Underwater sonar signal enhancement method based on rls adaptive filter

Abstract

The application relates to an underwater sonar signal enhancement method based on an RLS adaptive filter, which comprises the following steps: step S1, receiving an original sonar signal through an RLS adaptive filter module and performing real-time filtering processing; step S2, collecting underwater environment parameters by using an environment perception module to generate an environment feature vector; step S3, dynamically adjusting a forgetting factor lambda and an initial covariance matrix P(0) of the RLS adaptive filter module according to the environment feature vector through a parameter feedback adjustment module; and step S4, updating the weight of the RLS adaptive filter module by using the adjusted parameters, performing residual noise processing on the filtered signal, and outputting an enhanced signal. Through the introduction of a feedback adjustment mechanism based on environment perception data, the application realizes the dynamic optimization of key parameters of the RLS algorithm, thereby improving the signal enhancement capability and stability of an underwater communication system in a complex environment.

Description

Technical Field

[0001] This invention relates to the field of sonar signal enhancement technology, and in particular to an underwater sonar signal enhancement method based on an RLS adaptive filter. Background Technology

[0002] Underwater communication environments are unique and complex. Sound waves propagating in seawater are significantly affected by hydrological parameters such as temperature, salinity, and pressure, leading to uneven sound velocity distribution and severe propagation attenuation. Simultaneously, they face various complex noise interferences, including marine biological noise, turbulent water flow noise, and mechanical vibration noise, making sonar signals highly susceptible to distortion and attenuation. Traditional sonar signal processing methods often employ fixed-parameter filter structures, whose filter parameters cannot be dynamically adjusted according to the environment. In complex scenarios, it is difficult to balance convergence speed and steady-state error, resulting in poor signal enhancement.

[0003] Adaptive filtering techniques have garnered significant attention in sonar signal processing due to their ability to dynamically adjust filter parameters to adapt to environmental changes. Among these, the Recursive Filtering (RLS) algorithm, which recursively updates filter weights by minimizing the sum of squared weighted errors, boasts advantages such as fast convergence and good tracking performance, and has been applied in areas such as channel estimation and noise suppression. Current improvements to the RLS algorithm primarily focus on the field of wireless communication.

[0004] For example, Chinese invention patent with publication number CN101001218A discloses a blind method for channel estimation in wireless communication systems. It uses time-varying step size and ideal decision weighting techniques to improve the convergence speed and estimation accuracy of the RLS algorithm. However, its application scenario is wireless communication systems such as OFDM and CDMA, which are designed for electromagnetic wave transmission environments and do not consider the special characteristics of underwater acoustic wave transmission.

[0005] For example, Chinese invention patent with publication number CN101001219A proposes a fast-convergence adaptive blind estimation method for feature parameters. It improves the tracking ability of the RLS algorithm for time-varying features by using time-varying step size and soft decision weighting. However, its core is for feature parameter estimation in wireless communication and does not involve the correlation mechanism between underwater environmental parameters and algorithm parameters.

[0006] However, due to the special nature of underwater sonar applications, parameters such as seawater temperature, salinity, and pressure can significantly alter the propagation speed and attenuation characteristics of sound waves, resulting in stronger time-varying channel characteristics and more complex models, which differ significantly from the stationary channel characteristics of wireless communication. Underwater noise includes marine biological noise (such as whale calls and fish disturbances), water flow noise (turbulence and eddies), and mechanical noise (vibration of submersible engine), which has a wide spectral distribution, large intensity fluctuations, and nonlinear and impulse characteristics.

[0007] Therefore, designing an underwater sonar signal enhancement method that combines environmental perception, dynamic parameter optimization, and multi-dimensional noise suppression has become the key to improving the performance of underwater communication systems. Summary of the Invention

[0008] To address the technical problems existing in the prior art, the present invention aims to provide an underwater sonar signal enhancement method based on an RLS adaptive filter. By introducing an environmental perception module and a parameter feedback adjustment mechanism, the system can dynamically adjust the RLS algorithm parameters to adapt to the complex and ever-changing underwater environment. Combining nonlinear noise suppression and hardware acceleration optimization, the system achieves efficient real-time signal processing while maintaining low power consumption, providing a stable and reliable technical solution for underwater communication.

[0009] To achieve the above-mentioned objectives, this invention provides a method for enhancing underwater sonar signals based on an RLS adaptive filter, comprising the following steps:

[0010] Step S1: Receive the raw sonar signal through the RLS adaptive filter module and perform real-time filtering processing;

[0011] Step S2: Collect underwater environmental parameters using the environmental perception module and generate an environmental feature vector;

[0012] Step S3: The parameter feedback adjustment module dynamically adjusts the forgetting factor λ and the initial covariance matrix P(0) of the RLS adaptive filter module according to the environmental feature vector.

[0013] Step S4: Update the weights of the RLS adaptive filter module using the adjusted parameters, process the residual noise of the filtered signal, and output the enhanced signal.

[0014] According to one technical solution of the present invention, the environmental sensing module includes:

[0015] A multi-dimensional sensor array, including at least a temperature sensor, a pressure sensor, a salinity meter, and a noise detector;

[0016] High-precision analog-to-digital converters are used to convert analog signals into digital data;

[0017] The data fusion unit extracts features from the normalized environmental parameters to generate environmental feature vectors.

