Anti-delay frequency offset estimation method for NTN scenarios

Abstract

This application relates to the field of wireless communication technology, and in particular to a delay-resistant frequency offset estimation method for NTN scenarios. The method includes: locating and grouping multiple DMRS time-domain segments based on the non-zero sample positions in the local reference waveform and calculating the sampling interval; performing frequency domain transformation on each segment to obtain received and local frequency-domain sequences; multiplying the received frequency-domain sequence and the local frequency-domain sequence conjugately to obtain a frequency-domain matched response; selecting at least two sets of matched responses, performing frequency-point-by-frequency conjugate multiplication and accumulating along the frequency domain to extract the common phase difference; calculating the folded frequency offset based on the common phase difference and the sampling interval, and constructing a candidate frequency offset set based on the maximum residual frequency offset search range; performing a consistency decision on multiple candidate sets to obtain the final residual frequency offset estimation result. This method can effectively resist timing deviations and noise interference in NTN scenarios with large delays, low signal-to-noise ratios, and large residual frequency offsets, improving the accuracy and stability of frequency offset estimation.

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, and in particular to a delay-resistant frequency offset estimation method for NTN scenarios. Background Technology

[0002] With the development of 5G-Advanced and 6G communication technologies, non-terrestrial networks (NTNs), as an important supplement and extension of terrestrial cellular networks, can achieve seamless global coverage and meet the needs of communication in remote areas, aviation and maritime communications, and emergency communications. In NTN uplink communication scenarios, the uplink signal sent by the terminal needs to travel a long distance to reach the satellite or ground gateway. Due to the combined effects of factors such as the high-speed movement of the satellite platform, the movement of the terminal itself, the high carrier frequency of the satellite-to-ground link, and local oscillator errors, the receiver will observe significant Doppler frequency shift and residual frequency offset. Especially in the LEO satellite scenario, the satellite's relative speed to the ground is high, and the Doppler frequency shift changes rapidly; at higher carrier frequencies such as the Ka band, the same relative speed will result in a larger absolute frequency offset. Therefore, even if the system completes initial synchronization during the random access phase through methods such as PRACH, timing advance, and pre-compensation, a certain residual frequency offset may still exist in the uplink shared channel PUSCH. If the residual frequency offset cannot be accurately estimated and compensated, it will disrupt the orthogonality between OFDM subcarriers, causing common phase rotation and inter-subcarrier interference, which in turn will affect PUSCH data demodulation, DMRS channel estimation and subsequent equalization performance.

[0003] In existing communication receivers, the cyclic prefix (CP) is often used for coarse frequency offset estimation. The CP frequency offset estimation method utilizes the repetition relationship between the cyclic prefix and the symbol tail in OFDM symbols. By performing conjugate correlation on the two repetitive sample segments, the phase difference is extracted and converted to obtain the frequency offset estimate. This method is simple to implement and has low computational cost, making it suitable for coarse frequency offset estimation at the receiver front end. However, the CP frequency offset estimation method is quite sensitive to timing position and propagation delay. Its validity depends on a good cyclic repetition relationship between the CP samples selected at the receiver and the corresponding OFDM symbol tail samples. When the receiving timing deviation is small and the multipath delay spread is within the CP protection range, the CP correlation results are usually relatively stable; however, in NTN scenarios, there may be significant propagation delay and timing errors in the link. Due to factors such as large propagation delay, timing advance compensation errors, satellite link delay variations, and low signal-to-noise ratio, the CP correlation window at the receiver may experience a large shift. When the CP correlation window deviates from the ideal position, or the proportion of effective repeating samples decreases significantly, the point-to-point correspondence between the CP and the symbol tail is disrupted, the correlation amplitude decreases, and the correlation phase is contaminated by interference from adjacent symbols, noise, and erroneous correlation terms, ultimately leading to an increase in frequency offset estimation error. This is especially true when the time delay deviation is large, as the performance of CP frequency offset estimation deteriorates rapidly. In engineering practice, when the received timing deviation reaches a certain proportion of the CP length, or even approaches or exceeds half the CP length, the number of samples in the CP correlation that can maintain an effective cyclic repetition relationship decreases significantly, the proportion of erroneous correlation terms increases, and the correlation phase no longer only reflects the phase rotation caused by the true frequency offset, but also includes phase components caused by timing offset, multipath leakage, inter-symbol interference, and noise disturbances.

