A GI-free Joint Decoding-Iterative Frequency Domain Equalization Method, System, Device and Medium for FTN Signals

Abstract

A joint decoding-iterative frequency-domain equalization method, system, device and medium for GI-free FTN signals. Method: Divide the bit sequence to be transmitted, concatenate all the coded and interleaved bit sequences, perform constellation mapping and FTN shaping processing; The received signal passes through a matched filter and a downsampler to obtain the FTN received signal at the symbol rate; Divide according to the FFT sliding window of length N; Judge the current number of iterations; Use IFFT operation to generate the time-domain equalization result; Send the time-domain equalization result to the eraser to obtain the data not affected by inter-symbol interference and inter-block interference; Soft demodulation; Judge whether to perform the next iteration; The system, device and medium are used to implement a joint decoding-iterative frequency-domain equalization method for GI-free FTN signals; The present invention removes GI from the existing FTN joint decoding-iterative frequency-domain equalization algorithm and deletes the interfered part through an eraser, so as to effectively improve the system spectrum utilization rate on the premise of maintaining the system bit error rate performance, thereby improving its communication performance.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communication, and particularly relates to a joint decoding-iterative frequency domain equalization method, system, device and medium for FTN signals without GI. Background Art

[0002] With the rapid development of communication technologies, a large number of wireless devices such as smart phones and tablets have emerged. People's demand for high-definition multimedia data and ultra-high wireless transmission rates continues to increase. How to further improve the frequency band utilization rate for baseband data transmission systems on the premise of scarce spectrum resources in modern society is a key issue in the 5th Generation Mobile Communication Technology (5G). As a non-orthogonal data transmission method, the Faster Than Nyquist (FTN) technology can transmit more information in the same time at the cost of artificially introducing certain inter-symbol interference (ISI), and can provide a higher channel capacity and throughput than the traditional Nyquist system transmission technology. It is a candidate technology for 5G communication.

[0003] The Faster Than Nyquist (FTN) technology can transmit more data symbols in the same time through non-orthogonal data transmission, and can provide a higher channel capacity and throughput than the traditional Nyquist transmission system. Compared with Nyquist signals, the transmission shaping pulse interval of FTN signals is less than the minimum symbol interval corresponding to the Nyquist criterion without inter-symbol interference (ISI). The time-domain shaping pulses at the transmitter are no longer orthogonal, inevitably introducing interference between signals, which is equivalent to the transmitted signal passing through a frequency-selective channel with severe fading. If the channel environment of the FTN system is a multipath Rayleigh channel, the transmitted signal is affected by both the inherent ISI and the multipath channel. How to design an equalizer to eliminate interference is an important issue in the development of FTN technology.

[0004] The frequency-domain equalization scheme is a technique widely used in the detection of FTN signals. Sugiura et al. first proposed applying the frequency-domain equalization technique to the FTN system. The frequency-domain equalization technique is applicable to the case where the ISI taps are long. The computational complexity is proportional to the data block length, and the complexity does not increase significantly with the increase of the modulation order. To improve the throughput of the FTN system, some literatures add a channel coding and decoding module on the basis of frequency-domain equalization, and extend the hard-decision MMSE frequency-domain equalization to soft-decision frequency-domain equalization, jointly constituting a multi-stage cascaded iterative frequency-domain equalization system with the decoder. Subsequently, some scholars proposed a three-stage cascaded joint decoding-iterative equalization. This scheme can further improve the system bit error rate performance on the basis of the two-stage cascade. However, all these iterative frequency-domain equalization algorithms require inserting a sufficiently long cyclic prefix (CP) or a specific training sequence as the GI before the data. Moreover, the length of the inserted sequence is usually required to be greater than the maximum multipath delay of the channel. This will lead to transmission redundancy and reduce the spectral efficiency. Summary of the Invention

[0005] In order to overcome the above deficiencies of the prior art, the purpose of the present invention is to provide a method, system, device and medium for joint decoding-iterative frequency-domain equalization of FTN signals without GI, which removes the GI in the existing FTN joint decoding-iterative frequency-domain equalization algorithm and deletes the interfered part by adding a deleter, so as to effectively improve the system spectral utilization rate on the premise of maintaining the system bit error rate performance, thereby improving its communication performance.

[0006] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0007] A method for joint decoding-iterative frequency-domain equalization of FTN signals without GI, comprising the following steps:

[0008] Step 1, segment the original bit sequence to be transmitted, send it into the encoder block by block, then perform in-block interleaving, and then concatenate all the encoded and interleaved bit sequences together, perform constellation mapping and FTN shaping processing to obtain the FTN transmission signal;

[0009] Step 2, after passing the FTN transmission signal obtained in Step 1 through the channel, obtain the FTN received signal. The FTN received signal passes through a matched filter and a downsampler to obtain the FTN received signal at the symbol rate;

[0010] Step 3, divide the FTN received signal at the symbol rate obtained in Step 2 according to an FFT sliding window of length N. The first sliding window of the data stream is selected from the first p symbols before the start position of the received data, and the start position of the next sliding window is M symbols after the left boundary of the previous sliding window;

[0011] Step 4, determine the current number of iterations; if it is the first iteration, the number of iterations \(l = 0\), execute Step 4.1, perform parallel frequency-domain equalization on the received signal within the FFT window based on the minimum mean square error (MMSE) criterion to obtain the equalization result of the first iteration; otherwise, execute Step 4.2, generate the forward filter coefficients \(\{C p \}\) and feedback filter coefficients \(\{B p \}\) according to the detection data of the \((l - 1)\)th time and the filter design criterion, and perform parallel frequency-domain equalization to obtain the equalization result of the \(l\)th iteration;

