Underwater positioning method and positioning system based on TF-GSC sound signal enhancement

Abstract

The present application relates to underwater orientation detection and positioning technology field, especially underwater positioning method and positioning system based on TF-GSC sound signal enhancement. Including the following steps: several underwater transducer arrays are laid in the periphery of the target object area, and the reference sound signal is periodically actively emitted; the target object receives the emitted signal, carries out preliminary spherical positioning according to the time difference, and obtains the rough position estimation of the target object; each underwater transducer array carries out STFT decomposition on the echo signal of the target object, divides the time-frequency unit on the "time-frequency" two-dimensional plane, and executes TF-GSC algorithm on each time-frequency unit; the SNR of the enhanced signal is used as feedback to dynamically adjust the array beam direction and weight, and the receiving effect is further optimized; the judgment of global convergence is carried out by cross-iteration cycle. It can improve the target extraction ability and interference suppression effect of underwater positioning, so as to improve the accuracy of underwater positioning.

Description

Technical Field

[0001] This invention relates to the field of underwater positioning and location technology, and in particular to an underwater positioning method and system based on TF-GSC acoustic signal enhancement. Background Technology

[0002] Underwater acoustic positioning is a key technology that utilizes the stable propagation speed, strong penetration, and resistance to electromagnetic interference of sound waves in water to achieve target positioning. It possesses three core advantages: high precision potential, long-distance adaptability, and environmental robustness. It is the only positioning method in underwater environments that can balance accuracy and range, and cannot be replaced by land-based technologies such as GPS. Therefore, it has become an indispensable technical support in many fields: In the field of marine resource development, it can enable underwater robots to conduct centimeter-level inspections of oil and gas pipelines, and provide dynamic navigation for deep-sea mining equipment, solving the problem of precision control for equipment operations in complex waters; in the field of marine science and environmental protection, it can locate deep-sea underwater observation stations and track the migration trajectories of marine organisms; in the field of underwater engineering construction, it can calibrate underwater immersed tube docking and solve positioning deviation and water noise interference problems through acoustic signal enhancement technology, and so on.

[0003] Underwater acoustic positioning is often affected by multipath effects, marine environmental interference, and signal attenuation. Traditional methods based on time difference of arrival (TDOA) and ultra-short baseline positioning show a significant decrease in positioning accuracy under low signal-to-noise ratio (SNR) conditions. Although array beamforming technology can enhance the signal, traditional minimum variance distortionless response (MINOCR) or fixed-weight algorithms are difficult to adapt to broadband non-stationary signals and dynamic interference sources.

[0004] Current underwater acoustic positioning technology is developing towards higher precision, intelligence, and modularity. While multi-sensor fusion can improve robustness, it is still limited by the anti-interference capability of the core acoustic signal. The three mainstream baseline systems in the industry all have significant drawbacks: Long baseline positioning (LBL) systems, although achieving centimeter-level accuracy, require the deployment of transponder arrays ranging from hundreds of meters to several kilometers on the seabed. The system is complex, costly, and dependent on a pre-defined operating area. It is also susceptible to seabed topography, which can increase errors in shallow water. Short baseline positioning (SBL) systems are easier to install than LBL systems, but deep-water operations require a baseline length exceeding 40 meters to ensure accuracy, resulting in high costs for hull modification. Ultra-short baseline positioning (USBL) systems have become the mainstream for civilian use due to their simplicity and low cost, but they are greatly affected by sound velocity gradients, have stringent installation and calibration requirements, and their signals are easily submerged in low signal-to-noise ratio environments, making stable positioning impossible. Summary of the Invention

[0005] The purpose of this invention is to overcome the above-mentioned defects in the existing technology and propose an underwater positioning method and system based on TF-GSC acoustic signal enhancement, so as to achieve efficient enhancement and interference suppression of broadband non-stationary signals in complex environments, improve the target extraction capability and interference suppression effect of underwater positioning, and thus improve the accuracy of underwater positioning.

[0006] The technical solution of this invention is: an underwater positioning method based on TF-GSC acoustic signal enhancement, comprising the following steps:

[0007] S1. Deploy several underwater transducer arrays around the area where the target is located, and periodically and actively transmit reference acoustic signals;

[0008] S2. The target receives the transmitted signal and performs preliminary spherical positioning based on the time difference of arrival to obtain a rough estimate of the target's position.

[0009] S3. Each underwater transducer array performs STFT decomposition on the target echo signal, divides the time-frequency unit on the "time-frequency" two-dimensional plane, and executes the TF-GSC algorithm on each time-frequency unit to achieve target signal enhancement and interference suppression.

[0010] S4. Using the SNR of the enhanced signal as feedback, dynamically adjust the array beam direction and weight to further optimize the reception effect;

[0011] S5. Use cross-iteration cycles to determine global convergence.

[0012] In this invention, during the spherical positioning process in step S2, the distance from each acoustic element to the target is calculated based on the time delay experienced by the sound wave as it travels from the acoustic element to the target. Each distance value defines a sphere centered on the acoustic element. When at least three spheres in three-dimensional space intersect, their intersection point is the uniquely determined specific location of the target.

