Maritime communication anti-fading optimization method and system based on signal data processing

Abstract

This invention relates to the field of wireless communication transmission monitoring technology, specifically to a method and system for optimizing anti-fading in maritime communication based on signal data processing. The scheme includes: extracting pilot sequences from segmented baseband signals; performing a Fast Fourier Transform on the pilot sequences to extract frequency-domain Doppler broadening feature vectors, and performing time-domain cross-correlation operations to extract time-delay power spectrum feature vectors; concatenating the two feature vectors according to a preset dimension to generate a two-dimensional channel fading state matrix; and using the aforementioned order and step size factor to jointly eliminate multipath interference and frequency shift distortion. This ensures real-time matching between the equalizer order and the two-dimensional fading state of the channel, eliminating redundant tap noise and truncation errors. The nonlinear step size mechanism based on matrix distance avoids the divergence risk of fixed step sizes during channel fluctuations and the slow convergence problem during stable conditions, improving the tracking and compensation capability for non-stationary, fast-time-varying channels.

Description

Technical Field

[0001] This invention discloses a method and system for optimizing anti-fading in maritime communication based on signal data processing, which relates to the field of wireless communication transmission monitoring technology. Background Technology

[0002] In the field of anti-fading technology for maritime communication, existing technologies generally employ time-domain adaptive equalizers to process the received baseband signal. The conventional approach involves the receiver segmenting the baseband signal to obtain a known pilot sequence. This pilot sequence is then cross-correlated with a locally stored reference pilot sequence in the time domain to extract the arrival time and relative amplitude of each multipath component, which is used as the multipath delay estimation result. Based on this delay estimation result, the system sets a fixed filter order during the initialization phase and constructs a decision feedback equalizer. In the equalizer coefficient iteration stage, the conventional approach uses the minimum mean square error algorithm, updating the tap coefficients with a constant step size or a variable step size factor calculated solely based on the equalizer output error signal. Throughout the entire processing flow, the equalizer's topology and step size adjustment are based solely on the time-domain delay parameters.

[0003] Maritime communication environments are affected by wave fluctuations and the relative motion of ships, resulting in a coupled characteristic of large-scale Doppler frequency shift broadening and dynamic changes in multipath delay. Existing technologies treat delay estimation and frequency shift suppression separately, employing fixed-order or variable-step-size mechanisms that rely solely on time-domain parameters. This leads to a mismatch between the equalizer topology and the actual two-dimensional fading characteristics of the channel. In scenarios where the number of multipath paths dynamically increases or decreases while Doppler frequency shift deteriorates, existing technologies cannot adjust the equalizer order based on the physical quantity of frequency-domain broadening. This causes the equalizer to introduce redundant tap noise when facing sparse scattering in the channel and truncation errors when facing dense multipath paths, ultimately leading to the failure of joint fading suppression. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for optimizing anti-fading in maritime communication based on signal data processing, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for optimizing anti-fading in maritime communication based on signal data processing includes: segmenting the baseband signal at the receiving end and extracting the pilot sequence from each segment; The pilot sequence is subjected to Fast Fourier Transform to extract the frequency domain Doppler broadening feature vector, and the pilot sequence is subjected to time domain cross-correlation operation to extract the time delay power spectrum feature vector; The frequency domain Doppler broadened feature vector and the time delay power spectrum feature vector are concatenated according to a preset dimension to generate a two-dimensional channel fading state matrix. The feedforward filter update order of the decision feedback equalizer is determined based on the extreme value of the frequency domain Doppler broadening eigenvector in the two-dimensional channel fading state matrix, and the feedback filter update order is determined based on the number of multipath components in the time delay power spectrum eigenvector. Calculate the Euclidean distance between the current two-dimensional channel fading state matrix and the previous two-dimensional channel fading state matrix, and use the nonlinear function of the Euclidean distance as the step size factor of the minimum mean square error algorithm at the current time. The combined elimination of multipath interference and frequency shift distortion in the baseband signal is achieved by using the updated order of the feedforward filter, the updated order of the feedback filter, and the step size factor.

[0006] Preferably, the step of extracting the frequency domain Doppler broadening feature vector by performing a fast Fourier transform on the pilot sequence includes: performing a fast Fourier transform on the pilot sequence to obtain a frequency domain sequence; Calculate the power spectral density at each frequency point in the frequency domain sequence and construct a power spectral density curve; Determine the noise floor mean in the power spectral density curve, and multiply the noise floor mean by a preset factor to obtain a decision threshold; Extract the set of frequency points above the decision threshold from the power spectral density curve, calculate the frequency difference between the maximum and minimum frequency points in the set, and combine the frequency difference and the power value of the set to generate the frequency domain Doppler broadening feature vector.

[0007] Preferably, the step of performing time-domain cross-correlation operation on the pilot sequence to extract the time-delay power spectrum feature vector includes: performing cross-correlation operation on the pilot sequence and the locally stored reference pilot sequence to obtain the time-delay cross-correlation function curve; Envelope detection is performed on the time-delay cross-correlation function curve to extract the envelope peak value; Based on the envelope peak value, a peak decision threshold is set, and multiple peak points higher than the peak decision threshold are selected from the time delay cross-correlation function curve; Record the time delay value and relative amplitude value corresponding to each peak point, remove peak points whose amplitude value is lower than the preset interference removal threshold, and arrange the time delay value and relative amplitude value of the remaining peak points in chronological order to generate the time delay power spectrum feature vector.

[0008] Preferably, the step of concatenating the frequency domain Doppler broadened feature vector and the time delay power spectrum feature vector according to a preset dimension to generate a two-dimensional channel fading state matrix includes: obtaining the first dimension length of the frequency domain Doppler broadened feature vector and the second dimension length of the time delay power spectrum feature vector; When the length of the first dimension is inconsistent with the length of the second dimension, the feature vector of the shorter dimension is padded with zero values ​​until the lengths of the two dimensions are equal. The padded frequency domain Doppler broadened eigenvectors are used as the row vectors of the matrix, and the padded time delay power spectrum eigenvectors are used as the column vectors of the matrix. The two-dimensional channel fading state matrix is ​​generated by vector outer product operation.

