Noise estimation method and device for orthogonal frequency division multiplexing communication system

Abstract

The invention discloses a noise estimation method and device for an orthogonal frequency division multiplexing communication system, and the method comprises the following steps: S1, calculating respective initial noise estimation matrixes on subcarriers based on a received training sequence; s2, on the basis of the initial noise estimation matrix, a measurement parameter vector used for representing noise characteristics of each subcarrier is obtained through calculation, and the dimension of the measurement parameter vector is equal to the number of the subcarriers; s3, dynamically and adaptively dividing all the subcarriers into one or more continuous frequency segments according to the measurement parameter vector, and determining the dividing boundary by the change characteristic of the measurement parameter vector in the frequency domain; and S4, for each frequency segment obtained by division, based on the initial noise estimation matrix corresponding to the subcarriers contained in the segment, calculating a noise covariance matrix of the segment, so that colored noise can be more effectively converted into white noise, the detection error of the signal is remarkably reduced, and the receiving performance and demodulation reliability of the whole communication system are improved.

Description

A method and apparatus for noise estimation in an orthogonal frequency division multiplexing communication system Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a noise estimation method for Orthogonal Frequency Division Multiplexing (OFDM) communication systems, which is especially suitable for communication systems with complex and time-varying noise environments, such as power line carrier communication. Background Technology

[0002] Orthogonal frequency division multiplexing (OFDM) technology is widely used in various communication systems, such as wireless communication and power line communication (PLC), due to its high spectral efficiency and resistance to multipath fading. At the receiver, for accurate signal detection and demodulation, such as using the minimum mean squared error (MMSE) equalization algorithm, it is necessary to accurately estimate the statistical characteristics of the noise, typically represented as a noise covariance matrix. An accurate noise covariance matrix is ​​crucial for subsequent whitening processing, which converts spatially colored noise or interference into spatially white noise, thereby simplifying the detection algorithm and optimizing system performance.

[0003] The noise characteristics of power line communication channels differ significantly from those of traditional additive white Gaussian noise (AWGN) channels. Their noise typically exhibits strong frequency selectivity and time dependence, manifested as significant differences in noise power and interference levels across different frequency subcarriers, and dynamic changes in noise characteristics over time. Existing technologies typically employ two strategies to calculate the noise covariance matrix: one is to calculate a global noise covariance matrix across the entire communication frequency band; the other is to pre-divide the frequency band into several uniform sub-bands and then calculate the noise covariance matrix separately within each fixed sub-band.

[0004] However, the first strategy cannot accurately describe the richly varied noise characteristics in the frequency domain. When noise power changes drastically in the frequency domain, using a single global matrix will severely degrade the whitening effect and subsequent demodulation performance. While the second strategy partially improves the problem, its partitioning is fixed and cannot adapt to the real-time, dynamically changing noise conditions of the current channel. For example, in frequency bands with flat noise characteristics, fixed partitioning may lead to unnecessary segmentation, increasing computational overhead; while in frequency bands with drastically changing noise characteristics, fixed partitioning may not provide sufficient resolution, resulting in a segment still containing noise with drastically different characteristics, and the calculated covariance matrix remaining inaccurate.

[0005] Therefore, there is an urgent need in this field for a noise estimation method that can adapt to real-time noise characteristics and dynamically divide the frequency domain in order to calculate a more accurate noise covariance matrix. Summary of the Invention

[0006] This invention aims to address the shortcomings of the prior art by providing a noise estimation method for orthogonal frequency division multiplexing (OFDM) communication systems. The core technical problem this invention addresses is how to overcome the rigidity of fixed frequency band division methods and achieve a method that adaptively and dynamically divides subcarriers into one or more continuous frequency segments based on the actual distribution characteristics of current channel noise in the frequency domain. This allows the noise covariance matrix calculated for each segment to more accurately characterize the noise characteristics within that segment, thereby improving the performance of subsequent whitening processing and signal detection.

[0007] The technical problem of the present invention is solved by the following technical solution: In a first aspect, the present invention provides a noise estimation method for an orthogonal frequency division multiplexing communication system, comprising the following steps: S1, calculating the initial noise estimation matrix for each subcarrier based on the received training sequence; S2, calculating a metric parameter vector characterizing the noise characteristics of each subcarrier based on the initial noise estimation matrix, wherein the dimension of the metric parameter vector is equal to the number of subcarriers; S3, dynamically and adaptively dividing all subcarriers into one or more continuous frequency segments according to the metric parameter vector, wherein the boundary of the division is determined by the variation characteristics of the metric parameter vector in the frequency domain; S4, for each frequency segment obtained by the division, calculating the noise covariance matrix of the segment based on the initial noise estimation matrix corresponding to the subcarriers contained in the segment.