[0013] ,

[0014] in, This represents a specific acoustic array element. Represents the position coordinates of the acoustic array elements. Indicates the speed of sound. Indicates the sound propagation time;

[0015] For a target of unknown depth, when there is no error in the slant distance between the measured target and the acoustic array elements, the four spheres can determine a unique intersection point, and the four equations of its four-element array can be expressed as follows:

[0016] ,

[0017] ,

[0018] ,

[0019] ,

[0020] After rearranging the four equations of the quaternion matrix, we can obtain the following system of linear equations:

[0021] ,

[0022] make:

[0023] ,

[0024] ,

[0025] ,

[0026] get The only solution:

[0027] ,

[0028] Where P represents the three-dimensional spatial coordinate vector of the underwater target to be located, including the x-axis coordinate, y-axis coordinate, and z-axis coordinate of the target, and is the core result of the final solution for spherical positioning.

[0029] The specific implementation process of step S3 is as follows:

[0030] S3.1, Time-Frequency Transformation;

[0031] S3.2, Reference beam construction;

[0032] S3.3, Modeling the interference subspace;

[0033] S3.4, Sidelobe Suppression and Weight Optimization.

[0034] In step S3.1,

[0035] A window function is added to the target echo signal received by the underwater transducer array and a short-time Fourier transform is performed to extract the complex spectral coefficients corresponding to each time-frequency unit, thereby realizing the joint representation of the signal in time-frequency-space.

[0036] Select the window function type, window length, sampling points, and sampling rate, and set the frame shift for a signal of length N. , No. The formula for calculating the STFT of a frame is:

[0037] ,

[0038] in, Represents the complex spectral coefficients of the time-frequency unit; L represents the window length; R represents the frame shift; Indicates the window function; Indicates frequency index; Represents the imaginary unit; This represents the time-domain discrete sequence of the target echo signal acquired by the underwater transducer array, which is the original input signal for STFT decomposition.

[0039] The target echo signal is converted into a multidimensional complex spectrum by short-time Fourier transform, forming a three-dimensional tensor input for subsequent beam processing.

[0040] In step S3.2, let the coarse positioning result be the coordinates of the target object in space. The coordinates of the underwater transducer array are Then the direction vector of the target object relative to the underwater transducer array can be calculated through spatial geometric relationships;

[0041] ,

[0042] in, d represents the azimuth and elevation angles that the reference beam needs to point to; d represents the straight-line distance between the target and the underwater transducer array.

[0043] The target echo signal received by the underwater transducer array is decomposed by STFT and divided into several time-frequency units. Each unit corresponds to a specific time segment and frequency component. Construct a separate reference beam, and the reference beam vector. The calculation formula is:

[0044] ,

[0045] Where M represents the number of array elements; H represents the conjugate transpose; , representing the array manifold vector of the target direction;

[0046] To satisfy the distortion-free constraint in the above formula, the reference beam vector adopts the minimum norm solution.

[0047] ,

[0048] in, The array manifold vector representing the target direction; This represents the inner product of the array manifold vectors.

[0049] The specific implementation process of step S3.3 is as follows:

[0050] S3.3.1 Extract the multi-channel signals of the time-frequency unit. After completing the short-time Fourier transform, each time-frequency unit... The multi-channel signal is represented as:

[0051] ,

[0052] Where M represents the number of acoustic array elements in the underwater transducer array; Indicates the first The complex spectral coefficients of each acoustic array element in this time-frequency unit; T represents the transpose;

[0053] For the Time-frequency unit, covariance matrix The estimation formula is:

[0054] ,

[0055] in, For the mathematical expectation operator, the average of multiple sets of observations in the same time-frequency unit is taken; H represents the conjugate transpose.

[0056] According to the signal model, the covariance matrix is ​​decomposed as follows:

[0057] ,

[0058] in, Represents the covariance component of the target signal; Indicates the target signal power; Represents the covariance matrix of the interference plus noise;

[0059] S3.3.2 Constructing the interference subspace through eigenvalue decomposition, the eigenvalue decomposition formula is as follows:

[0060] ,

[0061] in, , representing the covariance matrix of the interference plus noise; The eigenvector matrix representing the interference subspace is calculated using the following formula:

[0062] ,

[0063] Where K represents the number of interference sources, and these feature vectors span the space where the interference signal is located, containing the directional information of all interference signals; Represents a diagonal matrix;

[0064] eigenvector matrix The space spanned by the column vectors is the interference subspace, denoted as:

[0065] ,

[0066] This subspace contains all possible directions of interference signals, providing a spatial reference for the next step of orthogonal separation of the target signal and interference signals.

[0067] In step S3.4, the principle of generalized sidelobe suppression is used to calculate the weight vector under the constraints. The optimal value ensures that the beam is aligned with the target direction. The gain is maximized on the upper direction, while suppressing interference signals from other directions;

[0068] Weight vector The optimal value is calculated by solving a constrained least squares problem, i.e. The constraints are:

[0069] ,

[0070] in, This represents the target guidance vector, describing the frequency k and the target direction. The spatial phase characteristics, corresponding to the array manifold, satisfy... ;

[0071] Represents the weight vector Norm, This is equivalent to suppressing sidelobe energy, and its calculation formula is:

[0072] ;

[0073] In the process of sidelobe suppression and weight optimization, the projected energy of the interference subspace is minimized, i.e. The optimal beam weight vector can be solved.

[0074] By using orthogonal projection spatial filtering, the target signal space is mapped to the orthogonal direction of the interference subspace, thereby suppressing non-target signals while ensuring the distortion-free output of the target signal. Finally, after TF-GSC processing, the enhanced output signal is:

[0075] .

