Frequency offset estimation method and device

Abstract

The present invention discloses a frequency offset estimation method and device, which are applied to the field of communications. The receiving end extracts a power delay sub-spectrum in a noise window from a first power delay spectrum of a reference signal, determines the noise power based on a first power value in the power delay sub-spectrum, and determines a weight coefficient of the reference signal based on the noise power. Since the weight coefficient is negatively correlated with the noise power, and the noise power is negatively correlated with the signal-to-noise ratio of the reference signal, the weight coefficient is positively correlated with the signal-to-noise ratio of the reference signal, thereby adaptively increasing the weight of the reference signal with a higher signal-to-noise ratio. Then, based on the weight coefficients of multiple reference signals, the first power delay spectra of multiple reference signals are weighted to obtain a second power delay spectrum, which can ensure that the determined second power delay spectrum conforms to the changes in the channel and has better real-time performance. This can ensure that the target time domain channel response value based on the first delay index in the second power delay spectrum has a higher accuracy in the determined frequency offset.

Description

Technical Field

[0001] The present application relates to the field of communications, and in particular to a frequency offset estimation method and device. Background Art

[0002] The Doppler effect between the transmitter and receiver in an orthogonal frequency division multiplexing (OFDM) system will cause a carrier frequency offset (CFO) (hereinafter referred to as frequency offset) in the signals of the two ends. The frequency offset will cause signal amplitude attenuation, phase distortion, and destroy the orthogonality between subcarriers, thereby causing inter-carrier interference (ICI), which reduces the signal-to-noise ratio of the OFDM system and poses a huge challenge to signal demodulation at the receiving end.

[0003] Therefore, frequency offset estimation is required in OFDM systems, and accurately estimating and compensating for frequency offset is an important prerequisite for leveraging the advantages of OFDM technology. Summary of the invention

[0004] The present invention provides a frequency offset estimation method and device, the scheme is as follows:

[0005] On the one hand, a frequency offset estimation method is provided, the method comprising:

[0006] Acquire multiple reference signals received in the current period, and determine a first power delay spectrum of each reference signal;

[0007] For each reference signal, extract a power delay sub-spectrum within a noise window from a first power delay spectrum of the reference signal, determine the noise power based on a first power value in the power delay sub-spectrum, and determine a weight coefficient of the reference signal based on the noise power, wherein the weight coefficient is negatively correlated with the noise power;

[0008] Performing weighted processing on the first power delay profiles of the multiple reference signals based on the weight coefficients of the multiple reference signals to obtain a second power delay profile;

[0009] The frequency offset is determined based on a target time-domain channel response value of a first delay index of multiple reference signals, wherein the first delay index is a delay index corresponding to a maximum power value in a second power delay spectrum.

[0010] Optionally, determining the noise power based on the first power value in the power delay sub-spectrum includes:

[0011] The noise power is determined based on a first ratio of a sum of a plurality of first power values ​​in the power delay sub-spectrum to a first window length of the noise window.

[0012] Optionally, the starting position of the noise window is obtained by performing a modulo operation on the number of fast Fourier transform (FFT) points of the reference signal based on the first value, and the ending position of the noise window is obtained by performing a modulo operation on the number of FFT points based on the second value;

[0013] Among them, the first value is determined based on the sum of the second delay index and the offset of the noise window, the second value is determined based on the sum of the second delay index, the offset of the noise window and the first window length, and the second delay index is the delay index corresponding to the maximum power value in the first power delay spectrum.

[0014] Optionally, the noise window includes multiple noise sub-windows, and the power delay sub-spectrum includes multiple power sub-sequences corresponding one-to-one to the multiple noise sub-windows; determining the noise power based on the first power value in the power delay sub-spectrum includes: for each power sub-sequence, determining the reference noise power based on the sum of the first power values ​​in the power sub-sequence and a second ratio of the second window length of the noise sub-window corresponding to the power sub-sequence; and determining the noise power from multiple reference noise powers.

[0015] Optionally, determining the noise power from a plurality of reference noise powers includes: determining a minimum value among the plurality of reference noise powers as the noise power.

[0016] Optionally, determining a first power delay spectrum of each reference signal includes:

[0017] Determine the time domain channel response of each reference signal; perform a modular square operation on the time domain channel response of the reference signal to obtain a first power delay spectrum of the reference signal.

[0018] Optionally, determining the frequency offset based on a target time domain channel response value of a first delay index of multiple reference signals includes:

[0019] Based on the target time domain channel response value of the first delay index of at least two reference signals in the same time slot, the phase change values ​​of the at least two reference signals are determined, and a third ratio of the average of the at least two phase change values ​​to the target parameter is determined, wherein the target parameter is determined based on the length of the cyclic prefix and the number of FFT points of the reference signal; the frequency deviation is determined based on the product of the third ratio and the subcarrier spacing; wherein the frequency deviation is positively correlated with each phase change value.

[0020] Optionally, determining the frequency offset based on a target time domain channel response value of a first delay index of multiple reference signals includes:

[0021] Input target time domain channel response values ​​of multiple reference signals at a first delay index into a frequency offset estimation model to obtain a fourth ratio output by the frequency offset estimation model; and determine the frequency offset based on the product of the fourth ratio and the subcarrier spacing.

[0022] Optionally, the frequency offset estimation model includes at least two first sub-models and a second sub-model; inputting target time domain channel response values ​​of multiple reference signals at the first delay index into the frequency offset estimation model to obtain a fourth ratio output by the frequency offset estimation model, including:

[0023] Inputting target time domain channel response values ​​of at least two reference signals in the same time slot at a first delay index into a first sub-model to obtain a frequency deviation characteristic representation output by the first sub-model;

[0024] The frequency deviation characteristic representations of at least two first sub-model outputs are input into the second sub-model to obtain a fourth ratio of the second sub-model output.

[0025] Optionally, the method further includes:

[0026] Acquire multiple sample data; train the multiple sample data to obtain a frequency deviation estimation model;

[0027] Each sample data includes a sample feature and a sample label corresponding to the sample feature, the sample feature includes: a sample target time domain channel response value of a plurality of sample reference signals in a sample period at a sample first delay index, the sample label includes a sample ratio, and the sample ratio is a ratio of a sample frequency deviation to a sample subcarrier spacing;

[0028] The sample first delay index is determined based on the corresponding sample second power delay spectrum, the sample second power delay spectrum is obtained by weighting the sample first power delay spectrum of multiple sample reference signals based on the sample weight coefficients of the multiple sample reference signals, and the sample weight coefficient of the sample reference signal is negatively correlated with the sample noise power of the sample reference signal;

[0029] The sample weight coefficient of each sample reference signal is determined based on the sample noise power, which is determined based on the sample first power value in the sample power delay sub-spectrum, and the sample power delay sub-spectrum is extracted from the sample noise window in the sample first power delay spectrum.

[0030] Optionally, the method further includes:

[0031] If the current cycle is after the first cycle, the second power delay spectrum and the historical second power delay spectrum are weighted to obtain a reference power delay spectrum, where the historical second power delay spectrum is determined based on a reference signal in a historical cycle before and adjacent to the current cycle; the second power delay spectrum is updated to the reference power delay spectrum.

[0032] Optionally, the first weighting coefficient of the second power delay spectrum before updating is determined according to a fifth ratio of the first reference power value to the total reference power value; the second weighting coefficient of the historical second power delay spectrum is determined according to a sixth ratio of the second reference power value to the total reference power value;

[0033] Among them, the first reference power value is the maximum power value in the second power delay spectrum before updating, the second reference power value is the maximum power value in the historical second power delay spectrum, and the total reference power value is determined according to the sum of the first reference power value and the second reference power value.

[0034] On the other hand, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the frequency offset estimation method of the above aspect is implemented.

[0035] On the other hand, a chip is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the frequency offset estimation method of the above aspect is implemented.

[0036] On the other hand, a receiving end is provided, comprising the chip of the above aspect.

[0037] On the other hand, a frequency offset estimation device is provided, the device comprising:

[0038] An acquisition module, used to acquire multiple reference signals received in the current cycle, wherein the multiple reference signals are at different positions in the current cycle;

[0039] A first determining module, configured to determine a first power delay spectrum of each reference signal;

[0040] a second determination module, configured to extract, for each reference signal, a power delay sub-spectrum within a noise window from a first power delay spectrum of the reference signal, determine the noise power based on a first power value in the power delay sub-spectrum, and determine a weight coefficient of the reference signal based on the noise power, wherein the weight coefficient is negatively correlated with the noise power;

[0041] A third determination module is used to perform weighted processing on the first power delay profiles of the multiple reference signals based on the weight coefficients of the multiple reference signals to obtain a second power delay profile;

[0042] The fourth determination module is used to determine the frequency offset based on the target time domain channel response value of multiple reference signals at the first delay index, wherein the first delay index is the delay index corresponding to the maximum power value in the second power delay spectrum.