[0009] Preferably, the step of determining the feedforward filter update order of the decision feedback equalizer based on the extreme value of the frequency domain Doppler broadening feature vector in the two-dimensional channel fading state matrix, and determining the feedback filter update order based on the number of multipath components in the time delay power spectrum feature vector includes: extracting the maximum and minimum values ​​corresponding to the row vectors in the two-dimensional channel fading state matrix, and calculating the Doppler broadening range; The Doppler widening range is input into a pre-constructed piecewise function of feedforward order mapping, and the updated order of the feedforward filter is output. The number of non-zero elements corresponding to the column vectors in the two-dimensional channel fading state matrix is ​​used as the number of multipath components. The number of the multipath components is directly assigned as the update order of the feedback filter.

[0010] Preferably, the step of calculating the Euclidean distance between the current two-dimensional channel fading state matrix and the previous two-dimensional channel fading state matrix, and using the nonlinear function of the Euclidean distance as the step size factor of the minimum mean square error algorithm at the current time, includes: taking the difference between the elements at the same position in the current two-dimensional channel fading state matrix and the previous two-dimensional channel fading state matrix and calculating the sum of squares, and taking the square root of the sum of squares to obtain the Euclidean distance. The Euclidean distance is input into a nonlinear mapping function containing an exponential decay term. The nonlinear mapping function is used to output a value that approaches a preset maximum initial step size when the Euclidean distance increases and a value that approaches a preset minimum convergence step size when the Euclidean distance decreases. The output value of the nonlinear mapping function is used as the step size factor.

[0011] Preferably, before performing a fast Fourier transform on the pilot sequence to obtain a frequency domain sequence, the method further includes: applying a Hamming window function to the pilot sequence for windowing truncation. The length of the Hamming window function is dynamically set according to the coherence time of the maritime communication link. The coherence time is obtained by reciprocal calculation of the absolute value of the Doppler frequency shift measured in advance by the receiver. By applying the Hamming window function to suppress the truncation effect at both ends of the pilot sequence, the interference of sidelobe energy in the frequency domain sequence on the power spectral density curve is reduced, so that the division of the decision threshold avoids the frequency band corresponding to the sidelobe energy.

[0012] Preferably, after outputting the feedforward filter update order and assigning the feedback filter update order, the method further includes: determining whether the feedforward filter update order or the feedback filter update order output at the current time is greater than the historical order corresponding to the previous time. If the current output of the feedforward filter update order or the feedback filter update order is greater than the corresponding historical order at the previous time, then all filter tap coefficients corresponding to the historical order at the previous time are retained, and the coefficients of the newly added tap positions are initialized to zero. If the current feedforward filter update order or feedback filter update order is less than the corresponding historical order at the previous time, then the tap coefficient at the center position of the historical order at the previous time is taken as the initial coefficient at the current time.

[0013] Preferably, after using the output value of the nonlinear mapping function as the step size factor, the method further includes: calculating the sum of the absolute values ​​of all elements in the two-dimensional channel fading state matrix at the current time as the channel state energy value; The channel state energy value is compared with a pre-stored stable channel energy threshold. If the channel state energy value is greater than the stable channel energy threshold, then the step size factor is restricted to between the preset upper limit of the first step size and the lower limit of the first step size. If the channel state energy value is less than or equal to the stable channel energy threshold, the step size factor is restricted to a preset second step size upper limit value and a second step size lower limit value, wherein the first step size upper limit value is less than the second step size upper limit value.

[0014] A marine communication anti-fading optimization system based on signal data processing includes: a signal interception device for segmenting the baseband signal at the receiving end and extracting the pilot sequence from each segment of the signal; The feature extraction device is used to perform a fast Fourier transform on the pilot sequence to extract the frequency domain Doppler broadening feature vector, and to perform a time domain cross-correlation operation on the pilot sequence to extract the time delay power spectrum feature vector. A matrix construction device is used to concatenate the frequency domain Doppler broadened feature vector and the time delay power spectrum feature vector according to a preset dimension to generate a two-dimensional channel fading state matrix. The order mapping device is used to determine the update order of the feedforward filter of the decision feedback equalizer based on the extreme value of the frequency domain Doppler broadening feature vector in the two-dimensional channel fading state matrix, and to determine the update order of the feedback filter based on the number of multipath components in the time delay power spectrum feature vector. A step size generation device is used to calculate the Euclidean distance between the current two-dimensional channel fading state matrix and the previous two-dimensional channel fading state matrix, and to use the nonlinear function of the Euclidean distance as the step size factor of the minimum mean square error algorithm at the current time. An equalization processing device is used to jointly eliminate multipath interference and frequency shift distortion in the baseband signal by using the updated order of the feedforward filter, the updated order of the feedback filter, and the step size factor.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention extracts the frequency domain Doppler broadening feature vector and the time-delay power spectrum feature vector, concatenates them along a preset dimension, and generates a two-dimensional channel fading state matrix using vector outer product. Based on the range of the row vectors and the number of non-zero elements in the column vectors of this matrix, the update orders of the feedforward filter and the feedback filter are dynamically determined. This processing method directly introduces the frequency shift physical quantity in the frequency domain into the topology decision of the time-domain equalizer, enabling the equalizer order to form a real-time matching relationship with the actual number of multipath components and the degree of frequency shift broadening in the maritime channel. This eliminates the truncation error and redundant tap noise caused by a fixed order when the channel state changes dynamically, and solves the problem of joint fading suppression failure caused by the mismatch between the equalizer structure and the two-dimensional fading characteristics.