[0008] In some embodiments, the following technical features are also included: Step S1 specifically includes the following steps: S11, performing least squares channel estimation on the training sequences received on different orthogonal frequency division multiplexing symbols to obtain multiple channel response estimation matrices; S12, subtracting the least squares estimation results of different subcarriers of adjacent OFDM symbols one by one to obtain the initial noise estimation matrices of different subcarriers.

[0009] In some embodiments, in step S2, the metric parameter vector is a scalar set that reflects the noise power or interference level of each subcarrier, calculated based on the initial noise estimation matrix; the calculation of the metric parameter vector specifically includes the following steps: S21, averaging the multiple initial noise estimation matrices along the symbol dimension; S22, averaging the main diagonal elements of the averaged matrix to generate the metric parameter vector.

[0010] In some embodiments, in step S3, all subcarriers are dynamically and adaptively divided into one or more continuous frequency segments according to the metric parameter vector, using a sliding window growth algorithm. Specifically, this includes the following steps: S31A: Initialize the current frequency segment starting from the first subcarrier; S32A: Iteratively attempt to add subsequent subcarriers to the current frequency segment; S33A: In each iteration, calculate the statistical dispersion of the metric parameters of the subcarriers within the current frequency segment; the statistical dispersion is one of variance, standard deviation, or root mean square deviation; S34A: If the statistical dispersion is less than a first predetermined threshold, and the number of subcarriers in the current segment is less than a second predetermined threshold, then the subcarrier is formally added to the current frequency segment; otherwise, the previous subcarrier is used as the end boundary of the current frequency segment, and a new frequency segment begins with the current subcarrier.

[0011] In some embodiments, in step S3, all subcarriers are dynamically and adaptively divided into one or more continuous frequency segments according to the metric parameter vector, using a recursive segmentation algorithm. Specifically, this includes the following steps: S31B: If the statistical dispersion of the metric parameters of the subcarriers in the current frequency band is greater than a predetermined threshold, then an optimal segmentation point is found in the frequency band; S32B: The current frequency band is divided into two sub-frequency bands using the optimal segmentation point; S33B: The segmentation steps are recursively performed on the two sub-frequency bands respectively until the statistical dispersion of all sub-frequency bands is less than or equal to the predetermined threshold.

[0012] In some embodiments, before calculating the noise covariance matrix in step S4, the following steps are also included: S41, detecting the element values ​​in the initial noise estimation matrix; S42, setting or removing elements whose values ​​exceed a predetermined anomaly threshold to suppress the influence of impulse noise or single-frequency interference on the calculation of the covariance matrix.

[0013] In some embodiments, the orthogonal frequency division multiplexing communication system is a power line carrier communication system.

[0014] In a second aspect, the present invention provides a communication device, comprising: one or more processors; a memory for storing one or more programs; and, when the one or more programs are executed by the one or more processors, causing the communication device to implement the above-described method.

[0015] Thirdly, the present invention provides a chip including a processing circuit configured to perform the above-described method.

[0016] Fourthly, the present invention provides a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implements the above-described method.

[0017] Fifthly, the present invention provides a receiver in a wireless communication system that uses the above-described method for noise estimation and whitening.

[0018] The beneficial effects of this invention compared to existing technologies include: This invention achieves a match between noise estimation and the actual frequency domain characteristics of the current channel noise by first calculating the initial noise estimation matrix of each subcarrier and generating a metric parameter vector accordingly, then dynamically and adaptively dividing the frequency into segments based on this vector, and finally calculating the noise covariance matrix segment by segment. Firstly, by analyzing the frequency domain variation characteristics of the metric parameter vector, regions of continuous subcarriers with relatively stable noise characteristics can be identified, thus dividing these subcarriers into the same segment. Within this segment, due to the consistent noise characteristics, the covariance matrix calculated by statistically averaging the initial noise estimation matrices of all subcarriers within the segment can more realistically and accurately reflect the noise statistical characteristics of the local frequency band. This accurate representation means that subsequent whitening processing can more effectively convert colored noise into white noise, thereby providing higher-quality input for subsequent detection modules such as the MMSE equalizer, ultimately significantly reducing signal detection errors and improving the reception performance and demodulation reliability of the entire communication system. This top-level effect is achieved through the synergistic effect of two intermediate effects: "dynamic partitioning" to ensure consistency within the partition, and "partition statistics" to improve estimation accuracy.

