…,h_q (t)). In view of the complexity of the actual communication environment, the communication system must have resistance to multipath. Thus adding a guard band of length l before and after transmitting the information, respectively, i.e

x^prefix＝(x_-l,…,x_-1,x₀,x₁,…,x_j,…,x_n-1,x_n,…,x_n+l-1)

Where (x _-l,…,x_-1) is the front guard band and (x _n,…,x_n+l-1) is the back guard band. The front and back guard bands may be any sequence consisting of + -1, which may be different in length, but should have a length of p+q or more. For convenience, let the front and back guard bands have the same length, i (i.e., p+q) and are all +1 vectors. The mathematical expression of the received value at the t time is known by the parameters

Wherein w _t is a sampling value of normal distribution with a mean value of 0 and a variance of w-t. For convenience of description of the comb filter principle, the reception vector r _j＝(r_j-p,…,r_j,…,r_j+q at the j (0. Ltoreq.j < n) th time is defined herein.

Step 2, the receiving end adopts comb filtering Turbo equalization algorithm to equalize and decode the received signal, and outputs estimated code wordThe method comprises the following steps:

a) The comb filter principle is as follows:

The core part of the invention is a comb filter, and has the capability of separating multipath aliasing signals path by path and extracting clear and hierarchical user information. This section will describe the working principle of the comb filter in detail. Assuming that the user soft information vector of the decoder output is L ^out (x) (this information is not completely correct), here, the j-th mapping symbol is taken as an example to illustrate how to extract the equivalent output of each path. The extraction method comprises the following steps:

1. Constructing average information based on soft information vector L ^out (x) output by decoder And variance information

Wherein the method comprises the steps of

Wherein,

Wherein, tan h represents hyperbolic tangent; The size of (3) reflects the decoder's estimate of symbol x _j, when the estimate is sufficiently accurate Indicating the degree of deviation (dispersion) of the evaluation result of the symbol x _j ifThe larger indicates a greater degree of dispersion.

2. Constructing an analog output vector at the j (0. Ltoreq. J < n) th time

Order theCalculating the analog output at the j-p, …, j, … j+q times, respectively

3. Construct the j (0.ltoreq.j < n) th time r _j andIs a difference vector d _j＝(d_j,-p,…,d_j,0,…,d_j,q of (a)

It can be seen from the above expression that the mathematical expression of all differences has the same form, and is composed of three parts.

Wherein the first part is a received value obtained by transmitting (fading) the jth bit information of the user via a certain path, for example, h _-p(j-p)x_j represents a received value obtained by x _j via a channel component h _-p (j-p); h ₀(j)x_j denotes a received value obtained by x _j via the channel component h ₀ (j); h _q(j+q)x_j denotes a reception value obtained by x _j via the channel component h _q (j+q), and the like.

The second part is the interference of the other paths to the first term (useful signal), while it can also be seen from the expression that the more reliable the average information output by the decoder, the less this interference.

The third part is the noise sample value introduced by the channel.

The second portion and the third portion are interfering with respect to the first portion. Each difference d _j,k should therefore be regarded as a random variable.

The first part is the ideal value of d _j,k (i.e., the value that should be obtained in the absence of interference). The sum of the second and third portions is the error (degree of deviation) between d _j,k and the ideal value.

The ideal value is related only to the transmission symbol x _j and the channel characteristic h _k (j+k) at the j+k-th time. X _j is an unknown quantity for the comb filter, and the value of the symbol x _j is not known when the log-likelihood ratio of the jth mapping symbol is calculated, so that the channel characteristic h _k (j+k) at the jth+k time can be taken as an ideal value (average value), namely

Average in generalSum of variancesThe method comprises the following steps of:

4. Traversing the sequence index variable j, (0 < = j < n), a difference matrix D ^dispart can be obtained:

at average information In the very accurate and very small channel superposition noise, the j-th column of the difference matrix D ^dispart is equivalent to the transmission result of the symbol x _j through the path h _-p(j-p),…,h₀(j),…,h_q (j+q), respectively (this can be seen from the mathematical expression of D _j,k), namely:

Based on this, the difference matrix D ^dispart is referred to as a path separation matrix.