[0043] Optionally, the second determination module is used to determine the noise power based on a first ratio of a sum of multiple first power values ​​in the power delay sub-spectrum to a first window length of the noise window.

[0044] Optionally, the starting position of the noise window is obtained by performing a modulo operation on the number of FFT points of the reference signal based on the first value, and the ending position of the noise window is obtained by performing a modulo operation on the number of FFT points based on the second value;

[0045] Among them, the first value is determined based on the sum of the second delay index and the offset of the noise window, the second value is determined based on the sum of the second delay index, the offset of the noise window and the first window length, and the second delay index is the delay index corresponding to the maximum power value in the first power delay spectrum.

[0046] Optionally, the noise window includes a plurality of noise sub-windows, and the power delay sub-spectrum includes a plurality of power sub-sequences corresponding one-to-one to the plurality of noise sub-windows; and the second determining module is used to:

[0047] For each power subsequence, determining a reference noise power based on a sum of first power values ​​in the power subsequence and a second ratio of a second window length of a noise subwindow corresponding to the power subsequence;

[0048] A noise power is determined from a plurality of reference noise powers.

[0049] Optionally, the second determination module is used to: determine a minimum value among multiple reference noise powers as the noise power.

[0050] Optionally, a fourth determination module is used to: determine the phase change values ​​of at least two reference signals based on the target time domain channel response value of the first delay index of at least two reference signals in the same time slot, and determine a third ratio of the average of the at least two phase change values ​​to a target parameter, wherein the target parameter is determined based on the length of the cyclic prefix and the number of FFT points; determine the frequency deviation based on the product of the third ratio and the subcarrier spacing; wherein the frequency deviation is positively correlated with each phase change value.

[0051] Optionally, a fourth determination module is used to: input multiple target time domain channel response values ​​into the frequency offset estimation model to obtain a fourth ratio output by the frequency offset estimation model; and determine the frequency offset based on the product of the fourth ratio and the subcarrier spacing.

[0052] Optionally, the frequency offset estimation model includes at least two first sub-models and a second sub-model; a fourth determination module is used to: input the target time domain channel response values ​​of at least two reference signals in the same time slot at the first delay index into a first sub-model to obtain a frequency offset feature representation output by the first sub-model;

[0053] The frequency deviation characteristic representations of at least two first sub-model outputs are input into the second sub-model to obtain a fourth ratio of the second sub-model output.

[0054] Optionally, the device also includes: a fifth determination module, used to: if the current cycle is received after the first cycle, weightedly process the second power delay spectrum and the historical second power delay spectrum to obtain a reference power delay spectrum, the historical second power delay spectrum is determined based on a reference signal in a historical cycle before and adjacent to the current cycle; update the second power delay spectrum to the reference power delay spectrum.

[0055] Optionally, the first weighting coefficient of the second power delay spectrum before updating is determined according to a fifth ratio of the first reference power value to the total reference power value; the second weighting coefficient of the historical second power delay spectrum is determined according to a sixth ratio of the second reference power value to the total reference power value;

[0056] Among them, the first reference power value is the maximum power value in the second power delay spectrum before updating, the second reference power value is the maximum power value in the historical second power delay spectrum, and the total reference power value is determined according to the sum of the first reference power value and the second reference power value.

[0057] In summary, the embodiment of the present invention provides a frequency offset estimation method and device. Since the time interval between different reference signals is relatively long, the channel between the transmitting end and the receiving end may change during the process, resulting in different signal-to-noise ratios of different reference signals. Therefore, after determining the first power delay spectrum of each reference signal, the receiving end extracts the power delay sub-spectrum in the noise window from the first power delay spectrum of the reference signal, determines the noise power based on the first power value in the power delay sub-spectrum, and determines the weight coefficient of the reference signal based on the noise power. Since the weight coefficient is negatively correlated with the noise power, and the noise power is negatively correlated with the signal-to-noise ratio of the reference signal, the weight coefficient is positively correlated with the signal-to-noise ratio of the reference signal, thereby adaptively increasing the weight of the reference signal with a higher signal-to-noise ratio. Then, based on the weight coefficients of multiple reference signals, the first power delay spectrum of multiple reference signals is weighted to obtain a second power delay spectrum, which can ensure that the determined second power delay spectrum conforms to the changes in the channel and has good real-time performance. This ensures that the first delay index determined based on the maximum power value in the second power delay spectrum has a high accuracy, and the frequency offset determined based on the target time domain channel response values ​​of multiple reference signals at the first delay index has a high accuracy.

[0058] In addition, since the frequency deviation at the first delay index corresponding to the maximum power value in the second power delay spectrum is the smallest compared to the actual frequency deviation, the reliability of the maximum power value is the highest. Therefore, determining the frequency deviation based on the target time domain channel response value of the first delay index based on multiple reference signals can improve the accuracy of frequency deviation estimation and obtain a more accurate frequency deviation.

[0059] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 is a flow chart of a frequency offset estimation method provided by an embodiment of the present invention;

[0061] Figure 2 is a flow chart of another frequency offset estimation method provided by an embodiment of the present invention;

[0062] Figure 3 is a schematic diagram of resource configuration of a reference signal provided by an embodiment of the present invention;

[0063] Figure 4 is a schematic diagram of a noise window provided by an embodiment of the present invention;

[0064] Figure 5 is a schematic diagram of another noise window provided by an embodiment of the present invention;

[0065] Figure 6 is a schematic diagram of a sample data set provided by an embodiment of the present invention;

[0066] Figure 7 is a schematic diagram of an initial model provided by an embodiment of the present invention;

[0067] Figure 8 is a schematic diagram of a first initial sub-model provided by an embodiment of the present invention;

[0068] Fig. 9 is a schematic diagram of a second initial sub-model provided by an embodiment of the present invention;

[0069] Fig.10 1 is a schematic diagram for comparing the RMSE of the frequency deviation determined by using different schemes in a scenario where the frequency deviation is 0 Hz, provided by an embodiment of the present invention;

[0070] Fig.11 1 is a schematic diagram for comparing RMSE of frequency deviation determined by different schemes in a scenario where the frequency deviation is 500 Hz, provided by an embodiment of the present invention;

[0071] Fig.12 1 is a schematic diagram for comparing RMSE of frequency deviation determined by different schemes in a scenario where the frequency deviation is 1500 Hz, provided by an embodiment of the present invention;

[0072] Fig.13 It is a schematic diagram of the structure of a chip provided by an embodiment of the present invention;

[0073] Fig.14It is a block diagram of a frequency offset estimation device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0074] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0075] The fifth generation mobile communication technology (5G) uses OFDM technology for downlink. OFDM is a modulation scheme widely used in wireless broadband systems. It can efficiently utilize spectrum resources and better resist interference. The Doppler effect between the transmitter and the receiver in the OFDM system will cause the carrier frequency deviation of the signals of both ends. This frequency deviation will cause signal amplitude attenuation, phase distortion, and destroy the orthogonality between subcarriers, thereby causing ICI, which reduces the signal-to-noise ratio of the OFDM system and brings great challenges to signal demodulation at the receiving end. Therefore, frequency deviation estimation is required in the OFDM system, and accurately estimating and compensating for frequency deviation is an important prerequisite for leveraging the advantages of OFDM technology.

[0076] Figure 1 is a flow chart of a frequency offset estimation method provided by an embodiment of the present invention, which is applied to a receiving end, such as Figure 1 As shown, the method includes:

[0077] Step 101: Acquire multiple reference signals received in the current period, and determine a first power delay spectrum of each reference signal.

[0078] The receiving end can obtain multiple reference signals received in the current period and determine the first power delay spectrum of each reference signal. The reference signal can be a tracking reference signal (TRS), or a demodulation reference signal (DMRS), or a reference signal constructed into a multi-column pilot structure, such as the primary synchronization signals (PSS) and the secondary synchronization signals (SSS) in the synchronization signal block (synchronization signal and pbch block, SSB).

[0079] Step 102: For each reference signal, extract a power delay sub-spectrum within a noise window from a first power delay spectrum of the reference signal, determine the noise power based on a first power value in the power delay sub-spectrum, and determine a weight coefficient of the reference signal based on the noise power.

[0080] After determining the first power delay spectrum of each reference signal, the receiving end can extract the power delay sub-spectrum within the noise window from the first power delay spectrum of the reference signal for each reference signal, determine the noise power based on the first power value in the power delay sub-spectrum, and determine the weight coefficient of the reference signal based on the noise power. The weight coefficient is negatively correlated with the noise power. Since the noise power is negatively correlated with the signal-to-noise ratio of the reference signal, the weight coefficient is positively correlated with the signal-to-noise ratio of the reference signal.

[0081] Step 103: Perform weighted processing on the first power delay profiles of the multiple reference signals based on the weight coefficients of the multiple reference signals to obtain a second power delay profile.

