A method for receiving underwater acoustic communication based on inverse discrete cosine transform

Abstract

The application discloses a kind of based on inverse discrete cosine transform underwater acoustic communication receiving method, by alternating matching to the up-sweep frequency and down-sweep frequency LFM sequence in received signal, according to the arrangement relation of correlation peak effective frame is identified and multiple frame boundary is reconstructed;By LFM template search under different time scales, waveform compression or expansion is estimated, and received signal is resampled correction;Joint cycle prefix correlation index and pilot matching index correct symbol starting position, use pilot amplitude and phase information to carry out channel response interpolation, and carry out frequency domain equalization according to data symbol and effective subcarrier;According to the discrete cosine base and frequency dimension transformation order used by sending end, transpose inverse mapping is carried out to equalization data matrix, and data symbol before frequency dimension orthogonal expansion is recovered;Reduce the accumulation and amplification of front-stage parameter estimation deviation in orthogonal recovery process, improve data recovery stability and demodulation reliability under time-varying multipath underwater acoustic channel.

Description

Technical Field

[0001] This invention relates to the field of communications, and more particularly to an underwater acoustic communication receiving method based on inverse discrete cosine transform. Background Technology

[0002] Due to the combined effects of low-speed sound propagation, complex multipath structures, and the relative motion of the transceiver platforms, underwater acoustic multicarrier signals often exhibit problems such as large time delay spread, time-varying Doppler distortion, and limited spectrum resources at the receiver. The receiver needs to perform signal detection, frame structure recovery, time scale adjustment, frequency synchronization, channel reconstruction, and symbol decision from continuously sampled data. These operations are not independent: synchronization position deviations change the truncation range of the cyclic prefix and effective symbols; resampling errors introduce residual time drift; and insufficient frequency compensation causes inter-carrier interference and continuous phase rotation. These mismatches may be further amplified after channel estimation and equalization, causing the received data to deviate from the structural conditions required by the precoding transform, ultimately making it difficult to accurately recover the frequency-dimensional discrete cosine orthogonal domain.

[0003] Existing orthogonal precoding reception methods often treat the inverse discrete cosine transform (ICT) as an independent processing module, assuming that frame boundaries, time scales, and carrier frequency offsets have been accurately corrected. However, in practical underwater acoustic communication, non-integer sampling drift, inter-frame interval fluctuations, and rapid pilot phase changes often coexist. Without jointly utilizing synchronization correlation peaks, cyclic prefix matching indices, pilot phase information, and one-sided discrete cosine mapping constraints, directly performing the inverse transform can easily lead to constellation rotation, noise enhancement, and bit error propagation.

[0004] On the other hand, discrete cosine precoding (DCT) reception has strict consistency requirements regarding the order of the transform matrix, the normalization method, the basis function arrangement, and the effective subcarrier index. If the DCT matrix definitions used by the transmitter and receiver are inconsistent, or if an unnormalized discrete cosine matrix is ​​used, the invertibility of the frequency-dimensional orthogonal mapping may be destroyed, leading to systematic demodulation errors. Therefore, it is necessary to construct a complete underwater acoustic reception process that strictly matches the discrete cosine basis, unifying the design of alternating LFM chain detection, time-scale compensation, phase-aware pilot interpolation, regularized channel equalization, and frequency-dimensional discrete cosine inverse mapping to improve synchronization stability, numerical reliability, and data recovery performance under complex time-varying underwater acoustic channels. Summary of the Invention

[0005] The following is an overview of the topics described in detail in this article.

[0006] The purpose of this application is to at least partially solve one of the technical problems existing in the related technologies. The embodiments of this application provide an underwater acoustic communication receiving method based on discrete cosine inverse transform, which reduces the accumulation and amplification of the pre-stage parameter estimation deviation in the orthogonal recovery process, and improves the data recovery stability and demodulation reliability under time-varying multipath underwater acoustic channels.

[0007] An embodiment of this application provides an underwater acoustic communication receiving method based on inverse discrete cosine transform, comprising:

[0008] Obtain the real-valued passband sequence and transmitting parameters collected by the hydrophone;

[0009] Alternate matching is performed on the up-sweep and down-sweep linear frequency modulated (LFM) signal sequences in the real-value passband sequence, and valid frames are identified and multi-frame boundaries are reconstructed based on the arrangement of relevant peaks.

[0010] By searching the LFM template at different time scales, the waveform compression or expansion caused by the relative motion between the transmitter and receiver is estimated, and the received signal is resampled and corrected.

[0011] The symbol start position is corrected by combining the cyclic prefix correlation index and the pilot matching index. Channel response interpolation is performed using pilot amplitude and phase information. Frequency domain equalization is performed by a minimum mean square error single-tap equalizer with regularization term based on data symbols and effective subcarriers.

[0012] Based on the discrete cosine basis and frequency-dimensional transformation order used at the transmitting end, the equalized data matrix is ​​transposed and inversely mapped to restore the data symbols before the frequency-dimensional orthogonal expansion.

[0013] According to certain embodiments of this application, prior to the alternating matching step, the method further includes:

[0014] Establish the input interface between received data and system parameters using the following steps:

[0015] Read the received data file;

[0016] Check the number and data types of variables in the received data file, and verify the integrity of each configuration in the parameter file;

[0017] Check the matching relationship between the orthogonal transformation matrix and the data grid, the number of effective subcarriers, and the frequency dimension transformation order, and calculate the orthogonality error of the transformation matrix;

[0018] When the number of variables, parameter dimensions, or orthogonality error does not meet the preset conditions, the subsequent processing is terminated and the corresponding abnormal information is output.

[0019] According to certain embodiments of this application, the alternating matching of the up-scan and down-scan LFM sequences in the received signal, identifying valid frames and reconstructing multi-frame boundaries based on the arrangement of correlation peaks, includes:

[0020] The received signal is normalized and matched with the upper-sweep LFM template and the lower-sweep LFM template respectively to obtain two types of synchronization correlation sequences and their corresponding candidate peak positions;

[0021] Based on the pre-set alternation pattern of up-scan and down-scan frequencies at the transmitting end, candidate peaks are combined in time sequence and filtered group by group.