[0016] 2. This invention calculates the Euclidean distance between the two-dimensional channel fading state matrices of adjacent time points, uses its nonlinear function as a step size factor, and applies upper and lower limits to the step size factor in different intervals based on the comparison between the channel state energy value and the stationary channel energy threshold. This avoids the divergence risk of a constant step size during channel fluctuations and the slow convergence problem during stationary periods. During order switching, when the order increases, historical tap coefficients are retained and new positions are initialized to zero; when the order decreases, the center position coefficient is truncated as the initial value, avoiding signal distortion and prolonged convergence period caused by coefficient resetting. By applying a window function dynamically set according to the coherence time before extracting the pilot sequence, sidelobe energy interference caused by truncation effect is suppressed, improving the purity of feature extraction. Attached Figure Description

[0017] Figure 1 The main flowchart is shown below for the anti-fading optimization method for maritime communication based on signal data processing. Figure 2 Flowchart for extracting feature vectors for frequency domain Doppler stretching; Figure 3 Here is a flowchart for extracting the feature vector of the time-delay power spectrum; Figure 4 Flowchart for constructing and mapping the order of a two-dimensional channel fading state matrix; Figure 5 Here is a flowchart of the step size factor generation and constraint process; Figure 6 This is a block diagram of a marine communication anti-fading optimization system based on signal data processing. Detailed Implementation

[0018] Please refer to Figure 1 This embodiment provides a method and system for optimizing anti-fading in maritime communication based on signal data processing. The receiving end acquires a baseband signal down-converted after transmission through a maritime wireless channel. The baseband signal is encapsulated using a preset frame structure. A known pilot sequence is embedded in the frame header of each data frame. This pilot sequence is a pseudo-random sequence with sharp autocorrelation peaks to reduce noise interference during channel estimation. The received baseband signal is segmented, with the segmentation operation performed according to the frame period of the baseband signal. The time length of each segment is consistent with the time length of a single data frame to ensure that each segment contains a complete frame header structure and the corresponding pilot sequence. After segmentation, a frame synchronization operation is performed on each segment. The frame synchronization operation uses a sliding correlation method, performing a sliding cross-correlation operation between the received segmented baseband signal and a locally stored reference pilot sequence. When the peak value obtained from the cross-correlation operation exceeds a preset synchronization threshold, the start position of the frame header is detected. Based on the frame header start position obtained from frame synchronization detection, a signal segment with the same length as the preset pilot sequence is extracted from the segmented signal. This signal segment is the pilot sequence corresponding to the segment.

[0019] Two parallel feature extraction operations are performed on the extracted pilot sequence. The first type of feature extraction operation targets the Doppler broadening feature in the frequency domain. It performs a Fast Fourier Transform on the pilot sequence to transform the time-domain pilot sequence into the frequency domain, obtaining the corresponding frequency-domain sequence. The frequency-domain Doppler broadening feature vector is then extracted based on the frequency-domain sequence. The second type of feature extraction operation targets the multipath delay feature in the time domain. It performs a time-domain cross-correlation operation on the pilot sequence, cross-correlates the pilot sequence with the locally stored reference pilot sequence, obtains the time-delay cross-correlation function, and extracts the time-delay power spectrum feature vector based on the time-delay cross-correlation function.

[0020] The obtained frequency domain Doppler broadening feature vector and time delay power spectrum feature vector are subjected to dimension matching processing. The two feature vectors are processed according to the preset dimension requirements to make the dimension length of the two feature vectors consistent. The dimension-matched frequency domain Doppler broadening feature vector and time delay power spectrum feature vector are concatenated, and a two-dimensional channel fading state matrix is ​​generated by vector outer product operation. The row dimension of the two-dimensional channel fading state matrix corresponds to the frequency domain Doppler broadening feature, and the column dimension corresponds to the time domain multipath delay feature, realizing the joint representation of the two-dimensional fading features in the frequency domain and time domain of the maritime channel.

[0021] Based on the generated two-dimensional channel fading state matrix, a dynamic decision operation on the filter order of the decision feedback equalizer is performed. The extreme values ​​corresponding to the row vectors in the two-dimensional channel fading state matrix are extracted, and the dynamic range of Doppler broadening is calculated based on these extreme values. The update order of the feedforward filter in the decision feedback equalizer is determined according to the dynamic range of Doppler broadening. The number of non-zero elements corresponding to the column vectors in the two-dimensional channel fading state matrix is ​​counted. This number of non-zero elements corresponds to the number of multipath components in the maritime channel, and the counted number of multipath components is used as the update order of the feedback filter in the decision feedback equalizer.

[0022] Perform the adaptive step size factor generation operation. Obtain the two-dimensional channel fading state matrix generated at the current time and the two-dimensional channel fading state matrix stored at the previous time, and calculate the Euclidean distance between the two matrices. The Euclidean distance is used to characterize the degree of change in the maritime channel state between adjacent time points. Input the calculated Euclidean distance into a preset nonlinear function to obtain the output value of the nonlinear function. Use this output value as the step size factor of the minimum mean square error algorithm at the current time. The output characteristics of the nonlinear function are: when the input Euclidean distance increases, the output value approaches the preset maximum initial step size; when the input Euclidean distance decreases, the output value approaches the preset minimum convergence step size.

[0023] Based on the obtained update orders of the feedforward filter and the feedback filter, as well as the step size factor, the coefficients of the decision feedback equalizer are updated and the signal is equalized, achieving joint elimination of multipath interference and frequency shift distortion in the baseband signal. The decision feedback equalizer consists of two branches: a feedforward filter and a feedback filter. The number of taps in the feedforward filter is set to the obtained update order, and the number of taps in the feedback filter is set to the obtained update order. Using the least mean square error algorithm, the tap coefficients of the feedforward and feedback filters are iteratively updated based on the obtained step size factor. The received baseband signal is input into the coefficient-updated decision feedback equalizer. The feedforward filter compensates for frequency shift distortion in the baseband signal, and the feedback filter eliminates multipath interference, ultimately yielding the equalized baseband signal.

[0024] Table 1. Parameter Configuration Table for Segmenting and Extracting Baseband Signals in Marine Communication

[0025] Table 1 shows the configuration of baseband signal segmentation parameters for different maritime channel scenarios. The configuration of segment length and pilot sequence length is determined based on the dynamic changes of the channel. In scenarios with high channel dynamics, a longer pilot sequence length is used to improve the accuracy of channel feature extraction. The starting offset of the segmentation is set to 0, corresponding to the starting position of the frame header, to ensure that the segmented pilot sequence is synchronized with the preset reference pilot sequence.

[0026] This embodiment achieves data source acquisition for maritime channel feature extraction by segmenting and extracting the baseband signal and pilot sequence; it obtains two types of feature vectors characterizing channel Doppler broadening and multipath delay through parallel feature extraction in the frequency and time domains; it achieves joint characterization of two-dimensional channel fading features by constructing a two-dimensional channel fading state matrix; it completes dynamic decision-making of the order of feedforward and feedback filters by using the row and column vector features of the matrix; it generates an adaptive step size factor by using the Euclidean distance of the channel matrix at adjacent time points; and finally, it achieves joint elimination of multipath interference and frequency shift distortion in the baseband signal through a decision feedback equalizer, enabling the equalizer's topology and iteration step size to match the dynamic fading characteristics of the maritime channel in real time.