[0082] For each reference signal, after determining the weight coefficient of the reference signal based on the noise power, the receiving end may perform weighted processing on the first power delay profiles of the multiple reference signals based on the weight coefficients of the multiple reference signals to obtain a second power delay profile.

[0083] Step 104: determine a frequency offset based on a target time-domain channel response value of a plurality of reference signals at a first delay index.

[0084] The receiving end performs weighted processing on the first power delay profiles of the multiple reference signals based on the weight coefficients of the multiple reference signals, and after obtaining the second power delay profile, the frequency offset can be determined based on the target time domain channel response values ​​of the multiple reference signals at the first delay index. The first delay index is the delay index corresponding to the maximum power value in the second power delay profile.

[0085] In summary, the embodiment of the present invention provides a frequency offset estimation method. Since the time interval between different reference signals is long, the channel between the transmitting end and the receiving end may change during the process, resulting in different signal-to-noise ratios of different reference signals. Therefore, after determining the first power delay spectrum of each reference signal, the receiving end extracts the power delay sub-spectrum in the noise window from the first power delay spectrum of the reference signal, determines the noise power based on the first power value in the power delay sub-spectrum, and determines the weight coefficient of the reference signal based on the noise power. Since the weight coefficient is negatively correlated with the noise power, and the noise power is negatively correlated with the signal-to-noise ratio of the reference signal, the weight coefficient is positively correlated with the signal-to-noise ratio of the reference signal, thereby adaptively increasing the weight of the reference signal with a higher signal-to-noise ratio. Then, based on the weight coefficients of multiple reference signals, the first power delay spectrum of multiple reference signals is weighted to obtain a second power delay spectrum, which can ensure that the determined second power delay spectrum conforms to the change of the channel and has good real-time performance. Therefore, it can be ensured that the accuracy of the first delay index determined based on the maximum power value in the second power delay spectrum is high, and the accuracy of the frequency offset determined based on the target time domain channel response value of the first delay index of multiple reference signals is high.

[0086] In addition, since the frequency deviation at the first delay index corresponding to the maximum power value in the second power delay spectrum is the smallest compared to the actual frequency deviation, the reliability of the maximum power value is the highest. Therefore, determining the frequency deviation based on the target time domain channel response value of the first delay index based on multiple reference signals can improve the accuracy of frequency deviation estimation and obtain a more accurate frequency deviation.

[0087] Figure 2 is a flow chart of another frequency offset estimation method provided by an embodiment of the present invention, which is applied to a receiving end, such as Figure 2 As shown, the method may include:

[0088] Step 201: Acquire multiple reference signals received in the current period.

[0089] After receiving the communication signal of the current period, the receiving end can demodulate the communication signal to obtain multiple reference signals. Among them, the communication signal can be sent by the transmitting end, and the multiple reference signals have different symbol positions in the current period. The multiple reference signals can all be frequency domain signals, and can all be complex numbers or real numbers. Each reference signal can be TRS, DMRS, or a reference signal that can construct a multi-column pilot structure (such as PSS and SSS in SSB).

[0090] In the embodiment of the present invention, the multiple reference signals may include at least two reference signal groups, each reference signal group may include at least two reference signals, and the at least two reference signal groups are respectively located in at least two consecutive time slots. In the time domain, at least two reference signals in each reference signal group are located in the same time slot, and the symbol interval of the symbols occupied by the at least two reference signals is a preset number of symbols, and by way of example, the preset number may be 4.

[0091] For example, if the reference signal is a TRS, the multiple reference signals may include two reference signal groups, each reference signal group may include two reference signals, that is, the multiple reference signals are four. The two reference signal groups are located in two consecutive time slots.

[0092] The symbol position of each reference signal can be represented by a time slot index and a symbol index. For example, the symbol position of each reference signal can be represented by express, It represents the index of the time slot. It represents the index of the symbol. For example, ,Right now The value can be 0 or 1. The value can be 4 or 8.

[0093] It should be noted that, assuming that the reference signal is TRS, the reference signal can be sent in a burst manner. A burst includes 2 time slots, and each time slot includes 2 TRS symbols. The reference signal is a TRS symbol. When using TRS to estimate the frequency offset, it is necessary to use all TRS symbols in 2 consecutive time slots (that is, a total of 4 TRS symbols).

[0094] In OFDM systems, multiple reference signals are usually inserted discretely in the time-frequency domain for channel estimation. Assume that there are four reference signals, and the distribution of the reference signals in the time-frequency domain is as follows: Figure 3 As shown, the gray filled square represents the reference signal. Figure 3 The horizontal axis is the time domain, and the vertical axis is the frequency domain. Figure 3 The length of a cycle in the time domain may include a first time slot s1 (i.e. s = 0 and the second time slot s2 (i.e. s =1), including multiple subcarriers in the frequency domain. Figure 3 As shown, in each time slot, It represents a signal with a symbol index of 4, and when the symbol index is 4, a plurality of gray filled pattern blocks represent a reference signal. It represents the signal with symbol index 8.

[0095] Step 202: For each reference signal, determine a frequency domain channel response based on the reference signal and a local frequency domain sequence corresponding to the reference signal, and convert the frequency domain channel response into a time domain channel response.

[0096] After acquiring multiple reference signals, the receiving end may determine a frequency domain channel response for each reference signal based on the reference signal and a local frequency domain sequence corresponding to the reference signal, and convert the frequency domain channel response into a time domain channel response, thereby obtaining multiple time domain channel responses, and the multiple time domain channel responses correspond to the multiple reference signals one by one. The receiving end may pre-store a local frequency domain sequence corresponding to each reference signal.

[0097] For the local frequency domain sequence corresponding to each reference signal, the receiving end may perform conjugate processing on the local frequency domain sequence, and determine the frequency domain channel response based on the first product of the reference signal and the local frequency domain sequence after the conjugate processing. Optionally, the receiving end may determine the first product as the frequency domain channel response.

[0098] The index of the time slot is , and the index of the symbol is As an example, the frequency domain channel response determined based on the reference signal is Frequency domain channel response value Can satisfy: .

[0099] in, is the first frequency domain values, It represents the first The value after conjugation of the frequency domain values ​​is In It represents the conjugate treatment.

[0100] The time domain channel response value of each time domain path in the time domain channel response can be a complex number or a real number, and the time domain channel response n The time domain channel response value of the time domain path Can satisfy:

[0101] .

[0102] in, is greater than or equal to 0, and less than or equal to integer, is the number of points of the inverse Fourier transform of the reference signal, It represents the Convert from frequency domain to time domain, It represents an imaginary unit.

[0103] In the embodiment of the present invention, the reference signal is converted from the frequency domain to the time domain, thereby improving the ability to resist noise interference.

[0104] Step 203: Determine a first power delay spectrum of the reference signal based on the time domain channel response of the reference signal.

[0105] For each reference signal, the receiving end determines the frequency domain channel response based on the reference signal and the local frequency domain sequence corresponding to the reference signal, and after converting the frequency domain channel response into a time domain channel response, the first power delay spectrum of the reference signal can be determined based on the time domain channel response of the reference signal.

[0106] In the embodiment of the present invention, the receiving end may perform a modular square operation on the time domain channel response of the reference signal to obtain a first power delay spectrum of the reference signal.

[0107] The index for the time slot is , and the index of the symbol is The reference signal of the first power delay spectrum of the reference signal n The power value of the delay index Can satisfy: , It means that Perform modular square operation.

[0108] Step 204: Obtain a weight coefficient of a reference signal based on the first power delay spectrum.

[0109] After determining the first power delay spectrum of the reference signal based on the time domain channel response of the reference signal, the receiving end can obtain a weight coefficient of the reference signal based on the first power delay spectrum, wherein the weight coefficient is positively correlated with the signal-to-noise ratio of the reference signal.

[0110] In an optional implementation manner of the embodiment of the present invention, the receiving end may determine the signal power from the first power delay spectrum of the reference signal, and determine the weight coefficient of the reference signal based on the ratio of the signal power to the target signal power.

[0111] The signal power is the maximum power value in the first power delay spectrum, and the target signal power is determined based on the sum of the signal powers corresponding to the multiple reference signals. Optionally, the target signal power may be the sum of the signal powers corresponding to the multiple reference signals.

[0112] Optionally, the receiving end may determine the ratio of the signal power to the target signal power as the weight coefficient of the reference signal.

[0113] The index of the time slot is , and the index of the symbol is The weight coefficient of the reference signal Can satisfy: .in, is the signal power, that is, the maximum power value in the first power delay spectrum of the reference signal. is a second delay index, and the second delay index is a delay index corresponding to the maximum power value in the first power delay spectrum. , It means from Select the delay index corresponding to the maximum power value.