[0022] When the following conditions are met: the peak amplitude reaches the detection threshold, the relative difference of peaks within the group is within the allowable range, the interval between adjacent peaks conforms to the frame structure constraints, and the number of detected peaks is consistent with the number of frames contained in a complete playback, the peak group is determined as a valid LFM frame chain, and the initial boundary of each frame in this playback is determined.

[0023] Isolated peaks, out-of-order peaks, and incomplete peak groups that do not meet the aforementioned joint conditions are removed.

[0024] According to certain embodiments of this application, waveform compression or expansion caused by relative motion between transmit and receive signals is estimated through LFM template search at different time scales, and resampling correction is performed on the received signal, including:

[0025] By using LFM template search at different time scales, coarse search sets and fine search sets for candidate time scales are established around zero scale bias.

[0026] Scaling the LFM template for each candidate time scale and using the median of the LFM correlation peaks of all frames of the correct type as the score, we estimate the waveform compression or expansion caused by the relative motion between the transmitter and receiver.

[0027] Select the time scale with the highest score;

[0028] When the deviation of the time scale with the largest score from 1 exceeds a set threshold, reverse linear interpolation correction is performed on the entire received waveform of the received signal.

[0029] According to certain embodiments of this application, the joint cyclic prefix correlation index and pilot matching index correct the symbol start position, including:

[0030] The load segment is extracted based on the independently detected LFM start point of each frame. The load segment is multiplied by the carrier complex exponent and resampled to the baseband sampling rate.

[0031] Candidate offsets are searched around the prediction starting point, and OFDM symbol boundaries are selected using a weighted average of the cyclic prefix correlation index and the pilot matching index to correct the symbol starting position.

[0032] According to certain embodiments of this application, after correcting the symbol start position of the joint cyclic prefix correlation index and pilot matching index, the method includes:

[0033] The correlation between the cyclic prefix and the corresponding tail sample of the OFDM symbol in a frame is accumulated to estimate the residual common frequency offset and correct the residual common frequency offset.

[0034] After removing the cyclic prefix, a Fast Fourier Transform is performed to retain the valid subcarriers.

[0035] According to certain embodiments of this application, channel response interpolation is performed using pilot amplitude and phase information, including:

[0036] Extract the received values ​​of the pilot OFDM symbols on each effective subcarrier, and perform a complex ratio operation between the received values ​​and the corresponding known pilot symbols to obtain the discrete channel frequency response at the pilot time.

[0037] For data symbols located between adjacent pilot symbols, the channel responses at both ends are decomposed into amplitude and phase components. The phase sequence is unwrapped, and the amplitude and continuous phase are interpolated along the OFDM symbol direction to reconstruct the complex channel response corresponding to each data symbol.

[0038] For subcarriers with isolated abrupt changes in frequency direction or with strong estimated noise, the complex channel response is locally smoothed.

[0039] According to certain embodiments of this application, the regularization coefficient of the regularization term of the minimum mean square error single-tap equalizer is the product of the median channel power and a fixed proportion or is adaptively determined by the variance of the inactive subcarrier noise, and is limited to a preset upper and lower bound.

[0040] According to certain embodiments of this application, the equalization data matrix is ​​transposed and inversely mapped according to the discrete cosine basis and frequency-dimensional transformation order used at the transmitting end, including:

[0041] Based on the discrete cosine basis and frequency-dimensional transform order used at the transmitting end, the equalized data symbols are arranged into an equalized data matrix with effective subcarriers as rows and data OFDM symbols as columns. Balanced data matrix The conjugate transpose of the DCT-II matrix yields the data grid recovered by inverse mapping.

[0042] Among them, the frequency-dimensional discrete cosine orthogonal extension of the transmitting end satisfies ;

[0043] The receiver inverse mapping satisfies ;

[0044] The normalized DCT-II matrix satisfies the unitary constraint.

[0045] In the formula, This represents the QPSK data grid to be sent. This represents the time-frequency data grid after discrete cosine orthogonal expansion in the frequency dimension. Denotes an L-order normalized DCT-II matrix. Indicates that after subcarrier-by-subcarrier MMSE equalization, according to OK, A balanced data matrix arranged in columns. express The conjugate transpose of . This represents the data grid recovered by inverse mapping. express Identity matrix.

[0046] According to certain embodiments of this application, the received recovery grid obtained by inverse mapping is expanded into a complex symbol sequence by column, the QPSK common phase is estimated, decision-guided gain normalization is performed, and then QPSK hard decision, block deinterleaving and pseudo-random sequence descrambling are performed; when the transmitter uses convolutional coding, Viterbi decoding is performed.

[0047] For multiple complete received records corresponding to the same transmitted waveform, a no-reference quality evaluation is performed based on frame synchronization correlation strength, cyclic prefix correlation, pilot reconstruction consistency, and equalization error vector magnitude. At least one received result is selected based on the quality evaluation results, or multiple received results are fused.

[0048] The above-mentioned solution has at least the following beneficial effects:

[0049] (1) Ensure consistency of orthogonal mapping. The transmitting and receiving ends use the same normalized unitary matrix to complete the forward mapping and the conjugate transpose inverse mapping, respectively, and specify the order, normalization coefficients and element arrangement of the discrete cosine transform matrix. This can avoid data grid misalignment and persistent demodulation errors caused by inconsistent transformation scale, basis function order or symbol definition.

[0050] (2) Improve the reliability of multi-frame synchronization. By utilizing the alternating arrangement of the upper-sweep frequency LFM and the time-reversed lower-sweep frequency LFM, the type, number, amplitude, and spacing of related peaks are jointly constrained, thereby eliminating isolated spurious peaks caused by noise, multipath, or transient interference. Each frame of data is independently extracted based on the actual detection position of the corresponding LFM, without relying on a fixed frame interval, and can adapt to changes in silence length and local sampling loss during playback and recording.