[0027] In a preferred embodiment, reference Figure 2 Before performing a Fast Fourier Transform on the pilot sequence, a windowing truncation process is first applied. This windowing truncation uses a Hamming window function to suppress the truncation effect at both ends of the pilot sequence, reducing the interference of sidelobe energy in the frequency domain sequence on subsequent power spectral density calculations. The length of the Hamming window function is dynamically set based on the coherence time of the maritime communication link. This coherence time is obtained by reciprocal calculation of the absolute value of the Doppler frequency shift measured beforehand at the receiver. The calculation of the coherence time satisfies… ,in For the coherence time of maritime communication links, This represents the Doppler frequency shift value measured at the receiver. The length of the Hamming window function is set to be no greater than the number of symbols corresponding to the coherence time, to ensure that the channel state corresponding to the windowed pilot sequence remains stable within the coherence time, avoiding the influence of channel time-varying characteristics on the feature extraction results. The discrete form of the Hamming window function satisfies:

[0028] in, For the first The window function values ​​corresponding to each discrete time point. The length of the Hamming window function. This is a discrete-time index. The extracted pilot sequence is multiplied by a Hamming window function to obtain a windowed pilot sequence. By applying the Hamming window function, the signal amplitude at both ends of the pilot sequence is smoothly attenuated, suppressing the Gibbs effect caused by rectangular window truncation and reducing the amplitude of sidelobe energy in the frequency domain sequence. This allows the decision threshold of the noise floor in the subsequent power spectral density curve to avoid the frequency band corresponding to the sidelobe energy, thus improving the purity of frequency domain feature extraction.

[0029] Perform a Fast Fourier Transform (FFT) on the windowed pilot sequence to extract the frequency domain Doppler broadening feature vector. Point Fast Fourier Transform, The value of is consistent with the length of the pilot sequence. Converting the time-domain pilot sequence to a frequency-domain sequence, the discrete Fourier transform operation satisfies:

[0030] in, This is the pilot sequence after windowing. For the first The frequency domain sequence values ​​corresponding to each frequency point. For frequency point index, The length of the pilot sequence. The imaginary unit is used. Based on the obtained frequency domain sequence, the power spectral density corresponding to each frequency point is calculated. The power spectral density calculation is implemented using the periodogram method, and the operation satisfies:

[0031] in, For the first The power spectral density value corresponding to each frequency point Frequency domain sequence The magnitude of the power spectral density is calculated. Based on the power spectral density values ​​at all frequencies, a power spectral density curve is constructed with the frequency index on the x-axis and the power spectral density value on the y-axis.

[0032] Further, the noise floor mean in the power spectral density curve is determined. Power spectral density values ​​corresponding to a predetermined proportion of frequency points at both ends of the curve are taken (the predetermined proportion is set to 10%). The arithmetic mean of these power spectral density values ​​is calculated and used as the noise floor mean. The noise floor mean is multiplied by a predetermined factor to obtain the power spectral density decision threshold. The predetermined factor is set based on the noise level of the maritime channel, with a default value of 3. All frequency points with power spectral density values ​​higher than the decision threshold are extracted from the power spectral density curve to form an effective frequency point set. The frequency difference between the maximum and minimum frequency points in the effective frequency point set is calculated; this frequency difference corresponds to the Doppler broadening width of the maritime channel. The calculated frequency difference and the power spectral density values ​​corresponding to each frequency point in the effective frequency point set are arranged in ascending order of frequency point index to generate a frequency domain Doppler broadening feature vector.

[0033] Perform time-domain cross-correlation on the pilot sequence to extract the time-delay power spectrum feature vector. Cross-correlate the extracted pilot sequence with the locally stored reference pilot sequence to obtain the time-delay cross-correlation function curve. The time-domain cross-correlation operation satisfies:

[0034] in, For delay index The corresponding cross-correlation function value, The pilot sequence extracted by the receiving end. The reference pilot sequence is stored locally. This indicates that the complex conjugation operation is performed on the sequence. For delay index, denoted as the length of the pilot sequence. Envelope detection is performed on the obtained time-delay cross-correlation function curve using Hilbert transform. The envelope of the time-delay cross-correlation function curve is extracted, resulting in the envelope curve. All local peaks in the envelope curve are extracted to obtain the envelope peak set. The rule for determining local peaks is: if the envelope value at a certain position is greater than the envelope values ​​of its two adjacent positions, then that position is determined to be a local peak point.

[0035] Further, based on the maximum peak value in the envelope peak set, a peak decision threshold is set. This threshold is the maximum envelope peak value multiplied by a preset scaling factor, which defaults to 0.1. In the envelope curve, all peak points with envelope values ​​higher than the peak decision threshold are selected. The time delay value and relative amplitude value corresponding to each peak point are recorded. The relative amplitude value is the ratio of the envelope value of that peak point to the maximum envelope peak value. A preset interference rejection threshold is set, which is lower than the peak decision threshold (default value is 0.05 of the maximum envelope peak value). Peak points with relative amplitude values ​​lower than the interference rejection threshold are rejected to eliminate spurious peaks caused by noise. The time delay values ​​and relative amplitude values ​​corresponding to the remaining peak points are arranged in ascending order of time delay value to generate a time delay power spectrum feature vector.

[0036] The frequency-domain Doppler broadened feature vector and the time-delay power spectrum feature vector are concatenated according to a preset dimension to generate a two-dimensional channel fading state matrix. The first dimension length of the frequency-domain Doppler broadened feature vector and the second dimension length of the time-delay power spectrum feature vector are obtained. The first and second dimension lengths are compared. When the two dimension lengths are inconsistent, zero-value padding is performed on the shorter feature vector, adding zero-value elements to the end of the feature vector until the two feature vectors have equal dimension lengths, resulting in two dimension-matched feature vectors. The padded frequency-domain Doppler broadened feature vector is used as a row vector, and the padded time-delay power spectrum feature vector is used as a column vector. A two-dimensional channel fading state matrix is ​​generated through a vector outer product operation, where the vector outer product operation satisfies:

[0037] in, The generated two-dimensional channel fading state matrix, The padded frequency domain Doppler broadened feature vector has a dimension of . , This is the filled time-delay power spectrum feature vector, with dimension . , for The transpose of the vector, with dimension , The dimension of the generated two-dimensional channel fading state matrix is ​​the length of the feature vector after dimension matching. .