[0114] It represents the sum of the signal powers corresponding to multiple reference signals. It represents the index of the time slot. It represents the index of the symbol. For example, , . It means that the index of the time slot is , the index of the symbol is The signal power corresponding to the reference signal. The index of this time slot is , the index of the symbol is The delay index of the maximum power value in the first power delay spectrum of the reference signal.

[0115] In another optional implementation of the embodiment of the present invention, the receiving end may extract a power delay sub-spectrum within a noise window from a first power delay spectrum of a reference signal, determine the noise power based on a first power value in the power delay sub-spectrum, and determine a weight coefficient of the reference signal based on the noise power. The weight coefficient is negatively correlated with the noise power.

[0116] The starting position of the noise window is obtained by performing a modulo operation on the number of FFT points of the reference signal based on the first value, and the ending position of the noise window is obtained by performing a modulo operation on the number of FFT points based on the second value. There may be multiple first power values ​​in the power delay sub-spectrum, and the minimum delay index corresponding to the multiple first power values ​​in the power delay sub-spectrum is the starting position of the noise window, and the maximum delay index corresponding to the multiple first power values ​​in the power delay sub-spectrum is the ending position of the noise window.

[0117] The number of FFT points is equal to the number of inverse Fourier transform points , the number of FFT points can also be marked with The first value may be determined based on the sum of the second delay index and the offset of the noise window, and the second value may be determined based on the sum of the second delay index, the offset of the noise window and the first window length. The second delay index is the delay index corresponding to the maximum power value in the first power delay spectrum, and the offset and the first window length may be pre-stored in the receiving end.

[0118] Optionally, the first value may be the sum of the second delay index and the offset of the noise window, and the second value may be the sum of the second delay index, the offset of the noise window and the first window length.

[0119] refer to Figure 4 , the starting position of the noise window Can satisfy: , mod() means modulus, For example, the offset Can satisfy: .

[0120] The end position of the noise window Can satisfy: , is the first window length. For example, the first window length Can satisfy: .

[0121] Optionally, the receiving end may determine the weight coefficient of the reference signal based on the inverse of the noise power. For example, the receiving end may determine the inverse of the noise power as the weight coefficient of the reference signal. That is, the index of the time slot is , and the index of the symbol is The weight coefficient of the reference signal Can satisfy: , is the noise power corresponding to the reference signal.

[0122] As an optional implementation, the noise window may be one, and the receiving end may determine the noise power based on a first ratio of a sum of multiple first power values ​​in the power delay sub-spectrum to a first window length of the noise window.

[0123] Optionally, the receiving end may determine the first ratio as the noise power. Wherein, the index of the time slot is , and the index of the symbol is The noise power corresponding to the reference signal Can satisfy .

[0124] As another optional implementation, the noise window may include multiple noise sub-windows, and the power delay sub-spectrum may include multiple power sub-sequences corresponding to the multiple noise sub-windows one by one. Any two adjacent noise sub-windows in the multiple noise sub-windows may not overlap or may partially overlap. Furthermore, the second window lengths of any two noise sub-windows in the multiple noise sub-windows may be the same or different.

[0125] refer to Figure 5 , the noise window can include K Noise sub-window, the K There is no overlap between any two adjacent noise sub-windows in the noise sub-windows.

[0126] For each power subsequence, the receiving end may determine the reference noise power based on the sum of the first power values ​​in the power subsequence and the second ratio of the second window length of the noise subwindow corresponding to the power subsequence, thereby obtaining multiple reference noise powers corresponding to the multiple power subsequences one by one. Then, the receiving end may determine the noise power from the multiple reference noise powers.

[0127] Optionally, the receiving end may determine the second ratio as a reference noise power, and the receiving end may determine a minimum value among multiple reference noise powers as the noise power.

[0128] The index of the time slot is , and the index of the symbol is The noise power corresponding to the reference signal Can satisfy: .in, Based on z The reference noise power is determined by the power subsequence within the noise subwindow.

[0129] In this implementation, by dividing the noise window into multiple noise sub-windows, the signal energy in the noise sub-window can be effectively reduced, and the minimum value of multiple reference noise powers is determined as the noise power, which can effectively improve the accuracy of noise power determination.

[0130] Step 205: Perform weighted processing on the first power delay profiles of the multiple reference signals based on the weight coefficients of the multiple reference signals to obtain a second power delay profile.

[0131] After obtaining the weight coefficient of each reference signal, the receiving end may perform weighted processing on the first power delay profiles of the multiple reference signals based on the weight coefficients of the multiple reference signals to obtain the second power delay profile.

[0132] Optionally, the receiving end may perform weighted summation on the first power delay profiles of the multiple reference signals based on weight coefficients of the multiple reference signals to obtain the second power delay profile.

[0133] Optionally, the receiving end may determine a weight coefficient of each reference signal and a target product of the first power delay spectrum of the reference signal, and sum the target products corresponding to multiple reference signals to obtain a second power delay spectrum. n The power value of the delay index Can satisfy:

[0134] .

[0135] Assumptions , then the first n The power value of the delay index Can satisfy: .

[0136] In an embodiment of the present invention, if the current cycle is after the first cycle, the receiving end may perform weighted processing on the second power delay spectrum and the historical second power delay spectrum to obtain a reference power delay spectrum, and may update the second power delay spectrum to the reference power delay spectrum. The historical second power delay spectrum is determined based on a reference signal in a historical cycle before and adjacent to the current cycle. The receiving end may store the historical second power delay spectrum.

[0137] Optionally, the receiving end may perform weighted summation on the second power delay spectrum and the historical second power delay spectrum to obtain a reference power delay spectrum.

[0138] In the embodiment of the present invention, the first weighting coefficient of the second power delay profile before updating is determined according to the fifth ratio of the first reference power value to the total reference power value. The second weighting coefficient of the historical second power delay profile is determined according to the sixth ratio of the second reference power value to the total reference power value.

[0139] Among them, the first reference power value is the maximum power value in the second power delay spectrum before updating, the second reference power value is the maximum power value in the historical second power delay spectrum, and the total reference power value is determined according to the sum of the first reference power value and the second reference power value.

[0140] Optionally, the first weighting coefficient may be the fifth ratio, the second weighting coefficient may be the sixth ratio, and the total reference power value may be the sum of the first reference power value and the second reference power value.

[0141] In the updated second power delay spectrum, n The power value of the delay index Can satisfy: .

[0142] in, is the first weighting coefficient, is the first power delay spectrum before updating. n The power value of the delay index, is the second weighting coefficient, The second highest power delay spectrum in history n The power value of the delay index.

[0143] The first weighting factor Can satisfy: .

[0144] The second weighting factor Can satisfy: .

[0145] in, is the first reference power value, is the delay index corresponding to the maximum power value in the second power delay spectrum before updating, is the second reference power value, It is the delay index corresponding to the maximum power value in the historical second power delay spectrum.

[0146] Due to the constraints on the period of the reference signal in the protocol, taking the reference signal as TRS as an example, the interval between TRSs is at least 10 milliseconds (ms). If the time is too long, the frequency deviation adjustment cannot be performed in time. Therefore, it is generally necessary to estimate a group of reference signals once within a cycle. Through continuous filtering between cycles, the weighting coefficient corresponding to the reference signal with a high signal-to-noise ratio level in two adjacent cycles is larger, thereby enhancing the confidence of the first delay index selected based on the second power delay spectrum, and eliminating the impact of poor burst reference signals within the cycle. Among them, the first delay index is the delay index corresponding to the maximum power value in the second power delay spectrum.

[0147] Step 206: Determine a frequency offset based on a target time-domain channel response value of a plurality of reference signals at a first delay index.

[0148] After the receiving end performs weighted processing on the first power delay profiles of the multiple reference signals based on the weight coefficients of the multiple reference signals to obtain the second power delay profile, the frequency offset can be determined based on the target time domain channel response values ​​of the multiple reference signals at the first delay index. The first delay index is the delay index corresponding to the maximum power value in the second power delay profile.

[0149] In this step, the second power delay spectrum may be the second power delay spectrum before updating determined in step 205, or may be the second power delay spectrum after updating.

[0150] Taking the second power delay spectrum as the updated second power delay spectrum as an example, the first delay index Can satisfy: , It means from Select the delay index corresponding to the maximum power value.

[0151] The index of the time slot is , and the index of the symbol is The reference signal, the target time domain channel response value of the first delay index of the time domain channel response of the reference signal satisfy: .

[0152] In an optional implementation of the embodiment of the present invention, which is hereinafter referred to as solution S1, a process in which a receiving end determines a frequency offset based on a target time domain channel response value of a first delay index of multiple reference signals may include the following steps A1 to A3:

[0153] Step A1: Determine phase change values ​​of at least two reference signals in the same time slot based on target time domain channel response values ​​of the at least two reference signals at a first delay index.

[0154] The receiving end may determine the phase change values ​​of the at least two reference signals based on the target time domain channel response values ​​of the at least two reference signals in the same time slot at the first delay index. For example, the multiple reference signals may be four, and the at least two reference signals in the same time slot may be two.