[0051] (3) Enhance the adaptability to time scale changes caused by motion. A scale estimation method combining coarse search and fine search is adopted to reconstruct the LFM reference template under different time scaling ratios, and the optimal scale is determined by robust statistics of multi-frame matching results. This method can estimate non-integer sampling scale deviation and recover the time reference of the received waveform through resampling; after scale correction, band filtering is performed, which can reduce the mismatch between the filter frequency response and the band of the scaling signal, and reduce the impact of time compression or expansion on the orthogonality of OFDM subcarriers.

[0052] (4) Improve the stability of equalization under time-varying fading conditions. After obtaining the discrete channel response based on the pilot position, inter-symbol interpolation is performed on the channel amplitude and unwrapped phase to maintain the continuity of channel reconstruction under rapid phase changes. The equalization stage adopts a minimum mean square error single-tap equalizer with regularization term to avoid excessive noise amplification caused by direct inversion at deep fading subcarriers, thereby improving the stability of equalization symbols.

[0053] (5) A complete and feasible receiving and processing flow is formed. The complete frame chain detection, time scale compensation, symbol fine timing, frequency offset correction, pilot channel estimation, regularized equalization, orthogonal inverse mapping, common phase correction, deinterleaving, and channel decoding are incorporated into a unified processing chain. Each processing stage has clear input, output, and verification conditions, and can be directly used for the reception and recovery of replay, multi-frame OFDM, and discrete cosine orthogonal precoded underwater acoustic communication signals. Attached Figure Description

[0054] The accompanying drawings are used to provide a further understanding of the technical solutions of this application and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of this application and do not constitute a limitation on the technical solutions of this application.

[0055] Figure 1 This is a flowchart illustrating the steps of an underwater acoustic communication receiving method based on inverse discrete cosine transform.

[0056] Figure 2 This is a flowchart illustrating the steps involved in establishing an input interface between received data and system parameters.

[0057] Figure 3 This is a step diagram of alternating matching of the up-scan and down-scan LFM sequences in the received signal, identifying valid frames and reconstructing multi-frame boundaries based on the arrangement of related peaks;

[0058] Figure 4 This diagram illustrates the steps involved in estimating waveform compression or expansion caused by relative motion between the transmitter and receiver using LFM template search at different time scales, and then resampling and correcting the received signal.

[0059] Figure 5This is a step diagram illustrating the process of correcting the starting position of the symbol for the combined cyclic prefix correlation index and pilot matching index;

[0060] Figure 6 This is a step diagram of the residual common frequency offset correction and FFT process;

[0061] Figure 7 This is a flowchart illustrating the steps of channel response interpolation using pilot amplitude and phase information.

[0062] Figure 8 This is a step diagram of the residual phase normalization and bit recovery steps;

[0063] Figure 9 This is a step diagram of the steps for fusing and receiving results. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0065] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, or the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0066] The characters and symbols appearing in this application are explained in Table 1.

[0067] Table 1. Explanation of the meanings of characters and symbols

[0068]

[0069] To address the challenges of stable identification of multi-frame boundaries, accurate compensation of time scale offsets, and susceptibility of orthogonal precoding data recovery to synchronization and channel estimation errors in underwater acoustic channels, embodiments of this application provide an underwater acoustic communication receiving method based on inverse discrete cosine transform. This method sequentially performs complete frame chain identification, time scale compensation, OFDM parameter recovery, pilot-assisted equalization, orthogonal inverse mapping, and bit decoding at the underwater acoustic communication receiver.

[0070] The embodiments of this application will be further described below with reference to the accompanying drawings.

[0071] Reference Figure 1 A method for receiving underwater acoustic communication based on inverse discrete cosine transform includes the following steps:

[0072] Step S100: Obtain the real-valued passband sequence and transmitting parameters collected by the hydrophone;

[0073] Step S200: Alternately match the up-sweep and down-sweep linear frequency modulation (LFM) signal sequences in the real-value passband sequence, identify valid frames and reconstruct multi-frame boundaries based on the arrangement of related peaks;

[0074] Step S300: Through LFM template search at different time scales, estimate the waveform compression or expansion caused by the relative motion between the transmitter and receiver, and perform resampling correction on the received signal;

[0075] Step S400: Combine the cyclic prefix correlation index and pilot matching index to correct the symbol start position, use pilot amplitude and phase information to perform channel response interpolation, and use a minimum mean square error single-tap equalizer with regularization term to perform frequency domain equalization based on data symbols and effective subcarriers.

[0076] Step S500: According to the discrete cosine basis and frequency dimension transformation order used by the transmitting end, the equalization data matrix is ​​transposed and inversely mapped to restore the data symbols before the frequency dimension orthogonal expansion.

[0077] In this embodiment, by alternately matching the up-scanning and down-scanning LFM sequences in the received signal, valid frames are identified and multi-frame boundaries are reconstructed based on the arrangement of related peaks.

[0078] Subsequently, LFM template search at different time scales is used to estimate waveform compression or expansion caused by relative motion between the transmitter and receiver, and resampling correction is applied to the received signal. The symbol start position is corrected by combining cyclic prefix correlation features and pilot observations, channel response interpolation is performed using pilot amplitude and phase information, and the minimum mean square error criterion is used to suppress noise enhancement on deep fading subcarriers. After synchronization compensation and frequency domain equalization, the receiver performs a transpose inverse mapping on the equalized data matrix according to the discrete cosine basis and frequency-dimensional transform order used by the transmitter to recover the data symbols before frequency-dimensional orthogonal expansion.

[0079] Since the discrete cosine transform matrix used is a real orthogonal matrix, the inverse mapping can be directly implemented by its transpose without constructing a general complex inverse matrix, and can preserve the signal energy before and after the transformation under ideal conditions. By integrating frame detection, scale compensation, fine synchronization, channel interpolation, equalization, and discrete cosine inverse mapping into a single design, the accumulation and amplification of the pre-stage parameter estimation bias in the orthogonal recovery process can be reduced, thereby improving the data recovery stability and demodulation reliability in time-varying multipath underwater acoustic channels.