[0019] Furthermore, in some embodiments, the following beneficial effects are also achieved: by employing adaptive algorithms such as sliding window growth or recursive segmentation, the method of the present invention can flexibly cope with different noise scenarios. When the noise is uniformly distributed in the frequency domain, the algorithm tends to generate fewer segments or even one segment, avoiding unnecessary computational complexity; when the noise is colorful and drastically changing in the frequency domain, the algorithm automatically segments at the boundary of change, generating more segments to provide higher resolution, thereby achieving an optimal balance between computational complexity and performance while ensuring estimation accuracy.

[0020] By introducing outlier detection and removal steps, the negative impact of impulse noise or strong single-frequency interference on the covariance matrix calculation is effectively suppressed, further improving the robustness and accuracy of noise estimation.

[0021] The method described in this invention is not limited to power line communication systems; its core concept is also applicable to any OFDM system with colored noise in the frequency domain, such as wireless communication systems, and has wide applicability.

[0022] Integrating the above methods into communication devices, chips, or receivers can enable these devices to have superior noise suppression and signal processing capabilities.

[0023] Other beneficial effects of the embodiments of the present invention will be further described below. Attached Figure Description

[0024] Figure 1 is a flowchart of a noise estimation method for an orthogonal frequency division multiplexing communication system according to an embodiment of the present invention.

[0025] Figure 2 is a flowchart of an implementation method for calculating the dynamic segmented noise covariance matrix in an embodiment of the present invention.

[0026] Figure 3 is a flowchart of a method for dynamically segmenting noise covariance calculation in an embodiment of the present invention.

[0027] Figure 4 is a schematic diagram of the existing technology of using fixed uniform frequency segmentation for noise covariance matrix calculation.

[0028] Figure 5 is a schematic diagram of the use of dynamic frequency segmentation for noise covariance matrix calculation in an embodiment of the present invention. Detailed Implementation

[0029] The present invention will be further described below with reference to the accompanying drawings and preferred embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0030] Implementation concept: The core of this invention lies in dynamically and adaptively dividing the entire frequency band into one or more continuous frequency segments based on the noise characteristics of each subcarrier estimated in real time at the OFDM system receiver, so that the subcarriers in each segment have similar noise statistical characteristics, thereby calculating a more accurate noise covariance matrix for each segment for subsequent signal processing.

[0031] This invention proposes an adaptive segmented calculation method for the noise covariance matrix at the receiver of a power line OFDM communication system. Based on the noise characteristics of different subcarrier regions, the subcarriers are adaptively divided into continuous, non-uniform segments. Different noise covariance matrices are then generated for each segmented region for whitening processing (converting spatial colored noise (or interference) in the received signal into spatial white noise, thereby simplifying subsequent detection algorithms, optimizing their performance, and reducing computational complexity). The segmented calculation method involves: calculating the correlation matrix of all subcarriers; segmenting the subcarriers using the obtained matrix information, with segmentation methods including, but not limited to, the sliding window growth method: dynamically comparing the deviation between the current subcarrier correlation matrix metric and the metric corresponding to the current segment, iteratively determining the segment boundaries; and calculating the noise covariance matrix for each of the different regions obtained from the final segmentation for subsequent whitening processing. This method mainly considers that the noise in actual power line channels is no longer additive white Gaussian noise, but colored noise that varies with frequency. Therefore, calculating a single noise covariance matrix for the entire communication frequency band or dividing the frequency band into several fixed segments may not accurately describe the noise situation in different frequency bands.

[0032] Example 1: This example provides a noise estimation method for an orthogonal frequency division multiplexing (OFDM) communication system, as shown in Figure 1, including the following steps: S1. Calculate the initial noise estimation matrix for each subcarrier based on the received training sequence. The training sequence includes a preamble sequence (PREAM), a training field (TF), and reference signals such as the Channel State Information Reference Signal (CSI-RS). Step S1 specifically includes the following steps: S11. Perform least-squares channel estimation on the training sequences received on different OFDM symbols to obtain multiple channel response estimation matrices. Specifically, the channel response estimation matrix of the current channel is obtained by performing least-squares (LS) estimation using the frequency domain transformation of PREAM or TF symbols in a power line communication system. For PREAM or TF frequency domain The symbol of the first Data received in the frequency domain of each subcarrier. For the known first If data is transmitted in the frequency domain of the nth subcarrier, then different transmit and receive antennas will pair with the nth subcarrier. The symbol of the first Least squares estimation of subcarriers for

[0033] S12. Subtract the least squares estimation results of different subcarriers of adjacent OFDM symbols one by one to obtain the initial noise estimation matrices of different subcarriers. Specifically, assume that a total of OFDM symbols and Each subcarrier. The initial noise estimation matrix for different transmit / receive antenna pairs is calculated using the frequency response of the same subcarrier in different OFDM symbols: Initial noise estimation matrix The dimension is . and These represent the number of transmitting antennas and the number of receiving antennas, respectively.