Meanwhile, a mean matrix M ^dispart and a variance matrix V ^dispart after path separation can be obtained:

The dimensions of matrix D ^dispart、M^dispart、V^dispart are (p+q+1) x n.

The above 4 steps are similar to "combing" the aliased signal r with a "comb", extracting from the time domain the signal (viewed from the columns) after each symbol has been transmitted (faded) via the various paths. It can also be seen that the comb filter converts the aliased multipath time-varying channels into parallel (viewed from the row) separate channels. The purpose of the comb filter is to separate the multipath aliased signal from the time domain to obtain multiple noisy versions of the user's useful signal, similar to the use of a special "comb" to "comb" the multipath signal to extract a clear, hierarchical user signal.

B) The signal separation algorithm is as follows:

In general, the multipath channel parameter h (t) = (h _-p(t),…,h₀(t),…,h_q (t)), gaussian white noise variance is known Soft information vector L ^out(x)＝(L^out(x₀),…,L^out(x_j),…,L^out(x_n-1 output by the decoder), the received value r= (r _-p,…r_-1,r₀,…,r_n-1,…,r_n+q-1) after transmission via the multipath channel; the object is: the aliased signal is separated path by path. The specific algorithm for path-by-path separation of the aliased signal is as follows:

1. Constructing average information based on soft information vector L ^out (x) output by decoder Sum of variance informationWherein,

2. Let j=0. The following loop is entered when j < n:

2.1. order the Calculating the analog output vector at the j-th momentWherein the method comprises the steps of

2.2. Constructing a received vector r _j＝(r_j-p,…,r_j,…,r_j+q) at a j-th time and a difference vector d _j＝(d_j,-p,…,d_j,0,…,d_j,q), wherein,

2.3. Calculating the mean value of d _j,k And variance of

2.4. Adding 1 to the variable j, and jumping to the sub-step 2;

3. Constructing a path separation matrix D ^dispart:

constructing a mean matrix M ^dispart and a variance matrix V ^dispart:

c) Multipath soft information fusion

After the received aliased signal is separated path by path to obtain the separation matrix D ^dispart, a corresponding log-likelihood ratio L (x _j,k) is calculated based on the mean matrix M ^dispart and the variance matrix V ^dispart.

Because the log-likelihood ratio vector corresponding to each path contains the same information, the log-likelihood ratio vector of any path can be independently input to a decoder for decoding. It should be noted that, although the log likelihood ratio vectors corresponding to each path contain the same information, their reliability (reliability) is different, and some are high and others are low. Intuitively, the reliability is related to the corresponding channel coefficient, and the reliability of high channel coefficient is high.

However, the log likelihood ratio vectors of the paths are typically superimposed/fused for better decoding performance. Multipath soft information fusion generally adopts a direct addition method, and the following formula is adopted:

d) Comb filtering Turbo equalization algorithm

And taking the fused soft information as the input of a decoder to perform a new round of decoding, thereby obtaining a more accurate soft information vector L ^out (x), and feeding back to a comb filter to separate the aliasing signals again. Thus there is the following comb filter Turbo equalization algorithm:

It is known that: multipath channel parameter h (t) = (h _-p(t),…,h₀(t),…,h_q (t)) and gaussian white noise variance w to;

an initial soft information vector L ^out(x)＝(L^out(x₀),…,L^out(x_j),…,L^out(x_n-1);

The received value r= (r _-p,…r_-1,r₀,…,r_n-1,…,r_n+q-1) after transmission over the multipath channel.

The object is: turbo equalization of received multipath signals to obtain codeword estimates

0. Initializing, namely enabling an iteration variable l=0, and setting the maximum iteration number L of joint decoding;

1. Constructing a path separation matrix D ^dispart, a mean matrix M ^dispart and a variance matrix V ^dispart based on a signal separation algorithm;

2. calculating the log-likelihood ratio L (x _j,k) of each equivalent received value

3. Soft information adaptive fusion yields a log-likelihood ratio vector L (x) = (L (x ₀),…,L(x_j),…,L(x_n-1)), where,

4. Sending L (x) into a decoder for decoding, and outputting soft information L ^out (x);

Let the soft information vector output by the decoder be L ^out(x)＝(L^out(x₀),…,L^out(x_j),…,L^out(x_n-1)). Wherein symbol L ^out(x_j) represents the log-likelihood ratio of the decoder output with respect to symbol x _j, which is defined as Where r _j is the received value at time j. If L ^out(x_j) is not less than 0, judging the j bit as 0; on the contrary, it is judged as 1. Obviously, the larger |l ^out(x_j) | is, the more reliable the result of the determination is.