[0080] Reference Figure 2 At the receiving end, the first step is to establish an input interface between the received data and system parameters, including the following steps:

[0081] Step S11: Read the received data file;

[0082] Step S12: Check the number and data type of variables in the received data file, and verify the integrity of each configuration in the parameter file;

[0083] Step S13: Check the matching relationship between the orthogonal transformation matrix and the data grid, the number of effective subcarriers and the frequency dimension transformation order, and calculate the orthogonality error of the transformation matrix;

[0084] Step S14: When the number of variables, parameter dimensions, or orthogonality error does not meet the preset conditions, terminate the subsequent processing and output the corresponding abnormal information.

[0085] Specifically, the received data file only retains the received signal sequence to be processed, while the parameter file centrally stores the received sampling rate, carrier bandwidth, OFDM modulation parameters, up-and-down sweep LFM reference template, pilot symbols, frame structure configuration, and discrete cosine transform matrix. After reading, the number and data types of variables in the received data file are checked, and the completeness of each configuration in the parameter file is verified. Further checks are performed on the matching relationship between the orthogonal transform matrix and the data grid, the number of effective subcarriers, and the frequency-dimensional transform order, while calculating the orthogonality error of the transform matrix. If the number of variables, parameter dimensions, or orthogonality error does not meet the preset conditions, subsequent processing is terminated and corresponding abnormal information is output, thereby avoiding imperceptible interface mismatch caused by inconsistencies in transmit and receive parameters.

[0086] For step S100, the real-valued passband sequence acquired by the hydrophone is obtained. And the sending parameters.

[0087] In real-valued passband sequences middle, This is the discrete sampling sequence number.

[0088] The sending parameters should include at least the receiving sampling rate. Nominal emission sampling rate Baseband sampling rate Carrier center frequency Frequency-dimensional DCT-II transform order Effective number of subcarriers Number of OFDM symbols per frame FFT length , length of cyclic prefix Upscan frequency LFM template LFM template with downscan frequency Known pilot frequency And orthogonal matrices.

[0089] It should be noted that the number of effective subcarriers is denoted as... Furthermore, since DCT-II operates on all effective subcarriers, it should be explicitly stipulated that... .

[0090] The steps for performing front-end resampling and bandpass limiting. Depend on Resampling to ,get ; Use passband coverage Finite impulse response filters suppress out-of-band noise. If If the ratio cannot be approximated to a small integer, then anti-aliasing multiphase resampling is used.

[0091] Reference Figure 3 For step S200, the up-scan and down-scan LFM sequences in the received signal are alternately matched, and valid frames are identified and multi-frame boundaries are reconstructed based on the arrangement of related peaks, including the following steps:

[0092] Step S210: Normalize and match the received signal with the upper frequency sweep LFM template and the lower frequency sweep LFM template respectively to obtain two types of synchronization correlation sequences and corresponding candidate peak positions;

[0093] Step S220: Based on the pre-set alternation pattern of up-scanning and down-scanning frequencies at the transmitting end, the candidate peaks are combined in time sequence and filtered group by group.

[0094] Step S230: When the peak amplitude reaches the detection threshold, the relative difference of peaks within the group is within the allowable range, the interval between adjacent peaks conforms to the frame structure constraints, and the number of detected peaks is consistent with the number of frames included in a complete playback, the peak group is determined as a valid LFM frame chain, and the initial boundary of each frame in this playback is determined.

[0095] Step S240: Remove isolated peaks, out-of-order peaks, and incomplete peak groups that do not meet the joint conditions.

[0096] Among them are: ;

[0097] In the formula, Indicates the up-scan or down-scan template. For candidate delay The normalized correlation coefficient at the point, within the summation range This is the LFM sample point number.

[0098] Reference Figure 4 For step S300, the waveform compression or expansion caused by the relative motion between the transmitter and receiver is estimated by searching the LFM template at different time scales, and the received signal is resampled and corrected, including the following steps:

[0099] Step S310: Through LFM template search at different time scales, establish a coarse search set and a fine search set for candidate time scales around zero scale bias;

[0100] Step S320: Scale the LFM template for each candidate time scale and use the median of the LFM correlation peaks of all frames of the correct type as the score to estimate the waveform compression or expansion caused by the relative motion between the transmitter and receiver.

[0101] Step S330: Select the time scale with the highest score;

[0102] Step S340: When the deviation of the time scale with the largest score relative to 1 exceeds a set threshold, reverse linear interpolation correction is performed on the entire received waveform of the received signal.

[0103] Specifically, a coarse search set and a fine search set are established around zero-scale bias. The LFM template is scaled for each candidate scale, and the median of the LFM correlation peaks of all frames with correct types is used as the score. The scale with the highest score is selected. When its deviation from 1 exceeds a set threshold, reverse linear interpolation correction is performed on the entire received waveform.

[0104] ;

[0105] in, For the candidate timescale set, For the first Frame uses scale The correct LFM template related sequence.

[0106] ;

[0107] in, This represents a linear interpolation operator. For non-integer input positions; This is the scale-corrected sequence.

[0108] Reference Figure 5 For step S400, the joint cyclic prefix correlation index and pilot matching index correct the symbol start position, including the following steps:

[0109] Step S411: Extract the load segment based on the independently detected LFM start point of each frame, multiply the load segment by the carrier complex exponent and resample to the baseband sampling rate;

[0110] Step S412: Search for candidate offsets around the prediction start point, and select OFDM symbol boundaries using the weighted values ​​of the cyclic prefix correlation index and pilot matching index to correct the symbol start position.

[0111] Specifically, the load segment is truncated based on the independently detected LFM start point of each frame, multiplied by the carrier complex exponent, and resampled. Candidate offsets are searched around the prediction starting point, and OFDM symbol boundaries are selected using a weighted average of the cyclic prefix correlation index and the pilot matching index.

[0112] ;

[0113] in, For the first Frame load at The starting sample points in the sample, This is the down-conversion complex envelope.

[0114] ;

[0115] in, For fine timing candidate offsets, This is a normalized cyclic prefix correlation index.

[0116] Reference Figure 6 After correcting the symbol start position using the joint cyclic prefix correlation index and pilot matching index, the residual common frequency offset correction and FFT steps are performed, including the following steps: Step S421, the correlation between the cyclic prefix and the corresponding tail sample of the OFDM symbol in a frame is accumulated to estimate the residual common frequency offset and correct the residual common frequency offset; Step S422, after removing the cyclic prefix, a fast Fourier transform is performed to retain the effective subcarrier.