[0038] Table 2. Hamming window length and windowing performance parameters corresponding to different coherence times.

[0039] Table 2 shows the performance parameters of the Hamming window length configuration and windowing processing for different coherence times. The coherence time is inversely proportional to the absolute value of the Doppler frequency shift. The Hamming window length is set to half of the coherence time to ensure that the pilot sequence after windowing processing is within the coherence time of the channel. The sidelobe suppression ratio of the Hamming window is stable at 43dB, which can effectively suppress the sidelobe energy interference caused by the truncation effect. The main lobe width is maintained at 2 frequency points, which will not have a significant broadening effect on the spectral characteristics of the effective signal.

[0040] This embodiment suppresses sidelobe energy interference caused by truncation by windowing the pilot sequence using a Hamming window function with dynamically set length based on coherence time. The extraction process of the frequency domain Doppler broadening feature vector is refined through Fast Fourier Transform and power spectral density calculation, achieving accurate characterization of channel Doppler broadening characteristics. The extraction process of the time delay power spectral feature vector is refined through time domain cross-correlation and envelope detection, achieving accurate extraction of channel multipath delay characteristics. Finally, the construction of a two-dimensional channel fading state matrix is ​​completed through dimension matching and vector outer product operations, realizing a joint characterization of the two-dimensional fading characteristics in the frequency and time domains of the maritime channel, improving the completeness and accuracy of the channel state description.

[0041] In a preferred embodiment, reference Figure 3 Based on the generated two-dimensional channel fading state matrix, the order mapping operation of the feedforward filter and feedback filter of the decision feedback equalizer is performed. Row vectors are extracted from the two-dimensional channel fading state matrix, and these row vectors correspond to the padded frequency domain Doppler broadening feature vectors. The maximum and minimum values ​​in the row vectors are extracted, and the difference between the maximum and minimum values ​​is calculated to obtain the Doppler broadening range. The calculation of the Doppler broadening range satisfies the following:

[0042] in, To widen the range for Doppler, This represents the maximum value corresponding to the row vector in the two-dimensional channel fading state matrix. This represents the minimum value corresponding to the row vector in the two-dimensional channel fading state matrix. The Doppler broadening range is used to characterize the dynamic range of the Doppler frequency shift in the maritime channel. The larger the range, the more severe the Doppler broadening of the channel and the stronger the time-varying characteristics of the channel.

[0043] Furthermore, the calculated Doppler broadening range is input into a pre-constructed piecewise feedforward order mapping function, which outputs the updated order of the feedforward filter. The piecewise feedforward order mapping function satisfies:

[0044] in, Let this be the update order of the feedforward filter. , , The preset segmentation threshold, , , , The order of the feedforward filter corresponding to the Doppler widening range interval, and satisfying... The segment threshold is set based on the range of Doppler frequency shift variation in the maritime communication scenario. When the Doppler widening range increases, a higher order of feedforward filter is output to improve the equalizer's ability to compensate for frequency shift distortion. When the Doppler widening range decreases, a lower order of feedforward filter is output to reduce noise interference introduced by redundant taps.

[0045] Column vectors are extracted from the two-dimensional channel fading state matrix, corresponding to the padded time-delay power spectrum feature vector. The number of non-zero elements in the column vectors is counted. A non-zero element is defined as an element with a value greater than a preset zero threshold. The zero threshold is set based on the channel noise level, with a default value of 0.01 (the maximum value of the column vector), used to eliminate false non-zero elements caused by noise. The count of non-zero elements corresponds to the number of effective multipath components in the maritime channel. This count is directly assigned to the update order of the feedback filter, ensuring that the order of the feedback filter matches the number of multipath components in the channel in real time. This guarantees that the feedback filter can cover all effective multipath components while avoiding noise introduced by redundant taps.

[0046] Furthermore, after outputting the updated order of the feedforward filter and assigning the updated order of the feedback filter, the tap coefficient initialization operation is performed during filter order switching. The historical orders of the feedforward filter and feedback filter stored at the previous time step are retrieved. The updated order of the feedforward filter output at the current time step is compared with the historical order of the feedforward filter at the previous time step, and the updated order of the feedback filter assigned at the current time step is compared with the historical order of the feedback filter at the previous time step. If the updated order of the feedforward filter output at the current time step is greater than the historical order of the feedforward filter at the previous time step, all tap coefficients corresponding to the historical order of the feedforward filter at the previous time step are retained, and zero-value elements equal to the number of newly added orders are added to the end of the tap coefficient vector as the initial tap coefficients of the feedforward filter at the current time step. If the updated order of the feedforward filter output at the current time step is less than the historical order of the feedforward filter at the previous time step, the coefficient segment with the same length as the current updated order is extracted from the center position of the tap coefficient vector corresponding to the historical order of the feedforward filter at the previous time step and used as the initial tap coefficients of the feedforward filter at the current time step.

[0047] Similarly, if the updated order of the feedback filter at the current time is greater than the historical order of the feedback filter at the previous time, all tap coefficients corresponding to the historical order of the feedback filter at the previous time are retained, and zero-value elements equal to the number of newly added orders are added to the end of the tap coefficient vector as the initial tap coefficients of the feedback filter at the current time. If the updated order of the feedback filter at the current time is less than the historical order of the feedback filter at the previous time, the coefficient segment with the same length as the current updated order is extracted from the center position of the tap coefficient vector corresponding to the historical order of the feedback filter at the previous time, and used as the initial tap coefficients of the feedback filter at the current time. The center position of the tap coefficient vector corresponds to the main path component of the channel. The main path component has the highest energy proportion. Extracting the coefficients at the center position can preserve the equalizer's processing capability for the main path component and avoid signal distortion and prolonged convergence period during order switching.

[0048] Table 3. Mapping Relationship between Doppler Broadening Range and Feedforward Filter Update Order

[0049] Table 3 shows the segmented mapping relationship between the Doppler widening range and the feedforward filter update order. The segmentation intervals are determined based on the range of Doppler frequency shift dynamic changes in the maritime communication scenario. The order of the feedforward filter increases with the increase of the Doppler widening range to match the enhanced time-varying characteristics of the channel. A lower order is used in low dynamic scenarios to reduce computational complexity and tap noise, while a higher order is used in high dynamic scenarios to improve the compensation capability for frequency shift distortion.