[0012] Step 5, perform IFFT operation on the equalization result \(U (l) obtained in Step 4 to generate the time-domain equalization result;

[0013] Step 6, send the time-domain equalization result \(u (l) to the eraser to obtain the data free from inter-symbol interference and inter-block interference;

[0014] Step 7, perform soft demodulation on the data free from inter-symbol interference and inter-block interference to obtain the bit LLR information, then perform code group segmentation on the bit LLR information, perform deinterleaving and channel decoding block by block to obtain the LLR extrinsic information of the coded bits and the estimated value \(\hat{b} n ;

[0015] Step 8, determine whether to perform the next iteration; if the iteration is completed, execute Step 8.1, output the initial estimated value \(\hat{b} n of the transmitted bit sequence; if the next iteration is still required, execute Step 8.2, send the LLR extrinsic information of the coded bits to the interleaver and then perform QPSK soft mapping to obtain the expected information of the transmitted modulation symbols, and then continue from Step 4 to Step 8 until the iteration is completed, and output the estimated value \(\hat{b} n of the transmitted bit sequence.

[0016] The specific method of the above-mentioned Step 1 is as follows:

[0017] The above-mentioned FTN shaping process includes an upsampler and a shaping filter;

[0018] The sampling factor of the above-mentioned upsampler is less than the upsampling multiple of the shaping filter. The compression factor of the FTN signal is adjusted by changing the sampling factor of the upsampler and the upsampling multiple of the shaping filter. The specific relationship is that the compression factor is equal to the sampling factor of the upsampler divided by the upsampling multiple of the shaping filter. If the sampling factor of the upsampler and the upsampling multiple of the shaping filter are equal, the compression factor is equal to 1, realizing Nyquist transmission.

[0019] The specific method of the above-mentioned Step 4.1 is as follows:

[0020] Set the number of iterations \(l = 0\) in the initial state, perform parallel frequency-domain equalization on the received signal within the FFT window based on the MMSE criterion, and obtain the initial iteration equalization result;

[0021] Step 4.1.1, design the forward filter coefficients based on the MMSE criterion Simplify the matrix \(Q_{E}[n,n]\) H Q H to the diagonal matrix \(\varPhi\) n , considering the simplified correlation of colored noise, the forward filter coefficients for the initial iteration are:

[0022]

[0023] where \(H\) p is the channel frequency response, \(\varPhi\) n [p] is the \(p\)-th element on the diagonal of the diagonal matrix \(\varPhi\) n , which can be expressed as:

[0024]

[0025] where \(g = [g(-vT),\cdots,g(0),\cdots,g(vT)]\in R\) 2v+1 is the channel impulse response corresponding to the combined action of the transmit filter, the transmission channel, and the receive filter, and the channel length is \(N\) h = 2v + 1;

[0026] Step 4.1.2, multiply the received frequency-domain signal \(R\) p element-wise with the filter coefficients \(\{C\) p \}\) to obtain the forward filter output vector \(Z\) (l) , and the specific expression is:

[0027]

[0028] where are the forward filter coefficients; \(R\) p is the received frequency-domain signal; \(Z\) (l) is the forward filter output vector;

[0029] Step 4.1.3, the equalization result \(U\) (l) is equal to the forward filter output vector \(Z\) (l) . Since this is the initial iteration and no detection data is available, the feedback filter is turned off, and the equalization result is directly obtained from the output of the forward filter, which is expressed as:

[0030] U( l) = Z (l) ;

[0031] The specific method of the said Step 4.2 is:

[0032] For the iteration number \(l > 0\), generate the forward filter parameters \(\{C\) p \} and the feedback filter parameters \(\{B\) p \} according to the detection data of \((l - 1)\) times and the filter design criterion, and perform parallel frequency-domain equalization;

[0033] Step 4.2.1, design the forward filter and the feedback filter based on the detection data of \((l - 1)\) times and the MMSE design criterion. Assume that the statistical characteristics of the time-domain detection signal and the transmitted signal are the same, and obtain the forward filter coefficients \(C\) k and the feedback filter coefficients \(B\) k as follows:

[0034]

[0035]

[0036] In the coefficient expressions described above, \(\kappa\) is a normalization parameter used to control the constant amplitude of the equalized output symbol, that is

[0037] In the coefficient expressions described above, is represented by the credibility \(\rho\), and the credibility \(\rho\) can be estimated by a simple method, which is expressed as follows:

[0038]

[0039] where \(\rho\) n is the credibility of the \(n\)th detection symbol, and can be obtained from the in-phase component credibility \(\rho\) n,I and the quadrature component credibility \(\rho\) n,Q The calculation method is:

[0040]

[0041]

[0042]

[0043] In the formula, the value ranges of the in-phase credibility and the quadrature credibility are both If the equalization is at the 0th iteration, initialize

[0044] Step 4.2.2, multiply the received frequency-domain signal \(R\) p element-wise with the forward filter coefficients \(\{C\) p \} to obtain the forward filter output vector \(Z\) (l) , which is expressed as follows:

[0045]

[0046] Among them, are the forward filter coefficients; R p is the received frequency-domain signal; Z (l) is the forward filter output vector;

[0047] Step 4.2.3, convert the detection data of the (l-1) iteration to the frequency domain to obtain Then, multiply the frequency-domain detection signal element-wise with the feedback filter coefficients {B p} to obtain the feedback filter output Y (l) , and the expression is as follows:

[0048]

[0049] Step 4.2.4, accumulate the forward filter output Z (l) and the feedback filter output vector Y (l) to obtain the equalization output U (l) ,

[0050] U (l) =Z (l) +Y (l) .