[0117] Specifically, the steps for residual common frequency offset correction and FFT are as follows: The correlation between the cyclic prefix and the corresponding tail sample of all OFDM symbols within a frame is accumulated to estimate and correct the residual common frequency offset; after removing the cyclic prefix, the following steps are performed: Point FFT, retain One effective subcarrier.

[0118] ;

[0119] in, Intra-frame OFDM symbol sequence number, Represents the complex phase angle. This is the estimated value of the residual frequency offset.

[0120] ;

[0121] in, For the first The OFDM symbol of the first FFT output at each frequency point.

[0122] Reference Figure 7Channel response interpolation using pilot amplitude and phase information includes the following steps:

[0123] Step S431: Extract the received values ​​of the pilot OFDM symbols on each effective subcarrier, and perform a complex ratio operation between the received values ​​and the corresponding known pilot symbols to obtain the discrete channel frequency response at the pilot time.

[0124] For data symbols located between adjacent pilot symbols, the channel responses at both ends are decomposed into amplitude and phase components. The phase sequence is unwrapped, and the amplitude and continuous phase are interpolated along the OFDM symbol direction to reconstruct the complex channel response corresponding to each data symbol.

[0125] For subcarriers with isolated abrupt changes in frequency direction or with strong estimated noise, the complex channel response is locally smoothed.

[0126] Specifically, the steps for time-varying channel reconstruction are as follows: Extract the received values ​​of the pilot OFDM symbols on each effective subcarrier, and perform a complex ratio operation with the corresponding known pilot symbols to obtain the discrete channel frequency response at the pilot time.

[0127] For data symbols located between adjacent pilot symbols, the channel responses at both ends are first decomposed into amplitude and phase components. The phase sequence is then unwrapped, and the amplitude and continuous phase are interpolated along the OFDM symbol direction to reconstruct the complex channel response corresponding to each data symbol.

[0128] For subcarriers with isolated abrupt changes in frequency direction or with strong estimated noise, local smoothing can be further applied to the complex channel response to suppress pilot noise propagation while preserving the main characteristics of channel amplitude and phase changes with time and frequency.

[0129] ;

[0130] in, This refers to the pilot OFDM symbol number. For the first Known pilots on each effective subcarrier, Channel estimation for pilot positions.

[0131]

[0132] in, and Data symbols Pilot numbers before and after, , and These represent the channel amplitude and unwrapped phase obtained through interpolation, respectively.

[0133] A minimum mean square error (MMS) single-tap equalizer with a regularization term is used for frequency domain equalization based on data symbols and effective subcarriers. The regularization coefficient of the MMS single-tap equalizer is the product of the median channel power and a fixed proportion, or is adaptively determined by the variance of the inactive subcarrier noise, and is limited to preset upper and lower bounds.

[0134] Specifically, the steps for performing regularized MMSE equalization are as follows: Single-tap MMSE equalization is performed based on data symbols and effective subcarriers. The regularization coefficient can be the product of the median channel power and a fixed proportion, or it can be adaptively determined by the variance of the inactive subcarrier noise and limited to preset upper and lower bounds.

[0135] ;

[0136] in, For balanced output, For the first The regularization coefficient of each data symbol.

[0137] For step S500, according to the discrete cosine basis and frequency dimension transformation order used by the transmitting end, the equalization data matrix is ​​transposed and inversely mapped to restore the data symbols before the frequency dimension orthogonal expansion.

[0138] For the steps of inverse discrete cosine mapping, according to the discrete cosine basis and frequency-dimensional transformation order used at the transmitting end, the equalized data symbols are arranged into an equalized data matrix with effective subcarriers as rows and data OFDM symbols as columns. Balanced data matrix The conjugate transpose of the DCT-II matrix yields the data grid recovered by inverse mapping.

[0139] Among them, the frequency-dimensional discrete cosine orthogonal extension of the transmitting end satisfies ;

[0140] The receiver inverse mapping satisfies ;

[0141] The normalized DCT-II matrix satisfies the unitary constraint.

[0142] In the formula, This represents the QPSK data grid to be sent. This represents the time-frequency data grid after discrete cosine orthogonal expansion in the frequency dimension. Denotes an L-order normalized DCT-II matrix. Indicates that after subcarrier-by-subcarrier MMSE equalization, according to OK, A balanced data matrix arranged in columns. express The conjugate transpose of . This represents the data grid recovered by inverse mapping. express Identity matrix.

[0143] The construction and inverse mapping of discrete cosine orthogonal matrices are explained below.

[0144] The transmitting end uses a normalized discrete cosine transform matrix to orthogonally extend the frequency dimension of the data grid and maps the transformed symbols to each effective subcarrier.

[0145] After the receiver completes channel estimation and equalization, it applies the conjugate transpose of the discrete cosine transform matrix to the equalization symbol matrix, following the same frequency-dimensional transform order and matrix construction rules as the transmitter, to eliminate orthogonal spread and restore the original data grid. Since the normalized discrete cosine transform matrix is ​​a real orthogonal matrix, its conjugate transpose is the same as its transpose and equal to the inverse of the matrix.

[0146] Therefore, inverse mapping does not require calculating the numerical inverse of a general complex matrix, which can reduce the complexity of matrix operations at the receiver while preserving signal energy and reduce the impact of matrix inversion error on the data recovery result.

[0147] Transmitter frequency-dimensional DCT-II orthogonal spread:

[0148] ;

[0149] Receiver inverse mapping:

[0150] ;

[0151] Unitary / orthogonal constraints of normalized DCT-II matrices:

[0152] ;

[0153] because For real orthogonal matrices, we have:

[0154] ;

[0155] This constraint ensures that the total energy remains unchanged before and after the mapping, and allows the receiver to perform a lossless inverse mapping using the conjugate transpose.

[0156] ;

[0157] The normalization coefficient is:

[0158]

[0159] in, For the transformation order, For row and column indexes; when hour ,when hour This yields the real orthogonal DCT-II matrix. .