[0018] According to a technical solution of the present invention, step S2, generating an environmental feature vector, specifically includes:

[0019] Time-domain feature extraction is achieved by calculating the signal variance, as shown below:

[0020]

[0021] in, This represents the variance, and N represents the total number of samples. Indicates sample value, This represents the mean;

[0022] Frequency domain feature extraction is achieved by extracting spectral components using Discrete Fourier Transform, as shown below:

[0023]

[0024] Where x(n) represents a discrete sampling sequence in the time domain with a length of N, and a frequency domain index with a value range of 0 to N. 1. For N discrete frequency points, X(k) is a discrete sampling sequence in the frequency domain, containing the amplitude and phase information of the signal;

[0025] Feature clustering analysis is used to pre-set the initial covariance matrix P(0).

[0026] According to a technical solution of the present invention, in step S3, based on the environmental feature vector, a nonlinear function mapping method is used to dynamically adjust the forgetting factor and the initial covariance matrix in the RLS adaptive filter module;

[0027] The forgetting factor is adaptively adjusted based on the background noise intensity and the rate of environmental change;

[0028] The initial value of the covariance matrix is ​​preset through environmental feature clustering analysis;

[0029] A sliding window mechanism is introduced to weightedly fuse historical environmental states;

[0030] The adjustment process employs a low-latency control strategy to ensure that parameter updates are synchronized with the signal processing process.

[0031] According to one technical solution of the present invention, step S4 specifically includes:

[0032] Step S41: Update the weights of the RLS adaptive filter module using the adjusted forgetting factor and the initial covariance matrix to optimize convergence performance and steady-state error;

[0033] Step S42: Perform residual noise processing on the filtered signal and output the enhanced sonar signal;

[0034] Steps S41 and S42 are executed using hardware acceleration.

[0035] According to one technical solution of the present invention, step S41 specifically includes:

[0036] Step S411: Define the cost function as the sum of the weighted sum of squared errors and the regularization term;

[0037] Step S412: Calculate the gain vector ;

[0038] Step S413: Update the inverse correlation matrix ;

[0039] Step S414: Calculate the prior estimation error ;

[0040] Step S415: Update the weight coefficients of the RLS adaptive filter module. .

[0041] According to one technical solution of the present invention, step S42 specifically includes:

[0042] Step S421: Use a Volterra filter to extend and capture the nonlinear echo path. The output is:

[0043]

[0044] Where y(n) represents the output signal and x(n) represents the input signal. The p-th order kernel coefficients are represented by M, and the memory length is represented by M.

[0045] Step S422: When impulse noise is present, the M-estimation method is used instead of the least squares method, and the objective function is defined as:

[0046]

[0047] in, Represents sample data, This represents the loss function.

[0048] According to one aspect of the present invention, an underwater sonar signal enhancement system based on an RLS adaptive filter is proposed, comprising:

[0049] The RLS adaptive filter module is configured to receive the raw sonar signal and output the filtered signal.

[0050] The environmental sensing module is configured to collect underwater environmental parameters in real time, including temperature, pressure, and background noise intensity.

[0051] The parameter feedback adjustment module is configured to dynamically adjust the forgetting factor λ and the initial covariance matrix P(0) of the RLS algorithm based on the environmental parameters.

[0052] The nonlinear noise suppression module uses a neural network to suppress residual noise in the filtered signal.

[0053] According to one aspect of the present invention, an electronic device is provided, comprising: one or more processors, one or more memories, and one or more computer programs; wherein the processor is connected to the memory, and the one or more computer programs are stored in the memory, and when the electronic device is running, the processor executes the one or more computer programs stored in the memory to cause the electronic device to perform the underwater sonar signal enhancement method based on RLS adaptive filter as described in any of the above technical solutions.

[0054] According to one aspect of the present invention, a computer-readable storage medium is provided for storing computer instructions, which, when executed by a processor, implement the underwater sonar signal enhancement method based on an RLS adaptive filter as described in any of the above technical solutions.

[0055] Compared with the prior art, the present invention has the following advantages:

[0056] This invention proposes an underwater sonar signal enhancement method based on RLS adaptive filters. By introducing a feedback adjustment mechanism based on environmental perception data, the key parameters of the RLS algorithm are dynamically optimized, thereby improving the signal enhancement capability and stability of the underwater communication system in complex environments.