[0076] In step S4, one iteration cycle includes several rounds of iteration. After each round of iteration, the enhanced signal is... To calculate the SNR value, assume the target direction beam output power is... The average power in the direction of interference is ,but:

[0077] .

[0078] It is obtained by performing a sum of squares operation on the beam output signal in the target direction. It is obtained by averaging the sum of squares of the beam output signals in all directions within the interference subspace;

[0079] like If the current beam direction or weight configuration has sufficiently suppressed interference, then the next step of processing can be carried out directly.

[0080] when If the current beam direction or weight configuration fails to adequately suppress interference, adaptive array control is performed to adjust the target direction, beamwidth, and weight vector of each array beamformer, and the SNR value is recalculated until the result obtained in the current iteration is obtained. Compared to the previous iteration calculation The improvement rate is greater than 1%.

[0081] In step S5, if the SNR improvement rate of the positioning result obtained in two consecutive iteration cycles is <1%, it indicates that the positioning result has become stable, the iteration process is determined to have converged, and the positioning result is output.

[0082] If convergence is not achieved, repeat steps S3 to S4 until the positioning result stabilizes and is output.

[0083] This application also includes a positioning system for implementing the above-described underwater positioning method, comprising:

[0084] Multiple underwater transducer array modules are used to transmit signals and receive signals fed back from the target.

[0085] Spherical positioning module for coarse positioning of the target object;

[0086] The short-time Fourier transform module is used to perform time-frequency decomposition on the target echo signal received by the underwater transducer array to obtain the complex spectral coefficients of the time-frequency unit;

[0087] Reference beamforming module for forming a distortion-free response channel in the target direction;

[0088] The interference subspace modeling module is used to estimate and separate interference signals from non-target directions;

[0089] The sidelobe suppression and weight optimization module is used to achieve target enhancement and interference suppression.

[0090] The beneficial effects of this invention are:

[0091] (1) This application introduces the TF-GSC beamforming algorithm into the underwater three-dimensional positioning method. By combining time-frequency analysis with generalized sidelobe suppression, under the same main lobe gain, the sidelobe energy in the interference direction decreases by about 20 dB, and the output signal-to-noise ratio is improved by about 3–4 dB under low signal-to-noise ratio conditions. This is significantly better than the traditional MVDR and DSB beamforming methods, effectively enhancing the fidelity of broadband, non-stationary target signals and significantly suppressing multipath and interference signals. This scheme can maintain high positioning accuracy under low signal-to-noise ratio and strong interference conditions, and improve the system's ability to detect and locate the orientation of objects in complex marine acoustic environments.

[0092] (2) By adopting an adaptive beam control mechanism based on signal-to-noise ratio feedback, this application can automatically adjust the beam direction and array weight parameters according to the real-time calculated signal-to-noise ratio, so as to realize the dynamic response to the target motion state and changes in the marine environment. Compared with the traditional fixed parameter beamforming, this mechanism significantly improves the robustness and dynamic tracking capability of the positioning system, and can still maintain stable output when the target direction changes or the environmental noise changes abruptly.

[0093] (3) An active-passive hybrid working mechanism and energy consumption control strategy are proposed, which automatically switches the working mode according to whether the target has the ability to emit sound: when the target emits sound actively, the system enters the passive listening mode, thereby significantly reducing energy consumption; when the target is silent or the echo is weak, the active sound source is automatically activated to achieve continuous positioning. This strategy enables the system to effectively extend the endurance of low-energy platforms such as underwater robots while ensuring high-precision positioning performance, meeting the application requirements of long-term tasks and unmanned deployment.

[0094] In summary, this application uses TF-GSC technology to suppress water noise and multipath interference, solving the positioning failure problem in USBL / LBL noisy environments; it eliminates the need for deploying seabed transponder arrays (LBL) or long baseline transducer arrays (SBL), reducing system complexity and deployment costs; and through time-frequency domain signal processing optimization, it reduces the impact of sound velocity gradient and ship attitude changes on positioning results, improving positioning stability in deep water areas and dynamic operation scenarios. Attached Figure Description

[0095] Figure 1 This is a flowchart of the method described in this application;

[0096] Figure 2 This is a flowchart of step S3 in this application. Detailed Implementation

[0097] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0098] Specific details are set forth in the following description to provide a full understanding of the invention. However, the invention can be practiced in many ways other than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0099] The underwater positioning method based on TF-GSC acoustic signal enhancement described in this application includes the following steps, the flowchart of which is shown below. Figure 1 As shown.

[0100] The first step is multi-array deployment and active transmission: multiple underwater transducer arrays are deployed around the area where the target is located, and reference acoustic signals are periodically and actively transmitted.

[0101] Target objects may include underwater robots, biological tracking beacons, side-scan sonar, towing robots, and equipment carriers. At least three sets of underwater transducer arrays are deployed around the target area, each array comprising 8 to 16 acoustic elements. These arrays are arranged in an equidistant linear array structure at different locations within the target area, creating spatial geometric constraints to improve TDOA positioning accuracy and beamforming directional resolution. The spacing between the array elements is designed to be half the wavelength of the acoustic wave corresponding to the dominant frequency of the received signal. This is to suppress spatial aliasing and ensure that the array has sufficient beam focusing capability.