5. Adding 1 to the iteration variable L, if L is less than L, entering the step 1, otherwise, entering the step 6;

6. judging the soft information output by the decoder to obtain a decoded codeword The judgment criterion is

The decoder cannot learn any soft information about the user at the initial time of iteration, nor can it provide any useful information to the build average information module, typically the initial soft information vector is set to a full 0 vector, i.e., L ^out (x) = (0, …,0, …, 0).

It should be emphasized and pointed out that the present invention only uses BPSK modulation mode as an example to illustrate the comb filtering Turbo equalization algorithm and the main idea, and the idea of comb filtering is equally applicable to other modulation modes (such as QPSK, 8PSK, etc.).

Performance simulation

In order to verify the performance of the comb filtering Turbo time domain equalization algorithm provided by the invention, the LDPC code is used as an error correction code to simulate the BPSK modulation under the baseband. The signal to noise ratio in the simulation is defined asThe unit is dB. Where E _s is the average symbol energy received, Is the gaussian white noise variance. E _s =1 is typically given in the simulation. The simulation mainly looks at the following 3 points:

A. The performance of the comb filtering Turbo time domain equalization algorithm in a multipath environment;

B. the iteration times influence the equalization performance of the comb filtering Turbo time domain;

C. The influence of the number of the extraction paths on the overall performance is extracted.

All simulation parameters were set as follows:

LDPC code: selecting LDPC code (516,1032) adopted by mobile broadband wireless access standard (IEEE 802.16 e), wherein the code rate is 12×24, and the expansion factor is 43; the LDPC decoding algorithm is Sum-product decoding algorithm (Sum-Product Algorithm, SPA), and the iteration number is 50;

baud rate (transmit symbol rate): 2000 symbols/sec.

The time-invariant channel means that the characteristics of the channel are the same at each moment and do not change with time;

A time-varying channel means that the channel characteristics are different at each time, but the characteristics at the front and rear time have a certain correlation. The dual-path 2ms means that two channels exist and a front-back interval of 2ms is formed; fading 1Hz refers to doppler shift due to the relative motion of the transmit and receive two-shot, the doppler spectrum being 1Hz wide, and the wider the width the more intense the channel variation.

Here, the following simulation focuses on the performance of the Turbo equalization algorithm, so it is assumed that the receiving end knows the parameters of the multipath channel, and in practical application, the channel estimation algorithm may be used to obtain the parameters related to the multipath channel.

Simulation A

The performance of the comb filtering Turbo time domain equalization algorithm under channel 1 and channel 2 is examined. Meanwhile, for comparison, paper Turbo Equalization is also presented in the figure: PRINCIPLES AND NEW Results describes the performance of the MMSE Turbo equalization algorithm with a transversal filter order of 25. The joint iteration times of the two algorithms are 5 times; the simulation performance is shown in fig. 2.

It can be seen from fig. 2 that both equalization algorithms have nearly identical performance in both time-invariant and time-variant channels. For example, when ber=10 ^-5, the required signal-to-noise ratio of the MMSE Turbo equalization algorithm and the comb filter Turbo equalization algorithm under the time-invariant channel is 2.13dB and 2.16dB, respectively, as shown in fig. 2 (a); the required signal-to-noise ratio under the time-varying channel is 12.2dB and 11.9dB, respectively, as shown in fig. 2 (b). However, the MMSE Turbo equalization process involves inversion operation of the channel cyclic matrix, and the calculation amount increases with the increase of the order of the transversal filter, and the complexity and calculation amount are far higher than those of the comb filtering Turbo equalization algorithm. Therefore, the comb filtering Turbo equalization algorithm can greatly reduce the complexity of engineering realization on the premise of almost not losing performance.