[0051] The specific method of the said Step 5 is:

[0052] Let the time-domain equalization result be u (l) , then it can be expressed as:

[0053]

[0054] Among them, Q is the FFT matrix, and the element in its k-th row and j-th column is H represents the Hermitian operation.

[0055] The specific method of the said Step 6 is:

[0056] Use a deleter to delete the symbol segments with a length of p at the head and a length of q at the tail of the equalization result, and only keep the equalization output data u' (i) of the i-th FFT window that is not disturbed;

[0057] The said deletion process can be expressed by the formula:

[0058] u' (i) =(0 M×p ,I M ,0 M×q )u (l)

[0059] Among them, 0 m×n is a zero matrix with m rows and n columns, IM is an identity matrix with M rows and M columns.

[0060] The specific method of step 7 is as follows:

[0061] Soft-demodulate the data that is not affected by inter-symbol interference and inter-block interference. When using QPSK modulation, the bit LLR information can be expressed as:

[0062]

[0063]

[0064] In the formula, is the time-domain symbol after equalization The mean square error of can be estimated as:

[0065]

[0066] s′ is the equalized symbol The result of hard decision to the constellation diagram can be expressed as:

[0067]

[0068] The channel decoding module used for the channel decoding is a soft input soft output channel decoder SISO. After decoding, the extrinsic information of the nth symbol of the coded bits is obtained. Assume that the LLR soft information of the two bits of the nth symbol in the coded bits given by the channel decoder SISO are respectively The LLR information is defined as:

[0069]

[0070]

[0071] In the formula, the symbol respectively represent the LLR information The corresponding bits.

[0072] The specific method of step 8 is as follows:

[0073] Step 8.1, complete the iteration and output the estimated value b n ;

[0074] Step 8.2, another iteration is required. Send the LLR extrinsic information of the coded bits into the interleaver and then perform QPSK soft mapping to obtain the expected information of the transmitted modulation symbols, and then continue with steps 4 to 8 until the iteration is completed and the initial estimated value b n ;

[0075] In the QPSK soft mapping described above, the symbol sn Desired information can be obtained from the in-phase component and the quadrature component as follows:

[0076] The mean information of the in-phase component and the mean information of the quadrature component can be expressed as:

[0077]

[0078]

[0079] Obtain the desired information of the transmitted modulation symbol, that is, after detecting the data, continue steps 4 to 8 using the detected data until the iteration is completed, and output the initial estimate value b of the transmitted bit sequence n .

[0080] The present invention also provides a system for a joint decoding-iterative frequency-domain equalization method of FTN signals without GI, including:

[0081] A preprocessing module: used to preprocess the FTN received signal. The received signal stream is divided according to an FFT sliding window of length N to obtain parallel data blocks to be processed. The first sliding window of the data stream is selected from the first p symbols before the start position of the received data, and the start position of the subsequent sliding window is shifted M symbols to the right of the left boundary of the previous sliding window;

[0082] A frequency-domain equalization module: used to calculate the forward filter parameters {C p} and the feedback filter parameters {B p}, perform parallel frequency-domain equalization to obtain the equalization result of the l-th iteration. When the number of iterations is 0, the feedback filter is turned off, and the frequency-domain equalization module degenerates into a forward filter module, and the output of the forward filter is the equalization result; this module is also used to generate a time-domain equalization result by using IFFT processing;

[0083] A deleter module: used to delete the symbol segments with a length of p at the head and a length of q at the tail of the equalization result, and only retain the equalized output data u′ of the i-th FFT window that is not disturbed (i) ;

[0084] A soft demodulation module: used to calculate the bit LLR soft information according to the equalized output data;

[0085] A soft mapping module: used to perform QPSK soft mapping according to the interleaved coded bit soft information to obtain the data symbol expectation, that is, the detected data;

[0086] Output module: used to determine the current iteration count and decide whether the iteration is completed. If completed, it outputs the estimated value b of the transmitted bit sequence. n If not completed, this module does not perform any operation.

[0087] The present invention also provides a joint decoding - iterative frequency - domain equalization device for FTN signals without GI, including:

[0088] A memory for storing computer programs;

[0089] A processor for implementing the joint decoding - iterative frequency - domain equalization method for FTN signals without GI when executing the computer program.

[0090] A computer - readable storage medium stores a computer program, and when the computer program is executed by a processor, it can implement a joint decoding - iterative frequency - domain equalization method for FTN signals without GI.

[0091] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0092] 1. Aiming at the problem that the existing frequency - domain equalization scheme requires a sufficiently long guard interval, which will lead to transmission redundancy and reduce spectral efficiency, the present invention provides a joint decoding - iterative equalization method for FTN signals without GI. Based on the existing joint decoding - iterative frequency - domain equalization algorithm, this method removes the GI and deletes the interfered part through an adder - subtractor, which can effectively improve the frequency - offset utilization rate of the system without degrading the system bit - error rate performance.

