Underwater sound target identification method and system based on expert hybrid model

Abstract

The invention discloses an underwater acoustic target identification method and system based on an expert hybrid model, and the method comprises the steps: introducing an unbalanced bandwidth Mel spectrogram for solving the problems that the low-frequency features of underwater acoustic signals are not fully extracted and the frequency distribution is unknown because the distribution rules of the underwater acoustic signals of different frequency bands are different; therefore, the learning ability for different frequency band features is improved, and the time-frequency feature enhancement for an underwater acoustic target recognition task is realized. In order to solve the problem that the learning ability of a model for different frequency band features is limited when different sub-band features are identified, an expert hybrid model for different sub-band features is designed, and the learning ability of a UBM-MoER network for different frequency band features is improved. In order to further explore how to improve the feature extraction capability of the model under the expert hybrid model, a hybrid domain attention mechanism module is designed, and the sensitivity of each expert hybrid model to key features is expected to be improved through a channel and space attention mechanism, so that the overall performance of the model is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of underwater acoustic target recognition, and relates to an underwater acoustic target recognition method and system based on an expert hybrid model. Background Art

[0002] Passive recognition of underwater acoustic targets is a key link in underwater sound source management, key sea area protection and future marine information construction. However, underwater acoustic target recognition is a very difficult task. The underwater acoustic targets themselves are pseudo-random and are affected by the ocean environmental noise. The differences between targets are not obvious, and the targets themselves have diverse motion states, which can easily lead to misjudgments or conflicting results in recognition. Traditional recognition methods cannot meet the increasingly complex and demanding recognition needs. Deep learning methods provide the possibility of realizing underwater acoustic target recognition.

[0003] Traditional underwater acoustic target recognition combines signal processing with machine learning. The characteristics of underwater acoustic signals are generally divided into time domain characteristics and spectrum characteristics. Time domain characteristics mainly include mean, zero crossing rate, peak value, etc. The characteristics of these features are that they do not require complex feature extraction, but due to the complex composition of underwater acoustic target radiation noise, the descriptiveness of this type of features is poor. Therefore, it is very important to apply frequency domain features in target recognition technology. Time-frequency features can characterize the change of signal frequency over time and can more comprehensively describe the characteristics of the signal. Short-time Fourier transform, wavelet transform and other methods can be used to achieve time-frequency representation. However, underwater acoustic signals have frequency absorption during propagation. High-frequency signals are gradually absorbed by seawater at close range, and low-frequency signals propagate farther. Therefore, in the task of performing underwater acoustic target signal recognition, due to the lack of prior knowledge of the target, the target distance corresponding to the received signal is unknown, and the main frequency band of the signal is also unknown. Machine learning methods based on statistical classification can achieve good results in some ideal environments, but machine learning can only roughly fit the distribution of samples and cannot mine the deep features of input data. The public research on the feature extraction or recognition of underwater passive targets by deep learning networks has been less than ten years so far, but it has shown obvious advantages over traditional methods. Because it can automatically explore the features required for the task from complex raw data, it has made great achievements in fields such as computer vision and natural language processing. Convolutional neural networks have undergone a series of improvements, such as AlexNet, VGGNet, ResNet and DenseNet, and have obtained strong feature extraction capabilities. Due to their superiority in feature extraction, many studies have applied them to underwater acoustic target recognition. However, the accuracy rate has shown a downward trend in actual recognition tasks. The main reasons for this phenomenon are as follows: (1) Due to the differences in the distribution laws of underwater acoustic signals in different frequency bands, the low-frequency features of underwater acoustic signals are not sufficiently extracted and the frequency distribution is unknown, which affects the recognition efficiency; (2) Due to the complexity and diversity of underwater acoustic signals, the network model is prone to insufficient extraction of their time-frequency features; (3) Ship noise with large intra-class differences and high inter-class similarity cannot be effectively recognized, resulting in recognition errors. Summary of the invention

[0004] The purpose of the present invention is to solve the problem of low recognition rate in underwater acoustic recognition in the prior art, and to provide an underwater acoustic target recognition method and system based on an expert mixture model.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] The present invention proposes an underwater acoustic target recognition method based on an expert mixture model, comprising:

[0007] Acquire an underwater acoustic signal sample, pre-process the underwater acoustic signal sample, and acquire an unbalanced bandwidth Mel-spectrogram time-frequency diagram;

[0008] The front-end backbone network is used to convert the unbalanced bandwidth Mel-spectrogram into a fixed-dimensional time-frequency graph to obtain deep features;

[0009] The deep features are input into the pre-built expert mixture model to obtain the underwater acoustic target recognition results.