[0034] S2. Based on the initial noise estimation matrix, a metric parameter vector characterizing the noise characteristics of each subcarrier is calculated, the dimension of which is equal to the number of subcarriers. The metric parameter vector is a scalar set reflecting the noise power or interference level of each subcarrier, calculated from the initial noise estimation matrix. The calculation of the metric parameter vector specifically includes the following steps: S21. First, the multiple initial noise estimation matrices are averaged along the sign dimension.

[0035] Specifically, using the calculation in step S12 matrix Calculate matrix Regarding the above... Each dimension is matrix The OFDM symbol is averaged along this dimension.

[0036] S22. Then, the diagonal elements of the averaged matrix are averaged to generate the metric parameter vector. Specifically, for matrix The average of the diagonals is then calculated to obtain the final result. vector This refers to the metric parameter vector. This metric parameter vector reflects... The noise condition of each subcarrier. The metric parameter, i.e., the vector. The The elements are The metric parameter vector is composed of the metric parameters of all subcarriers.

[0037] S3. Based on the metric parameter vector, dynamically and adaptively divide all subcarriers into one or more continuous frequency segments, wherein the boundary of the division is determined by the variation characteristics of the metric parameter vector in the frequency domain.

[0038] Specifically, based on the metric parameter vector, all subcarriers are dynamically and adaptively divided into one or more consecutive frequency segments, implemented using a sliding window growth algorithm, specifically including the following steps: S31A, starting from the first subcarrier, initialize the current frequency segment. Specifically, using the above... vector ,right The subcarriers are segmented, and the specific segmentation method is as follows: Initialize the frequency band index. Subcarrier index , .

[0039] S32A, iteratively attempts to add subsequent subcarriers to the current frequency segment.

[0040] S33A. In each iteration, calculate the statistical dispersion of the metric parameters of the subcarriers within the current frequency segment; the statistical dispersion is one of variance, standard deviation, or root mean square deviation.

[0041] Specifically, for subcarrier indexing Accumulate. Calculate the subcarrier index arrive Average noise power and the variance of that segment .

[0042] S34A. If the statistical dispersion is less than a first predetermined threshold and the number of subcarriers in the current segment is less than a second predetermined threshold, then the subcarrier is formally added to the current frequency segment; otherwise, the previous subcarrier is used as the end boundary of the current frequency segment, and a new frequency segment begins with the current subcarrier. Specifically, it is determined whether the variance calculated above is less than a preset first predetermined threshold. .

[0043] If the variance is less than the first predetermined threshold and the number of subcarriers in the current segment is less than the second predetermined threshold. ,Right now or If so, the range of this frequency band will be further expanded, and the process will proceed to step S32A.

[0044] If the variance value is greater than or equal to the first predetermined threshold and If the subcarrier is not placed in the current frequency band, the current frequency band segmentation ends; and a new frequency band segmentation is performed to initialize the next frequency band. , The first predetermined threshold It can be set to K times the overall variance (Var(v)) of the metric parameter vector, where K is an empirical coefficient, typically ranging from 0.1 to 2.0. The second predetermined threshold can be set as a function of the total number of subcarriers (N), for example, max(10, N / 50), to ensure that each segment has sufficient statistical samples.

[0045] Then repeat step S32A until all subcarriers are included in a certain segment.

[0046] S4. For each frequency segment obtained by division, calculate the noise covariance matrix of the segment based on the initial noise estimation matrix corresponding to the subcarriers contained in the segment.

[0047] Specifically, the noise covariance matrix is ​​calculated for each segment based on the segmentation results of step S3 above. Assume that step five generates a total of... Each frequency segment is divided into several frequency segments, and the set of subcarriers for each frequency segment is as follows: Using step three Each dimension is matrix Perform the corresponding average, that is, the first The noise covariance matrix for each frequency segment is calculated as follows: .