[0004] Besides CP frequency offset estimation, existing systems can also utilize DMRS (demodulation reference signal) for frequency offset estimation. The receiver can estimate the residual frequency offset based on the phase relationship between the locally generated DMRS reference sequence and the received DMRS. Compared to CP, DMRS has a known sequence structure and a closer mapping relationship with actual data resources, thus theoretically it can be used for more refined frequency offset estimation. However, traditional DMRS frequency offset estimation methods still have shortcomings in NTN scenarios. NTN uplink coverage distances are long, and received signal power may be low, making low signal-to-noise ratio (SNR) scenarios common. Under low SNR conditions, if two DMRS segments are directly multiplied conjugate in the time domain, noise samples will directly participate in the phase difference calculation, leading to significant jitter in the phase estimation. Since frequency offset estimation essentially depends on the phase difference, when the phase difference is disturbed by noise, the frequency offset estimate will exhibit random shifts. If the residual frequency offset is close to the boundary of the unambiguous estimation range, noise may also cause the phase to cross positive and negative boundaries, resulting in a frequency offset estimation reversal phenomenon.

[0005] Therefore, it is necessary to propose an enhanced frequency offset estimation method for NTN scenarios, so that it can avoid the strong dependence of traditional CP frequency offset estimation on timing position under the conditions of large propagation delay, low signal-to-noise ratio and large residual frequency offset, and overcome the problem of large phase jitter of traditional DMRS phase difference estimation under low signal-to-noise ratio. Summary of the Invention

[0006] Therefore, it is necessary to provide a delay-resistant frequency offset estimation method for NTN scenarios that can avoid the strong dependence of traditional CP frequency offset estimation on timing position and overcome the problem of large phase jitter in traditional DMRS phase difference estimation under low signal-to-noise ratio, and large residual frequency offset.

[0007] A delay-enhanced frequency offset estimation method for NTN scenarios, the method comprising: A local time-domain reference waveform is generated based on the PUSCH DMRS configuration. The length of the received uplink waveform is normalized to obtain a normalized received waveform. Based on the non-zero sample locations in the local time-domain reference waveform, multiple independent DMRS time-domain segments are located and grouped, and the sampling interval between each group of DMRS time-domain segments is calculated. Frequency domain transformation is performed on the normalized received waveform segment and the local reference waveform segment corresponding to each DMRS time-domain segment to obtain the received DMRS frequency domain sequence and the local reference DMRS frequency domain sequence. Multiply each received DMRS frequency domain sequence by the conjugate of the corresponding local reference DMRS frequency domain sequence on a frequency-by-frequency basis to obtain the frequency domain matching response corresponding to each DMRS time domain segment; Select at least two different DMRS time domain segments and their corresponding frequency domain matching responses, perform frequency-point conjugate multiplication, and then accumulate all valid DMRS frequency domain points along the frequency domain dimension to obtain the frequency domain correlation quantity. Extract the phase of the frequency domain correlation quantity as the common phase difference between the two DMRS time domain segments. Based on the common phase difference and corresponding sampling interval between the time domain segments of each group of DMRS, the folded frequency offset corresponding to each group of DMRS is calculated, and a candidate frequency offset set corresponding to each group of DMRS is constructed based on the maximum residual frequency offset search range preset by the system. Consistency judgment is performed on multiple candidate frequency offset sets to select the frequency offset candidate combination with the smallest sum of differences. The final residual frequency offset estimation result is then calculated based on the frequency offset candidate combination.