[0057] This invention, by constructing a closed-loop processing system of "environmental perception - parameter feedback adjustment - RLS adaptive filtering - nonlinear noise suppression", achieves a comprehensive improvement in sonar signal enhancement performance in the complex environment of underwater sonar signal propagation. Compared with existing technologies, its core technological advantages are: significantly enhanced environmental adaptability; real-time acquisition of key underwater environmental parameters and dynamic adjustment of the core parameters of the RLS algorithm, solving the problem of poor environmental adaptability caused by parameter adjustment lag in traditional methods, maintaining stable filtering performance even in scenarios with sudden environmental changes; comprehensive improvement in noise suppression capabilities, effectively suppressing underwater nonlinear echoes and impulse noise with targeted strategies while preserving target signal energy, significantly improving the signal-to-noise ratio of the output signal; a balance between convergence speed and steady-state accuracy, achieving a balance between fast convergence and low steady-state error through time-varying parameter adjustment strategies and refined weight update mechanisms; real-time performance meeting underwater communication requirements, employing hardware acceleration design, supporting multi-channel signal parallel processing, and adaptable to the real-time communication requirements of mobile platforms; and excellent system compatibility and scalability, with modular design supporting seamless integration with existing sonar systems, and a multi-node collaborative mechanism improving long-distance transmission stability, enabling wide application in diverse underwater scenarios.

[0058] Furthermore, the environmental perception module achieves comprehensive and accurate perception of underwater environmental parameters through multi-dimensional sensor arrays, high-precision conversion, and multi-dimensional feature extraction. Compared with single-parameter perception schemes, it more completely depicts environmental characteristics, providing a rich data foundation for subsequent parameter adjustment. Preprocessing and normalization improve data reliability, and feature fusion and cluster analysis make the initial parameter presets more in line with the actual environment, avoiding performance fluctuations of traditional experience-based presets.

[0059] Meanwhile, the parameter feedback adjustment module achieves dynamic optimization of RLS parameters through nonlinear mapping, sliding window fusion, and low-latency control. The forgetting factor adaptively adjusts with environmental changes, improving the response speed to channel mutations. The precise preset of the initial covariance matrix balances convergence speed and steady-state accuracy. The sliding window mechanism effectively filters out transient outliers, ensuring stable parameter adjustment. The real-time guarantee mechanism avoids performance loss caused by adjustment lag. RLS weight update and residual noise processing improve signal enhancement quality through refined processes and multi-dimensional strategies. The weight update process is efficient, avoiding numerical instability and reducing computational redundancy. The nonlinear noise suppression strategy specifically addresses multipath distortion and impulse interference. Hardware acceleration design solves the real-time bottleneck, reduces power consumption, and improves processing throughput.

[0060] In addition, the nonlinear noise suppression module uses a lightweight neural network to achieve intelligent suppression of residual noise. It has the ability to generalize and suppress complex noise through training with real data. The online learning mechanism can track changes in noise characteristics. The low-latency design meets the requirements of real-time communication and can be directly deployed on resource-constrained platforms.

[0061] This invention is applicable to mobile platforms such as underwater unmanned underwater vehicles (AUVs) and can greatly improve the transmission quality of long-distance underwater sonar signals, enabling adaptive response to complex underwater noise environments. Attached Figure Description

[0062] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly described below. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0063] Figure 1 This is a flowchart illustrating the underwater sonar signal enhancement method based on an RLS adaptive filter in an embodiment of the present invention.

[0064] Figure 2 This is a schematic diagram of the underwater sonar signal enhancement system based on the RLS adaptive filter in an embodiment of the present invention. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] Example 1

[0067] This embodiment details the complete implementation process of an underwater sonar signal enhancement method based on an RLS adaptive filter, which is implemented using an underwater sonar signal enhancement system. (Refer to...) Figure 2 The hardware environment of the underwater sonar signal enhancement system based on the RLS adaptive filter in this embodiment can include, but is not limited to, servers, industrial control computers, or embedded devices. Its core consists of one or more processors and a memory. The memory stores computer program instructions for implementing the method of this invention. The system is logically divided into multiple functional modules, which work together to complete the intelligent diagnostic task.

[0068] like Figure 2 As shown, the underwater sonar signal enhancement system based on the RLS adaptive filter mainly includes: an RLS adaptive filter module, an environmental perception module, a parameter feedback adjustment module, and a nonlinear noise suppression module.

[0069] The RLS adaptive filter module is configured to receive the raw sonar signal and output the filtered signal.

[0070] The RLS adaptive filter module is built based on the Recursive Least Squares (RLS) algorithm, receiving raw underwater sonar signals and performing real-time filtering. The filter's weight vector is dynamically updated by minimizing the sum of squared weighted errors, ensuring rapid response to time-varying channels and noisy environments. This module supports multi-channel input, enabling it to process signals from multiple hydrophones and improving spatial noise immunity. The RLS algorithm uses an exponential weighting factor to control the influence range of historical data, improving adaptability in non-stationary environments.

[0071] Meanwhile, the RLS adaptive filter module's internal embedded parameter interface supports the dynamic modification of the forgetting factor and covariance matrix by an external feedback adjustment mechanism.

[0072] The environmental sensing module is configured to collect underwater environmental parameters in real time, including temperature, pressure, and background noise intensity.

[0073] The environmental sensing module integrates multi-dimensional sensors to collect key environmental parameters in real time, such as underwater temperature, pressure, salinity, and background noise intensity. A high-precision analog-to-digital converter (ADC) converts the analog signals into digital data, followed by preliminary filtering and normalization. The collected environmental data undergoes feature extraction using a data fusion algorithm to generate environmental feature vectors for parameter adjustment.