[0102] The array deployment strategy employs a "non-coplanar-asymmetric" multi-base station topology. By combining different orientations and distances, it ensures good geometric accuracy factors during spherical intersection positioning and effectively eliminates positioning blind spots. In actual deployment, the distance between each array group is planned according to the target activity area and sea depth, ensuring signal coverage while avoiding multipath overlap and signal attenuation caused by excessively long acoustic propagation paths.

[0103] In active mode, the array actively transmits reference pulses at a set frequency, suitable for scenarios where the target has no spontaneous acoustic signature or is in initial localization. If a spontaneous acoustic signal from the target is detected within a specific time window, it automatically switches to passive monitoring mode. In passive mode, the array shuts down its own transmitting device, retaining only the receiving and processing components. Switching conditions include energy mutation detection, spectral correlation analysis, and signal classifier identification, ensuring that the probability of false switching is less than 1%. The underwater transducer array automatically and dynamically selects the mode based on the target's behavioral characteristics, maximizing the equipment's endurance.

[0104] The second step is target signal reception and TDOA estimation: the target receives multiple transmitted signals, and preliminary spherical positioning is performed based on the signal arrival time difference to obtain a rough estimate of the target's position.

[0105] The target receives multiple transmitted signals. The arrival timestamps of each pulse signal are extracted using correlation detection methods. The time difference between any two arrays is calculated, and preliminary spherical positioning is performed based on the arrival time difference to obtain a rough estimate of the target's position. This position is used for subsequent beam reference direction construction.

[0106] Spherical positioning is a positioning method based on time delay measurement. Its core principle is to calculate the distance from each acoustic element to the target object by measuring the time delay experienced by sound waves traveling from the acoustic element to the target object. Each distance value defines a sphere centered on the acoustic element; that is, a geometric surface formed by all points satisfying that distance, representing any possible location of the target object on this sphere. When at least three spheres of nodes in three-dimensional space intersect, their intersection point uniquely determines the specific location of the target object.

[0107] ,

[0108] in, This represents a specific acoustic array element. Represents the position coordinates of the acoustic array elements. Indicates the speed of sound. Indicates the sound propagation time.

[0109] For targets of unknown depth, when there is no error in the slant distance between the measured target and the acoustic array elements, the intersection of two spheres forms a circle, the intersection of three spheres forms two points, and the fourth sphere can determine a unique point. Taking a four-element array as an example, its equations are as follows:

[0110] ,

[0111] ,

[0112] ,

[0113] ,

[0114] After rearranging the four equations of the above quaternion matrix, we can obtain the following system of linear equations:

[0115] ,

[0116] make:

[0117] ,

[0118] ,

[0119] ,

[0120] Therefore, as long as If the inverse matrix exists, then the unique solution to P can be obtained:

[0121] ,

[0122] Where P represents the three-dimensional spatial coordinate vector of the underwater target to be located, including the x-axis coordinate, y-axis coordinate, and z-axis coordinate of the target, and is the core result of the final solution for spherical positioning.

[0123] To enhance the stability and robustness of spherical positioning, the following can be further introduced:

[0124] (1) Sound speed profile correction: Consider the influence of the actual sea area sound speed gradient change on propagation time;

[0125] (2) GDOP analysis: Select array combinations with good geometric distribution to reduce the degradation of positioning accuracy;

[0126] (3) Data fusion processing: The TDOA results are fused with other constraints (such as DOA) to improve accuracy;

[0127] (4) Weighted least squares: assign weights to different TDOA measurements to reduce the impact of outliers.

[0128] The spherical positioning results can provide direction for subsequent TF-GSC reference beam construction, serve as a starting point for iterative positioning and SNR feedback optimization, and also serve as a basis for positioning updates when the target actively emits sound.

[0129] The third step, time-frequency analysis and TF-GSC beamforming: Each underwater transducer array performs STFT decomposition on the target echo signal, dividing the time-frequency two-dimensional plane into the smallest independent processing unit, the time-frequency unit. The TF-GSC algorithm is then executed on each time-frequency unit to enhance the target echo signal and suppress interference. This specifically includes the following steps: Figure 2 As shown.

[0130] First, time-frequency transformation. A window function is added to the target echo signal received by the underwater transducer array and a short-time Fourier transform (STFT) is performed to extract the complex spectral coefficients corresponding to each time-frequency unit, thereby realizing the joint representation of the signal in time-frequency-space.

[0131] Select the window function type, window length, sampling points, and sampling rate, and set the frame shift: When performing STFT, the Hanning window is selected because it has good balance characteristics in both the time and frequency domains. For a signal of length N... , No. The formula for calculating the STFT of a frame is:

[0132] ,

[0133] in, Represents the complex spectral coefficients of the time-frequency unit; L represents the window length; R represents the frame shift; Indicates the window function; Indicates frequency index; The imaginary unit is the symbol used in complex number operations and satisfies mathematical relations. ; This represents the time-domain discrete sequence of the target echo signal acquired by the underwater transducer array, which is the original input signal for STFT decomposition.

[0134] The target echo signal is converted into a multidimensional complex spectrum through STFT calculation, forming a three-dimensional tensor input for subsequent beam processing.

[0135] In high-noise or high-interference frequency bands, the window length can be adaptively increased to improve frequency domain resolution and ensure the concentration of target signal energy. By introducing STFT preprocessing before beamforming, the system can effectively elevate time-domain spatial filtering to the time-frequency-space three-domain joint processing level, enhancing the ability to perceive non-steady-state signals.