[0008] The aforementioned delay-enhanced frequency offset estimation method for NTN scenarios replaces traditional CP correlation with inter-symbol common phase difference estimation based on DMRS frequency domain matched response. It leverages the characteristic that a fixed delay exhibits linear phase in the frequency domain and approximately cancels out in the conjugate multiplication of adjacent DMRS matched responses, fundamentally eliminating the strong dependence on the CP correlation window. This solves the problem of traditional CP frequency offset estimation where large receiving timing deviations or insufficient propagation delay compensation disrupt the cyclic repetition relationship between CP and symbol tails, leading to interference terms in the correlated phase and a sharp increase in estimation error. Furthermore, the process involves first converting the received DMRS and the local reference DMRS to the frequency domain before performing reference conjugate matching, removing the DMRS's own modulation phase so that the common phase caused by residual frequency offset is reflected in multiple frequency domains. By maintaining consistency between sampling points and then performing frequency domain conjugate multiplication and accumulation, effective phase coherence superposition and incoherent random noise superposition are achieved, significantly reducing phase jitter and frequency offset estimation variance under low signal-to-noise ratio. This solves the problem of severe estimation jitter and easy frequency offset reversal caused by the direct participation of noise samples in phase calculation in the traditional time-domain DMRS conjugate multiplication method. At the same time, by reusing the existing DMRS resources in PUSCH without adding an additional air interface reference signal, and by using local FFT, frequency domain conjugate multiplication, frequency domain accumulation, and comparison of a finite candidate set with adaptive size based on the maximum residual frequency offset range to replace the full-frequency domain blind search, the complexity of engineering implementation is controllable. Finally, high-precision and high-stability residual frequency offset estimation is achieved in the complex scenario of large propagation delay, low signal-to-noise ratio and large residual frequency offset in NTN. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating a delay-enhanced frequency offset estimation method for NTN scenarios in one embodiment; Figure 2 The residual frequency offset (RMSE) simulation diagram of different frequency offset estimation algorithms under random fractional time delay conditions in one embodiment is shown. Figure 3 The image shows the residual frequency offset (RMSE) simulation results of different frequency offset estimation algorithms under simulation conditions where the lead time is an integer delay of 1 / 2 CP length, as shown in one embodiment. Figure 4 The simulation diagram shows the residual frequency offset (RMSE) of different frequency offset estimation algorithms under simulation conditions where the lead time is an integer delay of 1 CP length, and the delay length is an integer delay of 1 CP length. Figure 5 This is a schematic diagram of the frequency offset CDF estimated by the single CP frequency offset estimation method under simulation conditions where the advance integer delay is one CP length, as shown in one embodiment. Figure 6This is a schematic diagram of the frequency offset CDF estimated by the multi-CP joint denoising frequency offset estimation method under simulation conditions where the advance integer delay is one CP length, as shown in one embodiment. Figure 7 This is a schematic diagram of the frequency offset CDF estimated by the anti-delay enhanced frequency offset estimation method under simulation conditions where the advance integer delay is 1 CP length, as shown in one embodiment. Figure 8 The image shows the residual frequency offset (RMSE) simulation results of different frequency offset estimation algorithms under simulation conditions, where the lead time is an integer delay and the delay length is 1 CP length + random decimal delay. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0011] In one embodiment, such as Figure 1 As shown, a delay-resistant frequency offset estimation method for NTN scenarios is provided, including the following steps: Step 1: Generate a local time-domain reference waveform according to the PUSCH DMRS configuration, and perform length normalization processing on the received uplink waveform to obtain a normalized received waveform.

[0012] The purpose of this step is to establish an accurate sample correspondence between the received waveform and the local reference waveform, enabling the subsequent extraction of a received segment from the received waveform that perfectly corresponds to the local DMRS segment. The specific process for generating the local time-domain reference waveform is as follows: In the frequency-domain resource grid, local DMRS symbols are filled in according to the DMRS port, DMRS symbol position, frequency-domain resource mapping, and reference sequence generation rules; non-DMRS positions are set to zero, resulting in the local DMRS resource grid; OFDM modulation is then applied to the local DMRS resource grid to generate the local time-domain reference waveform.

[0013] Since the received waveform may come from different truncation windows, its length may be shorter than the local reference waveform. To avoid out-of-bounds errors when trunculating DMRS segments according to the reference index later, this method performs length normalization processing on the received waveform. In one embodiment, the process of length normalizing the received uplink waveform includes: When the length of the received uplink waveform is less than the length of the local time domain reference waveform, zeros are added to the end of the received uplink waveform to the length of the local time domain reference waveform. When the length of the received uplink waveform is greater than or equal to the length of the local time-domain reference waveform, the first part of the received uplink waveform is truncated. Each sample point is used as a normalized received waveform, where The length of the local time-domain reference waveform.

[0014] Specifically, let the received waveform be... Length is Local reference waveform Length is The normalized received waveform is denoted as: ; ; when At that time, you can directly command: ; The above operations ensure that all received samples corresponding to the local reference fragment index set are available, and zero padding does not introduce any additional valid signals; it is only used to ensure index consistency and computational stability.

[0015] Step 2: Based on the non-zero sample positions in the local time-domain reference waveform, locate and group multiple independent DMRS time-domain segments, and calculate the sampling interval between each group of DMRS time-domain segments; perform frequency domain transformation on the normalized received waveform segment and the local reference waveform segment corresponding to each DMRS time-domain segment to obtain the received DMRS frequency domain sequence and the local reference DMRS frequency domain sequence.

[0016] In one embodiment, multiple independent DMRS time-domain segments are located and grouped, and the sampling interval between each group of DMRS time-domain segments is calculated, including: Extract the sample indexes with amplitude greater than zero from the local time-domain reference waveform to form a set of valid sample indexes; The effective sample point index set is decomposed into several mutually separate continuous index subsets, each of which corresponds to an independent DMRS time domain segment. Calculate the sampling interval between any two DMRS time-domain segments based on the starting index of each consecutive index subset.