[0074] The environmental sensing module has a self-calibration function, which can maintain measurement accuracy under different depths and temperature gradients. The data obtained by the environmental sensing module is synchronously transmitted with the parameter feedback adjustment module through a high-speed interface to ensure real-time requirements.

[0075] The parameter feedback adjustment module is configured to dynamically adjust the forgetting factor λ and the initial covariance matrix P(0) of the RLS algorithm based on the environmental parameters.

[0076] Based on the multi-dimensional feature vector output by the environmental perception module, a nonlinear function mapping method is used to dynamically adjust the forgetting factor and initial covariance matrix in the RLS algorithm. The forgetting factor is adaptively adjusted according to the background noise intensity and the rate of environmental change, improving the convergence performance of the algorithm under sudden disturbances. The initial value of the covariance matrix is ​​preset through environmental feature clustering analysis, enhancing the stability of the algorithm in the initial stage. A sliding window mechanism is introduced into the feedback adjustment mechanism to perform weighted fusion of historical environmental states, avoiding parameter oscillations caused by transient outliers. The adjustment process adopts a low-latency control strategy to ensure that parameter updates are synchronized with the signal processing process.

[0077] In this invention, based on a parameter feedback adjustment module, an environmental perception factor is introduced into the standard RLS algorithm to construct a weighted error function that reflects the dynamic characteristics of the current underwater environment. The weight update equation is improved to enhance its robustness in high-noise environments. A sparsity processing mechanism is introduced to reduce computational redundancy while maintaining filtering performance. The algorithm employs a fixed-point arithmetic optimization strategy to adapt to the low-power requirements of embedded platforms. Simultaneously, time-frequency analysis methods are combined to track the signal spectrum characteristics in real time, further improving the ability to suppress non-stationary noise.

[0078] The nonlinear noise suppression module uses a neural network to suppress residual noise in the filtered signal.

[0079] The nonlinear noise suppression module employs a neural network-based nonlinear mapping structure to further process the residual noise after RLS filtering. The network structure includes a multilayer perceptron (MLP) or a lightweight convolutional neural network (CNN), optimized using real underwater noise data during the training phase. The module supports an online learning mechanism, enabling dynamic adjustment of the noise model based on current environmental characteristics. The nonlinear processing module and the RLS filter form a cascaded structure, improving overall signal enhancement performance through a joint optimization strategy. This invention's nonlinear noise suppression module features low latency and is suitable for real-time underwater communication systems.

[0080] This invention relates to an underwater sonar signal enhancement system based on an RLS adaptive filter. To meet the real-time requirements of underwater communication, a lightweight RLS computational structure is designed, employing a matrix factorization optimization strategy to reduce the complexity of recursive computation. A parallel computing architecture is introduced to pipeline key computation modules, improving overall throughput. At the hardware implementation level, an FPGA platform is used to accelerate the execution of the RLS algorithm, supporting hardware-level parallel computing and low-power operation. A dedicated instruction set is designed to accelerate key operations, such as matrix inversion and vector multiplication. The system supports dynamic frequency adjustment, automatically adjusting the processing speed according to the current load, balancing performance and energy consumption.

[0081] Example 2

[0082] like Figure 1 As shown, this invention provides a method for enhancing underwater sonar signals based on an RLS adaptive filter, comprising the following steps:

[0083] Step S1: Receive the raw sonar signal through the RLS adaptive filter module and perform real-time filtering processing;

[0084] Specifically, the RLS adaptive filter module receives raw sonar signals from an underwater sonar array (such as a multi-hydrophone array). These signals include target echo signals, environmental noise (marine biological noise, current noise, etc.), and multipath interference signals. The filter performs real-time filtering based on the RLS algorithm. Its core principle is to minimize the sum of squared weighted errors by recursively updating the weight coefficients, thereby achieving adaptive tracking of the target signal and noise suppression. Initially, the filter weight coefficients w(0) are set to zero vectors, and the initial inverse correlation matrix P(0) is preset through environmental feature clustering analysis.

[0085] Step S2: Collect underwater environmental parameters using the environmental perception module and generate an environmental feature vector;

[0086] Step S3: The parameter feedback adjustment module dynamically adjusts the forgetting factor λ and the initial covariance matrix P(0) of the RLS adaptive filter module according to the environmental feature vector.

[0087] Step S4: Update the weights of the RLS adaptive filter module using the adjusted parameters, process the residual noise of the filtered signal, and output the enhanced signal.

[0088] By constructing a closed-loop processing flow of "RLS adaptive filtering - environmental perception - parameter feedback adjustment - residual noise processing", dynamic optimization of the entire process of underwater sonar signal enhancement is achieved. Compared with existing fixed-parameter RLS filtering methods, its core advantage lies in: capturing underwater environmental changes in real time through the environmental perception module, and dynamically adjusting the key parameters of the RLS algorithm (forgetting factor λ and initial covariance matrix P(0)) through the parameter feedback adjustment module, thus solving the problem of poor environmental adaptability caused by parameter adjustment lag in traditional methods. At the same time, combined with the residual noise processing step, the complex noise remaining after filtering is further suppressed, significantly improving the signal-to-noise ratio and clarity of the sonar signal. This method can adaptively match the dynamic changes of environmental parameters such as underwater temperature, pressure, and noise, and exhibits better signal enhancement stability and robustness in complex scenarios such as deep-sea exploration and underwater search and rescue.