[0136] Second, reference beam construction: Based on the coarse positioning results obtained in the second step, a reference beam signal pointing towards the target is constructed at each time frequency point and used as a distortion-free response channel.

[0137] Each time-frequency unit is treated as a local narrowband signal and processed independently using the TF-GSC beamforming algorithm.

[0138] Spherical positioning calculates the approximate spatial coordinates of a target object by using the time difference of multiple underwater transducer arrays. These coordinates are then used to infer the azimuth (horizontal direction) and elevation (vertical direction) of the target object relative to the underwater transducer array, which is the physical direction that the reference beam needs to point to.

[0139] Let the coarse localization result be the coordinates of the target sound source in space. The coordinates of the underwater transducer array are The direction vector of the target object relative to the underwater transducer array can then be calculated using spatial geometric relationships:

[0140] ,

[0141] in, represents the azimuth and elevation angles that the reference beam needs to point to; d is the straight-line distance between the target and the underwater transducer array, calculated from the coarse positioning results.

[0142] The target echo signal received by the underwater transducer array is decomposed by STFT and divided into multiple time-frequency units, each corresponding to a specific time segment and frequency component. Due to the significant differences in signal characteristics at different times and frequencies of underwater signals, it is necessary to define a time-frequency unit... A separate reference beam is constructed to suit its local stationary characteristics. Reference beam vector. This is the core of the "distortionless response channel" in the TF-GSC algorithm, which must satisfy the constraint of "distortionless response in the target direction", as shown in the formula:

[0143] ,

[0144] in, , indicating the first The reference beam weight vector of the time-frequency unit, where M represents the number of array elements and H is the conjugate transpose; , representing the array manifold vector of the target direction, i.e., the spatial direction in which the underwater transducer array points towards the target sound source, and the frequency Target azimuth Pitch angle Related. The "1" on the right side of the above equation indicates that the gain of the underwater transducer array in the spatial direction pointing towards the target sound source is 1, that is, there is no distortion constraint.

[0145] To satisfy the distortion-free constraint in the above formula, the reference beam vector adopts the minimum norm solution to ensure that the weight energy is minimized and noise is suppressed, that is:

[0146] ,

[0147] in, The inner product of the array manifold vectors is used to normalize the array manifold vectors, ensuring that the target signal, i.e., the signal along the spatial direction of the underwater transducer array pointing towards the target sound source, is not attenuated or distorted in the reference beam channel, thus providing a reliable "reference signal" for subsequent interference suppression.

[0148] Third, interference subspace modeling: construct an interference subspace for the remaining signals except for the spatial direction of the underwater transducer array pointing to the target sound source. This subspace can be obtained by estimating the covariance matrix of the multi-channel signal.

[0149] By analyzing the surrounding environmental signals, an interference subspace is modeled. The interference subspace contains all signal directions except the spatial direction of the underwater transducer array pointing to the target sound source, including non-target signals such as multipath effects, marine environmental noise, and dynamic interference sources, and is dedicated to separating and quantizing non-target signals.

[0150] Similar to the reference beam, the interference subspace is modeled independently in each time-frequency cell to accommodate the time-varying characteristics of broadband non-stationary underwater signals. During modeling, statistical analysis is performed on the multi-channel signals of the underwater transducer array to calculate the signal covariance matrix. Based on this, the spatial features of the interference signal are extracted to construct a subspace encompassing all interference directions.

[0151] First, the multi-channel signals of the time-frequency unit are extracted. After completing the short-time Fourier transform, each time-frequency unit... A multi-channel signal can be represented as:

[0152] ,

[0153] Where M represents the number of acoustic array elements in the underwater transducer array; Indicates the first The complex spectral coefficients of each acoustic array element in the time-frequency unit; T represents the transpose.

[0154] The modeling basis of the interference subspace is the covariance matrix of the multi-channel signal, which reflects the correlation between the acoustic array element signals in each underwater transducer array, including the joint estimation characteristics of the target, interference, and noise.

[0155] For the Time-frequency unit, covariance matrix The estimation formula is:

[0156] ,

[0157] in, The mathematical expectation operator is used to average multiple sets of observations for the same time-frequency unit; H represents the conjugate transpose, ensuring that the matrix is ​​a Hermitian matrix, i.e., symmetric with real numbers on the diagonal, which conforms to the characteristics of a covariance matrix.

[0158] covariance matrix The signal contains the target signal, interference signal, and noise. It is necessary to separate the components containing only the interference signal and noise. According to the signal model, the covariance matrix can be decomposed as follows:

[0159] ,

[0160] in, Represents the covariance component of the target signal; Indicates the target signal power; This represents the covariance matrix of the interference plus noise, which forms the basis of the interference subspace to be extracted.

[0161] Next, the interference subspace is constructed through eigenvalue decomposition. The eigenvalue decomposition formula is as follows:

[0162] ,

[0163] in, , where represents the covariance matrix of the interference plus noise. The eigenvector matrix representing the interference subspace is given by... The eigenvectors are composed of the first K largest eigenvalues, and their calculation formula is:

[0164] ,

[0165] Where K represents the number of interference sources, and these feature vectors span the space where the interference signal is located, containing the directional information of all interference signals.

[0166] This represents a diagonal matrix with diagonal elements as follows: The top K largest eigenvalues ​​are sorted in descending order. The magnitude of the eigenvalue reflects the power of the corresponding interference signal. The greater the power, the stronger the interference in that direction.

[0167] In summary, the eigenvector matrix The space spanned by the column vectors is the interference subspace, which can be represented as:

[0168] ,

[0169] This subspace contains all possible directions of interference signals, providing a spatial reference for the next step of orthogonal separation of the target signal and interference signals.

[0170] Fourth, sidelobe suppression and weight optimization: By minimizing the reference beam signal and the projected energy of the interference subspace, maximum suppression of interference components is achieved. The final beam output is the projection of the target signal space onto the orthogonal direction of the interference subspace, suppressing non-target signals while maintaining the amplitude of the target signal.

[0171] By orthogonally projecting the interference subspace onto the target signal space, i.e. projecting the target signal space onto the orthogonal direction of the interference subspace, the energy of the interference components can be reduced to the greatest extent while maintaining the amplitude of the target signal.

[0172] By utilizing the principle of generalized sidelobe suppression, the weight vector under constraints is calculated. The optimal value ensures that the beam is aligned with the target direction. The gain is maximized on the upper direction, while suppressing interference signals from other directions.

[0173] Specifically, the weight vector The optimal value is calculated by solving a constrained least squares problem, i.e. The constraints are:

[0174] ,

[0175] in, This represents the target guidance vector, describing the frequency k and the target direction. The spatial phase characteristics, corresponding to the array manifold, satisfy... .

[0176] Represents the weight vector Norm, This is equivalent to suppressing sidelobe energy, and its calculation formula is:

[0177] .

[0178] In the process of sidelobe suppression and weight optimization, the interference subspace To provide a basis for the "constraints", the projected energy of the disturbance subspace is minimized, i.e. The optimal beam weight vector can be solved to maximize the beam gain in the target direction and minimize the gain in the interference direction.

[0179] By using the above-mentioned orthogonal projection spatial filtering method, the target signal space is mapped to the orthogonal direction of the interference subspace, thereby suppressing non-target signals while ensuring the distortion-free output of the target signal. Finally, after TF-GSC processing, the enhanced output signal is:

[0180] .

[0181] in, The result obtained after the above STFT transformation is used as input at this time to achieve the enhancement of the target signal and the suppression of interference.

[0182] Step 4, Signal-to-noise ratio feedback and array adaptive adjustment: utilizing enhanced signals The SNR is used as feedback to dynamically adjust the array beam direction and weight, further optimizing the reception effect.

[0183] Each iteration cycle consists of several rounds of iterations. Within an iteration cycle, after each round of iterations, the enhanced signal is... The SNR value is calculated using the ratio of the target direction beam output power to the average power in the interference direction. Let the target direction beam output power be... The average power in the direction of interference is ,but:

[0184] .

[0185] In actual calculations, It is obtained by performing a sum of squares operation on the beam output signal in the target direction. It is obtained by averaging the sum of squares of the beam output signals in all directions within the interference subspace.

[0186] If the result obtained in the current iteration Compared to the previous iteration calculation If there is no significant improvement, it is determined that the current beam direction or weighting configuration has failed to adequately suppress interference. In this embodiment, the improvement rate is limited to 1%. When If the current beam direction or weight configuration has sufficiently suppressed interference, then the next step of processing can be carried out directly.

[0187] when At this point, the current beam orientation or weight configuration fails to adequately suppress interference. Next, adaptive array control is performed, adjusting the target orientation, beamwidth, and weight vector of each array beamformer, and recalculating the SNR value until the result obtained in the current iteration is reached. Compared to the previous iteration calculation The improvement rate is greater than 1%.

[0188] The fifth step is to determine the location, iteration, and convergence.

[0189] This step involves determining global convergence across iteration cycles. If the SNR improvement rate of the localization results obtained in two consecutive iteration cycles is less than 1%, it indicates that the localization results have stabilized, and further iterations will not bring effective improvement. At this point, the iteration process is considered to have converged, and the localization result is output.

[0190] If convergence is not yet achieved, repeat steps three and four, sequentially reconstructing the reference beam, reconstructing the interference subspace, forming the TF-GSC beam, calculating the SNR, and determining convergence, until the positioning result stabilizes and is output. This ensures that the system can achieve progressive convergence and highly stable target estimation output in complex and dynamic marine environments, reducing mispositioning and misidentification.

[0191] To verify the technical effectiveness of the underwater positioning method based on TF-GSC acoustic signal enhancement described in this invention, simulation tests were conducted on the MATLAB platform. The simulation employed an eight-element uniform linear array structure with an element spacing equal to half the wavelength, an array operating frequency of 20kHz, and a sampling rate of 96kHz. The target signal was a linear frequency modulated signal of 18–22kHz, with a target azimuth angle of 30° and an interference azimuth angle of [missing information]. The background noise is additive white Gaussian noise superimposed with random narrowband interference signals, and the input signal-to-noise ratios are respectively... , and .

[0192] (1) Comparison of beam direction between traditional MVDR algorithm and the method described in this application:

[0193] While maintaining a consistent main lobe gain at 30° in the target direction, this application significantly suppresses sidelobe energy at −20° in the interference direction, with a sidelobe level decrease of approximately 20dB. This demonstrates that this application can effectively suppress interference signals in the spatial dimension and improve the resolution of target signals.

[0194] (2) Energy distribution in the time-frequency domain before and after signal enhancement:

[0195] The original signal has a relatively dispersed time-frequency energy and contains obvious broadband noise and interference energy. However, after processing by the method described in this application, the target signal energy is concentrated in the main frequency band region, the background noise energy is significantly reduced, and the time-frequency texture structure of the target signal is clearer, demonstrating the ability of this application to jointly suppress noise and interference in the time-frequency domain.

[0196] (3) Comparison of the output signal-to-noise ratio of MVDR and the method described in this application under different input signal-to-noise ratios:

[0197] When the input signal-to-noise ratios are respectively , and In contrast, the traditional MVDR algorithm's output signal-to-noise ratio (SNR) gradually improves with increasing input SNR, while the method described in this application achieves a higher output SNR under the same conditions, with an average improvement of approximately 4–5 dB. This result demonstrates that the application can still achieve stable signal enhancement in low SNR environments. Furthermore, inputting the enhanced signal into a spherical positioning module for 3D position estimation reduces the positioning error by approximately 50% compared to the MVDR result. It can maintain sub-meter level accuracy under certain conditions, and its positioning accuracy can be stably achieved at the centimeter level under medium-high signal-to-noise ratio conditions.

[0198] In summary, the simulation results fully verify the robustness and superiority of the method disclosed in this application under conditions of low signal-to-noise ratio, strong interference, and broadband non-stationary signals. The TF-GSC acoustic signal enhancement method proposed in this application can significantly improve the signal-to-noise ratio of the target signal, suppress multi-source interference, and effectively improve the accuracy and stability of the underwater positioning system, demonstrating good engineering feasibility and application prospects.

[0199] This application also includes a positioning system that implements the above positioning method, comprising:

[0200] Multiple underwater transducer array modules are used to transmit signals and receive signals fed back from the target.

[0201] Spherical positioning module for coarse positioning of the target object;

[0202] The short-time Fourier transform module is used to perform time-frequency decomposition on the target echo signal received by the underwater transducer array to obtain the complex spectral coefficients of the time-frequency unit;

[0203] Reference beamforming module for forming a distortion-free response channel in the target direction;

[0204] The interference subspace modeling module is used to estimate and separate interference signals from non-target directions;

[0205] The sidelobe suppression and weight optimization module is used to achieve target enhancement and interference suppression.

[0206] The underwater positioning method and system based on TF-GSC acoustic signal enhancement provided by this invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. It should be noted that those skilled in the art can make various improvements and modifications to this invention without departing from the principles of this invention, and these improvements and modifications also fall within the protection scope of the claims of this invention. The above description of the disclosed embodiments enables those skilled in the art to implement or use this invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of this invention. Therefore, this invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An underwater positioning method based on TF-GSC acoustic signal enhancement, characterized in that, Includes the following steps: S1. Deploy several underwater transducer arrays around the area where the target is located, and periodically and actively transmit reference acoustic signals; S2. The target receives the transmitted signal and performs preliminary spherical positioning based on the time difference of arrival to obtain a rough estimate of the target's position. S3. Each underwater transducer array performs STFT decomposition on the target echo signal, dividing it into time-frequency units on the "time-frequency" two-dimensional plane. The TF-GSC algorithm is executed on each time-frequency unit to achieve target signal enhancement and interference suppression. The specific implementation process is as follows: S3.1 Time-frequency transformation; S3.2, Reference beam construction; S3.3, Modeling the interference subspace; S3.4, Sidelobe Suppression and Weight Optimization; S4. Using the SNR of the enhanced signal as feedback, dynamically adjust the array beam direction and weight to further optimize the reception effect; S5. Use cross-iteration cycles to determine global convergence; In step S3.2, let the coarse positioning result be the coordinates of the target object in space. The coordinates of the underwater transducer array are Then the direction vector of the target object relative to the underwater transducer array can be calculated through spatial geometric relationships; ， in, d represents the azimuth and elevation angles that the reference beam needs to point to; d represents the straight-line distance between the target and the underwater transducer array. The target echo signal received by the underwater transducer array is decomposed by STFT and divided into several time-frequency units. Each unit corresponds to a specific time segment and frequency component. Construct a separate reference beam, and the reference beam vector. The calculation formula is: ， Where M represents the number of array elements; H represents the conjugate transpose; , representing the array manifold vector of the target direction; To satisfy the distortion-free constraint in the above formula, the reference beam vector adopts the minimum norm solution. ， in, The array manifold vector representing the target direction; Represents the inner product of the array manifold vectors; The specific implementation process of step S3.3 is as follows: S3.3.1 Extract the multi-channel signals of the time-frequency unit. After completing the short-time Fourier transform, each time-frequency unit... The multi-channel signal is represented as: ， Where M represents the number of acoustic array elements in the underwater transducer array; Indicates the first The complex spectral coefficients of each acoustic array element in this time-frequency unit; T represents the transpose; For the Time-frequency unit, covariance matrix The estimation formula is: ， in, For the mathematical expectation operator, the average of multiple sets of observations in the same time-frequency unit is taken; H represents the conjugate transpose. According to the signal model, the covariance matrix is ​​decomposed as follows: ， in, Represents the covariance component of the target signal; Indicates the target signal power; Represents the covariance matrix of the interference plus noise; S3.3.2 Constructing the interference subspace through eigenvalue decomposition, the eigenvalue decomposition formula is as follows: ， in, , representing the covariance matrix of the interference plus noise; The eigenvector matrix representing the interference subspace is calculated using the following formula: ， Where K represents the number of interference sources, and these feature vectors span the space where the interference signal is located, containing the directional information of all interference signals; Represents a diagonal matrix; eigenvector matrix The space spanned by the column vectors is the interference subspace, denoted as: ， This subspace contains all possible directions of interference signals, providing a spatial reference for the next step of orthogonal separation of the target signal and interference signals; In step S3.4, the principle of generalized sidelobe suppression is used to calculate the weight vector under the constraints. The optimal value ensures that the beam is aligned with the target direction. The gain is maximized on the upper direction, while suppressing interference signals from other directions; Weight vector The optimal value is calculated by solving a constrained least squares problem, i.e. The constraints are: ， in, This represents the target guidance vector, describing the frequency k and the target direction. The spatial phase characteristics, corresponding to the array manifold, satisfy... ; Represents the weight vector Norm, This is equivalent to suppressing sidelobe energy, and its calculation formula is: ； In the process of sidelobe suppression and weight optimization, the projected energy of the interference subspace is minimized, i.e. The optimal beam weight vector can be solved. By using orthogonal projection spatial filtering, the target signal space is mapped to the orthogonal direction of the interference subspace, thereby suppressing non-target signals while ensuring the distortion-free output of the target signal. Finally, after TF-GSC processing, the enhanced output signal is: 。 2. The underwater positioning method based on TF-GSC acoustic signal enhancement according to claim 1, characterized in that, In the spherical positioning process of step S2, the distance from each acoustic element to the target is calculated by the time delay experienced by the sound wave from the acoustic element to the target. Each distance value defines a sphere centered on the acoustic element. When at least three spheres in three-dimensional space intersect, their intersection point is the uniquely determined specific location of the target. ， in, This represents a specific acoustic array element. Represents the position coordinates of the acoustic array elements. Indicates the speed of sound. Indicates the sound propagation time; For a target of unknown depth, when there is no error in the slant distance between the measured target and the acoustic array elements, the four spheres can determine a unique intersection point, and the four equations of its four-element array can be expressed as follows: ， ， ， ， After rearranging the four equations of the quaternion matrix, we can obtain the following system of linear equations: ， make: ， ， ， get The only solution: ， Where P represents the three-dimensional spatial coordinate vector of the underwater target to be located, including the x-axis coordinate, y-axis coordinate, and z-axis coordinate of the target, and is the core result of the final solution for spherical positioning.

3. The underwater positioning method based on TF-GSC acoustic signal enhancement according to claim 1, characterized in that, In step S3.1, A window function is added to the target echo signal received by the underwater transducer array and a short-time Fourier transform is performed to extract the complex spectral coefficients corresponding to each time-frequency unit, thereby realizing the joint representation of the signal in time-frequency-space. Select the window function type, window length, sampling points, and sampling rate, and set the frame shift for a signal of length N. , No. The formula for calculating the STFT of a frame is: ， in, Represents the complex spectral coefficients of the time-frequency unit; L represents the window length; R represents the frame shift; Indicates the window function; Indicates frequency index; Represents the imaginary unit; This represents the time-domain discrete sequence of the target echo signal acquired by the underwater transducer array, which is the original input signal for STFT decomposition. The target echo signal is converted into a multidimensional complex spectrum by short-time Fourier transform, forming a three-dimensional tensor input for subsequent beam processing.

4. The underwater positioning method based on TF-GSC acoustic signal enhancement according to claim 1, characterized in that, In step S4, after each iteration within an iteration cycle, the enhanced signal is... To calculate the SNR value, assume the target direction beam output power is... The average power in the direction of interference is ,but: ； It is obtained by performing a sum of squares operation on the beam output signal in the target direction. It is obtained by averaging the sum of squares of the beam output signals in all directions within the interference subspace; like If the current beam direction or weight configuration has sufficiently suppressed interference, then the next step of processing can be carried out directly. when If the current beam direction or weight configuration fails to adequately suppress interference, adaptive array control is performed to adjust the target direction, beamwidth, and weight vector of each array beamformer, and the SNR value is recalculated until the result obtained in the current iteration is obtained. Compared to the previous iteration calculation The improvement rate is greater than 1%.

5. The underwater positioning method based on TF-GSC acoustic signal enhancement according to claim 1, characterized in that, In step S5, if the SNR improvement rate of the positioning result obtained in two consecutive iteration cycles is <1%, it indicates that the positioning result has become stable, the iteration process is determined to have converged, and the positioning result is output. If convergence is not achieved, repeat steps S3 to S4 until the positioning result stabilizes and is output.

6. A positioning system for implementing any one of the underwater positioning methods based on TF-GSC acoustic signal enhancement described in 1-5, characterized in that, include: Multiple underwater transducer array modules are used to transmit signals and receive signals fed back from the target. Spherical positioning module for coarse positioning of the target object; The short-time Fourier transform module is used to perform time-frequency decomposition on the target echo signal received by the underwater transducer array to obtain the complex spectral coefficients of the time-frequency unit; Reference beamforming module for forming a distortion-free response channel in the target direction; The interference subspace modeling module is used to estimate and separate interference signals from non-target directions; The sidelobe suppression and weight optimization module is used to achieve target enhancement and interference suppression.

Images