[0010] Preferably, the step of obtaining the underwater acoustic signal sample and preprocessing the underwater acoustic signal sample to obtain the unbalanced bandwidth Mel-spectrogram time-frequency diagram is as follows:

[0011] The underwater acoustic signal samples are divided into multiple frequency bands by frames, each frame is windowed, and each frame group is Fourier transformed. Then the underwater acoustic signal samples are input into different filters for logarithmic operation to obtain sub-band Mel spectrograms. The sub-band Mel spectrograms are concatenated along the frequency direction to obtain Mel features, and the unbalanced bandwidth Mel spectrogram time-frequency diagram is obtained based on the Mel features.

[0012] Preferably, the front-end backbone network structure is based on ResNet, uses LayerNorm as the normalization method, and GELU as the activation function; specifically: the unbalanced bandwidth Mel-spectrogram time-frequency graph passes through the convolution layer and the LN layer, and the following four stages use the downsampling module and stacked ConvNeXt Blocks of different times to realize feature extraction, and finally passes through the global average pooling layer to output the deep features;

[0013] Among them, the downsampling module is composed of LN layer and convolution layer, and ConvNeXt Block is composed of convolution layer, LN layer, GELU layer and mixed domain attention mechanism module.

[0014] Preferably, the mixed-domain attention mechanism module combines a channel attention module and a spatial attention module;

[0015] The feature map input by the channel attention mechanism module is processed by global maximum pooling based on width and global average pooling based on height, and then operated through a shared fully connected layer. After summation and Sigmoid activation, the channel attention feature map is obtained;

[0016] The spatial attention mechanism takes the output of the channel attention module as input, performs global maximum pooling and global average pooling operations on the input feature map, then concatenates the results of the two, reduces the channel dimension to 1 through convolution operation, and generates a spatial attention feature map through Sigmoid.

[0017] Preferably, the expert mixture model is composed of a routing layer and an expert layer, the routing layer is composed of a softmax function, and the expert layer is composed of a multi-layer perceptron with two linear layers.

[0018] Preferably, the deep features are input into a pre-built expert mixture model to obtain the underwater acoustic target recognition result, specifically:

[0019] Set the batch size to n and the input unbalanced bandwidth Mel spectrogram to {x i}, and its corresponding label is set to {y i}, (i=1,2...n), the number of expert layers is set to m, and the expert level is set to E1(·),…E m (·), input unbalanced bandwidth Mel spectrogram time-frequency graph x i , after passing through the front-end backbone network, the deep feature r is obtained i , the deep feature r i Input into the routing layer S(·) to get the routing score The softmax function is used to normalize the routing score to the routing probability p i , routing probability p i represents the probability of assigning the sample to each expert; the deep feature r i Input it to the expert layer with the highest corresponding probability and get the category of the ship, which is the result of hydroacoustic target recognition.

[0020] The present invention proposes an underwater acoustic target recognition system based on an expert mixture model, comprising:

[0021] A data preprocessing module is used to obtain underwater acoustic signal samples, preprocess the underwater acoustic signal samples, and obtain an unbalanced bandwidth Mel-spectrogram time-frequency diagram;

[0022] A feature extraction module is used to convert the unbalanced bandwidth Mel-spectrogram time-frequency graph into a fixed-dimensional time-frequency graph using a front-end backbone network to obtain deep features;

[0023] The target recognition module is used to input deep features into the pre-built expert mixture model to obtain underwater acoustic target recognition results.

[0024] Preferably, the step of obtaining the underwater acoustic signal sample and preprocessing the underwater acoustic signal sample to obtain the unbalanced bandwidth Mel-spectrogram time-frequency diagram is as follows:

[0025] The underwater acoustic signal samples are divided into multiple frequency bands by frames, each frame is windowed, and each frame group is Fourier transformed. Then the underwater acoustic signal samples are input into different filters for logarithmic operation to obtain sub-band Mel spectrograms. The sub-band Mel spectrograms are concatenated along the frequency direction to obtain Mel features, and the unbalanced bandwidth Mel spectrogram time-frequency diagram is obtained based on the Mel features.

[0026] A terminal device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of an underwater acoustic target recognition method based on an expert hybrid model when executing the computer program.

[0027] A computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the steps of a method for underwater acoustic target recognition based on an expert mixture model.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] The present invention proposes an underwater acoustic target recognition method based on an expert mixture model. Due to the differences in the distribution patterns of underwater acoustic signals in different frequency bands, an unbalanced bandwidth Mel-spectrogram is introduced to address the problems of insufficient extraction of low-frequency features of underwater acoustic signals and unknown frequency distribution, thereby improving the learning ability of features in different frequency bands to achieve time-frequency feature enhancement for underwater acoustic target recognition tasks. In view of the limited learning ability of the model for features in different frequency bands when identifying features in different sub-bands, an expert mixture model for different sub-band features is designed to enhance the learning ability of the UBM-MoER network for features in different frequency bands. In order to further explore how to improve the feature extraction ability of the model under the expert mixture model, a mixed domain attention mechanism module is designed, hoping to improve the sensitivity of each expert mixture model to key features through channel and spatial attention mechanisms, thereby improving the overall performance of the model.