[0089] In some embodiments of the present invention, the environment sensing module includes:

[0090] A multi-dimensional sensor array, including at least a temperature sensor, a pressure sensor, a salinity meter, and a noise detector;

[0091] High-precision analog-to-digital converters are used to convert analog signals into digital data;

[0092] The data fusion unit extracts features from the normalized environmental parameters to generate environmental feature vectors.

[0093] The environmental sensing module is used to collect underwater environmental parameters in real time, including temperature, pressure, salinity, and background noise intensity. Data is updated at a frequency of 10Hz, and the collected data is normalized to generate environmental feature vectors, ensuring the accuracy and reliability of underwater environmental parameter acquisition and processing. A multi-dimensional sensor array covers key parameters such as temperature, pressure, salinity, and noise, achieving comprehensive perception of factors affecting underwater sound propagation. A high-precision analog-to-digital converter (e.g., 16-bit resolution) ensures high-precision conversion of analog signals to digital data, reducing signal quantization errors. The data fusion unit, through normalization and feature extraction, transforms the raw environmental parameters into feature vectors that can be directly used for parameter adjustment, providing high-quality input for subsequent parameter feedback and adjustment. Compared to existing single-parameter sensing or low-precision conversion solutions, this module can more comprehensively and accurately characterize underwater environmental features, providing a data foundation for the dynamic optimization of RLS algorithm parameters, thereby indirectly improving the overall performance of sonar signal enhancement.

[0094] Furthermore, multi-dimensional sensor arrays can achieve hydrological environment perception, while multibeam echo sounders, sound velocity meters and other equipment can be used to obtain information such as water depth and underwater sound velocity at the location of the sonar, thus achieving geographic environment perception; the undulation of the seabed topography will cause changes in the reflection and scattering characteristics of sound waves, thereby affecting the detection effect of sonar.

[0095] For example, the temperature sensor can be a high-precision PT1000 sensor (measurement range -50℃~150℃, accuracy ±0.05℃), the pressure sensor can be an MS5837-30BA (measurement range 0~300bar, accuracy ±0.1%FS), the salinity meter can be a HACH sension+ EC700 (measurement range 0~70psu, accuracy ±0.1psu), and the noise detector can be a B&K8104 (measurement range 40~120dB, frequency range 2Hz~20kHz), with a sampling frequency of 10Hz for all sensors.

[0096] High-precision analog-to-digital converters (ADCs) can use high-precision 16-bit ADCs (such as AD7606) to convert analog signals acquired by sensors into digital data with a conversion rate ≥1MHz, ensuring high-precision preservation of the original data.

[0097] In some embodiments of the present invention, step S2, generating an environmental feature vector, specifically includes:

[0098] Time-domain feature extraction is achieved by calculating the signal variance, as shown below:

[0099]

[0100] in, This represents the variance, and N represents the total number of samples. Indicates sample value, This represents the mean;

[0101] Frequency domain feature extraction is achieved by extracting spectral components using Discrete Fourier Transform, as shown below:

[0102]

[0103] Where x(n) represents a discrete sampling sequence in the time domain (usually a sequence of real numbers), with a length of N, and a frequency domain index, with a value range of 0 to N. 1. For N discrete frequency points, X(k) is a discrete sampling sequence (complex sequence) in the frequency domain, containing the amplitude and phase information of the signal;

[0104] Feature clustering analysis is used to pre-set the initial covariance matrix P(0).

[0105] By employing time-domain feature extraction (signal variance calculation), frequency-domain feature extraction (Discrete Fourier Transform), and feature clustering analysis, a multi-dimensional and in-depth construction of environmental feature vectors is achieved. Specifically, time-domain variance features reflect the intensity of environmental parameter fluctuations and can distinguish target materials (such as the difference in reflection between metal and rock), while frequency-domain DFT features capture the spectral distribution characteristics of parameters. The combination of these two features provides a comprehensive description of the dynamic characteristics of the environment. Feature clustering analysis, by classifying similar environmental features, provides a data-driven basis for presetting the initial covariance matrix P(0). Compared with existing schemes that rely solely on single-dimensional features or empirically pre-set initial parameters, this scheme generates richer and more discriminative environmental feature vectors, making the pre-set initial covariance matrix P(0) more closely aligned with the actual environment. This effectively improves the stability and convergence speed of the RLS algorithm in its initial stage and reduces the filtering performance degradation caused by initial parameter mismatch.

[0106] Specifically, for the filter coefficients: w(0) = 0 or initialized based on prior knowledge; for the inverse correlation matrix: P(0) = δ⁻¹I, where I is the identity matrix and δ is adjusted according to the noise level.

[0107] In some embodiments of the present invention, in step S3, based on the environmental feature vector, the forgetting factor and the initial covariance matrix in the RLS adaptive filter module are dynamically adjusted using a nonlinear function mapping method, thereby realizing the fine-grained and real-time adjustment of the RLS algorithm parameters (forgetting factor λ and initial covariance matrix P(0)).