[0155] In an embodiment of the present invention, the receiving end may perform conjugate processing on the target time domain channel response value of the first reference signal at the first delay index among at least two reference signals located in the same time slot, and determine the phase change value of the at least two reference signals based on the second product of the target time domain channel response value after the conjugate processing and the target time domain channel response value of the second reference signal at the first delay index among the at least two reference signals. The index of the symbol of the first reference signal is less than the index of the symbol of the second reference signal.

[0156] Optionally, the receiving end may determine the phase of the second product as a phase change value of at least two reference signals.

[0157] Assuming that there are four reference signals, and at least two reference signals in the same time slot are two, the time slot index of the first reference signal of the two reference signals is 0, and the symbol index is 4, and the time slot index of the second reference signal is 0, and the symbol index is 8, then the phase change values ​​of the two reference signals are Can satisfy: , It means The phase of In It represents the conjugate treatment.

[0158] Step A2: Determine a third ratio of an average of at least two phase change values ​​to a target parameter.

[0159] After determining at least two phase change values, the receiving end can determine a third ratio of the mean of the at least two phase change values ​​to the target parameter. The target parameter can be determined based on the length of the cyclic prefix and the number of FFT points. The third ratio is the normalized frequency deviation.

[0160] After demodulating the communication signal, the receiving end can also obtain the cyclic prefix, thereby the receiving end can determine the length of the cyclic prefix.

[0161] The receiving end may determine the sum of the length of the cyclic prefix and the number of FFT points, determine the third product of the symbol distance and the sum, and determine the ratio of the third product to the number of FFT points as the target parameter.

[0162] The target parameters Can satisfy: ; is the length of the cyclic prefix, It represents the signed distance, for example, the signed distance It can be 4.

[0163] Assume that there are four reference signals and at least two reference signals in the same time slot are two, the time slot index of the first reference signal of two of the four reference signals is 0, and the symbol index is 4, the time slot index of the second reference signal is 0, and the symbol index is 8, and the time slot index of the first reference signal of the other two reference signals is 1, and the symbol index is 4, and the time slot index of the second reference signal is 1, and the symbol index is 8, then the average of the two phase change values ​​is Can satisfy: , and then the third ratio can satisfy: . In It represents the conjugate treatment.

[0164] Step A3: Determine the frequency offset based on a fourth product of the third ratio and the subcarrier spacing.

[0165] After determining the third ratio, the receiving end may determine the frequency offset based on the fourth product of the third ratio and the subcarrier spacing. The frequency offset is positively correlated with each phase change value. Before determining the frequency offset, the receiving end may perform a cell search to determine the subcarrier spacing.

[0166] Optionally, the receiving end may determine the fourth product as a frequency offset. Can satisfy: , is the subcarrier spacing.

[0167] In an embodiment of the present invention, at least two groups of reference signals with a symbol distance of 4 are used to estimate at least two phase change values, and the frequency offset is determined based on the average of the at least two phase change values, thereby further improving the reliability and accuracy of the frequency offset estimation.

[0168] In another optional implementation of the embodiment of the present invention, which is hereinafter referred to as solution S2, the receiving end determines the frequency offset based on the target time domain channel response value of the first delay index of multiple reference signals, which may include the following steps B1 and B2:

[0169] Step B1: input multiple target time-domain channel response values ​​into a frequency offset estimation model to obtain a fourth ratio output by the frequency offset estimation model.

[0170] The receiving end may input the real numbers and imaginary numbers of the multiple target time domain channel response values ​​into the frequency offset estimation model to obtain a fourth ratio output by the frequency offset estimation model, wherein the fourth ratio is the normalized frequency offset.

[0171] In an embodiment of the present invention, the frequency deviation estimation model may be a machine learning model, and the machine learning model may be a neural network regression model, or other models, such as a random forest, a decision tree, a support vector machine, and LightGBM.

[0172] Step B2: determine the frequency offset based on the product of the fourth ratio and the subcarrier spacing.

[0173] The process of determining the frequency offset estimation model at the receiving end may include the following steps C1 to C3:

[0174] Step C1: Get a sample dataset.

[0175] In an embodiment of the present invention, a receiving end may obtain a sample data set, which includes multiple sample data. Each sample data may include a sample feature and a sample label corresponding to the sample feature, and the sample feature may include: a sample target time domain channel response value of multiple sample reference signals in a sample period at a sample first delay index. The sample label may include a sample ratio, which is a ratio of a sample frequency deviation to a sample subcarrier spacing, and the sample ratio is a normalized sample frequency deviation.

[0176] The sample first delay index is determined based on the corresponding sample second power delay spectrum, and the sample second power delay spectrum is obtained by weighting the sample first power delay spectrum of multiple sample reference signals based on the sample weight coefficients of the multiple sample reference signals. The sample weight coefficient of the sample reference signal is negatively correlated with the sample noise power of the sample reference signal.

[0177] Optionally, the sample weight coefficient of each sample reference signal is determined based on the sample noise power of the sample reference signal, and the sample noise power can be determined based on the sample first power value in the sample power delay sub-spectrum, and the sample power delay sub-spectrum is extracted from the sample noise window in the sample first power delay spectrum. The sample first power delay spectrum is determined based on the sample time domain channel response of the sample reference signal. The determination process of the sample first power delay spectrum of the sample reference signal can refer to the determination process of the first power delay spectrum of the above reference signal. The sample weight coefficient of each sample reference signal can refer to the implementation process of determining the weight coefficient of each reference signal in the above steps.

[0178] Optionally, the sample target time domain channel response value may include a sample real part and a sample imaginary part, and the sample target time domain channel response value may be a complex number or a real number. When the sample target time domain channel response value is a real number, the sample imaginary part is 0. Index of the time slot , and the index of the symbol is The sample target time domain channel response value of the sample reference signal at the sample first delay index Can satisfy: .in, for The real part of the sample, for The imaginary part of the sample .

[0179] The sample dataset can be Figure 6 As shown, the sample data set can be expressed as: Each sample feature can include 8 real values, refer to Figure 6 , No. Sample characteristics Can satisfy: ,in, , . . , . No. Sample labels Can satisfy: , is the sample ratio.

[0180] In this sample data set, c +1-dimensional features can be represented as , c is an integer. M It represents the number of sample data.

[0181] In the embodiment of the present invention, the receiving end may perform linear normalization processing on each dimension feature in the sample data set, thereby mapping the data to between 0 and 1.

[0182] After normalization c +1-dimensional features Can satisfy: , max refers to the maximum value, and min refers to the minimum value.

[0183] In an embodiment of the present invention, the receiving end can adjust the simulated frequency deviation between the simulated transmitting end and the receiving end based on the frequency deviation range through the simulated channel between the simulated transmitting end and the simulated receiving end (the simulated frequency deviation is the sample ratio), and when the frequency deviation between the simulated transmitting end and the receiving end is the simulated frequency deviation, the simulated transmitting end sends multiple sample reference signals within the sample period to the simulated receiving end, thereby determining the sample target time domain channel response value at the sample first delay index based on the multiple sample reference signals to construct a mapping relationship between the sample feature and the sample label, thereby obtaining a sample data set. The sample reference signal can be TRS, DMRS, or a reference signal that can construct a multi-column pilot structure (such as PSS and SSS in SSB).

[0184] The simulation parameters of the simulation channel may be as shown in Table 1. Referring to Table 1, the parameter value of the signal-to-noise ratio (SNR) in the simulation parameters is between -12 decibels (dB) and 12 dB (with a step size of 3 dB), and the SNRs of multiple sample reference signals within a sample period may be the same or different. The parameter value of the frequency deviation range is between -1.75 kHz and 1.75 kHz.

[0185] Table 1

[0186]

[0187] In an embodiment of the present invention, taking the subcarrier spacing of 15kHz in the simulation parameters as an example, multiple sample reference signals occupy a total of 4 resources, and each time slot occupies 2 resources. In the first time slot, the simulation reference signal is located on symbols indexed 4 and 8. Similarly, in the second time slot, it is located on symbols indexed 4 and 8. In the frequency domain dimension, the pilot density is 3, that is, 3 resource elements (RE) are allocated in each RB.

[0188] Step C2: Input multiple sample features of the training subset in the sample data set into the initial model to obtain multiple prediction ratios, and input the multiple prediction ratios and multiple sample ratios in the training subset into the loss function to obtain a loss value.

[0189] After obtaining the sample data set, the receiving end can input multiple sample features of the training subset in the sample data set into the initial model to obtain multiple prediction ratios, and input the multiple prediction ratios and multiple sample ratios in the training subset into the loss function to obtain a loss value. The loss function can be a mean squared error (MSE) loss function.

[0190] The MSE loss function Can satisfy: .