[0017] Specifically, since the local reference resource grid only fills in the reference signal at the DMRS location, after OFDM modulation, the non-zero or valid reference segments in the reference waveform can be used to locate the corresponding DMRS time-domain segments in the received waveform. Let the set of valid sample indexes in the reference waveform be: ; If a slot contains multiple DMRS symbols, then It can be decomposed into several mutually separate, consecutive subsets of indices: ; For the three groups of DMRS segments, they can be defined separately. , , . No. The local reference fragment and the received fragment are as follows: ; ;

[0018] If the first The starting index of the group fragment is The sampling interval between the three DMRS sets is: ; ; ; The key function of this step is to decompose the reference structure in the entire slot into independently processable DMRS segments and provide accurate time intervals between DMRS symbol segments for subsequent phase difference to frequency offset conversion.

[0019] The frequency domain transformation is implemented using Fast Fourier Transform (FFT). The FFT length can be equal to the length of the corresponding DMRS time-domain segment, or it can be zero-padded or truncated depending on the receiver design. This frequency domain transformation is not equivalent to OFDM demodulation of the entire slot, but rather a local frequency domain analysis of the located DMRS segment, used to enhance the stability of subsequent reference matching and phase estimation. Traditional time-domain DMRS conjugate multiplication is performed directly on noisy time-domain samples, and noise cross terms directly affect the relevant phase. This application first transforms the received DMRS segment and the local reference DMRS segment to the frequency domain, so that the local reference sequence can be used to match each frequency domain point, thereby avoiding time-domain noise directly interfering with phase calculation. Specifically, The process of converting the received DMRS segment and the local reference DMRS segment to the frequency domain includes: ; ; in, For the first DMRS segment length, To receive DMRS frequency domain sequences, For local reference DMRS frequency domain sequence, .

[0020] Step 3: Multiply each received DMRS frequency domain sequence by the conjugate of the corresponding local reference DMRS frequency domain sequence to obtain the frequency domain matching response corresponding to each DMRS time domain segment.

[0021] In one embodiment, each received DMRS frequency domain sequence is multiplied point-by-point by the conjugate of the corresponding local reference DMRS frequency domain sequence to obtain the frequency domain matching response corresponding to each DMRS time domain segment, including: Multiply each received DMRS frequency domain sequence by its conjugate with the corresponding local reference DMRS frequency domain sequence point by point to obtain the frequency domain matching response for each DMRS time domain segment. ; in, This is a linear phase term introduced in the frequency domain for time delay or timing deviation. For the first The DMRS time-domain segment in the first k Channel frequency response at each frequency domain point For the first The local reference DMRS frequency domain sequence corresponding to each DMRS time domain segment The number of FFT points in the system. For the first The frequency offset common phase corresponding to each DMRS time domain segment For the first The DMRS time-domain segment in the first k Additive white Gaussian noise at each frequency point For the first The complex conjugate of the local reference DMRS frequency domain sequence corresponding to each DMRS time domain segment This refers to the system's fixed timing deviation.

[0022] Specifically, CP frequency offset estimation relies on the repetition relationship between the CP segment and the symbol tail. Once a timing deviation causes the correlation window to shift, erroneous correlation terms will be mixed into the correlation phase. This application estimates the inter-symbol phase difference based on the DMRS frequency domain matched response, where the fixed delay mainly manifests as a linear phase in the frequency domain.

[0023] When there is a fixed timing deviation or a relatively stable path delay At that time, the first The DMRS frequency domain received signal can be written as: ; After frequency domain reference conjugate matching: ; in, This is the linear phase term introduced in the frequency domain by time delay or timing deviation. If the... The and the first If the time delays between DMRS symbols are essentially stable, then they contain approximately the same linear phase. When performing inter-symbol conjugate multiplication, we have: ; Therefore, the time-delayed linear phase is The phase is approximately canceled out, and the remaining phase is mainly the common phase difference of the frequency offset. ; This step further illustrates that, when the inter-symbol delay response is relatively stable, this application successfully cancels the linear phase change caused by the delay by multiplying the conjugate of the frequency domain matching responses between DMRS, which greatly reduces the impact of fixed delay on the frequency offset estimation phase.

[0024] Simultaneously, this step implements a delay resistance enhancement mechanism. This is effective when there is a fixed timing deviation or a relatively stable path delay. In the frequency domain, time delay is mainly manifested as a linear phase term. If the first The and the first If the time delays between DMRS symbols are essentially stable, then both symbols contain approximately the same linear phase. When performing inter-symbol conjugate multiplication, the linear phase terms will cancel each other out. The remaining phase is mainly the common phase difference of the frequency offset. Therefore, this method does not directly rely on the CP correlation window and can maintain excellent frequency offset estimation stability even with large system time delays.

[0025] Step 4: Select at least two different DMRS time domain segments and perform frequency-point conjugate multiplication of their corresponding frequency domain matching responses. Then, accumulate all valid DMRS frequency domain points along the frequency domain dimension to obtain the frequency domain correlation quantity. Extract the phase of the frequency domain correlation quantity as the common phase difference between the two DMRS time domain segments.