[0050] This embodiment calculates the Doppler broadening range by analyzing the extrema of the row vectors in the two-dimensional channel fading state matrix. Based on a piecewise mapping function, it dynamically determines the order of the feedforward filter, ensuring real-time matching between the feedforward filter order and the Doppler broadening of the channel. By statistically analyzing the number of non-zero elements in the column vectors of the two-dimensional channel fading state matrix, it dynamically sets the order of the feedback filter, ensuring real-time matching between the feedback filter order and the number of multipath components in the channel. During filter order switching, the retention of historical coefficients and the truncation of the center coefficients enable smooth initialization of the tap coefficients, avoiding signal distortion and prolonged convergence period caused by order adjustment. This also eliminates the truncation error and redundant tap noise generated by the fixed-order equalizer when the channel state changes dynamically.

[0051] In a preferred embodiment, reference Figures 4 to 6 Based on the generated two-dimensional channel fading state matrix, an adaptive step size factor generation and constraint operation is performed. The current time step is obtained. The generated two-dimensional channel fading state matrix and the previous moment Stored two-dimensional channel fading state matrix The dimensions of the two matrices are kept consistent, both being 1. Calculate the Euclidean distance between two matrices. The Euclidean distance is calculated by satisfying the following conditions:

[0052] in, The Euclidean distance between the two-dimensional channel fading state matrices at the current time and the previous time is given. The current time in the two-dimensional channel fading state matrix is ​​the th Line number Column elements, The 2D channel fading state matrix at the previous time step is the th Line number Column elements, Let be the dimension of the matrix. , These are the row and column indices of the matrix, respectively. The magnitude of the Euclidean distance characterizes the degree of change in the maritime channel state between adjacent moments; a larger Euclidean distance indicates more drastic changes in the channel state, while a smaller Euclidean distance indicates a more stable channel state.

[0053] Furthermore, the calculated Euclidean distance is input into a preset nonlinear mapping function containing an exponential decay term, and the output value of the nonlinear mapping function is obtained. This output value is used as the step size factor of the minimum mean square error algorithm at the current time. The nonlinear mapping function satisfies:

[0054] in, The step size factor at the current moment. The preset minimum convergence step size, The preset maximum initial step size, The preset attenuation coefficient, The Euclidean distance between the channel matrices at adjacent time points. is a natural constant. The output characteristic of this nonlinear mapping function is: when the input Euclidean distance... As the exponent increases, the exponent term approaches 0, and the function output value approaches the preset maximum initial step size. To improve the tracking speed of the equalizer coefficients when the channel changes drastically; when the input Euclidean distance As the exponent decreases, the exponent term approaches 1, and the function output value approaches the preset minimum convergence step size. This is to reduce the steady-state error of the equalizer coefficients when the channel is stable.

[0055] After obtaining the step size factor at the current moment, an interval constraint operation on the step size factor is performed. The channel state energy value of the two-dimensional channel fading state matrix at the current moment is calculated, and the calculation of the channel state energy value satisfies:

[0056] in, This represents the channel state energy value at the current moment. The current time in the two-dimensional channel fading state matrix is ​​the th Line number The absolute value of each column element. The calculated channel state energy value. The energy value is compared with a pre-stored stable channel energy threshold, the value of which is set based on the channel state energy statistics under stable maritime channels. If the channel state energy value... If the energy value is greater than the stable channel energy threshold, it indicates that the current channel is in a stable state and the channel energy is stable. Therefore, the step size factor at the current moment is limited to between the preset upper and lower limits of the first step size. If the channel state energy value... If the energy level is less than or equal to the stationary channel energy threshold, it indicates that the current channel is in a non-stationary fluctuating state with drastic energy changes. Therefore, the step size factor at the current moment is restricted to a preset upper and lower limit of the second step size, where the upper limit of the first step size is less than the upper limit of the second step size. The lower limits of the first and second step sizes can be set to the same or different values ​​depending on the scenario requirements. This range constraint on the step size factor further mitigates the divergence risk of a constant step size during channel fluctuations and the slow convergence problem during stationary periods.

[0057] Furthermore, based on the obtained feedforward filter update order, feedback filter update order, and step size factor, signal equalization processing and coefficient iterative update of the decision feedback equalizer are performed. The output signal of the decision feedback equalizer is calculated to satisfy:

[0058] in, for The output signal of the time equalizer for Time-forward filter The coefficient of each tap for The received baseband signal at any time for Time feedback filter The coefficient of each tap for The output symbol of the decision at any given moment. Let this be the update order of the feedforward filter. This is the update order of the feedback filter. For the output signal of the equalizer... Perform the judgment operation to obtain the judgment output symbol at the current moment. The decision operation is implemented using the minimum distance decision criterion, and the output signal is... Mapping to the nearest constellation point, the corresponding constellation symbol becomes the decision output symbol. Calculate the error signal. The error signal is defined as Based on the error signal and step size factor, the minimum mean square error algorithm is used to iteratively update the tap coefficients of the feedforward and feedback filters. The coefficient update satisfies the following:

[0059] in, for Time-forward filter The update coefficient for each tap. for Time feedback filter The update coefficient for each tap. for The complex conjugate of the constantly received baseband signal for The complex conjugate of the output symbol is determined at each moment. Through the above coefficient iterative update and equalization processing, the joint elimination of multipath interference and frequency shift distortion in the baseband signal is completed.