[0030] The present invention proposes an underwater acoustic target recognition system based on an expert hybrid model, which divides the system into a data preprocessing module, a feature extraction module and a target recognition module to obtain underwater acoustic target recognition results and realize underwater acoustic target recognition. The modularization concept is adopted to make each module independent of each other, which is convenient for unified management of each module. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0032] Figure 1 This is a flow chart of the underwater acoustic target recognition method based on the expert mixture model of the present invention.

[0033] Figure 2 This is the overall network architecture diagram of the present invention.

[0034] Figure 3 It is a diagram of the unbalanced bandwidth Mel-spectrogram feature extraction process of the present invention.

[0035] Figure 4 It is a structural diagram of the front-end backbone network module of the present invention.

[0036] Figure 5 It is a structural diagram of the hybrid domain attention mechanism module of the present invention.

[0037] Figure 6 This is a structural diagram of the channel attention mechanism module of the present invention.

[0038] Figure 7 This is a structural diagram of the spatial attention mechanism module of the present invention.

[0039] Figure 8 This is a structural diagram of the expert mixture model of the present invention.

[0040] Fig. 9 This is a diagram of the underwater acoustic target recognition system based on the expert mixture model of the present invention.

[0041] Fig.10 The figure is a schematic structural diagram of an electronic device of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0043] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0045] In the description of the embodiments of the present invention, it should be noted that if the terms "upper", "lower", "horizontal", "inner", etc. indicate an orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of the invention is usually placed when in use, it is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0046] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0047] The present invention proposes a method for underwater acoustic target recognition based on an expert mixture model. Figure 1 As shown, the following steps are included:

[0048] S1. Obtain an underwater acoustic signal sample, pre-process the underwater acoustic signal sample, and obtain an unbalanced bandwidth Mel-spectrogram time-frequency diagram;

[0049] The method of obtaining a water acoustic signal sample and preprocessing the water acoustic signal sample to obtain an unbalanced bandwidth Mel-spectrogram time-frequency diagram is as follows:

[0050] The underwater acoustic signal samples are divided into multiple frequency bands by frames, each frame is windowed, and each frame group is Fourier transformed. Then the underwater acoustic signal samples are input into different filters for logarithmic operation to obtain sub-band Mel spectrograms. The sub-band Mel spectrograms are concatenated along the frequency direction to obtain Mel features, and the unbalanced bandwidth Mel spectrogram time-frequency diagram is obtained based on the Mel features.

[0051] S2, using the front-end backbone network to convert the unbalanced bandwidth Mel-spectrogram time-frequency graph into a fixed-dimensional time-frequency graph to obtain deep features;

[0052] The front-end backbone network structure is based on ResNet, uses LayerNorm as the normalization method, and GELU as the activation function. Specifically, the unbalanced bandwidth Mel-spectrogram time-frequency graph passes through the convolution layer and the LN layer, and the following four stages use the downsampling module and stacked ConvNeXt Blocks of different times to realize feature extraction, and finally passes through the global average pooling layer to output the deep features;

[0053] Among them, the downsampling module is composed of LN layer and convolution layer, and ConvNeXt Block is composed of convolution layer, LN layer, GELU layer and mixed domain attention mechanism module.

[0054] The mixed domain attention mechanism module combines the channel attention module and the spatial attention module;

[0055] The feature map input by the channel attention mechanism module is processed by global maximum pooling based on width and global average pooling based on height, and then operated through a shared fully connected layer. After summation and Sigmoid activation, the channel attention feature map is obtained;

[0056] The spatial attention mechanism takes the output of the channel attention module as input, performs global maximum pooling and global average pooling operations on the input feature map, then concatenates the results of the two, reduces the channel dimension to 1 through convolution operation, and generates a spatial attention feature map through Sigmoid.

[0057] S3. Input the deep features into the pre-built expert mixture model to obtain the underwater acoustic target recognition results.

[0058] The expert mixture model consists of a routing layer and an expert layer. The routing layer is composed of a softmax function, and the expert layer is composed of a multi-layer perceptron with two linear layers.

[0059] The deep features are input into the pre-built expert mixture model to obtain the underwater acoustic target recognition result, specifically:

[0060] Set the batch size to n and the input unbalanced bandwidth Mel spectrogram to {x i}, and its corresponding label is set to {y i}, (i=1,2...n), the number of expert layers is set to m, and the expert level is set to E1(·),…E m (·), input unbalanced bandwidth Mel spectrogram time-frequency graph x i , after passing through the front-end backbone network, the deep feature r is obtained i , the deep feature r i Input into the routing layer S(·) to get the routing score The softmax function is used to normalize the routing score to the routing probability p i , routing probability p i represents the probability of assigning the sample to each expert; the deep feature r i Input it to the expert layer with the highest corresponding probability and get the category of the ship, which is the result of hydroacoustic target recognition.