[0026] In one embodiment, extracting the phase of the frequency domain correlation quantity as the common phase difference between the two sets of DMRS time domain segments includes: The phase of the frequency domain correlation quantity is extracted as the common phase difference between the two sets of DMRS time domain segments: ; ; in, For the first The and the first Common phase difference between DMRS time-domain segments For frequency domain correlation, This indicates the phase taking a complex value. For the first The frequency offset common phase corresponding to each DMRS time domain segment For the first The frequency offset common phase corresponding to each DMRS time domain segment For the first The DMRS time-domain segment in the first k Channel frequency response at each frequency domain point.

[0027] Specifically, after completing the frequency domain reference conjugate matching, it is not directly applied to a single... Instead of determining the phase, the matched responses of the two DMRS symbols are multiplied conjugately at each frequency point, and then accumulated along the frequency domain. ; in, Indicates the first The and the first The DMRS symbol in the first k The matching response conjugate product at each frequency domain point. Summing up all valid frequency domain points involved in the estimation yields the frequency domain correlation: ; in, This represents the set of frequency domain points involved in frequency offset estimation. This set comprises all valid DMRS frequency domain points. The common phase difference is given by the phase of this correlation quantity: ; To illustrate the physical meaning of this step, we substitute the matched response model. If the DMRS amplitude is normalized and the channel change between two DMRS symbols is slow, then: ; Then we have: ; ; in This is the noise term.

[0028] After accumulating the frequency domain points: ; Since the frequency offset phase is consistent in direction (i.e., consistent in sign) across all subcarriers, while the noise phase is random (random in sign), coherent accumulation can significantly suppress noise and reduce the noise term. The effect on phase results in high processing gain. Furthermore, due to the summation term... The weights are mainly non-negative real numbers, therefore the phase of the correlation quantity is approximately equal to the common phase difference caused by the frequency offset. ; This process utilizes multiple frequency domain points to participate in phase estimation, enabling the useful common phases to be coherently superimposed, while the noise components are mainly incoherently superimposed. Therefore, it has better low signal-to-noise ratio stability compared to single-point phase or direct conjugate multiplication in the time domain.

[0029] Step 5: Calculate the folded frequency offset of each group of DMRS based on the common phase difference and corresponding sampling interval between the time domain segments of each group of DMRS, and construct the candidate frequency offset set of each group of DMRS based on the maximum residual frequency offset search range preset by the system.

[0030] In one embodiment, the folded frequency offset corresponding to each group of DMRS is calculated based on the common phase difference between the time-domain segments of each group of DMRS and the corresponding sampling interval, including: ; ; ; in, For the first The and the first Initial frequency offset estimates between DMRS time-domain segments For the system subcarrier spacing, The number of FFT points in the system. For the first The and the first Sampling interval between DMRS time domain segments For the first The and the first Folded frequency offset corresponding to each DMRS time-domain segment For the first The and the first Unambiguous frequency offset boundaries corresponding to each DMRS time-domain segment For modulo operation, For the first The and the first Common phase difference between DMRS time-domain segments For the first The and the first The time interval between DMRS time domain segments.

[0031] In one embodiment, the process of constructing the candidate frequency offset set includes: Based on the system's preset maximum residual frequency offset search range Determine the integer The range of values ​​for is: ; in, ; Then construct the first The and the first Candidate frequency offset set corresponding to each DMRS time domain segment for: ; in, For the first The and the first Folded frequency offset corresponding to each DMRS time-domain segment For the first The and the first Unambiguous frequency offset boundaries corresponding to each DMRS time-domain segment.

[0032] Specifically, according to the frequency offset phase model, the common phase difference between two DMRS symbols and the residual frequency offset satisfy the following: ; Therefore, the first , The frequency offset between the DMRS is estimated as follows: ; Will Substituting, we get: ; Reuse This yields the calculation formula for directly adapting OFDM parameters: ; For the three sets of DMRS fragments, we can obtain: ; ; ; Ideally, the three sets of estimates should be close to the same true residual frequency offset: ; This consistency relationship forms the basis for subsequent multi-DMRS candidate decisions. If one set of estimates is shifted due to noise or phase folding, other interval estimates can be used for constraint and correction.

[0033] Since the phase function outputs the principal value, the phase difference... It can only be located within a limited range: ; Therefore, single-group DMRS phase difference estimation has an unambiguous frequency offset range. It can be seen that, in order to avoid the phase exceeding the principal value range, the following must be satisfied: ; ; Substitution ,get: ; Substitute again , define the first , The fuzzy boundary corresponding to the group DMRS interval is: ; When the true frequency offset is close to this boundary, the correlated phase is close to Low signal-to-noise ratio noise disturbances may cause phase overflow and folding. In this case, directly calculated... It may not be the true frequency offset, but rather the folded value of the true frequency offset within the unambiguous range.