[0191] in, is the sample ratio in the training subset, is the first The sample ratio, is the predicted ratio, For the The predicted ratio, is the number of sample data in the training subset.

[0192] The receiving end can randomly select several small batches of data from the sample data as training subsets.

[0193] refer to Figure 7 The initial model may include an input layer I1, at least two first initial sub-models 10 ( Figure 7 Only two first initial sub-models 10), one second initial sub-model 20 and an output layer O1 are shown. At least two first initial sub-models 10 serve as the first intermediate layer M1 of the initial model, and the second initial sub-model 20 serves as the second intermediate layer M2 of the initial model. Each sub-model can be a deep neural network.

[0194] refer to Figure 7 , a first initial sub-model 10 is used to process each sample feature The output sample frequency deviation feature representation is input into the second initial sub-model 20. The sample frequency deviation feature representation is a parameter containing frequency deviation information learned by the first initial sub-model 10.

[0195] Another first initial sub-model 10 is used to process each sample feature The second initial submodel 20 is used to process the two sample frequency deviation feature representations to obtain a prediction ratio .

[0196] refer to Figure 8, each first initial sub-model 10 may include five neural network layers, the first neural network layer L11 in the first initial sub-model 10 includes 4 neurons, the second neural network layer L12 includes 48 neurons, the third neural network layer L13 includes 24 neurons, the fourth neural network layer L14 includes 8 neurons, and the fifth neural network layer L15 includes 1 neuron.

[0197] refer to Fig. 9 The second initial sub-model 20 may include five neural network layers. The first neural network layer L21 in the second initial sub-model 20 includes 2 neurons, the second neural network layer L22 includes 8 neurons, the third neural network layer L23 includes 8 neurons, the fourth neural network layer L24 includes 4 neurons, and the fifth neural network layer L25 includes 1 neuron.

[0198] The nonlinear function in each sub-model can use the scaled exponential linear unit (SELU) activation function. Can satisfy: ; is a numerical value of the sample feature, is 1.0507, It is 1.6733.

[0199] Step C3: When the loss value is greater than or equal to the preset loss value, adjust the parameters in the initial model based on the loss value.

[0200] The receiving end can adjust the parameters in the initial model based on the loss value when the loss value is greater than or equal to the preset loss value. Afterwards, steps B2 and B3 can be performed again until the loss value is less than the preset loss value, so that the loss value is minimized, thereby obtaining a frequency offset estimation model. Among them, the parameters in the initial model can include a weight matrix and a deviation vector. Afterwards, the receiving end can use a validation subset in the sample data set to verify the performance of the frequency offset estimation model.

[0201] In an embodiment of the present invention, the frequency offset estimation model may include at least two first sub-models and a second sub-model. The receiving end may input the target time domain channel response value of at least two reference signals in the same time slot at the first delay index into a first sub-model to obtain the frequency offset feature representation output by the first sub-model, and input the frequency offset feature representation output by at least two first sub-models into the second sub-model to obtain a fourth ratio output by the second sub-model. The frequency offset feature representation is a parameter containing frequency offset information learned by the first sub-model.

[0202] In the related art (hereinafter referred to as Scheme S3), the receiving end can use the following formula to determine the frequency deviation: :

[0203]

[0204] in, Indicates The subcarrier is The received signal in the frequency domain on OFDM symbols is Indicates The subcarrier is OFDM symbols in the frequency domain.

[0205] for The corresponding local frequency domain sequence, for The corresponding local frequency domain sequence. and In denotes conjugate treatment.

[0206] Represents the symbol distance, that is, the difference between the symbol indices of two adjacent pilot symbols, represents the subcarrier spacing, Indicates the number of subcarriers in the frequency domain. .

[0207] In order to verify the performance of the above schemes S1, S2 and S3, simulations are performed under different parameter settings. The simulation parameters can be as shown in Table 1. The following simulation results ( Fig.10 , Fig.11 , Fig.12 ) is a performance comparison of the root mean squared error (RMSE) of frequency offset in AWGN channel and different frequency offset scenarios through 1000 simulations.

[0208] Fig.10 is a schematic diagram for comparing the RMSE of the frequency deviation determined by using different schemes in a scenario where the frequency deviation is 0 Hz, provided by an embodiment of the present invention. Fig.11 is a schematic diagram for comparing RMSE of frequency deviation determined by different schemes in a scenario where the frequency deviation is 500 Hz, provided by an embodiment of the present invention. Fig.12 It is a schematic diagram for comparing the RMSE of the frequency deviation determined by using different schemes in a scenario where the frequency deviation is 1500 Hz, provided by an embodiment of the present invention.

[0209] The horizontal axis of the comparison diagram is the signal-to-noise ratio, and the vertical axis is the RMSE of the frequency deviation. The comparison diagram may include a first curve, a second curve, and a third curve. The first curve is the RMSE of the frequency deviation determined by scheme S1, the second curve is the RMSE of the frequency deviation determined by scheme S2, and the third curve is the RMSE of the frequency deviation determined by scheme S3.

[0210] pass Figures 10 to 12 It can be seen that in the scenario of small frequency deviation (below 500 Hz), the RMSE performance of Scheme S1 and Scheme S2 is better than that of Scheme S3. When the RMSE is 50 Hz, the performance gain of Scheme S1 is about 2.8 dB, and the performance of Scheme S2 is the best, with a gain of about 9 dB.

[0211] In the large frequency offset scenario, especially under extremely low signal-to-noise ratio, the performance of solution S3 obviously drops sharply. Although solution S1 has a certain performance improvement, the RMSE indicator is still very poor. Solution S2 shows excellent performance under large frequency offset and extremely low signal-to-noise ratio. At -12dB, the RMSE is about 1 / 30 of that of solution S3, and only solution S2 can work stably at this time. This shows that the machine learning-based solution can better adapt to various frequency offset scenarios and is more robust. At the same time, the performance of solutions S1 and S2 is better than that of solution S3, indicating that the time domain-based method has strong anti-noise interference ability and has obvious performance advantages under extremely low signal-to-noise ratio. Based on the above performance comparison, solutions S1 and S2, especially solution S2, have stable algorithm performance and strong robustness, meet the requirements of 5G terminal frequency offset estimation, and have certain practical value.

[0212] In an embodiment of the present invention, by converting the reference signal from the frequency domain to the time domain, and constructing sample features according to the real and imaginary parts of the target time domain channel response value, each sample feature only includes 8 real value values ​​as input. Compared with using the original input signal (i.e., the communication signal in step 201), which leads to a problem that the input dimension increases significantly as the RB increases, the method provided by the embodiment of the present invention can greatly reduce the complexity of the neural network and facilitate neural network learning.

[0213] In summary, the embodiment of the present invention provides a frequency offset estimation method. Since the time interval between different reference signals is long, the channel between the transmitting end and the receiving end may change during the process, resulting in different signal-to-noise ratios of different reference signals. Therefore, after determining the first power delay spectrum of each reference signal, the receiving end extracts the power delay sub-spectrum in the noise window from the first power delay spectrum of the reference signal, determines the noise power based on the first power value in the power delay sub-spectrum, and determines the weight coefficient of the reference signal based on the noise power. Since the weight coefficient is negatively correlated with the noise power, and the noise power is negatively correlated with the signal-to-noise ratio of the reference signal, the weight coefficient is positively correlated with the signal-to-noise ratio of the reference signal, thereby adaptively increasing the weight of the reference signal with a higher signal-to-noise ratio. Then, based on the weight coefficients of multiple reference signals, the first power delay spectrum of multiple reference signals is weighted to obtain a second power delay spectrum, which can ensure that the determined second power delay spectrum conforms to the change of the channel and has good real-time performance. Therefore, it can be ensured that the accuracy of the first delay index determined based on the maximum power value in the second power delay spectrum is high, and the accuracy of the frequency offset determined based on the target time domain channel response value of the first delay index of multiple reference signals is high.

[0214] In addition, since the frequency deviation at the first delay index corresponding to the maximum power value in the second power delay spectrum is the smallest compared to the actual frequency deviation, the reliability of the maximum power value is the highest. Therefore, determining the frequency deviation based on the target time domain channel response value of the first delay index based on multiple reference signals can improve the accuracy of frequency deviation estimation and obtain a more accurate frequency deviation.

[0215] An embodiment of the present invention provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the frequency offset estimation method described in the above aspect is implemented.

[0216] Fig.13 It is a structural schematic diagram of a chip provided by an embodiment of the present invention, the chip includes a memory 1301, a processor 1302 and a computer program stored in the memory 1301 and executable on the processor 1302, and when the processor 1302 executes the computer program, the frequency deviation estimation method described in the above aspects is implemented.

[0217] An embodiment of the present invention provides a receiving end, comprising Fig.13 The chip shown.