[0060] A maritime communication anti-fading optimization system based on signal data processing is constructed. The system includes a signal interception device, a feature extraction device, a matrix construction device, an order mapping device, a step size generation device, and an equalization processing device. The input of the signal interception device is connected to the baseband signal output of the receiver, and it is used to segment the baseband signal of the receiver and extract the pilot sequence from each segment. The signal interception device internally includes a frame synchronization unit, a segmentation unit, and a pilot extraction unit. The frame synchronization unit detects the start position of the frame header in the baseband signal, the segmentation unit segments the baseband signal according to the frame period, and the pilot extraction unit extracts the pilot sequence based on the start position of the frame header. The input of the feature extraction device is connected to the output of the signal interception device. It performs a Fast Fourier Transform (FFT) on the pilot sequence to extract the frequency-domain Doppler broadened feature vector and performs a time-domain cross-correlation operation on the pilot sequence to extract the time-delay power spectrum feature vector. The feature extraction device internally includes a frequency-domain feature extraction unit and a time-domain feature extraction unit. The frequency-domain feature extraction unit performs the FFT and frequency-domain feature extraction, while the time-domain feature extraction unit performs the time-domain cross-correlation operation and time-delay feature extraction. The input of the matrix construction device is connected to the output of the feature extraction device. It concatenates the frequency-domain Doppler broadened feature vector and the time-delay power spectrum feature vector according to a preset dimension to generate a two-dimensional channel fading state matrix. The matrix construction device internally includes a dimension matching unit and a matrix generation unit. The dimension matching unit fills the two feature vectors with zeros to achieve dimension matching, while the matrix generation unit generates the two-dimensional channel fading state matrix through a vector outer product operation. The input of the order mapping device is connected to the output of the matrix construction device. It is used to determine the update order of the feedforward filter of the decision feedback equalizer based on the extreme value of the frequency domain Doppler broadening eigenvector in the two-dimensional channel fading state matrix, and to determine the update order of the feedback filter based on the number of multipath components in the time delay power spectrum eigenvector. The order mapping device is equipped with a feedforward order mapping unit and a feedback order mapping unit. The feedforward order mapping unit is used to obtain the update order of the feedforward filter based on the Doppler broadening range mapping, and the feedback order mapping unit is used to determine the update order of the feedback filter based on the number of multipath components. The input of the step size generation device is connected to the output of the matrix construction device. It is used to calculate the Euclidean distance between the two-dimensional channel fading state matrix at the current time and the two-dimensional channel fading state matrix at the previous time. The nonlinear function of the Euclidean distance is used as the step size factor of the minimum mean square error algorithm at the current time. The step size generation device is equipped with an Euclidean distance calculation unit, a nonlinear mapping unit and a step size constraint unit. The Euclidean distance calculation unit is used to calculate the Euclidean distance between the channel matrices at adjacent time times. The nonlinear mapping unit is used to generate an adaptive step size factor. The step size constraint unit is used to apply interval constraints to the step size factor based on the channel state energy value.The input of the equalization processing device is connected to the output of the order mapping device and the step size generation device, respectively. It is used to jointly eliminate multipath interference and frequency shift distortion in the baseband signal by updating the order of the feedforward filter, updating the order of the feedback filter, and using the step size factor. The equalization processing device is equipped with a decision feedback equalizer unit and a coefficient iteration update unit. The decision feedback equalizer unit is used to perform equalization processing of the baseband signal, and the coefficient iteration update unit is used to perform iterative update of the filter tap coefficients based on the minimum mean square error algorithm.

[0061] Table 4. Functions and Input / Output Parameters of Each Device in the Marine Communication Anti-Fading Optimization System

[0062] Table 4 shows the core functions and input / output parameters of each device in the marine communication anti-fading optimization system. The devices are connected sequentially according to the order of signal processing, realizing a complete processing link from baseband signal input to signal output after equalization processing. The output parameters of each device serve as the input parameters of the next level device, ensuring the continuity of system data flow and the closed-loop nature of the processing flow.

[0063] This embodiment achieves adaptive nonlinear adjustment of the step size factor of the minimum mean square error algorithm by using the Euclidean distance of the two-dimensional channel fading state matrix at adjacent time points, enabling the step size to match the degree of channel state change in real time. By comparing the channel state energy value with the stable channel energy threshold, upper and lower limits of different intervals are imposed on the step size factor, further optimizing the convergence characteristics and tracking performance of the step size. Through the output calculation and coefficient iterative update of the decision feedback equalizer, the joint elimination of multipath interference and frequency shift distortion in the baseband signal is achieved. At the same time, a complete maritime communication anti-fading optimization system is constructed. Through the collaborative work of multiple functional devices, adaptive suppression of dynamic fading in maritime communication channels is achieved, solving the problem of joint fading suppression failure caused by the mismatch between the equalizer structure and two-dimensional fading characteristics in the prior art.

Claims

1. A method for optimizing anti-fading in maritime communication based on signal data processing, characterized in that, include: The baseband signal at the receiving end is segmented and the pilot sequence in each segment is extracted. The pilot sequence is subjected to Fast Fourier Transform to extract the frequency domain Doppler broadening feature vector, and the pilot sequence is subjected to time domain cross-correlation operation to extract the time delay power spectrum feature vector; The frequency domain Doppler broadened feature vector and the time delay power spectrum feature vector are concatenated according to a preset dimension to generate a two-dimensional channel fading state matrix. The feedforward filter update order of the decision feedback equalizer is determined based on the extreme value of the frequency domain Doppler broadening eigenvector in the two-dimensional channel fading state matrix, and the feedback filter update order is determined based on the number of multipath components in the time delay power spectrum eigenvector. Calculate the Euclidean distance between the current two-dimensional channel fading state matrix and the previous two-dimensional channel fading state matrix, and use the nonlinear function of the Euclidean distance as the step size factor of the minimum mean square error algorithm at the current time. The multipath interference and frequency shift distortion in the baseband signal are jointly eliminated by updating the order of the feedforward filter, updating the order of the feedback filter, and using the step size factor.

2. The maritime communication anti-fading optimization method based on signal data processing according to claim 1, characterized in that, The step of extracting the frequency domain Doppler broadening feature vector by performing a fast Fourier transform on the pilot sequence includes: performing a fast Fourier transform on the pilot sequence to obtain a frequency domain sequence; Calculate the power spectral density at each frequency point in the frequency domain sequence and construct a power spectral density curve; Determine the noise floor mean in the power spectral density curve, and multiply the noise floor mean by a preset factor to obtain a decision threshold; Extract the set of frequency points above the decision threshold from the power spectral density curve, calculate the frequency difference between the maximum and minimum frequency points in the set, and combine the frequency difference and the power value of the set to generate the frequency domain Doppler broadening feature vector.