[0160] Reference Figure 8 The process of recovering the data symbols before frequency-dimensional orthogonal expansion includes the following residual phase normalization and bit recovery steps:

[0161] Step S510: Expand the received recovery grid obtained by inverse mapping into a complex symbol sequence by column, estimate the QPSK common phase, perform decision-guided gain normalization, and then perform QPSK hard decision, block deinterleaving and pseudo-random sequence descrambling; when the transmitter uses convolutional coding, perform Viterbi decoding.

[0162] Specifically, the received recovery grid is expanded into a complex symbol sequence column by column, the QPSK common phase is estimated, decision-guided gain normalization is performed, and then QPSK hard decision, block deinterleaving and pseudo-random sequence descrambling are completed; when the transmitter uses convolutional coding, Viterbi decoding is performed.

[0163] ;

[0164] in, The first after inverse mapping A QPSK symbol, Represents the sample mean. QPSK module / 2 Common Phase Estimation.

[0165] Reference Figure 9 The process of recovering the data symbols before frequency-dimensional orthogonal expansion also includes the following steps for fusing and receiving results:

[0166] Step S521: For multiple sets of complete received records corresponding to the same transmitted waveform, calculate the frame synchronization correlation strength, cyclic prefix correlation, pilot reconstruction consistency and equalization error vector magnitude as indicators, and combine the normalized indicators into a referenceless reception quality evaluation value.

[0167] Step S522: Sort the reception results according to the no-reference reception quality evaluation value and select the single demodulation result with the highest quality.

[0168] In this process, a bit-by-bit majority decision is performed on each decoded bit, or the common phase deviation between equalized symbols is eliminated first, and then symbol-level fusion is completed using the median, truncated mean, or weighted average method.

[0169] Specifically, for multiple complete received records corresponding to the same transmitted waveform, the frame synchronization correlation strength, cyclic prefix correlation, pilot reconstruction consistency, and equalization error vector magnitude are calculated separately. These indicators are then normalized and combined into a referenceless reception quality evaluation value. The received results are sorted according to the evaluation value, with the highest-quality single demodulation result selected as the output. To utilize the time diversity gain from repeated transmissions, a bit-by-bit majority decision can be performed on each decoded bit, or the common phase deviation between equalization symbols can be eliminated first, followed by symbol-level fusion using median, truncated mean, or weighted average methods to reduce the impact of abnormal received records on the recovery results. After processing, the final recovered data, error statistics for each and the fused data, quality evaluation results, and processing diagrams for synchronization, channel estimation, equalization, and constellation distribution are output.

[0170] Perform the above steps for multiple complete playbacks of the same transmitted content, and form a blind quality score based on the LFM correlation peak, cyclic prefix index, pilot matching index, and error vector amplitude, and select the result with the highest blind quality score; or perform a position-by-position robust averaging on the phase-aligned equalized QPSK symbols before demodulation.

[0171] An embodiment of this application provides an underwater acoustic communication receiving device, including a sampling interface, a memory, and a processor; the memory stores a program, up-sweep and down-sweep LFM templates, pilot signals, and discrete cosine matrix parameters; when the program is executed by the processor, it implements the underwater acoustic communication receiving method based on inverse discrete cosine transform as described above.

[0172] An embodiment of this application provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described underwater acoustic communication receiving method based on inverse discrete cosine transform.

[0173] In one specific embodiment, a mid-frequency underwater acoustic communication configuration is employed. The transmitting end has a sampling rate of 96 kHz, the receiving end has a raw sampling rate of 100 kHz, the operating frequency band is 8 to 16 kHz, the carrier center frequency is 12 kHz, and the baseband processing sampling rate is 16 kHz. Each orthogonal frequency division multiplexing symbol uses a 1024-point fast Fourier transform, the cyclic prefix length is 512 points, the data grid size is 512 rows and 48 columns, and the modulation method is orthogonal phase shift keying. Each frame uses two data symbols, with necessary intervals maintained between adjacent data symbols. A one-second linear frequency modulation (LFM) synchronization signal is set before each frame, and a 0.4-second guard interval is set between the LFM signal and the data segment. Odd-numbered frames use an up-sweep LFM signal, and even-numbered frames use a down-sweep signal obtained by time reversal of the up-sweep signal.

[0174] The transmitting end first encodes, interleaves, and modulates the bits to be transmitted to form a two-dimensional data grid. Then, orthogonal expansion is performed on the frequency dimension of the data grid using a normalized discrete cosine matrix, while the other dimension remains unchanged. After orthogonal mapping, pilot symbols and guard subcarriers are inserted at designated positions, followed by orthogonal frequency division multiplexing modulation, cyclic prefix addition, carrier modulation, and frame structure splicing to obtain the time-domain waveform used for underwater acoustic transmission.

[0175] The receiver first reads the received signal and the transmitter parameters, converting the 100 kHz sampled data into 96 kHz sampled data. Then, normalized correlation detection is performed using both upper and lower frequency sweep templates, and a complete frame chain is searched based on the alternating pattern of upper and lower frequency sweeps. Only when the correlation peak amplitude, inter-peak distance, arrangement order, and frame number simultaneously meet the set conditions is the corresponding signal segment determined as a valid reception record.

[0176] After frame chain identification, the receiver constructs linear frequency modulation (LFM) templates with different scaling ratios within a preset time scale range, and determines the optimal time scale parameters through coarse and fine searches. During the evaluation process, the correlation results of multiple frames of LFM signals are comprehensively utilized to reduce the impact of single-frame noise and multipath interference on the estimation results. Based on the estimated time scale parameters, the received signal is resampled and corrected to restore the received waveform to a time reference consistent with that of the transmitter.

[0177] After time-scale compensation, bandpass filtering, carrier down-conversion, and baseband conversion are performed on each frame of signal. The starting position of each orthogonal frequency division multiplexing (OFDM) symbol is further corrected using cyclic prefix correlation characteristics and pilot matching results, and the remaining carrier frequency offset is estimated. After frequency compensation, the cyclic prefix is ​​removed, and a fast Fourier transform is performed to obtain the received symbols on each effective subcarrier.