[0034] To address the phase folding problem in single-interval estimation, this step maps the single set of frequency offset estimates to an unambiguous range and constructs a candidate frequency offset set based on the ambiguity period. The specific process is as follows: First, define the folded frequency offset: ; The mapping result satisfies: ; The true frequency offset may equal the folded frequency offset plus an integer number of ambiguity periods, therefore the 1st... , The candidate frequency offset set corresponding to the group DMRS is defined as follows: ; To control the complexity of the engineering implementation, the search range can be determined based on the maximum residual frequency offset. Limited integer The range of values ​​for ; ; Therefore, the finite candidate set can be written as: ;

[0035] This approach avoids a large-scale search across the entire frequency domain while ensuring that the candidate set covers the maximum residual frequency offset range expected by the system.

[0036] Step 6: Perform consistency judgment on multiple candidate frequency offset sets, select the frequency offset candidate combination with the smallest sum of differences, and calculate the final residual frequency offset estimation result based on the frequency offset candidate combination.

[0037] Different DMRS intervals correspond to different fuzzy boundaries and fuzzy periods. If the true frequency offset is folded by a certain interval, its candidate set still contains candidate values ​​near the true frequency offset. By comparing the candidate sets obtained from different DMRS intervals, the most consistent frequency offset candidate can be selected from multiple sets.

[0038] In one embodiment, when two sets of DMRS time-domain segments are selected corresponding to the first candidate frequency offset set and the second candidate frequency offset set, the specific method for consistency determination is to traverse all candidate value combinations in the first candidate frequency offset set and the second candidate frequency offset set, calculate the absolute difference between the two candidate values ​​in each combination, select the candidate value combination with the smallest absolute difference, and use the average of the two candidate values ​​in the combination as the final residual frequency offset estimation result.

[0039] Specifically, for adjacent DMRS pairs and The goal of consistency judgment is: ; The final frequency offset can be taken as the average of the two: .

[0040] In one embodiment, when three sets of DMRS time-domain segments are selected corresponding to the first candidate frequency offset set, the second candidate frequency offset set, and the third candidate frequency offset set, the consistency decision process is to traverse all candidate value combinations in the three candidate frequency offset sets, calculate the sum of the absolute differences between any two candidate values ​​in each combination, select the candidate value combination with the smallest sum of absolute differences, and use the average of the three candidate values ​​in the combination as the final residual frequency offset estimation result.

[0041] Specifically, if used simultaneously Intervals can be used to construct a three-set consistency decision: ; ;

[0042] This multi-pilot candidate consensus decision can greatly reduce the impact of low correlation strength DMRS on the final frequency offset result of the column pair, making the estimation results more suitable for NTN scenarios with low signal-to-noise ratio (SNR) and strong channel frequency selectivity.

[0043] To verify the effectiveness of this application, an NTN PUSCH link-level simulation was performed. The system parameters are shown in Table 1. Table 1

[0044] Simulations compared the performance of the proposed method with single-CP frequency offset estimation methods and multi-CP joint denoising frequency offset estimation methods. The simulation results are as follows: Figure 2 , Figure 3 , Figure 4 , Figure 5 , Figure 6 , Figure 7 , Figure 8As shown in the figure, RMSE represents the root mean square error, SNR represents the signal-to-noise ratio, and F(X) represents the cumulative distribution function value. Simulation results show that the single-CP frequency offset estimation method is simple to implement, but its estimation accuracy deteriorates significantly at low signal-to-noise ratios (SNR < 0 dB) or with integer time delays, such as the length of one CP. RMSE increases dramatically, the cumulative distribution function (CDF) curve becomes dispersed and exhibits a long tail, the estimation bias deteriorates non-linearly with decreasing signal-to-noise ratio, and the frequency offset estimation range is wide with systematic bias, making it difficult to meet the needs of practical multipath or noisy environments. The multi-CP joint denoising frequency offset estimation method effectively reduces estimation jitter caused by noise through cross-symbol joint processing and denoising mechanisms. It exhibits better error suppression capabilities than the single-CP method in the low signal-to-noise ratio range (SNR < 0 dB), and CDF convergence is significantly improved. However, its performance is still affected by phase distortion introduced by integer time delays, and the estimation accuracy decreases significantly when symbol timing offsets are present.