[0048] In step S12, noise calculation can be implemented using other methods. One method is to subtract adjacent subcarriers of the same OFDM symbol to obtain the noise. However, it should be noted that because the subcarrier spacing specified by power line communication protocols is generally large (e.g., the specified subcarrier spacing for low-voltage power line Gaussian carrier communication protocol is 24.414 kHz), when the channel frequency selectivity is severe, the frequency responses of adjacent subcarriers of the same symbol will also differ, causing the noise calculated using this method to be larger than the actual noise. Another implementation method is to filter and denoise the least squares result (e.g., linear minimum mean square error (LMMS) filtering or transform-domain filtering), and then subtract the denoised result from the least squares result to obtain the noise.

[0049] The metric parameter used for segmentation in step S31A can also be set to other parameters, such as the average of the absolute values ​​of the off-diagonal elements, or a weighted average of the average of the absolute values ​​of the off-diagonal elements and the average of the diagonal elements. The off-diagonal elements of the noise covariance matrix reflect the interference situation.

[0050] Due to the presence of power line communication channel interference (such as single-frequency interference), before calculating the noise covariance matrix in step S4, the following steps are also included: S41, detecting the element values ​​in the initial noise estimation matrix; S42, setting or removing elements whose values ​​exceed a predetermined anomaly threshold to suppress the influence of impulse noise or single-frequency interference on the calculation of the covariance matrix. The predetermined anomaly threshold can be set to M times the median of the amplitude of the elements in the initial noise estimation matrix within the current segment, where M is typically 3 to 5 (based on the 3σ or 5σ principle of normal distribution).

[0051] Specifically, noise matrix can be used calculate In such cases, some elements exceeding a predetermined anomaly threshold can be set to zero to prevent a few outliers caused by interference from affecting the calculation of the overall covariance matrix. For example, if the third subcarrier experiences significant single-frequency interference (assuming a total of 3 OFDM symbols are involved), this leads to... , , The elements of these three matrices are much larger than the noise threshold, so these three matrices can be set to zero and not used in the calculation.

[0052] The segmentation method described in step S3 can also employ other segmentation methods. For example, a top-down recursive segmentation algorithm specifically includes the following steps: S31B: Starting from the entire metric parameter vector, if the variance (statistical dispersion) of the metric parameters of the subcarriers in the current frequency band is greater than a preset threshold, then find an optimal segmentation point in that frequency band; the optimal segmentation point is defined as: traversing all possible segmentation points k in the current frequency band, selecting the point k that minimizes the maximum variance of the two sub-frequency bands after segmentation, i.e. S32B: Divide the current frequency band into two sub-bands at the optimal dividing point; S33B: Recursively execute the dividing step S31B on each of the two sub-bands until the variance of all sub-bands is less than or equal to a preset threshold.

[0053] The method described in this embodiment is not limited to power line communication systems, but is also applicable to other communication systems (such as wireless communication systems).

[0054] Example 2: Addressing the characteristics of power line channel noise, this example proposes a method to dynamically segment the frequency band based on the current noise state, enabling the calculation of the noise covariance matrix within each band. The noise covariance matrix calculated for each band can represent the noise characteristics of that band as accurately as possible. The core innovation lies in dynamically segmenting the spectrum according to the real-time state of the power line channel and calculating the noise covariance matrix for each segment. The noise covariance matrix for each segment can accurately reflect the noise characteristics of the corresponding spectrum band.

[0055] This embodiment provides a method for dynamically segmenting frequency bands and calculating the noise covariance matrix based on the current power line communication channel noise state, as shown in Figure 2. The method includes the following steps: Step 1: Using the frequency domain transformation of the preamble sequence PREAM or training field TF symbols in the power line communication system, least squares estimation is performed to obtain a coarse channel estimation result (channel response estimation matrix) for the current channel. Since the PREAM or training field of the power line protocol is generally repeatedly transmitted in multiple identical OFDM symbols, the noise estimation module performs least squares estimation on all OFDM symbols.

[0056] Step 2: Subtract the least squares estimates of different subcarriers from the LS results obtained in Step 1 for adjacent OFDM symbols to obtain the noise of different subcarriers (initial noise estimation matrix). This method is based on the assumption that the channel changes slowly over a short period of time. To cope with possible rapid channel changes, a change detection mechanism can be introduced. For example, if the power of the calculated channel response difference (ΔH) between adjacent symbols exceeds a threshold determined based on historical noise levels, it can be determined that the channel is changing drastically at that moment. In this case, this method can be suspended, and a more robust noise estimation method (such as decision-oriented estimation based on data symbols) can be used instead, or the noise covariance matrix calculated in the previous stable period can be used directly. If MIMO transmission is involved, the above operation needs to be performed on each transmit / receive antenna pair.