[0178] At the pilot symbol location, the discrete channel frequency response is estimated based on the known relationship between the transmitted and received pilots. For data symbols between adjacent pilots, interpolation is performed on the channel amplitude and the unwrapped phase to obtain the channel response corresponding to each data symbol. If necessary, local complex smoothing can be applied to the channel estimation results along the subcarrier direction to suppress pilot noise and isolated outliers. Subsequently, a minimum mean square error equalization method with regularization is used to recover the data subcarriers, avoiding significant noise amplification at deep fading locations.

[0179] After frequency domain equalization, all data subcarriers are rearranged into a two-dimensional data grid. The receiver performs a conjugate transpose inverse mapping on the equalized data grid according to the discrete cosine matrix used by the transmitter and its frequency-dimensional transformation order, thereby eliminating the frequency-dimensional orthogonal spread applied by the transmitter. Since the normalized discrete cosine matrix is ​​a real orthogonal matrix, its inverse mapping can be directly achieved through matrix transpose, without needing to calculate the numerical inverse of a general complex matrix.

[0180] During the offline configuration phase of the system, the dimension and orthogonality of the orthogonal mapping matrix at the transmitting end and the inverse mapping matrix at the receiving end are checked to ensure that the matrix order, normalization method, basis function arrangement and symbol convention are consistent.

[0181] After inverse mapping, the symbols undergo orthogonal phase shift keying (QPSK) phase correction, common phase normalization, and symbol decision sequentially. Then, in the reverse order from the transmitter, deinterleaving, descrambling, and channel decoding are performed to finally recover the original information bits. For multiple complete reception records of the same transmitted signal, the highest quality single result can be selected based on synchronization correlation strength, cyclic prefix correlation, pilot consistency, and error vector amplitude. Alternatively, robust fusion of multiple equalized symbols can be performed after common phase alignment to improve data recovery reliability.

[0182] It should be noted that those skilled in the art can adjust the filter order, pilot configuration, modulation order, time scale search range, equalization regularization parameters, and repeated reception fusion method without changing the above-mentioned key technical relationships. All of the above equivalent substitutions are part of the technical concept of this application.

[0183] The underwater acoustic communication receiving method based on the above discrete inverse cosine transform achieves the following beneficial effects:

[0184] (1) Ensure consistency of orthogonal mapping. The transmitting and receiving ends use the same normalized unitary matrix to complete the forward mapping and the conjugate transpose inverse mapping, respectively, and specify the order, normalization coefficients and element arrangement of the discrete cosine transform matrix. This can avoid data grid misalignment and persistent demodulation errors caused by inconsistent transformation scale, basis function order or symbol definition.

[0185] (2) Improve the reliability of multi-frame synchronization. By utilizing the alternating arrangement of the upper-sweep frequency LFM and the time-reversed lower-sweep frequency LFM, the type, number, amplitude, and spacing of related peaks are jointly constrained, thereby eliminating isolated spurious peaks caused by noise, multipath, or transient interference. Each frame of data is independently extracted based on the actual detection position of the corresponding LFM, without relying on a fixed frame interval, and can adapt to changes in silence length and local sampling loss during playback and recording.

[0186] (3) Enhance the adaptability to time scale changes caused by motion. A scale estimation method combining coarse search and fine search is adopted to reconstruct the LFM reference template under different time scaling ratios, and the optimal scale is determined by robust statistics of multi-frame matching results. This method can estimate non-integer sampling scale deviation and recover the time reference of the received waveform through resampling; after scale correction, band filtering is performed, which can reduce the mismatch between the filter frequency response and the band of the scaling signal, and reduce the impact of time compression or expansion on the orthogonality of OFDM subcarriers.

[0187] (4) Improve the stability of equalization under time-varying fading conditions. After obtaining the discrete channel response based on the pilot position, inter-symbol interpolation is performed on the channel amplitude and unwrapped phase to maintain the continuity of channel reconstruction under rapid phase changes. The equalization stage adopts a minimum mean square error single-tap equalizer with regularization term to avoid excessive noise amplification caused by direct inversion at deep fading subcarriers, thereby improving the stability of equalization symbols.

[0188] (5) A complete and feasible receiving and processing flow is formed. The complete frame chain detection, time scale compensation, symbol fine timing, frequency offset correction, pilot channel estimation, regularized equalization, orthogonal inverse mapping, common phase correction, deinterleaving, and channel decoding are incorporated into a unified processing chain. Each processing stage has clear input, output, and verification conditions, and can be directly used for the reception and recovery of replay, multi-frame OFDM, and discrete cosine orthogonal precoded underwater acoustic communication signals.

[0189] The above is a detailed description of the preferred embodiments of this application, but this application is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A method for receiving underwater acoustic communication based on inverse discrete cosine transform, characterized in that, include: Obtain the real-valued passband sequence and transmitting parameters collected by the hydrophone; Alternate matching is performed on the up-sweep and down-sweep linear frequency modulated (LFM) signal sequences in the real-value passband sequence, and valid frames are identified and multi-frame boundaries are reconstructed based on the arrangement of relevant peaks. By searching the LFM template at different time scales, the waveform compression or expansion caused by the relative motion between the transmitter and receiver is estimated, and the received signal is resampled and corrected. The symbol start position is corrected by combining the cyclic prefix correlation index and the pilot matching index. Channel response interpolation is performed using pilot amplitude and phase information. Frequency domain equalization is performed by a minimum mean square error single-tap equalizer with regularization term based on data symbols and effective subcarriers. Based on the discrete cosine basis and frequency-dimensional transformation order used at the transmitting end, the equalized data matrix is ​​transposed and inversely mapped to restore the data symbols before the frequency-dimensional orthogonal expansion.

2. The underwater acoustic communication receiving method based on inverse discrete cosine transform according to claim 1, characterized in that, Prior to the alternating matching step, the method further includes: Establish the input interface between received data and system parameters using the following steps: Read the received data file; Check the number and data types of variables in the received data file, and verify the integrity of each configuration in the parameter file; Check the matching relationship between the orthogonal transformation matrix and the data grid, the number of effective subcarriers, and the frequency dimension transformation order, and calculate the orthogonality error of the transformation matrix; When the number of variables, parameter dimensions, or orthogonality error does not meet the preset conditions, the subsequent processing is terminated and the corresponding abnormal information is output.