3. The maritime communication anti-fading optimization method based on signal data processing according to claim 1, characterized in that, The step of performing time-domain cross-correlation operation on the pilot sequence to extract the time-delay power spectrum feature vector includes: performing cross-correlation operation on the pilot sequence and the locally stored reference pilot sequence to obtain the time-delay cross-correlation function curve; Envelope detection is performed on the time-delay cross-correlation function curve to extract the envelope peak value; Based on the envelope peak value, a peak decision threshold is set, and multiple peak points higher than the peak decision threshold are selected from the time delay cross-correlation function curve; Record the time delay value and relative amplitude value corresponding to each peak point, remove peak points whose amplitude value is lower than the preset interference removal threshold, and arrange the time delay value and relative amplitude value of the remaining peak points in chronological order to generate the time delay power spectrum feature vector.

4. The maritime communication anti-fading optimization method based on signal data processing according to claim 1, characterized in that, The step of concatenating the frequency domain Doppler broadened feature vector and the time delay power spectrum feature vector according to a preset dimension to generate a two-dimensional channel fading state matrix includes: obtaining the first dimension length of the frequency domain Doppler broadened feature vector and the second dimension length of the time delay power spectrum feature vector; The padded frequency domain Doppler broadened eigenvectors are used as the row vectors of the matrix, and the padded time delay power spectrum eigenvectors are used as the column vectors of the matrix. The two-dimensional channel fading state matrix is ​​generated by vector outer product operation.

5. The maritime communication anti-fading optimization method based on signal data processing according to claim 1, characterized in that, The step of determining the feedforward filter update order of the decision feedback equalizer based on the extreme value of the frequency domain Doppler broadening feature vector in the two-dimensional channel fading state matrix, and determining the feedback filter update order based on the number of multipath components in the time delay power spectrum feature vector, includes: extracting the maximum and minimum values ​​corresponding to the row vectors in the two-dimensional channel fading state matrix, and calculating the Doppler broadening range; The Doppler widening range is input into a pre-constructed piecewise function of feedforward order mapping, and the updated order of the feedforward filter is output. The number of non-zero elements corresponding to the column vectors in the two-dimensional channel fading state matrix is ​​used as the number of multipath components. The number of the multipath components is directly assigned as the update order of the feedback filter.

6. The maritime communication anti-fading optimization method based on signal data processing according to claim 1, characterized in that, The calculation of the Euclidean distance between the current two-dimensional channel fading state matrix and the previous two-dimensional channel fading state matrix, and the use of the nonlinear function of the Euclidean distance as the step size factor of the minimum mean square error algorithm at the current time, includes: taking the difference between the elements at the same position in the current two-dimensional channel fading state matrix and the previous two-dimensional channel fading state matrix and calculating the sum of squares, and taking the square root of the sum of squares to obtain the Euclidean distance. The Euclidean distance is input into a nonlinear mapping function containing an exponential decay term. The nonlinear mapping function is used to output a value that approaches a preset initial step size when the Euclidean distance increases and to output a value that approaches a preset minimum convergence step size when the Euclidean distance decreases. The output value of the nonlinear mapping function is used as the step size factor.

7. The maritime communication anti-fading optimization method based on signal data processing according to claim 2, characterized in that, Before performing a fast Fourier transform on the pilot sequence to obtain a frequency domain sequence, the method further includes: applying a Hamming window function to the pilot sequence for windowing truncation. The length of the Hamming window function is dynamically set according to the coherence time of the maritime communication link. The coherence time is obtained by reciprocal calculation of the absolute value of the Doppler frequency shift measured in advance by the receiver. By applying the Hamming window function, the truncation effect at the beginning and end of the pilot sequence is suppressed, thereby reducing the interference of sidelobe energy in the frequency domain sequence on the power spectral density curve.

8. The maritime communication anti-fading optimization method based on signal data processing according to claim 5, characterized in that, After outputting the updated order of the feedforward filter and assigning the updated order of the feedback filter, the method further includes: determining whether the updated order of the feedforward filter or the updated order of the feedback filter output at the current moment is greater than the historical order corresponding to the previous moment. If the current output of the feedforward filter update order or the feedback filter update order is greater than the corresponding historical order at the previous time, then all filter tap coefficients corresponding to the historical order at the previous time are retained, and the coefficients of the newly added tap positions are initialized to zero. If the current feedforward filter update order or feedback filter update order is less than the corresponding historical order at the previous time, then the tap coefficient at the center position of the historical order at the previous time is taken as the initial coefficient at the current time.

9. The maritime communication anti-fading optimization method based on signal data processing according to claim 6, characterized in that, After using the output value of the nonlinear mapping function as the step size factor, the method further includes: calculating the sum of the absolute values ​​of all elements in the two-dimensional channel fading state matrix at the current time as the channel state energy value; The channel state energy value is compared with a pre-stored stable channel energy threshold. If the channel state energy value is greater than the stable channel energy threshold, then the step size factor is restricted to between the preset upper limit of the first step size and the lower limit of the first step size. If the channel state energy value is less than or equal to the stable channel energy threshold, the step size factor is restricted to a preset second step size upper limit value and a second step size lower limit value, wherein the first step size upper limit value is less than the second step size upper limit value.

10. A marine communication anti-fading optimization system based on signal data processing, characterized in that, include: A signal interception device is used to segment the baseband signal at the receiving end and extract the pilot sequence from each segment of the signal; The feature extraction device is used to perform a fast Fourier transform on the pilot sequence to extract the frequency domain Doppler broadening feature vector, and to perform a time domain cross-correlation operation on the pilot sequence to extract the time delay power spectrum feature vector. A matrix construction device is used to concatenate the frequency domain Doppler broadened feature vector and the time delay power spectrum feature vector according to a preset dimension to generate a two-dimensional channel fading state matrix. The order mapping device is used to determine the update order of the feedforward filter of the decision feedback equalizer based on the extreme value of the frequency domain Doppler broadening feature vector in the two-dimensional channel fading state matrix, and to determine the update order of the feedback filter based on the number of multipath components in the time delay power spectrum feature vector. A step size generation device is used to calculate the Euclidean distance between the current two-dimensional channel fading state matrix and the previous two-dimensional channel fading state matrix, and to use the nonlinear function of the Euclidean distance as the step size factor of the minimum mean square error algorithm at the current time. An equalization processing device is used to jointly eliminate multipath interference and frequency shift distortion in the baseband signal by using the updated order of the feedforward filter, the updated order of the feedback filter, and the step size factor.

Images