[0108] The forgetting factor is adaptively adjusted according to the background noise intensity and the rate of environmental change; the nonlinear function mapping ensures the accurate correlation between environmental features and parameters, avoiding the limitations of linear mapping; the forgetting factor is adaptively adjusted with the background noise intensity and the rate of environmental change, which accelerates convergence during sudden disturbances and ensures steady-state accuracy in stable environments.

[0109] The forgetting factor λ ranges from 0.95 to 0.99, and its value is negatively correlated with the background noise intensity and the rate of environmental change (the stronger the noise and the faster the environmental change, the smaller λ becomes, thus accelerating the convergence speed). The mapping function is: in, Normalized noise intensity (0≤ ≤1), The normalized rate of environmental change (calculated by the Euclidean distance between eigenvectors of adjacent time points, 0 ≤ ... ≤1), , This is an adjustment coefficient. For example, when the background noise intensity increases sharply ( And the environment changes drastically. When the environment is stable, λ≈0.95, ensuring the filter quickly tracks channel changes; when the environment is stable ( And the noise level is low. When λ≈0.99, the steady-state error is reduced.

[0110] The initial value of the covariance matrix is ​​preset through environmental feature clustering analysis;

[0111] For example, a preset base value of P(0) is selected based on the clustering results in step S2, and then fine-tuned according to the distance between the real-time feature vector and the cluster center, as shown in the formula: in, The fundamental matrix corresponding to the cluster centers. The normalized distance between the real-time feature vector and the cluster center (0≤ ≤1), These are the fine-tuning coefficients. For example, if the real-time feature vector deviates significantly from the cluster center ( If P(0) is increased by 16%, the algorithm's adaptability to atypical environments will be enhanced.

[0112] Introducing a sliding window mechanism to weightedly fuse historical environmental states can effectively suppress parameter oscillations caused by transient outliers;

[0113] The adjustment process employs a low-latency control strategy to ensure that parameter updates are synchronized with the signal processing process, which helps to avoid adjustment lag.

[0114] Compared with existing schemes using fixed parameters or simple time-varying parameters, this scheme significantly improves the RLS algorithm's ability to track complex dynamic underwater environments, maintains good filtering performance in scenarios with sudden environmental changes, and reduces steady-state errors.

[0115] In this application, the adaptive filter module employs the recursive least squares (RLS) algorithm to receive multi-channel signals from the AUV sonar array. The active sonar equation is expressed as: Performance optimization was performed, including For detection threshold, At the sound source level, To spread the loss, For target strength, To achieve the equivalent plane wave reverberation level, the target echo signal is enhanced. With suppressing reverberation level Improve detection performance.

[0116] Sonar receiving signals y ( t This can be represented as:

[0117]

[0118] in, s ( t (The target echo) n ( t This refers to environmental noise. e ( t The system error is represented by (). The RLS algorithm iteratively updates the filter coefficients by minimizing the sum of squared errors. w ( t ):

[0119]

[0120] in x ( t ) represents the input signal vector (such as historical echo data).

[0121] In some embodiments of the present invention, step S4 specifically includes:

[0122] Step S41: Update the weights of the RLS adaptive filter module using the adjusted forgetting factor and the initial covariance matrix to optimize convergence performance and steady-state error. Specifically, this includes:

[0123] Step S411: Define the cost function as the sum of the weighted sum of squared errors and the regularization term, with the following formula:

[0124] ;

[0125] Step S412: Calculate the gain vector Then we have:

[0126]

[0127] in, This is the inverse correlation matrix of the previous time step, reflecting the autocorrelation characteristics of the input signal;

[0128] Step S413: Update the inverse correlation matrix Then we have:

[0129] ;

[0130] Step S414: Calculate the prior estimation error Then we have:

[0131]

[0132] in, The expected output;

[0133] Step S415: Update the weight coefficients of the RLS adaptive filter module. Then we have:

[0134] .

[0135] By standardizing the specific steps of RLS weight update (cost function definition, gain vector calculation, inverse correlation matrix update, prior estimation error calculation, and weight coefficient update), the rigor and efficiency of the RLS algorithm are ensured. Specifically, the introduction of a regularization term in the cost function effectively avoids numerical instability issues during matrix inversion; the recursive update of the gain vector and inverse correlation matrix reduces computational complexity and improves iteration efficiency; and the accurate calculation of the prior estimation error provides a reliable basis for weight coefficient updates. Compared to existing simplified or non-standardized RLS implementations, this invention strictly adheres to the mathematical logic of the RLS algorithm, ensuring the accuracy and convergence speed of weight updates, enabling the filter to quickly converge to the optimal state, reducing steady-state error, and thus improving the filtering quality of sonar signals.

[0136] Step S42: Perform residual noise processing on the filtered signal to output the enhanced sonar signal.