3. The underwater acoustic communication receiving method based on inverse discrete cosine transform according to claim 1, characterized in that, The step of alternately matching the up-scanning and down-scanning LFM sequences in the received signal, identifying valid frames and reconstructing multi-frame boundaries based on the arrangement of relevant peaks, includes: The received signal is normalized and matched with the upper-sweep LFM template and the lower-sweep LFM template respectively to obtain two types of synchronization correlation sequences and their corresponding candidate peak positions; Based on the pre-set alternation pattern of up-scan and down-scan frequencies at the transmitting end, candidate peaks are combined in time sequence and filtered group by group. When the following conditions are met: the peak amplitude reaches the detection threshold, the relative difference of peaks within the group is within the allowable range, the interval between adjacent peaks conforms to the frame structure constraints, and the number of detected peaks is consistent with the number of frames contained in a complete playback, the peak group is determined as a valid LFM frame chain, and the initial boundary of each frame in this playback is determined. Isolated peaks, out-of-order peaks, and incomplete peak groups that do not meet the aforementioned joint conditions are removed.

4. The underwater acoustic communication receiving method based on inverse discrete cosine transform according to claim 1, characterized in that, By searching LFM templates at different time scales, waveform compression or expansion caused by relative motion between the transmitter and receiver is estimated, and resampling correction is performed on the received signal, including: By using LFM template search at different time scales, coarse search sets and fine search sets for candidate time scales are established around zero scale bias. Scaling the LFM template for each candidate time scale and using the median of the LFM correlation peaks of all frames of the correct type as the score, we estimate the waveform compression or expansion caused by the relative motion between the transmitter and receiver. Select the time scale with the highest score; When the deviation of the time scale with the largest score from 1 exceeds a set threshold, reverse linear interpolation correction is performed on the entire received waveform of the received signal.

5. The underwater acoustic communication receiving method based on inverse discrete cosine transform according to claim 1, characterized in that, The joint cyclic prefix correlation index and pilot matching index correct the symbol start position, including: The load segment is extracted based on the independently detected LFM start point of each frame. The load segment is multiplied by the carrier complex exponent and resampled to the baseband sampling rate. Candidate offsets are searched around the prediction starting point, and OFDM symbol boundaries are selected using a weighted average of the cyclic prefix correlation index and the pilot matching index to correct the symbol starting position.

6. The underwater acoustic communication receiving method based on inverse discrete cosine transform according to claim 1, characterized in that, After correcting the symbol start position using the joint cyclic prefix correlation index and pilot matching index, the method includes: The correlation between the cyclic prefix and the corresponding tail sample of the OFDM symbol in a frame is accumulated to estimate the residual common frequency offset and correct the residual common frequency offset. After removing the cyclic prefix, a Fast Fourier Transform is performed to retain the valid subcarriers.

7. The underwater acoustic communication receiving method based on inverse discrete cosine transform according to claim 1, characterized in that, Channel response interpolation using pilot amplitude and phase information includes: Extract the received values ​​of the pilot OFDM symbols on each effective subcarrier, and perform a complex ratio operation between the received values ​​and the corresponding known pilot symbols to obtain the discrete channel frequency response at the pilot time. For data symbols located between adjacent pilot symbols, the channel responses at both ends are decomposed into amplitude and phase components. The phase sequence is unwrapped, and the amplitude and continuous phase are interpolated along the OFDM symbol direction to reconstruct the complex channel response corresponding to each data symbol. For subcarriers with isolated abrupt changes in frequency direction or with strong estimated noise, the complex channel response is locally smoothed.

8. The underwater acoustic communication receiving method based on inverse discrete cosine transform according to claim 1, characterized in that, The regularization coefficient of the regularization term of the minimum mean square error single-tap equalizer is the product of the median channel power and a fixed proportion or is adaptively determined by the variance of the inactive subcarrier noise, and is limited to preset upper and lower bounds.

9. The underwater acoustic communication receiving method based on inverse discrete cosine transform according to claim 1, characterized in that, Based on the discrete cosine basis and frequency-dimensional transform order used at the transmitting end, the equalization data matrix is ​​transposed and inversely mapped, including: Based on the discrete cosine basis and frequency-dimensional transform order used at the transmitting end, the equalized data symbols are arranged into an equalized data matrix with effective subcarriers as rows and data OFDM symbols as columns. The equalized data matrix is ​​then multiplied by the conjugate transpose of the normalized DCT-II matrix to obtain the inversely mapped data grid. Among them, the frequency-dimensional discrete cosine orthogonal extension of the transmitting end satisfies ; The receiver inverse mapping satisfies ; The normalized DCT-II matrix satisfies the unitary constraint. In the formula, This represents the QPSK data grid to be sent. This represents the time-frequency data grid after discrete cosine orthogonal expansion in the frequency dimension. Denotes an L-order normalized DCT-II matrix. Indicates that after subcarrier-by-subcarrier MMSE equalization, according to OK, A balanced data matrix arranged in columns. express The conjugate transpose of . This represents the data grid recovered by inverse mapping. express Identity matrix.

10. The underwater acoustic communication receiving method based on inverse discrete cosine transform according to claim 1, characterized in that, The data symbols before the frequency dimension orthogonal expansion are recovered, including: The received recovery grid obtained by inverse mapping is expanded into a complex symbol sequence by column, the QPSK common phase is estimated, decision-guided gain normalization is performed, and then QPSK hard decision, block deinterleaving and pseudo-random sequence descrambling are performed; when the transmitter uses convolutional coding, Viterbi decoding is performed. For multiple complete received records corresponding to the same transmitted waveform, a no-reference quality evaluation is performed based on frame synchronization correlation strength, cyclic prefix correlation, pilot reconstruction consistency, and equalization error vector magnitude. At least one received result is selected based on the quality evaluation results, or multiple received results are fused.

Images