[0057] The noise calculation in step two above can be implemented using other methods. For example, another method is to subtract adjacent subcarriers of the same OFDM symbol to obtain the noise. However, it should be noted that because the subcarrier spacing specified by power line communication protocols is generally large (e.g., the specified subcarrier spacing for low-voltage power line Gaussian carrier communication protocol is 24.414 kHz), when the frequency selectivity of the channel is severe, the frequency responses of adjacent subcarriers of the same symbol will also differ, causing the noise calculated using this method to be larger than the actual noise. Another implementation method is to filter and denoise the least squares result (e.g., linear minimum mean square error filtering or transform domain filtering), and then subtract the denoised and filtered result from the least squares result to obtain the noise.

[0058] Step 3: Perform further calculations using the noise matrix calculated in Step 2. For example, the initial noise estimation matrix for a single subcarrier is calculated as follows: Then calculate The above matrix The dimension is ,matrix The dimension is . and These represent the number of transmitting antennas and the number of receiving antennas, respectively. It is clear that the matrix... The diagonal elements are real numbers.

[0059] Assuming a total of OFDM symbols and There are [number] subcarriers. If the matrix described above is calculated by subtracting the same subcarriers from different OFDM symbols, a total of [number] subcarriers will be generated. Each dimension is matrix If the noise is calculated by subtracting adjacent subcarriers of the same OFDM symbol, a total of [number] noise will be generated. Each dimension is matrix Record the first The symbol of the first indivual matrix for .

[0060] It should be noted that when performing frequency segmentation of the covariance matrix and calculating the covariance matrix, it is not necessary to use all PREAM symbols or TF symbols. The specific number of symbols used depends on a series of factors such as hardware implementation speed and required result accuracy.

[0061] In the following description, unless otherwise specified, it is assumed that the matrices described above are calculated using the same subtraction method for different OFDM symbols.

[0062] Step 4: Regarding the above... Each dimension is matrix (Right now The OFDM sign dimension of the matrix is ​​taken as an arithmetic mean, and then... matrix The average of the diagonals is then calculated to obtain the final result. vector This is the metric parameter vector. This vector reflects... The noise condition of each subcarrier. The metric parameter, i.e., the vector. The The values ​​of each element are

[0063] Step 5: As shown in Figure 3, using the above... vector ,right Each subcarrier is segmented, and the specific segmentation method is as follows: (a) Initialize the frequency band index Subcarrier index , .

[0064] (b) Subcarrier indexing Accumulate. Calculate the subcarrier index arrive Average noise power and the variance of that segment

[0065] (c) Determine whether the variance value calculated above is less than a pre-set first predetermined threshold. .

[0066] If the variance is less than the first predetermined threshold and the number of subcarriers in the current segment is less than the second predetermined threshold. ,Right now or If so, the range of that frequency band will be further expanded, proceeding to step (b).

[0067] If the absolute value of the difference is greater than or equal to the first predetermined threshold and If the subcarrier is not placed in the current frequency band, the current frequency band segmentation ends, a new frequency band segmentation begins, and the next frequency band is initialized. , Then repeat step (b) until all subcarriers are included in a segment.

[0068] The above noise difference first predetermined threshold The specific values ​​were obtained through extensive simulations; the second predetermined threshold for the number of subcarriers mentioned above. The introduction of this is to ensure that a frequency band segment contains at least [missing information]. Using multiple subcarriers avoids generating too many frequency segments while ensuring that the statistics for each segment are not too few.

[0069] Step Six: Assuming that the above process generates a total of Each frequency segment is divided into several frequency segments, and the set of subcarriers for each frequency segment is as follows: It is obvious , This indicates the number of elements in the set. Using step three... Each dimension is matrix Perform the corresponding average, that is, the first The noise covariance matrix for each frequency segment is calculated as follows: .

[0070] This invention proposes a method for dynamically segmenting a frequency band and calculating the noise covariance matrix based on the real-time noise status of the current power line communication channel. This method can calculate the noise covariance matrix by segmentation, so that the covariance matrix can accurately describe the noise status within the frequency band. The specific effects are as follows: (1) By dynamically adjusting the calculation of the signal covariance matrix, the calculated matrix can better describe the noise status of the corresponding frequency band segment, ensuring the subsequent whitening and further demodulation performance; (2) When the noise in the allocated power line communication frequency band has the characteristics of white noise in the frequency domain (for example, when the allocated communication frequency band is narrow), that is, the noise power of each subcarrier in the communication frequency band is roughly the same, a noise covariance matrix can be calculated in this way, and the covariance matrix can reflect the noise situation in the entire communication frequency band; (3) When the noise in the allocated power line communication frequency band exhibits the characteristics of colored noise, that is, when the noise power spectrum is different in different frequency segments, the communication frequency band can be segmented using the method of this embodiment, and then a noise covariance matrix can be calculated for each segment for the subsequent whitening process.

