Underwater sound source target identification method and system based on dynamic selection integration technology, and storage medium

Abstract

The invention discloses an underwater sound source target identification method based on a dynamic selection integration technology, and the method comprises the steps: obtaining a sound source signal in real time through an underwater sensor, including acoustic features of a submarine target and a sound bait, extracting a time-frequency feature, a modulation feature and a statistical feature through short-time Fourier transform, and fusing the features through PCA (principal component analysis). The method comprises the following steps: generating a multi-dimensional feature vector, establishing a plurality of classifier models, evaluating the performance through cross validation, intelligently selecting an optimal classifier combination, inputting the fused feature vector into the optimal combination for target classification, outputting an identification result, and finally, continuously updating the classifier models by adopting an incremental learning method according to the identification result and an actual situation. And continuous optimization management of underwater sound source target identification is realized. Through a multi-dimensional feature fusion mechanism, time-frequency features, modulation features, statistical features and other information are comprehensively considered, the distinguishing capacity of the model for a target and an interferent is improved, in addition, through intelligent selection of an optimal combination in multiple classifiers, the distinguishing capacity of a submarine target and a sound bait is improved, and the recognition precision is effectively enhanced.

Description

Technical Field

[0001] This invention belongs to the field of data recognition, and in particular relates to a method, system and storage medium for underwater sound source target recognition based on dynamic selection integration technology. Background Technology

[0002] In underwater acoustic homing weapon systems, target identification technology is crucial for submarine detection, tracking, and attack decisions. Existing underwater target identification methods primarily rely on single classification models, which can achieve high accuracy in ideal environments. However, in complex underwater acoustic environments, especially with active or passive acoustic decoy interference, the identification capability of traditional methods is severely affected. The underwater environment contains various noises and interferences (such as marine life and ship noise), and many existing methods perform poorly under noisy conditions, easily leading to false identifications or missed detections, thus reducing system reliability.

[0003] Existing underwater target recognition methods rely on fixed feature existence and extraction classification rules, making them unable to adapt to dynamic changes in the underwater acoustic environment. They suffer from insufficient ability to distinguish against acoustic decoys and are easily affected by noise and environmental changes, leading to decreased recognition accuracy. Furthermore, many traditional methods lack adaptability to changes in the underwater environment. This results in a significant decline in model performance under different water conditions, temperatures, salinity, and other conditions, impacting recognition accuracy.

[0004] In summary, these shortcomings and deficiencies limit the effectiveness and application scope of existing underwater target identification methods, and improvements are urgently needed in areas such as noise robustness and system adaptability. Summary of the Invention

[0005] To address the shortcomings of the existing technology, this invention provides an underwater sound source target identification method based on dynamic selection integration technology, the method comprising the following steps: Step 1: Acquire sound source signals through underwater sensors. The sound source signals include at least the acoustic characteristics of the submarine target and the acoustic decoy. Step 2: Obtain the time-frequency characteristics, modulation characteristics, and statistical characteristics of the sound source signal based on the short-time Fourier transform; Step 3: Based on the principal component analysis (PCA) method, the time-frequency features, modulation features, and statistical features obtained above are fused to obtain a multidimensional feature vector; Step 4: Build multiple classifier models, evaluate the performance of each classifier under different feature combinations through cross-validation, and intelligently select the optimal classifier combination based on dynamic selection ensemble. Step 5: Input the fused multidimensional feature vector into the optimal classifier combination for target classification, and output the identification result of submarine target or acoustic decoy; Step 6: Based on the recognition results and the actual situation, update the classifier model using an incremental learning method, and continuously execute steps 1 to 6 to achieve continuous optimization and management of underwater sound source target recognition.

[0006] The original sound source signal x(t) is divided into N frames, each with a length of L, and the frame shift is set to M (M <L); The window function for each frame is calculated as follows:

[0007] The sound source signal is processed based on the window function, as follows:

[0008] Where, x w [n] represents the windowed signal; For each frame x w [n] Calculate the Fourier transform, expressed as:

[0009] Where X(t,f) represents the spectral information at time t and frequency f. is a complex exponential function, representing the fundamental frequency in the Fourier transform.

[0010] Extracting time-frequency features from the sound source signal includes: The energy spectrum is calculated and expressed as:

[0011] The short-time energy is calculated and expressed as:

[0012] Where En represents the short-time energy of the nth frame, and f is the frequency index; The spectral centroid is calculated as follows:

[0013] Where, ∑ f f |X(n,f)| 2 This represents the frequency-weighted sum of energy. Extracting modulation features from sound source signals includes: The modulation depth D is calculated based on the envelope analysis method and is expressed as:

[0014]

[0015] Where D is the modulation depth, and H[x(t)] is the Hilbert transform of the signal x(t); Extracting statistical features from sound source signals, including: The mean and variance of each frame are calculated and expressed as follows:

[0016]

[0017] Where, μ n Let σ be the mean of the nth frame. n 2 Let F be the variance of the nth frame, and F be the number of frequency points used to calculate the mean and variance. The skewness γ and kurtosis β are calculated as follows:

[0018]

[0019] Where |X(n,f)| represents the amplitude of the nth frame at frequency f.

[0020] Based on the time-frequency features, modulation features, and statistical features, a feature matrix X is constructed, represented as follows:

[0021] Where m is the number of samples, n is the number of features, each row represents a sample, and each column represents a feature; The feature matrix is ​​standardized and expressed as follows:

[0022] Where μ is the mean of the feature and σ is the standard deviation of the feature; The covariance matrix C of the standardized feature matrix X′ is calculated as follows: ; Perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues ​​λ and eigenvectors v. Arrange the eigenvalues ​​in descending order and select the first k eigenvalues ​​and their corresponding eigenvectors. The projection matrix W is constructed based on the selected eigenvectors, and is expressed as follows:

[0023] Projecting the standardized feature matrix X' onto the multidimensional feature vector, we can express it as follows:

[0024] Where Z is a multidimensional feature vector containing k principal components.

[0025] Among them, multiple classifiers are selected, including support vector machines (SVM), decision trees, random forests, k-nearest neighbors (k-NN) and / or neural networks, to build multiple classifier models; The time-frequency features, modulation features, and statistical features are used to generate multiple datasets based on a binary combination algorithm. The datasets are then divided into training sets and test sets, and the training sets and test sets are standardized. We choose fold number k to divide the training set into k subsets. For each subset i, subset i is used as the validation set, and the remaining k are used as the validation set. A subset is used as the training set to train the classifier and evaluate its performance on the validation set. Based on the performance evaluation results, multiple classifier combinations are selected and weighted voting is used to build an ensemble model. The entire training set is then used for training to obtain the optimal classifier combination.

[0026] The multidimensional feature vectors are input into the optimal classifier combination. Each base classifier classifies the multidimensional feature vectors and outputs a predicted recognition result. The output of each base classifier is input into the meta-learner based on the stacking method to generate the final recognition result.

[0027] Among them, based on the final identification results and actual data, the performance indicators of the classifier combination are obtained, including at least one of the following: accuracy, precision, recall, and F1 score; Determine whether an update is needed based on the assessment results; If so, new sample data are collected and cleaned and preprocessed. Based on batch incremental learning, the cleaned and preprocessed sample data is input into the trained classifier for incremental learning, resulting in an updated classifier model.

[0028] This invention provides an underwater sound source target identification system based on dynamic selection integration technology, the system comprising: Data acquisition module, signal processing module, feature fusion module, model building module, recognition module, and optimization module; The data acquisition module is used to acquire sound source signals through underwater sensors, and the sound source signals include at least the acoustic characteristics of the submarine target and the acoustic decoy; The signal processing module is used to obtain the time-frequency characteristics, modulation characteristics, and statistical characteristics of the sound source signal based on the short-time Fourier transform; The feature fusion module is used to fuse the time-frequency features, modulation features, and statistical features obtained above based on the principal component analysis (PCA) method to obtain a multidimensional feature vector; The model building module is used to build multiple classifier models, evaluate the performance of each classifier under different feature combinations through cross-validation, and intelligently select the optimal classifier combination based on dynamic selection ensemble. The identification module is used to input the fused multidimensional feature vector into the optimal classifier combination for target classification and output the identification result of submarine target or acoustic decoy; The optimization module is used to update the classifier model using an incremental learning method based on the recognition results and the actual situation, and continuously execute steps 1 to 6 to achieve continuous optimization management of underwater sound source target recognition.

[0029] This invention acquires sound source signals in real time using underwater sensors, including the acoustic characteristics of submarine targets and acoustic decoys. It extracts time-frequency features, modulation features, and statistical features using short-time Fourier transform, and fuses these features through principal component analysis (PCA) to generate a multi-dimensional feature vector. Multiple classifier models are established, and their performance is evaluated through cross-validation. The optimal classifier combination is intelligently selected, and the fused feature vector is input into the optimal combination for target classification, outputting the recognition result. Finally, based on the recognition result and the actual situation, an incremental learning method is used to continuously update the classifier model, achieving continuous optimization and management of underwater sound source target recognition. Through the multi-dimensional feature fusion mechanism, which comprehensively considers information such as time-frequency features, modulation features, and statistical features, the model's ability to distinguish between targets and interference is improved. Furthermore, by intelligently selecting the optimal combination among multiple classifiers, the ability to distinguish submarine targets from acoustic decoys is enhanced, effectively improving recognition accuracy. Attached Figure Description

[0030] The above and other objects, features, and advantages of exemplary embodiments of the present disclosure will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the present disclosure are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein: Figure 1 This is a flowchart illustrating an underwater sound source target identification method based on dynamic selection integration technology according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the structure of an underwater sound source target identification system based on dynamic selection integration technology according to an embodiment of the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0032] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0033] It should be understood that although the terms first, second, third, etc., may be used to describe... in the embodiments of the present invention, these... should not be limited to these terms. These terms are only used to distinguish... For example, first... may also be referred to as second... without departing from the scope of the embodiments of the present invention, and similarly, second... may also be referred to as first...

[0034] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0035] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0036] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device that includes said element.

[0037] Existing underwater acoustic target identification methods mostly rely on fixed features, lacking adaptability to changes in the underwater environment. Furthermore, they are prone to false positives or false negatives in the face of various noises and interferences (such as marine life and ship noise), reducing system reliability. Therefore, an effective method is needed to significantly improve the accuracy, robustness, and real-time performance of underwater target identification, enhance the system's ability to counter acoustic decoys, and ensure the stable and efficient operation of underwater acoustic homing weapons in the aforementioned complex environments.

[0038] like Figure 1As shown, this invention discloses an underwater sound source target identification method based on dynamic selection ensemble technology, the method comprising: Step 1: Acquire sound source signals using underwater sensors. The sound source signals include at least the acoustic characteristics of the submarine target and the acoustic decoy.

[0039] Select sensors suitable for underwater acoustic signal acquisition, such as underwater microphones (Hydrophones), sonar systems, and Doppler acoustic sensors, and set appropriate sampling frequencies to monitor and acquire sound source signals. Consider underwater environmental factors, such as water flow, temperature, and salinity, as these factors affect sound wave propagation characteristics, and choose appropriate placement locations.

[0040] The collected sound source signals are recorded in real time to the data storage device, and the collected sound source signals are denoised to facilitate subsequent feature extraction and analysis.

[0041] By following the steps above, the acoustic characteristics of submarine targets and acoustic decoys can be effectively acquired using underwater sensors. This process includes selecting appropriate sensors, their deployment, signal acquisition, preprocessing, feature extraction, and data management, providing a solid foundation for subsequent analysis and identification.

[0042] Step 2: Obtain the time-frequency characteristics, modulation characteristics, and statistical characteristics of the sound source signal based on the short-time Fourier transform.

[0043] Short-Time Fourier Transform (STFT) can simultaneously provide time and frequency information of a signal, making it suitable for processing non-stationary signals and clearly showing the frequency spectrum of a sound source signal as it changes over time. Furthermore, by extracting time-frequency features, modulation features, and statistical features, a rich feature set is formed. This multi-dimensional feature set can more comprehensively describe the characteristics of the sound source signal, enhancing its recognition capabilities.

[0044] In one embodiment, the original sound source signal x(t) is divided into N frames, each frame having a length of L, and a frame shift M (M <L); The window function for each frame is calculated as follows:

[0045] The sound source signal is processed based on the window function, as follows:

[0046] Where, x w [n] represents the windowed signal; For each frame x w [n] Calculate the Fourier transform, expressed as:

[0047] Where X(t,f) represents the spectral information at time t and frequency f. is a complex exponential function, representing the fundamental frequency in the Fourier transform.

[0048] The time-frequency features, modulation features, and statistical features extracted through short-time Fourier transform significantly enhance the underwater target recognition performance. Time-frequency features provide a comprehensive view of the signal in both time and frequency domains, making the identification of submarine targets and acoustic decoys more accurate. The introduction of modulation features enables the system to capture more complex signal patterns, helping to distinguish similar targets. Statistical features increase robustness to noise, improving recognition stability in complex underwater environments.

[0049] In one embodiment, extracting time-frequency features from the sound source signal includes: The energy spectrum is calculated and expressed as:

[0050] The short-time energy is calculated and expressed as:

[0051] Where En represents the short-time energy of the nth frame, and f is the frequency index; The spectral centroid is calculated as follows:

[0052] Where, ∑ f f |X(n,f)| 2 This represents the frequency-weighted sum of energy. In one embodiment, extracting modulation features from the sound source signal includes: The modulation depth D is calculated based on the envelope analysis method and is expressed as:

[0053]

[0054] Where D is the modulation depth, and H[x(t)] is the Hilbert transform of the signal x(t); In one embodiment, extracting statistical features from the sound source signal includes: The mean and variance of each frame are calculated and expressed as follows:

[0055]

[0056] Where, μ n Let σ be the mean of the nth frame. n 2Let F be the variance of the nth frame, and F be the number of frequency points used to calculate the mean and variance. The skewness γ and kurtosis β are calculated as follows:

[0057]

[0058] Where |X(n,f)| represents the amplitude of the nth frame at frequency f.

[0059] The integration of these multidimensional features lays a solid foundation for subsequent data classification and model training, promotes the improvement of real-time processing capabilities, ensures that the system can quickly respond to dynamic changes, and improves overall recognition efficiency.

[0060] Step 3: Based on the principal component analysis (PCA) method, the time-frequency features, modulation features, and statistical features obtained above are fused to obtain a multidimensional feature vector.

[0061] Principal Component Analysis (PCA) maps a high-dimensional feature space to a low-dimensional space through linear transformation, effectively reducing the number of features while preserving the main information of the data and reducing computational complexity. By combining time-frequency features, modulation features, and statistical features, PCA can generate a comprehensive multi-dimensional feature vector, allowing different types of features to complement each other and enhancing the overall descriptive power of the signal.

[0062] Principal component analysis enables effective feature fusion and dimensionality reduction, which not only improves computational efficiency and classification performance but also enhances the model's generalization ability and adaptability, providing a solid foundation for underwater target recognition systems.

[0063] In one embodiment, a feature matrix X is constructed based on the time-frequency features, modulation features, and statistical features, and is represented as follows:

[0064] Where m is the number of samples, n is the number of features, each row represents a sample, and each column represents a feature; The feature matrix is ​​standardized and expressed as follows:

[0065] Where μ is the mean of the feature and σ is the standard deviation of the feature; The covariance matrix C of the standardized feature matrix X′ is calculated as follows: ; Perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues ​​λ and eigenvectors v. Arrange the eigenvalues ​​in descending order and select the first k eigenvalues ​​and their corresponding eigenvectors. The projection matrix W is constructed based on the selected eigenvectors, and is expressed as follows:

[0066] Projecting the standardized feature matrix X' onto the multidimensional feature vector, we can express it as follows:

[0067] Where Z is a multidimensional feature vector containing k principal components.

[0068] By fusing time-frequency features, modulation features, and statistical features through Principal Component Analysis (PCA), the efficiency and effectiveness of feature representation are significantly improved. PCA effectively reduces dimensionality, removes redundant information, and retains the most representative features, thereby reducing computational burden and accelerating subsequent processing. Furthermore, the fused multidimensional feature vectors provide a more compact and comprehensive signal description, which not only improves the learning efficiency of the classifier but also enhances the model's generalization ability, making it more stable when facing different underwater environments and targets. Ultimately, this process improves overall recognition accuracy and the system's real-time response capability, providing strong support for underwater target identification.

[0069] Step 4: Build multiple classifier models, evaluate the performance of each classifier under different feature combinations through cross-validation, and intelligently select the optimal classifier combination based on dynamic selection ensemble.

[0070] By building multiple classifier models (such as decision trees, support vector machines, and random forests), diverse options are provided, enabling optimization for different feature combinations. Cross-validation is employed to effectively evaluate the model's performance on different datasets, reducing the risk of overfitting and ensuring the model's generalization ability on unknown data.

[0071] By analyzing the performance of each classifier under different feature combinations, the system can intelligently select the optimal classifier combination, fully utilizing the strengths of each model. Evaluating the effectiveness of different feature combinations ensures that the system can find the most suitable feature set for the current task, thereby improving overall performance.

[0072] In one embodiment, multiple classifiers are selected, including support vector machines (SVM), decision trees, random forests, k-nearest neighbors (k-NN), and / or neural networks, to build multiple classifier models; The time-frequency features, modulation features, and statistical features are used to generate multiple datasets based on a binary combination algorithm. The datasets are then divided into training sets and test sets, and the training sets and test sets are standardized. We choose fold number k to divide the training set into k subsets. For each subset i, subset i is used as the validation set, and the remaining k are used as the validation set. A subset is used as the training set to train the classifier and evaluate its performance on the validation set. Based on the performance evaluation results, multiple classifier combinations are selected and weighted voting is used to build an ensemble model. The entire training set is then used for training to obtain the optimal classifier combination.

[0073] By establishing multiple classifier models and employing cross-validation to evaluate their performance, the accuracy and reliability of underwater target identification were significantly improved. Cross-validation comprehensively evaluates the performance of each classifier under different feature combinations, ensuring the robustness and generalization ability of the models. Through dynamic selection of ensemble, the system can intelligently identify the optimal combination of classifiers, fully utilizing the advantages of each model and reducing the bias and error that may be introduced by a single model. This not only improves identification accuracy but also enhances the system's adaptability to complex underwater environments, making the target identification process more efficient and reliable, thus providing stronger support for practical applications.

[0074] Step 5: Input the fused multidimensional feature vector into the optimal classifier combination for target classification, and output the identification result of submarine target or acoustic decoy.

[0075] The fused multidimensional feature vectors are input into the optimal classifier combination to ensure efficient and accurate classification, adapting to complex underwater environments. By combining multiple optimized classifiers and integrating their individual judgments, the reliability and accuracy of the overall decision-making are enhanced.

[0076] In one embodiment, the multidimensional feature vector is input into the optimal classifier combination, each base classifier classifies the multidimensional feature vector and outputs a predicted recognition result, and the output of each base classifier is input into the meta-learner based on the stacking method to generate the final recognition result.

[0077] By inputting the fused multidimensional feature vectors into an optimal classifier combination for target classification, the efficiency and accuracy of submarine target and acoustic decoy identification are significantly improved. This process utilizes an optimized classifier combination to ensure rapid and accurate target identification in complex underwater environments. By integrating the rich feature information extracted in previous steps, the classifier can effectively distinguish different types of sound sources, reducing the risk of false positives and false negatives. Furthermore, the output identification results provide a reliable basis for subsequent decision-making, enhancing the system's real-time response capability and practicality, thereby promoting more efficient underwater monitoring and safety assurance in real-world applications.

[0078] Step 6: Based on the recognition results and the actual situation, update the classifier model using an incremental learning method, and continuously execute steps 1 to 6 to achieve continuous optimization and management of underwater sound source target recognition.

[0079] Incremental learning enables the model to be updated upon receiving new data, eliminating the need for retraining from scratch and saving computational resources and time. The system can quickly adjust the classifier based on real-time recognition results and feedback from actual conditions, improving the model's adaptability to new environments and data. Incremental learning preserves existing knowledge and experience, preventing information loss and ensuring the model maintains high performance during updates.

[0080] In one embodiment, the performance metrics of the classifier combination are obtained based on the final identification results and actual data, including at least one of the following: accuracy, precision, recall, and F1 score. Determine whether an update is needed based on the assessment results; If so, new sample data are collected and cleaned and preprocessed. Based on batch incremental learning, the cleaned and preprocessed sample data is input into the trained classifier for incremental learning, resulting in an updated classifier model.

[0081] This invention, based on dynamic selection ensemble technology, intelligently selects the optimal combination from multiple classifiers to improve the ability to distinguish submarine targets from acoustic decoys, effectively enhancing recognition accuracy. It introduces a multi-dimensional feature fusion mechanism, comprehensively considering information such as time-frequency features, modulation features, and statistical features to improve the model's ability to distinguish targets and interference. By combining the dynamic classifier selection mechanism with optimized feature extraction strategies, it effectively improves the accuracy, robustness, and real-time performance of underwater target recognition, enhances the system's ability to counter acoustic decoys, and ensures the stable and efficient operation of underwater acoustic homing weapons in the aforementioned complex environments. It significantly improves the intelligence level and environmental adaptability of underwater target recognition systems, providing more accurate and efficient target recognition and strike decision support for underwater weapon systems.

[0082] Corresponding to the underwater sound source target identification method based on dynamic selection ensemble technology provided by this invention, this invention also provides an underwater sound source target identification system based on dynamic selection ensemble technology. For example... Figure 2 As shown, the system includes: Data acquisition module, signal processing module, feature fusion module, model building module, recognition module, and optimization module; The data acquisition module is used to acquire sound source signals through underwater sensors, and the sound source signals include at least the acoustic characteristics of the submarine target and the acoustic decoy; The signal processing module is used to obtain the time-frequency characteristics, modulation characteristics, and statistical characteristics of the sound source signal based on the short-time Fourier transform; The feature fusion module is used to fuse the time-frequency features, modulation features, and statistical features obtained above based on the principal component analysis (PCA) method to obtain a multidimensional feature vector; The model building module is used to build multiple classifier models, evaluate the performance of each classifier under different feature combinations through cross-validation, and intelligently select the optimal classifier combination based on dynamic selection ensemble. The identification module is used to input the fused multidimensional feature vector into the optimal classifier combination for target classification and output the identification result of submarine target or acoustic decoy; The optimization module is used to update the classifier model using an incremental learning method based on the recognition results and the actual situation, and continuously execute steps 1 to 6 to achieve continuous optimization management of underwater sound source target recognition.

[0083] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0084] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0085] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0086] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0087] The units described in the embodiments of this disclosure can be implemented in software or hardware. The names of the units are not, in some cases, intended to limit the specific unit.

[0088] The preferred embodiments of the present invention have been described above to make the spirit of the present invention clearer and easier to understand, and are not intended to limit the present invention. All modifications, substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope summarized by the appended claims.

Claims

1. A method for underwater sound source target identification based on dynamic selection ensemble technology, comprising: Step 1: Acquire sound source signals through underwater sensors. The sound source signals include at least the acoustic characteristics of the submarine target and the acoustic decoy. Step 2: Obtain the time-frequency characteristics, modulation characteristics, and statistical characteristics of the sound source signal based on the short-time Fourier transform; Step 3: Based on the principal component analysis (PCA) method, the time-frequency features, modulation features, and statistical features obtained above are fused to obtain a multidimensional feature vector; Step 4: Build multiple classifier models, evaluate the performance of each classifier under different feature combinations through cross-validation, and intelligently select the optimal classifier combination based on dynamic selection ensemble. Step 5: Input the fused multidimensional feature vector into the optimal classifier combination for target classification, and output the identification result of submarine target or acoustic decoy; Step 6: Based on the recognition results and the actual situation, update the classifier model using an incremental learning method, and continuously execute steps 1 to 6 to achieve continuous optimization and management of underwater sound source target recognition.

2. The underwater sound source target identification method based on dynamic selection ensemble technology as described in claim 1, characterized in that, The original sound source signal x(t) is divided into N frames, each with a length of L, and the frame shift is set to M (M <L); The window function for each frame is calculated as follows: The sound source signal is processed based on the window function, as follows: Where, x w [n] represents the windowed signal; For each frame x w [n] Calculate the Fourier transform, expressed as: Where X(t,f) represents the spectral information at time t and frequency f. is a complex exponential function, representing the fundamental frequency in the Fourier transform.

3. The underwater sound source target identification method based on dynamic selection integration technology as described in claim 2, characterized in that, Extracting time-frequency features from sound source signals includes: The energy spectrum is calculated and expressed as: The short-time energy is calculated and expressed as: Where En represents the short-time energy of the nth frame, and f is the frequency index; The spectral centroid is calculated as follows: Where, ∑ f f |X(n,f)| 2 This represents the frequency-weighted sum of energy. Extracting modulation features from sound source signals includes: The modulation depth D is calculated based on the envelope analysis method and is expressed as: Where D is the modulation depth, and H[x(t)] is the Hilbert transform of the signal x(t); Extracting statistical features from sound source signals, including: The mean and variance of each frame are calculated and expressed as follows: Where, μ n Let σ be the mean of the nth frame. n 2 Let F be the variance of the nth frame, and F be the number of frequency points used to calculate the mean and variance. The skewness γ and kurtosis β are calculated as follows: Where |X(n,f)| represents the amplitude of the nth frame at frequency f.

4. The underwater sound source target identification method based on dynamic selection integration technology as described in claim 1, characterized in that, Based on the aforementioned time-frequency features, modulation features, and statistical features, a feature matrix X is constructed, represented as follows: Where m is the number of samples, n is the number of features, each row represents a sample, and each column represents a feature; The feature matrix is ​​standardized and expressed as follows: Where μ is the mean of the feature and σ is the standard deviation of the feature; The covariance matrix C of the standardized feature matrix X′ is calculated as follows: ； Perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues ​​λ and eigenvectors v. Arrange the eigenvalues ​​in descending order and select the first k eigenvalues ​​and their corresponding eigenvectors. The projection matrix W is constructed based on the selected eigenvectors, and is expressed as follows: Projecting the standardized feature matrix X' onto the multidimensional feature vector, we can express it as follows: Where Z is a multidimensional feature vector containing k principal components.

5. The underwater sound source target identification method based on dynamic selection integration technology as described in claim 1, characterized in that, Choose multiple classifiers, including support vector machines (SVM), decision trees, random forests, k-nearest neighbors (k-NN), and / or neural networks, to build multiple classifier models; The time-frequency features, modulation features, and statistical features are used to generate multiple datasets based on a binary combination algorithm. The datasets are then divided into training sets and test sets, and the training sets and test sets are standardized. We choose fold number k to divide the training set into k subsets. For each subset i, subset i is used as the validation set, and the remaining k are used as the validation set. A subset is used as the training set to train the classifier and evaluate its performance on the validation set. Based on the performance evaluation results, multiple classifier combinations are selected and weighted voting is used to build an ensemble model. The entire training set is then used for training to obtain the optimal classifier combination.

6. The underwater sound source target identification method based on dynamic selection integration technology as described in claim 1, characterized in that, The multidimensional feature vectors are input into the optimal classifier combination. Each base classifier classifies the multidimensional feature vectors and outputs the predicted recognition result. Based on the stacking method, the output of each base classifier is input into the meta-learner to generate the final recognition result.

7. The underwater sound source target identification method based on dynamic selection integration technology as described in claim 1, characterized in that, Based on the final identification results and actual data, obtain the performance metrics of the classifier combination, including at least one of the following: accuracy, precision, recall, and F1 score; Determine whether an update is needed based on the assessment results; If so, new sample data are collected and cleaned and preprocessed. Based on batch incremental learning, the cleaned and preprocessed sample data is input into the trained classifier for incremental learning, resulting in an updated classifier model.

8. An underwater sound source target identification system based on dynamic selection integration technology, comprising: Data acquisition module, signal processing module, feature fusion module, model building module, recognition module, and optimization module; The data acquisition module is used to acquire sound source signals through underwater sensors, and the sound source signals include at least the acoustic characteristics of the submarine target and the acoustic decoy; The signal processing module is used to obtain the time-frequency characteristics, modulation characteristics, and statistical characteristics of the sound source signal based on the short-time Fourier transform; The feature fusion module is used to fuse the time-frequency features, modulation features, and statistical features obtained above based on the principal component analysis (PCA) method to obtain a multidimensional feature vector; The model building module is used to build multiple classifier models, evaluate the performance of each classifier under different feature combinations through cross-validation, and intelligently select the optimal classifier combination based on dynamic selection ensemble. The identification module is used to input the fused multidimensional feature vector into the optimal classifier combination for target classification and output the identification result of submarine target or acoustic decoy; The optimization module is used to update the classifier model using an incremental learning method based on the recognition results and the actual situation, and continuously execute steps 1 to 6 to achieve continuous optimization management of underwater sound source target recognition.

9. An underwater sound source target identification device based on dynamic selection integration technology, comprising: At least one processor; as well as At least one memory including computer program code, The at least one memory and the computer program code are configured, together with the at least one processor, to cause the device to perform the method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, the computer program, when executed by a processor, implementing the method of any one of claims 1-7.

Images