[0218] Fig.14 is a block diagram of a frequency offset estimation device provided by an embodiment of the present invention, such as Fig.14 As shown, the device comprises:

[0219] An acquisition module 1401 is used to acquire multiple reference signals received in the current period;

[0220] A first determining module 1402, configured to determine a first power delay profile of each reference signal;

[0221] A second determination module 1403 is used to extract, for each reference signal, a power delay sub-spectrum within a noise window from a first power delay spectrum of the reference signal, determine the noise power based on a first power value in the power delay sub-spectrum, and determine a weight coefficient of the reference signal based on the noise power, where the weight coefficient is negatively correlated with the noise power;

[0222] A third determining module 1404 is configured to perform weighted processing on the first power delay profiles of the multiple reference signals based on the weight coefficients of the multiple reference signals to obtain a second power delay profile;

[0223] The fourth determination module 1405 is configured to determine the frequency offset based on a target time domain channel response value of multiple reference signals at a first delay index, wherein the first delay index is a delay index corresponding to a maximum power value in the second power delay spectrum.

[0224] Optionally, the second determination module 1403 is configured to determine the noise power based on a first ratio of a sum of multiple first power values ​​in the power delay sub-spectrum to a first window length of the noise window.

[0225] Optionally, the starting position of the noise window is obtained by performing a modulo operation on the number of FFT points of the reference signal based on a first value, and the ending position of the noise window is obtained by performing a modulo operation on the number of FFT points based on a second value; wherein the first value is determined based on the sum of the second delay index and the offset of the noise window, the second value is determined based on the sum of the second delay index, the offset of the noise window and the first window length, and the second delay index is the delay index corresponding to the maximum power value in the first power delay spectrum.

[0226] Optionally, the noise window includes multiple noise sub-windows, and the power delay sub-spectrum includes multiple power sub-sequences corresponding one-to-one to the multiple noise sub-windows; the second determination module is used to: for each power sub-sequence, determine the reference noise power based on the sum of the first power values ​​in the power sub-sequence and the second ratio of the second window length of the noise sub-window corresponding to the power sub-sequence; determine the noise power from multiple reference noise powers.

[0227] Optionally, the second determining module 1403 is configured to: determine a minimum value among multiple reference noise powers as the noise power.

[0228] Optionally, the fourth determination module 1405 is used to: determine the phase change values ​​of at least two reference signals based on the target time domain channel response value of the first delay index of at least two reference signals in the same time slot, and determine a third ratio of the average of the at least two phase change values ​​to a target parameter, wherein the target parameter is determined based on the length of the cyclic prefix and the number of FFT points of the reference signal; determine the frequency deviation based on the product of the third ratio and the subcarrier spacing; wherein the frequency deviation is positively correlated with each phase change value.

[0229] Optionally, the fourth determination module 1405 is used to: input multiple target time domain channel response values ​​into the frequency offset estimation model to obtain a fourth ratio output by the frequency offset estimation model; and determine the frequency offset based on the product of the fourth ratio and the subcarrier spacing.

[0230] Optionally, the frequency offset estimation model includes at least two first sub-models and a second sub-model; the fourth determination module 1405 is used to: input the target time domain channel response value of at least two reference signals in the same time slot at the first delay index into a first sub-model to obtain the frequency offset characteristic representation output by the first sub-model; input the frequency offset characteristic representations output by at least two first sub-models into the second sub-model to obtain a fourth ratio output by the second sub-model.

[0231] Optionally, the device further includes: a fifth determining module, configured to:

[0232] If the current cycle is after the first cycle, the second power delay spectrum and the historical second power delay spectrum are weighted to obtain a reference power delay spectrum, where the historical second power delay spectrum is determined based on a reference signal in a historical cycle before and adjacent to the current cycle; the second power delay spectrum is updated to the reference power delay spectrum.

[0233] Optionally, the first weighting coefficient of the second power delay spectrum before updating is determined based on the fifth ratio of the first reference power value to the total reference power value; the second weighting coefficient of the historical second power delay spectrum is determined based on the sixth ratio of the second reference power value to the total reference power value; wherein the first reference power value is the maximum power value in the second power delay spectrum before updating, the second reference power value is the maximum power value in the historical second power delay spectrum, and the total reference power value is determined based on the sum of the first reference power value and the second reference power value.

[0234] In summary, the embodiment of the present invention provides a frequency offset estimation device. Since the time interval between different reference signals is long, the channel between the transmitting end and the receiving end may change during the process, resulting in different signal-to-noise ratios of different reference signals. Therefore, after determining the first power delay spectrum of each reference signal, the receiving end extracts the power delay sub-spectrum in the noise window from the first power delay spectrum of the reference signal, determines the noise power based on the first power value in the power delay sub-spectrum, and determines the weight coefficient of the reference signal based on the noise power. Since the weight coefficient is negatively correlated with the noise power, and the noise power is negatively correlated with the signal-to-noise ratio of the reference signal, the weight coefficient is positively correlated with the signal-to-noise ratio of the reference signal, thereby adaptively increasing the weight of the reference signal with a higher signal-to-noise ratio. Then, based on the weight coefficients of multiple reference signals, the first power delay spectrum of multiple reference signals is weighted to obtain a second power delay spectrum, which can ensure that the determined second power delay spectrum conforms to the change of the channel and has good real-time performance. Therefore, it can be ensured that the accuracy of the first delay index determined based on the maximum power value in the second power delay spectrum is high, and the accuracy of the frequency offset determined based on the target time domain channel response value of the first delay index of multiple reference signals is high.

[0235] In addition, since the frequency deviation at the first delay index corresponding to the maximum power value in the second power delay spectrum is the smallest compared to the actual frequency deviation, the reliability of the maximum power value is the highest. Therefore, determining the frequency deviation based on the target time domain channel response value of the first delay index based on multiple reference signals can improve the accuracy of frequency deviation estimation and obtain a more accurate frequency deviation.

[0236] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection portion with one or more wirings (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0237] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0238] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0239] In the description of the present invention, it is to be understood that the terms “center”, “longitudinal”, “lateral”, “length”, “width”, “thickness”, “up”, “down”, “front”, “back”, “left”, “right”, “vertical”, “horizontal”, “top”, “bottom”, “inside”, “outside”, “clockwise”, “counterclockwise”, “axial”, “radial”, “circumferential”, etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0240] In addition, the terms "first", "second", etc. used in the embodiments of the present invention are only used for descriptive purposes and should not be understood as indicating or implying relative importance, or implicitly indicating the number of technical features indicated in the present embodiment. Therefore, the features defined by the terms "first", "second", etc. in the embodiments of the present invention can explicitly or implicitly indicate that the embodiment includes at least one of the features. In the description of the present invention, the word "multiple" means at least two or two or more, such as two, three, four, etc., unless otherwise clearly and specifically defined in the embodiments.

[0241] In the present invention, unless otherwise clearly specified or limited in the embodiments, the terms "installed", "connected", "connected" and "fixed" etc. in the embodiments should be understood in a broad sense. For example, the connection can be a fixed connection, a detachable connection, or an integrated connection. It can be understood that it can also be a mechanical connection, an electrical connection, etc.; of course, it can also be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal connection of two elements, or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to the specific implementation situation.

[0242] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may mean that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. Moreover, a first feature being "above", "above" or "above" a second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. A first feature being "below", "below" or "below" a second feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is lower in level than the second feature.

[0243] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A frequency offset estimation method, characterized in that: The method comprises: Acquire multiple reference signals received in the current period, and determine a first power delay spectrum of each of the reference signals; For each of the reference signals, extract a power delay sub-spectrum within a noise window from a first power delay spectrum of the reference signal, determine a noise power based on a first power value in the power delay sub-spectrum, and determine a weight coefficient of the reference signal based on the noise power, wherein the weight coefficient is negatively correlated with the noise power; Performing weighted processing on the first power delay profiles of the plurality of reference signals based on weight coefficients of the plurality of reference signals to obtain a second power delay profile; The frequency offset is determined based on target time-domain channel response values ​​of the multiple reference signals at a first delay index, wherein the first delay index is a delay index corresponding to a maximum power value in the second power delay spectrum.

2. The method according to claim 1, characterized in that Determining the noise power based on the first power value in the power delay sub-spectrum includes: The noise power is determined based on a first ratio of a sum of a plurality of first power values ​​in the power delay sub-spectrum to a first window length of the noise window.

3. The method according to claim 2, characterized in that The starting position of the noise window is obtained by performing a modulo operation on the number of FFT points of the reference signal based on a first value, and the ending position of the noise window is obtained by performing a modulo operation on the number of FFT points based on a second value; Among them, the first value is determined based on the sum of the second delay index and the offset of the noise window, the second value is determined based on the sum of the second delay index, the offset of the noise window and the first window length, and the second delay index is the delay index corresponding to the maximum power value in the first power delay spectrum.