[0071] Example 3: This example describes how the above noise estimation method is applied to various communication devices, chips, storage media, and receivers. Specifically, it includes a communication device, which may be a power line modem, router, base station, user terminal equipment, etc. The device includes: one or more processors (such as DSP, CPU, ASIC); and a memory (such as ROM, RAM, Flash) for storing one or more programs (containing instructions to execute the method described in Example 1 or 2 above); when the program is executed by the processor, the communication device implements the aforementioned noise estimation method. The communication device also includes necessary hardware modules such as a radio frequency front-end, analog-to-digital converter, and power line coupling circuit.

[0072] A chip, which may be a baseband processing chip, a communication SoC, etc., includes processing circuitry (such as a hardware accelerator, programmable logic, or a processor core), which is configured to perform the aforementioned noise estimation method (through hardware logic or by executing firmware / software).

[0073] A computer-readable storage medium, such as a USB flash drive, portable hard drive, read-only memory, random access memory, magnetic disk, or optical disk, etc., stores a computer program / instructions thereon, which, when executed by a processor, implements the aforementioned noise estimation method.

[0074] A receiver in a wireless communication system, such as a Wi-Fi receiver or a 5G NR receiver. This receiver employs the method described in Embodiment 1 or 2 above for noise estimation and whitening to improve reception performance in non-uniform noise environments.

[0075] Experimental Example: A method for dynamically adjusting the frequency segmentation of the noise covariance matrix based on the current noise state characteristics of the power line communication channel. This considers data transmission using 411 data subcarriers (from subcarrier index 80 to carrier index 490) in the State Grid HPLC band 0 (2 - 12 MHz). For ease of description, it is assumed that... MIMO system, TF has One OFDM symbol. A specific implementation is as follows: Step 1: Time-domain data of different symbols of the TF in power line carrier communication. Perform Fourier transform to receive data in the frequency domain The least-squares (LS) results of the frequency domain channel response for different OFDM symbols of the TF are obtained by dividing the known frequency domain transmitted data by the frequency domain received data. Note the above. For size The matrix.

[0076] Step 2: Considering the slow time-domain variation characteristic of general power line communication, the noise of the subcarrier is calculated using the least-squares LS results of the same subcarrier of adjacent OFDM symbols in the frequency domain response described above. .

[0077] Step 3: Use the noise matrix calculated in Step 2 Perform further calculations The above matrix The dimension is ,matrix The dimension is For time-varying wireless communication systems, the time-varying rate of the channel can be estimated first (e.g., by estimating the correlation through adjacent symbol channels). If the time-varying rate is below a certain threshold, the difference method is used; otherwise, other time-resistant noise estimation methods are used (as described in Example 1).

[0078] Step 4: For the matrix calculated above The average value obtained by averaging according to the following formula is: real row vectors ,vector The The calculation method for each element is as follows: .

[0079] Step 5: Using the above vector The 411 subcarriers are segmented, and the specific segmentation method is as follows: (5.a) Initialize the frequency band index Subcarrier index , .

[0080] (5.b) Subcarrier indexing Accumulate. Calculate the subcarrier index arrive Average noise power and the variance of that segment .

[0081] (5.c) Determine the variance value calculated above. Is it less than a pre-set first predetermined threshold? .

[0082] If the variance calculated above is less than the first predetermined threshold and the number of subcarriers in the current segment is less than the second predetermined threshold. ,Right now or If so, the range of that frequency band will be further expanded, proceeding to step (5.b).

[0083] If the variance value calculated above is greater than or equal to the first predetermined threshold and If the subcarrier is not placed in the current frequency band, the current frequency band segmentation ends, a new frequency band segmentation begins, and the next frequency band is initialized. , Then repeat step (5.b) until all 411 subcarriers are included in a segment.

[0084] Step Six: Assuming that the process described in Step Five above has generated a total of The frequency is divided into five segments, and the subcarrier range of each segment is as follows: Then use the result obtained in step three. Calculate the noise covariance matrix for each segment: .

[0085] Experimental Results: The existing technology uses fixed uniform frequency segmentation, as shown in Figure 4. This involves dividing the signal into three segments, and it can be seen that the noise power of each segment differs significantly, indicating that the segmentation is not ideal. In contrast, the experimental results of this embodiment show that the noise power difference between each segment is significantly smaller, demonstrating a more reasonable segmentation, as shown in Figure 5.