[0045] In comparison, the delay-resistant frequency offset estimation method proposed in this application exhibits superior performance under all simulation conditions, including random fractional delays, integer lead delays, and complex delay environments combining both. This method fundamentally eliminates the systematic interference of timing offset on frequency offset estimation through a delay-resistant enhancement mechanism. Even in harsh scenarios where integer lead delays (one CP length) are superimposed with random fractional delays, its RMSE remains at the lowest level, and its CDF curve is the steepest, with the estimation variance approaching the theoretical lower limit of error. Especially under conditions of low signal-to-noise ratio (-10dB) and large integer delays, this method maintains a highly consistent estimation distribution, without exhibiting the residual bias of multi-CP joint denoising methods or the severe tailing phenomenon of single-CP methods. In summary, the delay-resistant frequency offset estimation method proposed in this application combines noise robustness and delay insensitivity, making it particularly suitable for multipath fading and delay uncertainty environments in practical NTN transmission system deployments.

[0046] In the aforementioned anti-delay enhanced frequency offset estimation method for NTN scenarios, to address the issue of traditional CP frequency offset estimation being sensitive to timing deviations, this application first uses DMRS time-domain segment localization and frequency-domain transformation to convert the received signal and local reference to the frequency domain for conjugate matching, so that the fixed time delay only manifests as a linear phase in the frequency domain. Then, it performs frequency-point conjugate multiplication between the frequency-domain matching responses of different DMRS symbols and accumulates them along the frequency domain. By utilizing the mechanism of mutual cancellation of linear phase delays between adjacent symbols, the final extracted common phase difference is freed from the interference of timing offset, thereby achieving anti-delay frequency offset estimation without relying on the CP correlation window, significantly improving the estimation stability under conditions of large propagation delay and timing compensation error. Furthermore, this application first performs frequency domain transformation on the received and local DMRS segments separately, and removes or weakens the modulation phase of the reference sequence through reference conjugate matching, so that the common phase caused by residual frequency offset remains consistent at multiple frequency domain points. Then, it performs frequency-point conjugate multiplication on the frequency domain matching responses of at least two sets of DMRS and accumulates them along the frequency domain dimension, so that the effective common phase is coherently superimposed while random noise is incoherently averaged. This significantly reduces phase jitter and frequency offset estimation variance under low signal-to-noise ratio conditions, and obtains more accurate and stable estimation results, solving the problem of direct phase contamination by noise in traditional time-domain DMRS conjugate multiplication under low signal-to-noise ratio conditions. In addition, this application reuses the existing DMRS resources in PUSCH without adding an additional air interface reference signal. The entire processing only involves local FFT, frequency domain multiplication, frequency domain accumulation, and comparison of a finite candidate set based on adaptive scaling of the maximum residual frequency offset range, avoiding the increased complexity caused by full-frequency domain blind search, and achieving controllable engineering implementation complexity while ensuring performance.

[0047] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0048] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0049] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these modifications and improvements all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A delay-enhanced frequency offset estimation method for NTN scenarios, characterized in that, The method includes: A local time-domain reference waveform is generated based on the PUSCH DMRS configuration. The length of the received uplink waveform is normalized to obtain a normalized received waveform. Based on the non-zero sample locations in the local time-domain reference waveform, multiple independent DMRS time-domain segments are located and grouped, and the sampling interval between each group of DMRS time-domain segments is calculated. Frequency domain transformation is performed on the normalized received waveform segment and the local reference waveform segment corresponding to each DMRS time-domain segment to obtain the received DMRS frequency domain sequence and the local reference DMRS frequency domain sequence. Multiply each received DMRS frequency domain sequence by the conjugate of the corresponding local reference DMRS frequency domain sequence on a frequency-by-frequency basis to obtain the frequency domain matching response corresponding to each DMRS time domain segment; Select at least two different DMRS time domain segments and perform frequency-point conjugate multiplication of their corresponding frequency domain matching responses. Then, accumulate all valid DMRS frequency domain points along the frequency domain dimension to obtain the frequency domain correlation quantity. Extract the phase of the frequency domain correlation quantity as the common phase difference between the two DMRS time domain segments. Based on the common phase difference and corresponding sampling interval between the time domain segments of each group of DMRS, the folded frequency offset corresponding to each group of DMRS is calculated, and a candidate frequency offset set corresponding to each group of DMRS is constructed based on the maximum residual frequency offset search range preset by the system. Consistency judgment is performed on multiple candidate frequency offset sets to select the frequency offset candidate combination with the smallest sum of differences. The final residual frequency offset estimation result is then calculated based on the frequency offset candidate combination.