4. The method according to claim 1, characterized in that The noise window includes a plurality of noise sub-windows, and the power delay sub-spectrum includes a plurality of power sub-sequences corresponding one-to-one to the plurality of noise sub-windows; Determining the noise power based on the first power value in the power delay sub-spectrum includes: For each of the power subsequences, determining a reference noise power based on a sum of first power values ​​in the power subsequence and a second ratio of a second window length of a noise subwindow corresponding to the power subsequence; The noise power is determined from a plurality of the reference noise powers.

5. The method according to claim 4, characterized in that Determining the noise power from a plurality of the reference noise powers comprises: A minimum value among the plurality of reference noise powers is determined as the noise power.

6. The method according to any one of claims 1 to 5, characterized in that: Determining a first power delay spectrum of each of the reference signals comprises: determining a time domain channel response of each of the reference signals; A modular square operation is performed on the time domain channel response of the reference signal to obtain a first power delay spectrum of the reference signal.

7. The method according to any one of claims 1 to 5, characterized in that: Determining the frequency offset based on a target time domain channel response value of a plurality of reference signals at a first delay index comprises: Determine, based on target time domain channel response values ​​of at least two reference signals in the same time slot at the first delay index, phase change values ​​of at least two reference signals, and determine a third ratio of an average of at least two phase change values ​​to a target parameter, wherein the target parameter is determined based on a length of a cyclic prefix and a number of FFT points of the reference signal; The frequency offset is determined based on the product of the third ratio and the subcarrier spacing; wherein the frequency offset is positively correlated with each of the phase change values.

8. The method according to any one of claims 1 to 5, characterized in that: Determining the frequency offset based on a target time domain channel response value of a plurality of reference signals at a first delay index comprises: Inputting target time domain channel response values ​​of the plurality of reference signals at the first delay index into a frequency offset estimation model to obtain a fourth ratio output by the frequency offset estimation model; The frequency offset is determined based on a product of the fourth ratio and the subcarrier spacing.

9. The method according to claim 8, characterized in that The frequency offset estimation model includes at least two first sub-models and a second sub-model; inputting target time domain channel response values ​​of the multiple reference signals at the first delay index into the frequency offset estimation model to obtain a fourth ratio output by the frequency offset estimation model, including: Inputting the target time domain channel response values ​​of at least two reference signals in the same time slot at the first delay index into a first sub-model to obtain a frequency offset characteristic representation output by the first sub-model; The frequency deviation characteristic representations of at least two outputs of the first sub-model are input into the second sub-model to obtain a fourth ratio output by the second sub-model.

10. The method according to claim 9, characterized in that The method further comprises: Get multiple sample data; Training the plurality of sample data to obtain the frequency offset estimation model; Each of the sample data includes a sample feature and a sample label corresponding to the sample feature, the sample feature includes: a sample target time domain channel response value of a plurality of sample reference signals in a sample period at a sample first delay index, the sample label includes a sample ratio, and the sample ratio is a ratio of a sample frequency deviation to a sample subcarrier spacing; The sample first delay index is determined based on the corresponding sample second power delay spectrum, the sample second power delay spectrum is obtained by weighting the sample first power delay spectrums of the plurality of sample reference signals based on the sample weight coefficients of the plurality of sample reference signals, and the sample weight coefficient of the sample reference signal is negatively correlated with the sample noise power of the sample reference signal; The sample weight coefficient of each sample reference signal is determined based on the sample noise power, and the sample noise power is determined based on the sample first power value in the sample power delay sub-spectrum, and the sample power delay sub-spectrum is extracted from the sample noise window in the sample first power delay spectrum.

11. The method according to any one of claims 1 to 5, characterized in that: The method further comprises: If the current cycle is after the first cycle, weighting the second power delay spectrum and the historical second power delay spectrum to obtain a reference power delay spectrum, where the historical second power delay spectrum is determined based on a reference signal in a historical cycle before and adjacent to the current cycle; The second power delay profile is updated to the reference power delay profile.

12. The method according to claim 11, characterized in that The first weighting coefficient of the second power delay profile before updating is determined according to a fifth ratio of the first reference power value to the total reference power value; The second weighting coefficient of the historical second power delay profile is determined according to a sixth ratio of the second reference power value to the total reference power value; Among them, the first reference power value is the maximum power value in the second power delay spectrum before updating, the second reference power value is the maximum power value in the historical second power delay spectrum, and the total reference power value is determined according to the sum of the first reference power value and the second reference power value.

13. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the frequency offset estimation method according to any one of claims 1 to 12 is implemented.

14. A chip, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the frequency offset estimation method according to any one of claims 1 to 12 is implemented.

15. A receiving end, characterized in that: Comprising the chip as claimed in claim 14.

16. A frequency offset estimation device, characterized in that: The device comprises: An acquisition module, used for acquiring a plurality of reference signals received in the current period; A first determining module, configured to determine a first power delay spectrum of each of the reference signals; a second determination module, configured to extract, for each of the reference signals, a power delay sub-spectrum within a noise window from a first power delay spectrum of the reference signal, determine a noise power based on a first power value in the power delay sub-spectrum, and determine a weight coefficient of the reference signal based on the noise power, wherein the weight coefficient is negatively correlated with the noise power; A third determination module is used to perform weighted processing on the first power delay profiles of the plurality of reference signals based on weight coefficients of the plurality of reference signals to obtain a second power delay profile; A fourth determination module is used to determine the frequency offset based on a target time domain channel response value of a plurality of the reference signals at a first delay index, wherein the first delay index is a delay index corresponding to a maximum power value in the second power delay spectrum.

17. The device according to claim 16, characterized in that The second determining module is used to: The noise power is determined based on a first ratio of a sum of a plurality of first power values ​​in the power delay sub-spectrum to a first window length of the noise window.

18. The device according to claim 17, characterized in that The starting position of the noise window is obtained by performing a modulo operation on the number of FFT points of the reference signal based on a first value, and the ending position of the noise window is obtained by performing a modulo operation on the number of FFT points based on a second value; Among them, the first value is determined based on the sum of the second delay index and the offset of the noise window, the second value is determined based on the sum of the second delay index, the offset of the noise window and the first window length, and the second delay index is the delay index corresponding to the maximum power value in the first power delay spectrum.

19. The device according to claim 16, characterized in that The noise window includes a plurality of noise sub-windows, and the power delay sub-spectrum includes a plurality of power sub-sequences corresponding one-to-one to the plurality of noise sub-windows; and the second determining module is used to: For each of the power subsequences, determining a reference noise power based on a sum of first power values ​​in the power subsequence and a second ratio of a second window length of a noise subwindow corresponding to the power subsequence; The noise power is determined from a plurality of the reference noise powers.

20. The device according to claim 19, characterized in that The second determining module is used to: A minimum value among the plurality of reference noise powers is determined as the noise power.

21. The device according to any one of claims 16 to 20, characterized in that The fourth determining module is used to: Determine, based on target time domain channel response values ​​of at least two reference signals in the same time slot at the first delay index, phase change values ​​of at least two reference signals, and determine a third ratio of an average of at least two phase change values ​​to a target parameter, wherein the target parameter is determined based on a length of a cyclic prefix and a number of FFT points of the reference signal; The frequency offset is determined based on the product of the third ratio and the subcarrier spacing; wherein the frequency offset is positively correlated with each of the phase change values.

22. The device according to any one of claims 16 to 20, characterized in that The fourth determining module is used to: Inputting the plurality of target time-domain channel response values ​​into a frequency offset estimation model to obtain a fourth ratio output by the frequency offset estimation model; The frequency offset is determined based on a product of the fourth ratio and the subcarrier spacing.

23. The device according to claim 22, characterized in that The frequency offset estimation model includes at least two first sub-models and a second sub-model; the fourth determination module is used to: Inputting the target time domain channel response values ​​of at least two reference signals in the same time slot at the first delay index into a first sub-model to obtain a frequency offset characteristic representation output by the first sub-model; The frequency deviation characteristic representations of at least two outputs of the first sub-model are input into the second sub-model to obtain a fourth ratio output by the second sub-model.

24. The device according to any one of claims 16 to 20, characterized in that The device further includes: a fifth determining module, configured to: If the current cycle is after the first cycle, weighting the second power delay spectrum and the historical second power delay spectrum to obtain a reference power delay spectrum, where the historical second power delay spectrum is determined based on a reference signal in a historical cycle before and adjacent to the current cycle; The second power delay profile is updated to the reference power delay profile.

25. The device according to claim 24, characterized in that The first weighting coefficient of the second power delay profile before updating is determined according to a fifth ratio of the first reference power value to the total reference power value; The second weighting coefficient of the historical second power delay profile is determined according to a sixth ratio of the second reference power value to the total reference power value; Among them, the first reference power value is the maximum power value in the second power delay spectrum before updating, the second reference power value is the maximum power value in the historical second power delay spectrum, and the total reference power value is determined according to the sum of the first reference power value and the second reference power value.

Images