[0137] The filtered signal is subjected to nonlinear noise and impulse noise suppression to further improve signal purity. Specifically, this includes:

[0138] Step S421: Use a second-order Volterra filter to capture the nonlinear echo path (nonlinear distortion caused by multipath effects), and the output is:

[0139]

[0140] in, The output of the Volterra filter, i.e., the output signal, ( (represents the filter order) (Indicates memory length) These are the p-order Volterra kernel coefficients (recursively estimated using the RLS algorithm). This is the delay term of the input signal. This filter can effectively capture the nonlinear signal components caused by underwater multipath propagation;

[0141] Step S422: When impulse noise is present, the M-estimation method is used instead of the least squares method, and the objective function is defined as:

[0142]

[0143] in, The objective function value, For loss function, Representing sample data can reduce the impact of impulse noise on parameter estimation. The algorithm's robustness to impulse noise is improved by iteratively reweighting and minimizing this objective function.

[0144] This solution specifically addresses the issues of nonlinear noise and impulse noise in underwater sonar signals. Specifically, the Volterra filter captures nonlinear echo paths through a multi-order kernel function, effectively suppressing nonlinear distortion caused by multipath effects. The M-estimation method uses a logarithmically robust function instead of the traditional least squares criterion, providing stronger resistance to impulse noise and avoiding its severe impact on the filtering results. Compared to existing schemes that can only handle linear noise or have insufficient impulse noise suppression, this solution significantly improves the suppression of complex underwater noise (especially nonlinear and impulse noise), further reduces residual noise intensity, and results in a clearer, more prominent target feature output, thus improving the detection and recognition accuracy of sonar signals.

[0145] In some embodiments of the present invention, steps S41 and S42 are executed using hardware acceleration.

[0146] Steps S41 and S42 implement hardware acceleration through an FPGA platform (such as a Xilinx Zynq UltraScale+ MPSoC), and specific optimizations include:

[0147] • The matrix inversion operation is optimized using Cholesky decomposition, reducing the computational load by more than 30%;

[0148] • Map the multiplication operations in the weight update equation to a hardware-level pipeline structure, increasing throughput to 10 MSPS (million samples per second).

[0149] • Fixed-point numbers (Q15.16 format) are used instead of floating-point operations, reducing power consumption by 60%;

[0150] • A parallel computing architecture is introduced to support simultaneous processing of 8 channels of signals, meeting the real-time processing requirements of multiple hydrophone arrays.

[0151] By implementing weight updates, residual noise processing, and hardware acceleration in a step-by-step manner, this invention achieves high efficiency and high quality in sonar signal enhancement. Specifically, step S41 updates the RLS weights based on adjusted parameters, ensuring the filter's adaptability to the current environment and optimizing convergence performance. Step S42 performs specific processing on the residual noise after filtering, further improving signal purity. Hardware acceleration (e.g., FPGA implementation) meets the real-time processing requirements of underwater sonar signals (e.g., completing one iteration within 200μs). Compared to existing solutions that do not distinguish between weight updates and residual noise processing, or rely on software implementation leading to insufficient real-time performance, this invention significantly improves processing speed through hardware acceleration while maintaining filtering accuracy. Furthermore, the residual noise processing step specifically addresses complex noise suppression issues, resulting in an enhanced signal with a higher signal-to-noise ratio and stronger real-time performance, better suited to the high dynamic and real-time requirements of underwater communication.

[0152] Example 3

[0153] The underwater sonar signal enhancement system based on the RLS adaptive filter of this invention was deployed in the communication link between a certain type of underwater unmanned underwater vehicle (AUV) to enhance the transmission quality of long-range sonar signals. In a deep-sea communication test, the two AUVs were 10 km apart, at a water depth of 3000 m, with a background noise intensity of 85 dB and a temperature gradient of 4.2 °C / m.

[0154] During system operation, the environment perception module continuously monitors environmental changes, and the feedback adjustment module dynamically adjusts the RLS algorithm parameters to enable the filter to quickly adapt to channel changes.

[0155] Example 4

[0156] According to one aspect of the present invention, an electronic device is provided, comprising: one or more processors, one or more memories, and one or more computer programs; wherein the processor is connected to the memory, and the one or more computer programs are stored in the memory, and when the electronic device is running, the processor executes the one or more computer programs stored in the memory to cause the electronic device to perform the underwater sonar signal enhancement method based on RLS adaptive filter as described in any of the above technical solutions.

[0157] According to one aspect of the present invention, a computer-readable storage medium is provided for storing computer instructions, which, when executed by a processor, implement the underwater sonar signal enhancement method based on an RLS adaptive filter as described in any of the above technical solutions.

[0158] Computer-readable storage media can include any medium capable of storing or transmitting information. Examples of computer-readable storage media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, and so on. Code segments can be downloaded via computer networks such as the Internet and intranets.

[0159] Furthermore, it should be noted that the present invention can be provided as a method, apparatus, or computer program product. Therefore, embodiments of the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0160] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0161] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing terminal equipment to cause a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0162] It should also be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0163] Finally, it should be noted that the above description represents a preferred embodiment of the present invention. It should be pointed out that although preferred embodiments have been described, those skilled in the art, once they understand the basic inventive concept of the present invention, can make various improvements and modifications without departing from the principles described herein. These improvements and modifications should also be considered within the scope of protection of the present invention. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present invention.