[0086] Experimental verification shows that the advantages of this embodiment mainly include: (1) Dynamically segmenting the frequency band based on noise status to calculate the noise covariance matrix: This embodiment adjusts the frequency segmentation for calculating the covariance matrix based on the power line communication noise status to calculate the covariance matrix that can reflect the noise characteristics of the frequency band, thus providing a guarantee for subsequent demodulation performance; (2) Architectural versatility: This embodiment does not depend on a specific frequency, avoiding the need to develop independent algorithms for each frequency band and reducing deployment complexity.

[0087] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, several equivalent substitutions or obvious modifications can be made without departing from the concept of the present invention, and all such modifications, achieving the same performance or application, should be considered within the scope of protection of the present invention.

Claims

1. A noise estimation method for an orthogonal frequency division multiplexing communication system, characterized in that, The steps include: S1. Calculate the initial noise estimation matrix for each subcarrier based on the received training sequence; S2. Based on the initial noise estimation matrix, a metric parameter vector characterizing the noise characteristics of each subcarrier is calculated, wherein the dimension of the metric parameter vector is equal to the number of subcarriers; S3. According to the metric parameter vector, all subcarriers are dynamically and adaptively divided into one or more continuous frequency segments, wherein the boundary of the division is determined by the variation characteristics of the metric parameter vector in the frequency domain. S4. For each frequency segment obtained by division, calculate the noise covariance matrix of the segment based on the initial noise estimation matrix corresponding to the subcarriers contained in the segment.

2. The method according to claim 1, characterized in that, Step S1 specifically includes the following steps: S11, performing least squares channel estimation on the training sequences received on different orthogonal frequency division multiplexing symbols to obtain multiple channel response estimation matrices; S12, subtracting the least squares estimation results of different subcarriers of adjacent OFDM symbols one by one to obtain the initial noise estimation matrices of different subcarriers.

3. The method according to claim 1, characterized in that, In step S2, the metric parameter vector is a scalar set that reflects the noise power or interference level of each subcarrier, calculated based on the initial noise estimation matrix. The calculation of the metric parameter vector specifically includes the following steps: S21, averaging the multiple initial noise estimation matrices along the symbol dimension; S22, averaging the main diagonal elements of the averaged matrix to generate the metric parameter vector.

4. The method according to claim 1, characterized in that, In step S3, based on the metric parameter vector, all subcarriers are dynamically and adaptively divided into one or more continuous frequency segments using a sliding window growth algorithm. Specifically, this includes the following steps: S31A: Initialize the current frequency segment starting from the first subcarrier; S32A: Iteratively attempt to add subsequent subcarriers to the current frequency segment; S33A: In each iteration, calculate the statistical dispersion of the metric parameters of the subcarriers within the current frequency segment; the statistical dispersion is one of variance, standard deviation, or root mean square deviation; S34A: If the statistical dispersion is less than a first predetermined threshold, and the number of subcarriers in the current segment is less than a second predetermined threshold, then the subcarrier is formally added to the current frequency segment; otherwise, the previous subcarrier is used as the end boundary of the current frequency segment, and a new frequency segment begins with the current subcarrier.

5. The method according to claim 1, characterized in that, In step S3, based on the metric parameter vector, all subcarriers are dynamically and adaptively divided into one or more continuous frequency segments using a recursive segmentation algorithm. Specifically, this includes the following steps: S31B: If the statistical dispersion of the metric parameters of the subcarriers in the current frequency band is greater than a predetermined threshold, then an optimal segmentation point is found within that frequency band; S32B: The current frequency band is divided into two sub-frequency bands using the optimal segmentation point; S33B: The segmentation steps are recursively performed on the two sub-frequency bands until the statistical dispersion of all sub-frequency bands is less than or equal to the predetermined threshold.

6. The method according to claim 1, characterized in that, In step S4, before calculating the noise covariance matrix, the following steps are also included: S41, detecting the element values ​​in the initial noise estimation matrix; S42, setting or removing elements whose values ​​exceed a predetermined abnormal threshold to suppress the influence of impulse noise or single-frequency interference on the calculation of the covariance matrix.

7. The method according to any one of claims 1 to 6, characterized in that, The orthogonal frequency division multiplexing communication system is a power line carrier communication system.

8. A communication device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the communication device implements the method as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instruction is executed by the processor, it implements the method as described in any one of claims 1 to 7.

10. A receiver in a wireless communication system, characterized in that, Noise estimation and whitening are performed using the method described in any one of claims 1 to 7.

Images