2. The method according to claim 1, characterized in that, Each received DMRS frequency domain sequence is multiplied point-by-point by the conjugate of its corresponding local reference DMRS frequency domain sequence to obtain the frequency domain matched response for each DMRS time domain segment, including: Multiplying each received DMRS frequency domain sequence by its conjugate with the corresponding local reference DMRS frequency domain sequence point by point yields the frequency domain matching response for each DMRS time domain segment: in, This is a linear phase term introduced in the frequency domain for time delay or timing deviation. For the first The DMRS time-domain segment in the first k Channel frequency response at each frequency domain point For the first The local reference DMRS frequency domain sequence corresponding to each DMRS time domain segment The number of FFT points in the system. For the first The frequency offset common phase corresponding to each DMRS time domain segment For the first The DMRS time-domain segment in the first k Additive white Gaussian noise at each frequency point For the first The complex conjugate of the local reference DMRS frequency domain sequence corresponding to each DMRS time domain segment This refers to the system's fixed timing deviation.

3. The method according to claim 1, characterized in that, Extracting the phase of the frequency domain correlation quantity as the common phase difference between the two sets of DMRS time domain segments includes: The phase of the frequency domain correlation quantity is extracted as the common phase difference between the two sets of DMRS time domain segments: in, For the first The and the first Common phase difference between DMRS time-domain segments For frequency domain correlation, This indicates the phase taking a complex value. For the first The frequency offset common phase corresponding to each DMRS time domain segment For the first The frequency offset common phase corresponding to each DMRS time domain segment For the first The DMRS time-domain segment in the first k Channel frequency response at each frequency domain point.

4. The method according to claim 1, characterized in that, Based on the common phase difference and corresponding sampling interval between the time-domain segments of each DMRS group, the folded frequency offset corresponding to each DMRS group is calculated, including: in, For the first The and the first Initial frequency offset estimates between DMRS time-domain segments For the system subcarrier spacing, The number of FFT points in the system. For the first The and the first Sampling interval between DMRS time domain segments For the first The and the first Folded frequency offset corresponding to each DMRS time-domain segment For the first The and the first Unambiguous frequency offset boundaries corresponding to each DMRS time-domain segment For modulo operation, For the first The and the first Common phase difference between DMRS time-domain segments For the first The and the first The time interval between DMRS time domain segments.

5. The method according to claim 4, characterized in that, The construction process of the candidate frequency offset set includes: Based on the system's preset maximum residual frequency offset search range Determine the integer The range of values ​​for is: in, ; Then construct the first The and the first Candidate frequency offset set corresponding to each DMRS time domain segment for: in, For the first The and the first Folded frequency offset corresponding to each DMRS time-domain segment For the first The and the first Unambiguous frequency offset boundaries corresponding to each DMRS time-domain segment.

6. The method according to claim 1, characterized in that, When two sets of DMRS time-domain segments are selected corresponding to the first candidate frequency offset set and the second candidate frequency offset set, the specific method of the consistency decision is to traverse all candidate value combinations in the first candidate frequency offset set and the second candidate frequency offset set, calculate the absolute difference between the two candidate values ​​in each combination, select the candidate value combination with the smallest absolute difference, and take the average of the two candidate values ​​in the combination as the final residual frequency offset estimation result.

7. The method according to claim 1, characterized in that, When three sets of DMRS time-domain segments are selected corresponding to the first candidate frequency offset set, the second candidate frequency offset set, and the third candidate frequency offset set, the consistency decision process is to traverse all candidate value combinations in the three candidate frequency offset sets, calculate the sum of the absolute differences between any two candidate values ​​in each combination, select the candidate value combination with the smallest sum of absolute differences, and take the average of the three candidate values ​​in the combination as the final residual frequency offset estimation result.

8. The method according to claim 1, characterized in that, The process of generating the local time-domain reference waveform includes: In the frequency domain resource grid, local DMRS symbols are filled in according to the DMRS port, DMRS symbol position, frequency domain resource mapping and reference sequence generation rules, and non-DMRS positions are set to zero to obtain the local DMRS resource grid; OFDM modulation is performed on the local DMRS resource grid to generate the local time domain reference waveform.

9. The method according to claim 1, characterized in that, The process of length normalizing the received uplink waveform includes: When the length of the received uplink waveform is less than the length of the local time domain reference waveform, zeros are added to the end of the received uplink waveform to the length of the local time domain reference waveform. When the length of the received uplink waveform is greater than or equal to the length of the local time-domain reference waveform, the first part of the received uplink waveform is truncated. Each sample point is used as a normalized received waveform, where The length of the local time-domain reference waveform.

10. The method according to claim 1, characterized in that, Multiple independent DMRS time-domain segments are located and grouped, and the sampling interval between each group of DMRS time-domain segments is calculated, including: Extract the sample indexes with amplitude greater than zero from the local time-domain reference waveform to form a set of valid sample indexes; The effective sample point index set is decomposed into several mutually separate continuous index subsets, each of which corresponds to an independent DMRS time domain segment. Calculate the sampling interval between any two DMRS time-domain segments based on the starting index of each consecutive index subset.

Images