[0093] 2. Aiming at the situation where the FTN received signal has colored noise, step 4 of the present invention provides a simplified method for designing an equalization filter related to colored noise in the frequency domain, which can improve the equalization performance of FTN signals without increasing the algorithm complexity.

[0094] 3. Step 4 of the present invention adopts a block - based data structure and separates the forward filter and the feedback filter, which can achieve parallel processing and pipeline processing, improving the processing speed and throughput.

[0095] In summary, on the basis that the FTN signal improves the spectral efficiency compared with the Nyquist signal, the present invention can further improve the spectral efficiency of the system without degrading the system bit - error rate performance. In the context of extremely scarce spectral resources in modern society, this performance improvement is very valuable. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] Figure 1 It is a flowchart of the implementation method of the present invention.

[0097] Figure 2Schematic diagram of the structure of a GI - free joint decoding - iterative frequency - domain equalization system for FTN signals implemented according to the present invention.

[0098] Figure 3 Schematic diagram of the FFT window sliding according to the present invention.

[0099] Figure 4 Schematic diagram of the implementation of the eraser according to the present invention.

[0100] Figure 5 Schematic diagram of the receiver algorithm flow according to the present invention.

[0101] Figure 6 BER comparison between the GI - free joint decoding - iterative equalization method and the method with added GI under different compression factors and roll - off factors. Detailed implementation manners

[0102] In order to enable those skilled in the art to better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the present application.

[0103] The method of the present invention includes obtaining the initial extrinsic information of the encoded bits and the estimated values of the original bits through the first iteration, and in the iteration, the receiving end performs processing using frequency - domain iterative equalization. Specifically, it includes designing a forward filter and a feedback filter according to the information fed back by the soft - input soft - output decoder; deleting the disturbed parts at the head and tail of the equalization result; performing soft demapping to obtain the log - likelihood ratio (LLR) information of the bits; feeding it into the de - interleaver and the decoder to obtain the estimated values of the original bits and the extrinsic information of the encoded bits in the current iteration for the next iteration. The present invention can effectively solve the problem that adding a guard interval (GI) in a faster - than - Nyquist (FTN) system with frequency - domain equalization reduces the spectral efficiency, and realizes improving the spectral utilization rate of FTN signals without changing the system energy efficiency.

[0104] As Figure 1 、 Figure 2 shown, the present invention provides a GI - free joint decoding - iterative frequency - domain equalization method for FTN signals. Figure 5 is the algorithm flow chart at the receiving end. Combining Figure 1 、 Figure 2 with Figure 5 , the algorithm flow is described as follows:

[0105] As Figure 1As shown in the figure, in step 1, the sending - end process is as follows: The sending - end information source generates random binary bits, divides the original bit sequence to be sent, and feeds it into the channel encoder block - by - block to obtain the encoded bit sequence with error - control redundancy added. Subsequently, intra - block interleaving is performed to randomly scramble the bit order, and then all the interleaved encoded bit sequences are concatenated together. After constellation mapping, a symbol - level sequence is obtained. Then, through the FTN mapping module, an accelerated data symbol is obtained, that is, the FTN transmission signal is obtained.

[0106] The FTN shaping process described above includes two parts: an up - sampler and a shaping filter.

[0107] The sampling factor of the up - sampler should be less than the up - sampling multiple of the shaping filter so that the FTN signal can be generated. Moreover, the compression factor of the FTN signal can be adjusted by changing the sampling factor of the up - sampler and the up - sampling multiple of the shaping filter. The specific relationship is that the compression factor is equal to the sampling factor of the up - sampler divided by the up - sampling multiple of the shaping filter. If the sampling factor of the up - sampler and the up - sampling multiple of the shaping filter are equal, the compression factor is equal to 1, realizing Nyquist transmission.

[0108] In step 2, after the FTN transmission signal obtained in step 1 passes through the channel, the FTN received signal is obtained. The FTN received signal passes through a matched filter and a down - sampler to obtain the FTN received signal at the symbol rate.

[0109] The matched filter corresponds to the shaping filter at the transmitting end. Assuming ideal timing synchronization at the receiving end, the down - sampler decimates at an interval of τT s where T s is the symbol transmission interval under the Nyquist criterion.

[0110] As Figure 3 shown, in step 3, the FTN received signal at the symbol rate obtained in step 2 is divided according to an FFT sliding window of length N. Note that the first sliding window of the data stream needs to be selected starting from p symbols before the start position of the received data, and the start position of the next sliding window is M symbols shifted to the right of the left boundary of the previous sliding window, so as to ensure that the complete data stream can be obtained.

[0111] In step 4, judge the current iteration number. If it is the first iteration, the iteration number l = 0, and step 4.1 is executed. The received signal within the FFT window is subjected to parallel frequency - domain equalization based on the minimum mean - square error (MMSE) criterion to obtain the equalization result of the first iteration. Otherwise, step 4.2 is executed. According to the detection data of the (l - 1)th time and the filter design criterion, the forward filter coefficients {C p} and the feedback filter coefficients {B p} are generated, and parallel frequency - domain equalization is performed to obtain the equalization result of the lth iteration.

[0112] Step 4.1, initially set the number of iterations \(l = 0\), perform parallel frequency-domain equalization on the received signal within the FFT window based on the minimum MMSE criterion to obtain the initial iteration equalization result.

[0113] Step 4.1.1, design the forward filter coefficients based on the MMSE criterion Due to the existence of the non-orthogonal matched filter, the received noise in the FTN signal is no longer white noise, but colored noise with time-domain correlation \(E[n(mT)n * (nT)] = N 0 g(mT - nT)\), resulting in non-diagonal frequency-domain noise correlation \(Q_{E}[nn H Q H and a high computational complexity when substituting it into the MMSE equalization weight matrix. To reduce the high complexity brought by the autocorrelation of the colored noise, the matrix \(Q_{E}[nn H Q H is simplified to a diagonal matrix \(\varPhi n . At this time, the algorithm complexity is the same as that when the received noise is Gaussian white noise. Considering the simplified correlation of the colored noise, the forward filter coefficients for the initial iteration are:

[0114]

[0115] where \(H p is the channel frequency-domain response, \(\varPhi n [p]\) is the \(p\)-th element on the diagonal of the diagonal matrix \(\varPhi n and can be expressed as:

[0116]

[0117] where \(g = [g(-vT), \ldots, g(0), \ldots, g(vT)] \in \mathbb{R} 2v+1 is the channel impulse response corresponding to the combined action of the transmit filter, the transmission channel, and the receive filter, and the channel length is \(N h = 2v + 1\).

[0118] Step 4.1.2, multiply the received frequency-domain signal \(R p element-wise with the forward filter coefficients \(\{C p \}\) to obtain the forward filter output vector \(Z (l) . The specific expression is:

[0119]

[0120] where are the forward filter coefficients; \(R p is the received frequency-domain signal; \(Z (l) is the forward filter output vector.

[0121] Step 4.1.3, the equalization result U (l) is equal to the forward filter output vector Z (l) . Since this is the initial iteration and no detection data is available, the feedback filter is turned off, and the equalization result is directly obtained from the forward filter output, which is expressed as:

[0122] U (l) = Z (l) .

[0123] Step 4.2, when the iteration number l > 0, generate the forward filter parameters {C p} and the feedback filter parameters {B p} according to the (l - 1) - th detection data and the filter design criterion, and perform parallel frequency - domain equalization.

[0124] Step 4.2.1, design the forward filter and the feedback filter based on the (l - 1) - th detection data and the MMSE design criterion. Assume that the statistical characteristics of the time - domain detection signal are the same as those of the transmitted signal. Through theoretical derivation, the forward filter coefficient C k and the feedback filter coefficient B k are:

[0125]

[0126]

[0127] In the coefficient expressions described above, κ is a normalization parameter used to control the constant amplitude of the equalization output symbol, that is

[0128] In the coefficient expressions described above, is represented by the credibility ρ. The credibility ρ can be estimated by a simple method, which is expressed as follows:

[0129]

[0130] where ρ n is the credibility of the n - th detection symbol, which can be obtained from the in - phase component credibility ρ n,I and the quadrature component credibility ρ n,Q . The calculation method is:

[0131]

[0132]

[0133]

[0134] In the formula, the value ranges of both the in - phase credibility and the quadrature credibility are If the equalization is at the 0th iteration, initialize

[0135] Step 4.2.2, multiply the received frequency-domain signal R p element-wise with the forward filter coefficients {C p} to obtain the forward filter output vector Z (l) , which is expressed as follows:

[0136]

[0137] where, are the forward filter coefficients; R p is the received frequency-domain signal; Z (l) is the forward filter output vector.

[0138] Step 4.2.3, convert the detection data information of the (l - 1)th iteration to the frequency domain to obtain Multiply the frequency-domain detection signal element-wise with the feedback filter {B p} to obtain the feedback filter output Y (l) , and the expression is as follows:

[0139]

[0140] Step 4.2.4, accumulate the forward filter output Z (l) and the feedback filter output vector Y (l) to obtain the equalization output U (l) ,

[0141] U (l) = Z (l) + Y (l) .

[0142] Step 5, apply the IFFT operation to the equalization result U (l) obtained in Step 4 to generate the time-domain detection signal vector.

[0143] Let the time-domain equalization result be u (l) , then it can be expressed as:

[0144]

[0145] where Q is the FFT matrix, and the element in its k-th row and j-th column is H represents the Hermitian operation.

[0146] For example Figure 4As shown, in step 6, a deleter is used to delete the symbol segments with length p at the head and length q at the tail of the equalization result, so as to obtain data free from inter-symbol interference and inter-block interference. Since no GI is added, the head and tail of the frequency-domain equalized output data block will be affected by residual interference, and only the middle data is the part we expect to obtain.

[0147] The deletion process can be expressed by the following formula:

[0148] u′ (i) =(0 M×p ,I M ,0 M×q )u (l)

[0149] where 0 m×n is a all-zero matrix with m rows and n columns, and I M is an identity matrix with M rows and M columns.

[0150] In step 7, the data free from inter-symbol interference and inter-block interference is soft-mapped to obtain bit LLR information. Then the bit soft information is block-segmented, de-interleaved block by block, and channel decoded to obtain the LLR extrinsic information of the initial encoded bits and the estimated value b n of the transmitted bit sequence.

[0151] The specific method of step 7 is as follows:

[0152] The data free from inter-symbol interference and inter-block interference is soft-demodulated. When QPSK modulation is used, the bit LLR information can be expressed as:

[0153]

[0154]

[0155] In the formula, is the mean square error of the time-domain symbol after equalization, which can be estimated as:

[0156] s′ is the result of hard-decision of the equalized symbol to the constellation diagram, which can be expressed as

[0157]

[0158] The channel decoder used for the channel decoding is a soft-input soft-output channel decoder (Single Input Single Output, SISO). After decoding, the extrinsic information of the nth symbol of the encoded bits is obtained. Assume that the LLR soft information of the two bits of the nth symbol in the encoded bits given by the channel decoder SISO are respectively The LLR information is defined as:

[0159]

[0160]

[0161] In the formula, the symbols respectively represent the bits corresponding to the LLR information.

[0162] Step 8: Determine whether to perform the next iteration. If the iteration is completed, execute Step 8.1 to output the initial estimate b n of the transmitted bit sequence; if the next iteration is still required, execute Step 8.2. Send the extrinsic LLR information of the coded bits into the interleaver and then perform QPSK soft mapping to obtain the expected information of the transmitted modulation symbols, and then continue from Step 4 to Step 8 until the iteration is completed, and output the estimate b n of the transmitted bit sequence.

[0163] Step 8.1: Complete the iteration and output the estimate b n of the transmitted bit sequence obtained in Step 7;

[0164] Step 8.2: If the next iteration is still required, send the extrinsic LLR information of the coded bits into the interleaver and then perform QPSK soft mapping to obtain the expected information of the transmitted modulation symbols, and then continue from Step 4 to Step 8 until the iteration is completed, and output the initial estimate b n of the transmitted bit sequence.

[0165] In the QPSK soft mapping described above, the expected information n of the symbol s can be obtained from the in-phase component and the quadrature component .

[0166] The mean information of the in-phase component and the mean information of the quadrature component can be expressed as:

[0167]

[0168]

[0169] Obtain the expected information of the transmitted modulation symbols, that is, after detecting the data, then use the detected data to continue from Step 4 to Step 8 until the iteration is completed, and output the initial estimate b n of the transmitted bit sequence.

[0170] For example, Figure 6As shown, for the BER results obtained from building a single - carrier ultra - Nyquist transmission system based on the MATLAB platform, in terms of channel coding and decoding, a 1 / 2 convolutional code is used for channel coding, and the generator matrix is {'1 + x^2', '1 + x + x^2'}. In terms of data generation, a QPSK modulation method is adopted, and the constellation mapping table Average energy M of the modulated data s = 1. In the ultra - Nyquist system, the ISI taps generated by the shaping filter and the receiving filter are infinitely long. During simulation, the tap coefficients are truncated to v = 10. The channel is a 3 - tap frequency - selective channel. In terms of the iterative equalization algorithm at the receiving end, assuming that the ideal channel impulse response is known, the length of the FFT window of the data block is set to 512. In the method of the present invention, the overlapping lengths p = q = 20, and the length of the guard interval for adding GI is set to L = 10. The number of equalization iterations is 5.

[0171] The present invention also provides a system for a joint decoding - iterative frequency - domain equalization method of FTN signals without GI, which includes a pre - processing module, a frequency - domain equalization module, a deleter module, a soft demodulation module, a soft mapping module, and an output module.

[0172] The pre - processing module described above is used to pre - process the FTN received signal in step 3. The received signal stream is divided according to an FFT sliding window of length N to obtain parallel data blocks to be processed. Note that the first sliding window of the data stream needs to be selected starting from the first p symbols before the start position of the received data, and the start position of the subsequent sliding window is M symbols shifted backward from the left boundary of the previous sliding window.

[0173] The frequency - domain equalization module described above is used to calculate the forward filter parameters {C p} and the feedback filter parameters {B p} in step 4, perform parallel frequency - domain equalization to obtain the equalization result of the l - th iteration. When the number of iterations is 0, the feedback filter is turned off, and the frequency - domain equalization module degrades to a forward filter module, and the output of the forward filter is the equalization result. This module is also used to generate the time - domain equalization result by using IFFT processing in step 5.

[0174] The deleter module described above is used to delete the symbol segments with a length of p at the head and a length of q at the tail of the equalization result in step 6, and only retain the equalized output data u′ of the i - th FFT window that is not affected by interference (i) .

[0175] The soft demodulation module described above is used to calculate the bit LLR soft information according to the equalized output data in step 7.

[0176] The soft mapping module described above is used to perform QPSK soft mapping according to the interleaved coded - bit soft information in step 8 to obtain the data symbol expectation, that is, the detected data.

[0177] The output module described above is used to implement the determination of the current iteration count in step 8 to determine whether the iteration is completed. If completed, it outputs the estimated value b of the transmitted bit sequence. n If not completed, this module does not perform any operation.

[0178] The present invention also provides a joint decoding-iterative frequency domain equalization device for FTN signals without GI, including:

[0179] A memory for storing computer programs;

[0180] A processor for implementing the joint decoding-iterative frequency domain equalization method for FTN signals without GI when executing the computer program.

[0181] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it can implement a joint decoding-iterative frequency domain equalization method for FTN signals without GI.

[0182] In summary, although the embodiments of the present invention are described in combination with specific structural parameters and drawings, they are not used to limit the protection scope of the present invention. For those skilled in the art, without departing from the principles and ideas of the present invention, several adjustments and improvements can be made to the structural parameters in combination with specific application scenarios, and these should also be regarded as belonging to the protection scope of the present invention.

Claims

1. A joint decoding - iterative frequency - domain equalization method for GI - free FTN signals, Characterized in that: It includes the following steps: Step 1: Segment the original bit sequence to be transmitted, send it into the encoder block by block, then perform in - block interleaving, and then concatenate all the encoded and interleaved bit sequences together, perform constellation mapping and FTN shaping processing to obtain the FTN transmission signal; Step 2: After the FTN transmission signal obtained in Step 1 passes through the channel, obtain the FTN received signal. The FTN received signal passes through a matched filter and a down - sampler to obtain the FTN received signal at the symbol rate; Step 3: Divide the FTN received signal at the symbol rate obtained in Step 2 according to an FFT sliding window of length N. The first sliding window of the data stream is selected starting from the first p symbols before the start position of the received data, and the start position of the next sliding window is M symbols shifted to the right of the left boundary of the previous sliding window; Step 4: Determine the current iteration number; If it is the first iteration, the iteration number l = 0, execute Step 4.1, perform parallel frequency - domain equalization on the received signal within the FFT window based on the minimum mean - square error (MMSE) criterion to obtain the first - iteration equalization result; Otherwise, execute Step 4.2 to generate the forward filter coefficients {C p} and the feedback filter coefficients {B p} according to the (l-1)-th detection data and the filter design criterion, and perform parallel frequency-domain equalization to obtain the equalization result of the l-th iteration; Step 5: For the equalization result U obtained in Step 4 (l) Perform IFFT operation to generate the time-domain equalization result; Step 6: Feed the time-domain equalization result u (l) into a deleter to obtain data that is not subject to inter-symbol interference and inter-block interference; Step 7: Soft-demodulate the data that is not subject to inter-symbol interference and inter-block interference to obtain bit LLR information. Then, perform code block segmentation on the bit LLR information, perform deinterleaving and channel decoding by block to obtain the extrinsic LLR information of the coded bits and the estimated value b of the transmitted bit sequence n ; Step 8: Determine whether to perform the next iteration; If the iteration is completed, execute Step 8.1 to output the initial estimate b of the transmitted bit sequence n ; if another iteration is required, execute Step 8.2, send the extrinsic information of the coded bits to the interleaver and then perform QPSK soft mapping to obtain the expected information of the transmitted modulation symbols, and then continue from Step 4 to Step 8 until the iteration is completed and output the estimate b of the transmitted bit sequence n .

2. A joint decoding - iterative frequency - domain equalization method for GI - free FTN signals according to claim 1, Characterized in that: The specific method of Step 1 is: The FTN shaping processing includes an up - sampler and a shaping filter; The sampling factor of the up - sampler is less than the up - sampling multiple of the shaping filter. The compression factor of the FTN signal is adjusted by changing the sampling factor of the up - sampler and the up - sampling multiple of the shaping filter. The specific relationship is that the compression factor is equal to the sampling factor of the up - sampler divided by the up - sampling multiple of the shaping filter. If the sampling factor of the up - sampler and the up - sampling multiple of the shaping filter are equal, the compression factor is equal to 1, realizing Nyquist transmission.

3. A joint decoding - iterative frequency - domain equalization method for GI - free FTN signals according to claim 1, Characterized in that: The specific method of Step 4.1 is: In the initial state, set the iteration number l = 0, perform parallel frequency - domain equalization on the received signal within the FFT window based on the MMSE criterion to obtain the first - iteration equalization result; Step 4.1.1, design the forward filter coefficients based on the MMSE criterion Reduce the matrix QE[nn H Q H to the diagonal matrix Φ n , considering the reduced correlation of the colored noise, the forward filter coefficients for the first iteration are as follows: Among them, H p is the channel frequency-domain response, and Φ n [p] is a diagonal matrix Φ n The p-th element on the diagonal can be expressed as: where \(g = [g(-vT),\ldots,g(0),\ldots,g(vT)]\in\mathbb{R}\) 2v+1 is the channel impulse response corresponding to the combined action of the transmit filter, the transmission channel, and the receive filter, and the channel length is \(N\) h \(= 2v + 1\); Step 4.1.2, multiply the received frequency-domain signal R p by the elements of the filter coefficient {C p} to obtain the forward filter output vector Z (l) , and the specific expression is: Among them, are the forward filter coefficients; R p is the received frequency-domain signal; Z (l) is the forward filter output vector; Step 4.1.3, equalization result U (l) is equal to the output vector Z of the forward filter (l) , since this is the first iteration and no detection data is available, the feedback filter is turned off, and the equalization result is directly obtained from the output of the forward filter, expressed as: U (l) = Z (l) ; The specific method of Step 4.2 is: For the iteration number \(l > 0\), generate the forward filter parameters \(\{C\) p \} and the feedback filter parameters \(\{B\) p \} according to the \((l - 1)\)-th detection data and the filter design criterion, and perform parallel frequency-domain equalization; Step 4.2.1, design the forward filter and the feedback filter based on the (l-1)-th detection data and the MMSE design criterion. Assume that the statistical characteristics of the time-domain detection signal and the transmitted signal are the same, and obtain the forward filter coefficient C k and the feedback filter coefficient B k as follows: In the coefficient expression, κ is a normalization parameter used to control the constant amplitude of the equalized output symbol, that is In the coefficient expression, is represented by the credibility ρ, and the credibility ρ can be estimated by a simple method, which is expressed as follows: Among them, ρ n is the credibility of the nth detection symbol, which can be obtained from the in-phase component credibility ρ n,I and the quadrature component credibility ρ n,Q The calculation method is as follows: Wherein, the value ranges of the in-phase credibility and the quadrature credibility are both If the equalization is at the 0th iteration, initialize Step 4.2.2, multiply the received frequency-domain signal R p by the elements of the forward filter coefficient {C p} to obtain the forward filter output vector Z (l) , which is expressed as follows: Among them, is the forward filter coefficient; R p is the received frequency-domain signal; Z (l) is the forward filter output vector; Step 4.2.3, convert the detection data of the (l-1) iteration to the frequency domain to obtain Then, multiply the frequency domain detection signal element by element with the feedback filter coefficients {B p} to obtain the feedback filter output Y (l) , and the expression is as follows: Step 4.2.4, accumulate the output Z of the forward filter (l) and the output vector Y of the feedback filter (l) to obtain the equalized output U (l) , U (l) = Z (l) + Y (l) .

4. A joint decoding - iterative frequency - domain equalization method for GI - free FTN signals according to claim 1, Characterized in that: The specific method of Step 5 is: Let the time-domain equalization result be u (l) , then it can be expressed as: where Q is the FFT matrix, and the element in its k-th row and j-th column is H indicating the Hermitian operation.

5. A joint decoding - iterative frequency - domain equalization method for GI - free FTN signals according to claim 1, Characterized in that: The specific method of Step 6 is: Use a deleter to delete the symbol segments with a head length of p and a tail length of q from the equalization result, and only retain the equalization output data u′ of the i-th FFT window that is not disturbed (i) ; The deletion process is expressed by the formula: u ′(i) = (0 M×p , I M , 0 M×q )u (l) Among them, 0 m×n is a zero matrix with m rows and n columns, and I M is an identity matrix with M rows and M columns.

6. A joint decoding - iterative frequency - domain equalization method for GI - free FTN signals according to claim 1, Characterized in that: The specific method of Step 7 is: Soft - demodulate the data that is not affected by inter - symbol interference and inter - block interference. When using QPSK modulation, the bit LLR information can be expressed as: wherein, is the mean square error of the equalized time-domain symbol and can be estimated as: s′ is the equalization symbol The result of hard decision to the constellation diagram can be expressed as: The channel decoding module used for the channel decoding is a soft input soft output channel decoder SISO. After decoding, the extrinsic information of the nth symbol of the coded bits is obtained. Assume that the LLR soft information of the two bits of the nth symbol in the coded bits given by the channel decoder SISO are respectively The LLR information is defined as: In the formula, the symbol represents the LLR information respectively corresponding bits.

7. A joint decoding - iterative frequency - domain equalization method for GI - free FTN signals according to claim 1, Characterized in that: The specific method of step 8 is as follows: Step 8.1, complete the iteration and output the estimated value b of the transmitted bit sequence obtained in Step 7 n ; Step 8.2, another iteration is required. The extrinsic information of the coded bits in the form of LLR is fed into the interleaver and then undergoes QPSK soft mapping to obtain the desired information of the transmitted modulation symbols. Then, continue with Steps 4 to 8 until the iteration is completed, and output the initial estimate b of the transmitted bit sequence n ; In the QPSK soft mapping described above, the symbol s n expected information can be obtained from the in-phase component mean information and the quadrature component mean information as follows The in-phase component mean information and the quadrature component mean information can be expressed as: Obtain the expected information of the transmitted modulation symbols, that is, after detecting the data, continue from step 4 to step 8 using the detected data until the iteration is completed, and output the initial estimate b of the transmitted bit sequence n .

8. A system for implementing the joint decoding-iterative frequency-domain equalization method of GI-free FTN signals as claimed in claim 1, characterized in that: It includes: A preprocessing module: used to preprocess the FTN received signal. The received signal stream is divided according to an FFT sliding window of length N to obtain parallel data blocks to be processed. The first sliding window of the data stream is selected starting from the first p symbols before the start position of the received data, and the start position of the subsequent sliding window is M symbols behind the left boundary of the previous sliding window; Frequency domain equalization module: used to calculate the forward filter parameters {C p} and the feedback filter parameters {B p}, perform parallel frequency domain equalization to obtain the equalization result of the l-th iteration. When the number of iterations is 0, the feedback filter is turned off, and the frequency domain equalization module degrades to a forward filter module, and the output of the forward filter is the equalization result; this module is also used to generate the time domain equalization result by using IFFT processing; Deletion module: used to implement the deletion of the symbol segment with a head length of p and a tail length of q in the equalization result, and only retain the equalization output data u′ of the i-th FFT window that is not disturbed (i) ; A soft demodulation module: used to calculate the bit LLR soft information according to the equalization output data; A soft mapping module: used to perform QPSK soft mapping according to the interleaved coded bit soft information to obtain the data symbol expectation, that is, the detected data; Output module: used to determine whether the current iteration count indicates completion of the iteration. If completed, it outputs the estimated value b of the transmitted bit sequence. n If not completed, this module does not perform any operations.

9. A GI-free FTN signal joint decoding-iterative frequency-domain equalization device, characterized in that: It includes: A memory, used to store computer programs; A processor, used to implement a GI-free FTN signal joint decoding-iterative frequency-domain equalization method as claimed in any one of claims 1-8 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it can implement a GI-free FTN signal joint decoding-iterative frequency-domain equalization method.

Images