Claims

1. A method for underwater sonar signal enhancement based on RLS adaptive filter, characterized in that, The method comprises the following steps: Step S1, receiving original sonar signals through an RLS adaptive filter module and performing real-time filtering processing; Step S2, collecting underwater environmental parameters using an environmental perception module to generate an environmental feature vector; Step S3, dynamically adjusting a forgetting factor λ and an initial covariance matrix P(0) of the RLS adaptive filter module according to the environmental feature vector through a parameter feedback adjustment module; Step S4, updating the weight of the RLS adaptive filter module using the adjusted parameters, performing residual noise processing on the filtered signals, and outputting an enhanced signal, specifically comprising: Step S41, updating the weight of the RLS adaptive filter module using the adjusted forgetting factor and initial covariance matrix to optimize convergence performance and steady-state error; Step S42, performing residual noise processing on the filtered signals to output an enhanced sonar signal; The steps S41 and S42 are executed in a hardware acceleration mode; In step S3, the forgetting factor and the initial covariance matrix in the RLS adaptive filter module are dynamically adjusted based on the environmental feature vector using a nonlinear function mapping method; the forgetting factor is adaptively adjusted according to the background noise intensity and the environmental change rate; the initial value of the covariance matrix is preset through environmental feature clustering analysis; a sliding window mechanism is introduced to weight and fuse historical environmental states; a low-delay control strategy is adopted in the adjustment process to ensure that parameter updating and signal processing are synchronized.

2. The RLS adaptive filter based underwater sonar signal enhancement method according to claim 1, characterized in that, The environmental perception module comprises: A multi-dimensional sensor array including at least a temperature sensor, a pressure sensor, a salinity meter, and a noise detector; A high-precision analog-to-digital converter for converting analog signals into digital data; A data fusion unit for feature extraction on normalized environmental parameters to generate an environmental feature vector.

3. The RLS adaptive filter based method of underwater sonar signal enhancement according to claim 1, characterized in that In step S2, the environmental feature vector is generated, specifically comprising: Temporal feature extraction is completed by calculating signal variance, represented as: wherein, denotes the variance value, N denotes the total number of samples, denotes the sample value, denotes the mean value; Frequency domain feature extraction is completed by extracting frequency spectrum components through discrete Fourier transform, represented as: wherein x(n) represents a time-domain discrete sampling sequence, with a length of N, a frequency-domain index, and a value range of 0 to N 1. X(k) is a frequency-domain discrete sampling sequence corresponding to N discrete frequency points, containing amplitude and phase information of a signal; Feature clustering analysis is used to preset the initial covariance matrix P(0).

4. The RLS adaptive filter based method of underwater sonar signal enhancement according to claim 1, wherein, In step S41, specifically comprising: Step S411, defining the cost function as the sum of weighted error squares and regularization terms; Step S412, calculating the gain vector ; Step S413, updating the inverse correlation matrix ; Step S414, calculate the priori estimation error ; Step S415, updating the weight coefficients of the RLS adaptive filter module .

5. The RLS adaptive filter based method of underwater sonar signal enhancement according to claim 1, wherein, In step S42, specifically comprising: Step S421, using a Volterra filter to expand and capture nonlinear echo paths, with the output being: where y(n) represents an output signal, x(n) represents an input signal, represents a pth order kernel coefficient, and M represents a memory length. Step S422, when there is impulse noise, using M-estimation method instead of least squares method, defining the objective function as: wherein, represents sample data, represents a loss function.

6. An underwater sonar signal enhancement system based on RLS adaptive filter, characterized in that, Comprising: An RLS adaptive filter module configured to receive original sonar signals and output filtered signals; An environmental perception module configured to collect underwater environmental parameters in real time, including temperature, pressure, and background noise intensity; The parameter feedback adjustment module is configured to dynamically adjust a forgetting factor λ and an initial covariance matrix P(0) of the RLS algorithm based on the environmental parameter by using a nonlinear function mapping method; the forgetting factor is adaptively adjusted according to the background noise intensity and the environmental change rate; the initial value of the covariance matrix is preset by environmental feature clustering analysis; a sliding window mechanism is introduced to perform weighted fusion on the historical environmental state; a low-delay control strategy is adopted in the adjustment process to ensure that the parameter update is synchronized with the signal processing process; The nonlinear noise suppression module uses a neural network to suppress residual noise of the filtered signal, specifically including the following steps executed in a hardware acceleration manner: The adjusted forgetting factor and the initial covariance matrix are used to update the weight of the RLS adaptive filter module, so as to optimize the convergence performance and the steady-state error; the filtered signal is subjected to residual noise processing, and the enhanced sonar signal is output.

7. An electronic device, comprising: Comprise: One or more processors, one or more memories, and one or more computer programs; wherein the processor is connected with the memory, the one or more computer programs are stored in the memory, and when the electronic device is running, the processor executes the one or more computer programs stored in the memory, so that the electronic device executes the underwater sonar signal enhancement method based on the RLS adaptive filter as claimed in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, A computer program product for storing computer instructions, which are executed by a processor to implement the underwater sonar signal enhancement method based on the RLS adaptive filter as claimed in any one of claims 1 to 5.

Images