Simulation B

And (5) examining the influence of the iteration times on the equalization performance of the comb filtering Turbo. As shown in fig. 3, the effect of different iterations of the algorithm on performance under channel 1 is shown. From the figures it can be seen that:

1) The performance of the comb filtering Turbo equalization algorithm is gradually improved along with the increase of iteration times;

2) The convergence speed of the comb filtering Turbo equalization algorithm is high, and the performance stability state is basically achieved only by 2 iterations. For example, at bit error rate ber=10 ^-5, the required signal-to-noise ratios for 2 iterations and 5 iterations are 2.15 and 2.19dB, respectively, with only 0.04dB spacing between them.

Simulation C

As described above, the comb filter performs path-by-path separation on the multipath aliasing signal, extracts a plurality of noisy versions of the useful signal, and since the information contained in each path is the same, the log likelihood ratio vector of any path can be decoded by a separate input decoder. The simulation examines the influence of the number of different paths extracted under channel 1 on the overall performance. The number of extraction is 1 path (path 2), 2 paths (paths 1, 2) and all paths, respectively. The simulation results are shown in fig. 4. From the figures it can be seen that:

1) The greater the number of extraction paths, the better the performance. For example, at bit error rate ber=10 ^-5, the signal to noise ratios required to extract 1 path, extract 2 paths, and extract all paths are 4.40dB, 2.85dB, and 2.15dB, respectively;

2) The comb filter can effectively separate the aliased signals to obtain multiple noisy versions of the useful signal, achieving diversity effects. For example, the performance of combining the 2 paths extracted by the comb filter is far better than that of extracting 1 path, and the two paths have about 1.55dB difference when the bit error rate ber=10 ^-5. When all paths are extracted, the comb filtering Turbo equalization algorithm achieves optimal performance.

While the invention has been described in detail in this specification with reference to the general description and the specific embodiments thereof, it will be apparent to one skilled in the art that modifications and improvements can be made thereto. Accordingly, such modifications or improvements may be made without departing from the spirit of the invention and are intended to be within the scope of the invention as claimed.

Claims

1. The comb filtering Turbo equalization algorithm suitable for short wave communication is characterized by comprising the following steps of:

Step2 comprises the following sub-steps:

2. Comb filtering Turbo equalization algorithm suitable for short wave communication according to claim 1, characterized in that in step 1, the higher order modulation is BPSK modulation, QPSK modulation, 8PSK modulation.

3. The comb filter Turbo equalization algorithm for short-wave communication according to claim 1, wherein in step 1, the data symbol x ^prefix is:

x^prefix＝(x_-l,…,x_-1,x₀,x₁,…,x_j,…,x_n-1,x_n,…,x_n+l-1)

r_t＝h_-p(t)x_p+t+…+h_-1(t)x_1+t+h₀(t)x_t+h₁(t)x_-1+t+…+h_q(t)x_-q+t+w_t,(-p＜＝t＜n+q)

wherein w _t is t time obeying the mean value of 0 and the variance of 0 Is a normal distribution of sampled values.

4. Comb filtering Turbo equalization algorithm suitable for short wave communication according to claim 1, characterized in that substep 2.2 comprises the substeps of:

Substep 2.2.2.3 calculating the mean value of d _j,k And variance of

Sub-step 2.2.2.4, add 1 to variable j, jump to sub-step 2.2.2;

Sub-step 2.2.3, constructing a path separation matrix D ^dispart:

constructing a mean matrix M ^dispart and a variance matrix V ^dispart:

5. The comb-filter Turbo equalization algorithm for short-wave communication according to claim 4, wherein in substep 2.3, the log-likelihood ratio L (x _j,k) of each equivalent received value is:

6. The comb-filtered Turbo equalization algorithm for short-wave communication according to claim 5, wherein in sub-step 2.4, soft information adaptive fusion results in a log-likelihood ratio vector L (x) as:

L (x) = (L (x ₀),…,L(x_j),…,L(x_n-1)), wherein,

7. Comb filtering Turbo equalization algorithm suitable for short wave communication according to claim 1, characterized in that in sub-step 2.5, the output soft information L ^out (x) is:

L^out(x)＝(L^out(x₀),…,L^out(x_j),…,L^out(x_n-1))

8. Comb filtering Turbo equalization algorithm suitable for short wave communication according to claim 7, characterized in that substep 2.7, the estimated codewordWherein, the criterion of judgement is: