Sequence pilot frequency and channel estimation method for window OTFS in fractional Doppler channel

Abstract

The invention relates to the technical field of wireless communication, and provides a sequence pilot frequency and channel estimation method for a window OTFS in a fractional Doppler channel. The sequence pilot frequency is placed in the center of a DD domain along a Doppler axis, and the sequence pilot frequency is obtained by limiting a sequence pilot frequency generation condition through a particle swarm optimization algorithm, so that a certain autocorrelation of the sequence is ensured while the peak-to-average power ratio influenced by the sequence pilot frequency is reduced. In the channel estimation process, firstly, an approximate solution is obtained based on self-correlation estimation, and then the error between the estimated pilot response and the received pilot response is minimized from the global perspective, so that the search step length is reduced. The sequence pilot frequency autocorrelation influenced by the PAPR in the sequence pilot frequency design process is compensated by the two-stage channel estimation method.

Description

Technical Field

[0001] The present application relates to the technical field of wireless communication, and in particular to a sequence pilot and channel estimation method for window OTFS in a fractional Doppler channel. Background Art

[0002] Orthogonal Time-Frequency-Space (OTFS) modulation is a potential solution to maintain stable reliability and efficiency in high mobility environments. OTFS is a new method that maps information symbols to the Delay-Doppler (DD) domain instead of the traditional time-frequency domain. The channel representation in the Delay-Doppler (DD) domain is more sparse, which is conducive to channel estimation. However, due to the limitation of frame length or number of symbols, OTFS systems often encounter fractional Doppler frequency shifts in practical applications, making the channel response more complex and difficult to accurately estimate. Window processing in the TF domain improves the location of channel multipath in this multi-search DD domain. In related technologies, most of the Orthogonal Time-Frequency-Space (OTFS) modulations use pulse pilots, which will result in a relatively high peak-to-average power ratio (PAPR), especially for high-power pilots.

[0003] The following is a study on channel estimation of Orthogonal Time-Frequency-Space (OTFS) modulation in related technologies.

[0004] (1), R.Hadani, S.Rakib, M.Tsatsanis, A.Monk, AJGoldsmith, AFMolisch, andR.Calderbank, "Orthogonal time frequency space modulation," in 2017 IEEEWireless Communications and Networking Conference (WCNC), 2017, pp.1–6.

[0005] (2), N.Hashimoto, N.Osawa, K.Yamazaki, and S.Ibi, "Channel estimation and equalization for CP-OFDM-based OTFS in fractional Doppler channels," in 2021 IEEE International Conference on Communications Workshops (ICC Workshops), 2021, pp.1–7.

[0006] (3)、Q. Wang, Y. Liang, Z. Zhang, and P. Fan, “2D off-grid decomposition and SBL combination for OTFS channel estimation,” IEEE Transactions on Wireless Communications, pp. 1–1, 2022.

[0007] (4)、Z. Wei, W. Yuan, S. Li, J. Yuan, and D. W. K. Ng, “Transmitter and receiver window designs for orthogonal time-frequency space modulation,” IEEE Transactions on Communications, vol. 69, no. 4, pp. 2207–2223, 2021.

[0008] (5)、P. Raviteja, K. T. Phan, and Y. Hong, “Embedded pilot-aided channel estimation for OTFS in delay-Doppler channels,” IEEE Transactions on Vehicular Technology, vol. 68, no. 5, pp. 4906–4917, 2019.

[0009] (6)、X. He, P. Fan, and Q. Wang, “A two-stage channel estimation algorithm for OTFS in fractional Doppler channels,” IEEE Communications Letters, pp. 1–1, 2023.

[0010] (7)、I. A. Khan and S. K. Mohammed, “A low-complexity OTFS channel estimation method for fractional delay-Doppler scenarios,” IEEE Wireless Communications Letters, vol. 12, no. 9, pp. 1484–1488, 2023. Summary of the Invention

[0011] In order to solve the technical problems existing in the prior art, the present application provides a sequence pilot and channel estimation method for window OTFS in a fractional Doppler channel.

[0012] In a first aspect, the present application provides a sequence pilot for window OTFS in a fractional Doppler channel using the following technical solution:

[0013] A sequence pilot for windowed OTFS in fractional Doppler channels.

[0014] OTFS sends frames, including data symbols and sequence pilots;

[0015] The sequence pilot is designed based on the particle swarm optimization algorithm. The elements in the sequence pilot are expressed in the form of amplitude and phase as follows: The sequence pilot is a complex number, n i is the amplitude of the i-th pilot element, θ i is the corresponding phase, and the sequence pilot is expressed as:

[0016]

[0017] The design goal of the sequence pilot is expressed as:

[0018]

[0019] S p represents the form of the sequence pilot in the time domain, X p represents the pilot matrix, matrix X p Only include the length L seq sequence pilot, minimize max(S p ) represents minimizing the maximum value of the time domain signal corresponding to the sequence pilot;

[0020] β represents the maximum theoretical amplitude when the matrix energy is concentrated on one element. 2 is a preset value, indicating the total pilot power; ζ indicates the sidelobe level allowed by the sequence pilot response; r indicates the autocorrelation of the sequence pilot, and r(k) indicates the kth element in r. ||X p || 2 Represents the matrix X p The result of performing the two-norm operation.

[0021] In a preferred embodiment, the sequence pilot is placed at the center of the DD domain along the Doppler axis.

[0022] By adopting the above technical solution, the sequence pilot provided by the present application can reduce the peak-to-average power ratio affected by the sequence pilot based on the particle swarm algorithm and the autocorrelation design of the sequence pilot.

[0023] In the second aspect, the present application provides a channel estimation method for window OTFS in a fractional Doppler channel using the following technical solution:

[0024] A channel estimation method for windowed OTFS in fractional Doppler channels,

[0025] An OTFS transmission frame comprising a data symbol and a sequence pilot as claimed in claim 1;

[0026] The method comprises:

[0027] Based on calculating the point with the highest autocorrelation value in the fractional Doppler channel, the delay, channel gain and Doppler are fed back according to the autocorrelation result;

[0028] The Doppler and channel gains are re-determined by the quasi-Newton method.

[0029] In a preferred embodiment, based on calculating the point with the highest autocorrelation value in the fractional Doppler channel, and feeding back the delay, channel gain and Doppler according to the autocorrelation result, the method comprises:

[0030] Design Matrix A seq To perform the correlation operation between the sequence and the received pilot response, the matrix A seq The i-th line is:

[0031]

[0032] where k r =[k 0 -k max :k 0 +L seq -1+k max ] is the response area of ​​the pilot sequence, k max Represents the index of the maximum Doppler shift. The i-th row contains N elements. seq [k] represents the data of the corresponding sequence pilot in N elements, x seq [k 0 :(k 0 +L seq -1)] is a specific pilot sequence, k 0 is the Doppler index of the first element of the sequence; k i is a possible Doppler shift, k i =k υec (i), whose value range is k vec =[-k max :(1 / D):k max ], 1 / D is the search step length; k is the index number of the Doppler at the receiving end;

[0033] In the fractional Doppler channel, each delay is processed separately, that is, ι∈{0:1:ι max}, ι max is the maximum delay of the channel;

[0034] b=A seq *Y p [:,ι p +ι] H

[0035] H represents the conjugate transpose operation, taking the point with the highest autocorrelation value as the path of the channel, and the result of the autocorrelation as the channel gain of the path:

[0036]

[0037] Among them, the delay-Doppler grid resources are divided into M*N grid points. For Doppler, For delay, is the channel gain, val and index represent the value of the maximum element in vector b and the corresponding index number respectively, j is an imaginary number, and lp is the delayed index number where the sequence pilot is located.

[0038] In a preferred embodiment, when there are multiple paths with the same delay index, the information of the estimated path is saved and the received pilot response Y at the ith delay is used. p [:,ι] iteratively subtracts the estimated information.

[0039] In a preferred embodiment, the stopping condition of the iteration includes that the estimated channel gain is less than a threshold value 3σ or the current estimated channel path number exceeds the maximum allowed estimated channel path number, where σ is the standard deviation of the noise.

[0040] In a preferred embodiment, re-determining Doppler and channel gain by the quasi-Newton method includes minimizing the received pilot response and the pilot response constructed from the estimated information:

[0041]

[0042] in represents the estimated pilot response for delay index ι.

[0043] In a preferred embodiment, the estimated pilot response is calculated based on the estimated channel information.

[0044] By adopting the above technical solution:

[0045] 1. First, an approximate solution is obtained based on autocorrelation estimation, and the error between the estimated pilot response and the received pilot response is minimized from a global perspective, thereby reducing the error caused by the search step size and IDI.

[0046] 2. The sequence pilot proposed in this application is designed based on PAPR and autocorrelation, that is, PAPR is taken into consideration. And combined with the characteristics of the fractional Doppler channel, the designed sequence pilot is placed along the Doppler axis to estimate the fractional Doppler, which is also conducive to the receiving end to process the pilot response data in parallel according to the delay. In addition, unlike the correlation with the basis function, the length of the sequence pilot is adjustable, which provides freedom for the design of PAPR and autocorrelation. The improvement of the sequence autocorrelation performance is conducive to improving the estimation accuracy of the entire estimation algorithm.

[0047] 3. The autocorrelation of the sequence pilot will affect the accuracy of the channel estimation process. The better the autocorrelation, the more accurate the channel estimation result, and the autocorrelation will affect the final PAPR. In the process of designing the sequence pilot, in order to reduce the PAPR, the present application scheme does not take the autocorrelation as the only influencing factor, that is, in the process of designing the sequence pilot, the sidelobe level is limited, and the reduction of PAPR and good autocorrelation are considered at the same time. The sequence pilot autocorrelation affected by PAPR in the process of sequence pilot design is compensated by the two-stage channel estimation method. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the arrangement of pilot and data symbols in the DD domain in an embodiment of the present application. DETAILED DESCRIPTION

[0049] To make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the embodiments of the present invention will be clearly and completely described below in conjunction with the appendix of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them.

[0050] The present application is further described in detail below in conjunction with the accompanying drawings.

[0051] Reference Figure 1 ,In the entire OTFS system, the delay Doppler grid resources are divided into M*N grid points, and the position of each symbol in the delay Doppler grid can be determined by [k,l].

[0052] When rectangular pulse shaping is used in the time domain, the delay index of the pilot is greater than the index l of the maximum channel delay. max , the input-output relationship of OTFS in the DD domain is given by the following formula (see references (1) and (2)):

[0053]

[0054] Where z[k,l] represents Gaussian white noise. k and l represent the Doppler index and delay index of the received signal respectively. x[k',l'] is the signal in the delay Doppler domain. k' and l' represent the Doppler index and delay index of the transmitted signal respectively. The number of paths in the channel is P. and are the delay and Doppler of the i-th path respectively. i is the channel gain of the ith path. M is the number of carriers, and N is the number of time slots.

[0055] and are the basis functions in the Doppler domain and the delay domain, respectively. is a phase shift produced by rectangular pulse shaping and has the following form:

[0056]

[0057] If a rectangular window is used in the TF domain, the window coefficient α[0] = α[1], .., = α[N-1] = 1. Compared with a rectangular window, the application of a non-rectangular window coefficient α[n] can reduce the sidelobe level of the response.

[0058] In order to process the pilot response received by each delay box separately, the present application places the sequence pilot in the DD domain along the Doppler axis. The number of sequence elements is generally a complex number, and its mathematical expression can be written as: where η i is the amplitude of the i-th pilot element, θ i is the corresponding phase. seq is the sequence length. Specifically, the sequence pilot sequence is placed in the center of the DD domain. The delay index where the pilot is located is l p =round[(M-1) / 2], the Doppler index of the first element of the sequence pilot sequence is k L =round[(NL seq -1) / 2], the Doppler index of the last element is k R =round[(NL seq -1) / 2]+L seq -1. The protection intervals in the Doppler axis direction and the delay axis direction are 4k respectively. max and 2l max , where k max Indicates the index of the maximum Doppler shift.

[0059] The inventors have found that placing sequence pilots along the Doppler axis and placing sequence pilots along the time domain axis have differences in how the receiver processes pilot responses in a fractional Doppler channel. Placing the sequence along the Doppler axis is beneficial for the receiver to process pilot responses in a fractional Doppler channel, but it will result in a higher peak-to-average power ratio. The details are as follows.

[0060] The relationship from DD domain to time domain is:

[0061]

[0062] The matrix X tf represents the TF domain signal. FN and FM are standard Fourier matrices. H represents the conjugate transpose operation. G tx =diag[g tx (0), g tx (T / M), ..., g tx ((M-1)T / M)] is pulse shaping, where g tx (t) is the pulse shaping filter, T is the time of one time slot, FN is the standard Fourier matrix of N*N, and FM is the standard Fourier matrix of M*M.

[0063] Matrix X d and X p Structure and Matrix X DD Similarly, the DD domain transmits symbol X DD Mainly includes data symbol X d and pilot symbol X p , whose energy distribution determines the PAPR, is just the data matrix X d Only contains Figure 1 The data symbol in the p Only contains Figure 1 The matrix Xp contains only the pilot symbols of length L. seq The sequence pilot, L seq is a complex number. This application adopts rectangular pulse shaping, so the matrix G tx is the unit matrix, that is, G tx =I.

[0064] The PAPR is given by

[0065]

[0066] Where the time domain signal s = vec(S)∈C MN×1 , C represents the complex field, and s(n) represents the nth element in the vector s. d (n) represents vector s d The nth element in s p (n) represents vector s pThe time domain signal of the data is The time domain signal of the sequence pilot is According to Parseval's theorem, || || 2 is a 2-norm operation, for example

[0067] When the pilot power and data power are constant, the signal The average value of |s(n)| is a constant. In this case, PAPR depends on the maximum value of |s(n)|, which is composed of the time domain components of the pilot symbol and the data symbol. The data symbol is random during data transmission, but the pilot design is usually fixed. Therefore, this application analyzes the impact of the pilot time domain component of the high-power pilot on PAPR.

[0068] Matrix X p It only contains the pilot sequence. When the total pilot power remains unchanged, for different pilot designs, However, different pilot designs will affect |s p (n)|, resulting in different PAPRs. p The smaller the maximum value of (n)| is, the easier it is to obtain a lower PAPR.

[0069] In addition, the TF domain window will also affect the time domain signal s of the pilot p (n), thus affecting PAPR. According to the literature [1], when a window is added in the time axis direction of the TF domain, the relationship from the DD domain to the time domain is:

[0070]

[0071] where the diagonal matrix U is given by:

[0072] U=diag{a[0],a[1],…,a[N-1]}

[0073] In order to facilitate the description of the present application, s is set in the present application. d and Remain unchanged.

[0074] When the pilot power is very high, the PAPR mainly depends on the pilot symbol. For a more intuitive understanding, this application describes it by example. For example, the pulse pilot X i , the sequence pilot X placed along the delay axis delay and the sequence pilot X along the Doppler axis doppler Among them, the pilot element and x p is a complex number. The total power β2 of different pilots is equal (i.e. ). For the sake of convenience, the sequence pilot in this example only contains two pilot elements x p1 and x p2 , that is I seq =2, j is the imaginary unit.

[0075]

[0076] After each pilot signal is converted to the time domain, as shown in formula (2) and formula (3):

[0077]

[0078] where ω N =e -j2π / N .

[0079] Non-rectangular windows have coefficients greater than 1, namely:

[0080] PAPR(α[n]with rect)=1 <PAPR(α[n]with non-rect)

[0081] Where "rect" and "non-rect" represent rectangular and non-rectangular windows respectively. This will increase the size of the window. p (n)||. Therefore, the rectangular window can have a lower PAPR, that is,

[0082] PAPR(X i with rect) <PAPR(Xiwith non-rect)

[0083] Formula (3) shows that the pilot signal X doppler The energy of is distributed in a row of the matrix in the time domain. However, the pilot signal X delay The energy of may be distributed in multiple rows of the matrix. In formula (3), It can be seen as two vectors and Add in different directions. So the following relationship exists:

[0084]

[0085] That is to say, the matrix The maximum value of the element is less than the matrix This means that X delay The PAPR is less than X doppler PAPR. For a rectangular window U = I, the above energy distribution characteristics mean that the matrix The largest element of can be much smaller than the matrix The largest element of the matrix The largest element of is

[0086] According to Parseval's theorem, the matrix The energy of can be concentrated on one element at most, and its maximum amplitude can theoretically reach β. However, its actual maximum value still depends on the specific pilot sequence, and its maximum value can be reduced by pilot design. Therefore, X doppler and X i The peak-to-average power ratio relationship between them depends on the specific pilot sequence. max and k max Using normal guard interval, more data symbols will affect the pulse pilot |s d (n)+s p The maximum value of (n)| leads to a higher PAPR. Therefore, the following relationship can be obtained:

[0087] PAPR(X delay ) <PAPR(X i ) or PAPR(X doppler )

[0088] Through the above analysis, it is found that placing the sequence pilot sequence along the Doppler axis is beneficial for the receiver to process the pilot response in the fractional Doppler channel, but its PAPR needs to be further reduced.

[0089] In order to reduce the PAPR when placing the sequence pilot sequence along the Doppler axis, the present application designs the sequence pilot by using the particle swarm optimization algorithm. The elements in the sequence pilot are expressed in the form of amplitude and phase as follows: The sequence pilot is a complex number, n i is the amplitude of the i-th pilot element, θ i is the corresponding phase, and the sequence pilot is expressed as:

[0090]

[0091] The design goal of the sequence pilot is expressed as:

[0092]

[0093]

[0094] S p represents the form of the sequence pilot in the time domain, X p represents the pilot matrix, matrix X p Only include the length L seq sequence pilot, minimize max(S p ) represents minimizing the maximum value of the time domain signal corresponding to the sequence pilot;

[0095] β represents the maximum theoretical amplitude when the matrix energy is concentrated on one element, β2 is a preset value, which represents the total pilot power; ζ represents the sidelobe level allowed by the sequence pilot response; r represents the autocorrelation of the sequence pilot, and r(k) represents the kth element in r. ||X p || 2 Represents the matrix X p The result of performing the two-norm operation.

[0096] It should be noted that the autocorrelation of the sequence pilot will affect the accuracy of the channel estimation process. The better the autocorrelation, the more accurate the channel estimation result, and the autocorrelation will affect the final PAPR. In the process of designing the sequence pilot, in order to reduce the PAPR, the present application scheme does not take the autocorrelation as the only influencing factor, that is, in the process of designing the sequence pilot, the sidelobe level is limited, and the reduction of PAPR and good autocorrelation are considered at the same time.

[0097] After obtaining the sequence pilot, the present application uses a two-stage channel estimation algorithm to finely estimate the channel information. The sequence pilot autocorrelation affected by PAPR in the sequence pilot design process is compensated by the two-stage channel estimation method of the present application. The specific contents of the channel estimation method are as follows:

[0098]

[0099] Combining the channel estimation method can overcome the estimation performance loss caused by a certain decrease in autocorrelation.

[0100] Specifically, the two-stage channel estimation method is explained below.

[0101] The channel estimation process mainly includes:

[0102] Step 1:

[0103] Based on calculating the point with the highest autocorrelation value in the fractional Doppler channel, the delay, channel gain and Doppler are fed back based on the autocorrelation result. Step 2:

[0104] The Doppler and channel gains are re-determined by the quasi-Newton method.

[0105] In step 1, design matrix A seq To perform the correlation operation between the sequence and the received pilot response, the matrix A seq The i-th behavior of is:

[0106]

[0107] where k r =[k 0 -k max:k 0 +L seq -1+k max ] is the response area of ​​the pilot sequence, k max Represents the index of the maximum Doppler shift. The i-th row contains N elements. seq [k] represents the data of the corresponding sequence pilot in N elements, x seq =[k 0 :(k 0 +L seq -1)] is a specific pilot sequence, k 0 is the Doppler index of the first element of the sequence; k i is a possible Doppler shift, k i =k vec (i), whose value range is k vec =[-k max :(1 / D):k max ], 1 / D is the search step length; k is the index number of the Doppler at the receiving end.

[0108] In the fractional Doppler channel, each delay is processed separately, that is, ι∈{0:1;ι max}, ι max The maximum delay of the channel is:

[0109]

[0110] Among them, the delay-Doppler grid resources are divided into M*N grid points. For Doppler, For delay, is the channel gain, val and index represent the value of the maximum element in vector b and the corresponding index number respectively, j is an imaginary number, and lp is the delayed index number where the sequence pilot is located.

[0111] In a preferred embodiment, when there are multiple paths with the same delay index, the information of the estimated path is saved and the received pilot response Y at the ith delay is used. p [:,ι] iteratively subtracts the estimated information.

[0112] The iterative method is a method that has been disclosed in the relevant technology. For details, please refer to the literature (Z. Wei, W. Yuan, S. Li, J. Yuan, and DWK Ng, "Transmitter and receiver window designs for orthogonal time-frequency space modulation," IEEE Transactions on Communications, vol. 69, no. 4, pp. 2207–2223, 2021). This application does not provide a specific description.

[0113] In the present application, the iteration process has two stopping conditions. One condition is that the estimated channel gain is less than the threshold 3σ, where σ is the standard deviation of the noise. The other condition is that the current estimated channel path number exceeds the maximum allowed channel path number, i.e., p num > max Both of these iteration stop conditions are designed to avoid estimating too many paths with small amplitudes or paths generated by response side lobes, and to reduce unnecessary parameters.

[0114] In step 2, the received pilot response and the pilot response constructed from the estimated information are first minimized:

[0115]

[0116] where Y p [:,ι] and denote the received pilot response and the estimated pilot response for the delay index ι, respectively, Specifically, it is calculated by the OTFS input-output formula without considering noise, and the estimated pilot response is calculated based on the estimated channel information

[0117] The multipath estimated in step 1 is close to the true value, and the quasi-Newton method can be used to iteratively update the estimation result, similar to the reference (Z. Wei, W. Yuan, S. Li, J. Yuan, and DWK Ng, "Transmitter and receiver window designs for orthogonal time-frequency space modulation," IEEE Transactions on Communications, vol. 69, no. 4, pp. 2207–2223, 2021), which is not described in detail in this application.

[0118] By adopting the above technical solution, the sequence pilot proposed in this application is designed based on PAPR and autocorrelation, that is, PAPR is taken into consideration. And combined with the characteristics of the fractional Doppler channel, the designed sequence pilot is placed along the delay axis to estimate the fractional Doppler, which is also beneficial for the receiving end to process the pilot response data in parallel according to the delay. In addition, unlike the correlation with the basis function, the length of the sequence pilot is adjustable, which provides freedom for the design of PAPR and autocorrelation. Improving the performance of sequence autocorrelation is conducive to improving the estimation accuracy of the entire estimation algorithm.

[0119] The following are the references used in the description of this application:

[0120] (1), R.Hadani, S.Rakib, M.Tsatsanis, A.Monk, AJGoldsmith, AFMolisch, andR.Calderbank, "Orthogonal time frequency space modulation," in 2017 IEEEWireless Communications and Networking Conference (WCNC), 2017, pp.1–6.

[0121] (2), N.Hashimoto, N.Osawa, K.Yamazaki, and S.Ibi, "Channel estimation and equalization for CP-OFDM-based OTFS in fractional Doppler channels," in 2021 IEEE International Conference on Communications Workshops (ICC Workshops), 2021, pp.1–7.

[0122] The embodiments of this specific implementation method are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Therefore, all equivalent changes made based on the principles of the present application should be included in the protection scope of the present application.

Claims

1. A sequence pilot for windowed OTFS in a fractional Doppler channel, characterized in that: OTFS sends frames, including data symbols and sequence pilots; The sequence pilot is designed based on the particle swarm optimization algorithm. The elements in the sequence pilot are expressed in the form of amplitude and phase as follows: The sequence pilot is a complex number, n i is the amplitude of the i-th pilot element, θ i is the corresponding phase, and the sequence pilot is expressed as: The design goal of the sequence pilot is expressed as: S p represents the form of the sequence pilot in the time domain, X p represents the pilot matrix, matrix X p Only include the length L seq sequence pilot, minimize max(S p ) represents minimizing the maximum value of the time domain signal corresponding to the sequence pilot; β represents the maximum theoretical amplitude when the matrix energy is concentrated on one element. 2 is a preset value, indicating the total pilot power; ζ indicates the sidelobe level allowed by the sequence pilot response; r indicates the autocorrelation of the sequence pilot, and r(k) indicates the kth element in r. ||x p ||2 means for matrix X p The result of performing the two-norm operation.

2. The sequence pilot for window OTFS in a fractional Doppler channel according to claim 1, characterized in that: The sequence pilot is placed at the center of the DD domain along the Doppler axis. By adopting the above technical solution, the sequence pilot provided by the present application can reduce the peak-to-average power ratio affected by the sequence pilot based on the particle swarm algorithm and the autocorrelation design of the sequence pilot.

3. A channel estimation method for windowed OTFS in a fractional Doppler channel, characterized in that: An OTFS transmission frame comprising a data symbol and a sequence pilot as claimed in claim 1; The method comprises: Based on calculating the point with the highest autocorrelation value in the fractional Doppler channel, the delay, channel gain and Doppler are fed back according to the autocorrelation result; The Doppler and channel gains are re-determined by the quasi-Newton method.

4. A channel estimation method for window OTFS in a fractional Doppler channel according to claim 3, characterized in that: Based on calculating the point with the highest autocorrelation value in the fractional Doppler channel, feedback of delay, channel gain and Doppler based on the autocorrelation result includes: Design Matrix A seq To perform the correlation operation between the sequence and the received pilot response, the matrix A seq The i-th line is: where k r =[k0-k max :k0+L seq -1+k max ] is the response area of ​​the pilot sequence, k max Represents the index of the maximum Doppler shift. The i-th row contains N elements. seq [k] represents the data of the corresponding sequence pilot in N elements, x seq [k0:(k0+L seq -1)] is a specific pilot sequence, k0 is the Doppler index number of the first element of the sequence; k i is a possible Doppler shift, k i =k vec (i) Its value range is k vec =[-k max :(1 / D):k max ], 1 / D is the search step length; k is the index number of the Doppler at the receiving end; In the fractional Doppler channel, each delay is processed separately, i.e., l∈{0:1:l max }, l max is the maximum delay of the channel; b=A seq *Y p [:,l p +i] H H represents the conjugate transpose operation, taking the point with the highest autocorrelation value as the path of the channel, and the result of the autocorrelation as the channel gain of the path: Among them, the delay-Doppler grid resources are divided into M*N grid points. For Doppler, For delay, is the channel gain, val and index represent the value of the maximum element in vector b and the corresponding index number respectively, j is an imaginary number, and lp is the delayed index number where the sequence pilot is located.

5. A channel estimation method for window OTFS in a fractional Doppler channel according to claim 4, characterized in that: When there are multiple paths with the same delay index, the information of the estimated path is saved and the received pilot response Y on the ith delay is used. p [:,ι] iteratively subtracts the estimated information.

6. A channel estimation method for window OTFS in a fractional Doppler channel according to claim 5, characterized in that: The stopping condition of the iteration includes that the estimated channel gain is less than the threshold 3σ or the maximum estimated channel gain is P max path, where σ is the standard deviation of the noise.

7. A channel estimation method for window OTFS in a fractional Doppler channel according to claim 3, characterized in that: Re-determining the Doppler and channel gain by the quasi-Newton method involves minimizing the received pilot response and the pilot response constructed from the estimated information: where Y p [:,l] and represent the received pilot response and the estimated pilot response for the delay index ι respectively.

8. A channel estimation method for window OTFS in a fractional Doppler channel according to claim 6, characterized in that: Calculate the estimated pilot response based on the estimated channel information

Images