[0061] The following is a description with reference to the accompanying drawings:

[0062] Figure 2It is the overall network architecture diagram proposed in the present invention, which includes three parts. The signal is firstly obtained through time-frequency representation and non-evaluation filtering to obtain the unbalanced bandwidth Mel-spectrogram time-frequency diagram, and then input into the front-end backbone network for feature extraction, and finally the MoE network is used for classification and recognition.

[0063] Figure 3 It is a diagram of the time-frequency feature extraction process of the unbalanced bandwidth Mel spectrogram of the present invention. In order to achieve effective classification and recognition, the underwater acoustic signal is divided into multiple frequency bands by framing, and each frequency band usually overlaps with each other to eliminate the error caused by the segmentation method, and then each frame is windowed, and the window signal is a Hamming window, which can reduce the influence of the edge effect on the signal frequency component. Then, each frame group is Fourier transformed. Since the main acoustic characteristics of long-range targets in underwater sound are concentrated in the low frequency band, and the ocean background noise is also concentrated in the low frequency band, the effective identification of low frequency band information is particularly important. The sampling rate is set to 16000Hz. When the signal is subjected to short-time Fourier transform (STFT), the window length of STFT is set to 2048, and the frame shift is set to 512. Then the signal is input into different sub-band Mel filters (Sub-band Vary Mel Filtering). After taking the logarithm operation (Logarithm), the Mel spectrum is logarithmic, which is usually to match the nonlinear loudness perception of the human ear. The subbands are divided into 0-1000, 1000-4000 and 4000-8000 Hz and 128 Mel filters are set at 0-1000 Hz, 64 Mel filters are set at 1000-4000 Hz, and 32 Mel filters are set at 4000-8000 Hz. The three subband Mel spectrograms are connected in series along the frequency direction to obtain 224-dimensional Mel features, and finally an unbalanced bandwidth Mel spectrogram is obtained. The conversion formula between Mel scale and frequency is:

[0064] Mel(f)=2595lg(1+f / 700)

[0065] Where f represents frequency. The Mel-spectrum is divided into sub-bands, and different numbers of independent Mel filter banks are used in each sub-band, and then the obtained sub-band features are spliced.

[0066] Figure 4This is a front-end backbone network module structure diagram of the present invention. The front-end backbone network converts the input acoustic features into fixed-dimensional representations. The structure of the front-end backbone network includes two main parts, including the ConvNeXt network and the HAM module. ConvNeXt is based on ResNet. The design process draws on the idea of ​​Swin-Transformer, uses LayerNorm (LN) as the normalization method and GELU as the activation function to simplify the network structure, reduce parameters and calculations, and improve the generalization ability and performance of the model. The size of the input time-frequency graph is 3*224*224. It first passes through the convolution layer and the LN layer. The following four stages use the downsampling module (Downsample) and stacked ConvNeXt Blocks of different times to extract features. The downsampling module is composed of the LN layer and the convolution layer. The ConvNeXt Block includes the convolution layer, the LN layer, the GELU layer and the hybrid domain attention mechanism module (HAM) module. The HAM module is placed at the end of the ConvNeXt Block in order to enhance the network's ability to extract key features. Finally, it passes through the global average pooling layer to output deep features.

[0067] Figure 5 This is the module structure diagram of the hybrid domain attention mechanism of the present invention. The hybrid domain attention mechanism combines the channel attention module and the spatial attention module, and pays attention to the relevance of the channel to key information and the importance of the feature map in the spatial dimension. Compared with the attention mechanism that only focuses on one direction, it can extract richer deep features.

[0068] Figure 6 The structure diagram of the channel attention module of the present invention is shown in Figure 2. The input feature map F is processed by the global maximum pooling (AvgPool(F) based on width) and the global average pooling (MaxPool(F) based on height. Then it is operated through a shared fully connected layer (Shared MLP), and after summation and Sigmoid activation, the channel attention feature map is obtained. c (F) represents the output feature map of the channel attention mechanism.

[0069] M c (F)=σ(MLP(AvgPool(F)+MLP(MaxPool(F)))

[0070] Figure 7The structure diagram of the spatial attention module of the present invention is shown in Figure 2. The spatial attention mechanism uses the output of the channel attention module as its input. It performs global maximum pooling (MaxPool) and global average pooling (AvgPool) operations on the input feature map, then concatenates the results of the two, reduces the channel dimension to 1 through the convolution operation (Conv layer), and generates a spatial attention feature map through Sigmoid. s (F) represents the output feature map of the spatial attention mechanism.

[0071] M s (F) = σ(f 7×7 ([AvgPool(F); MaxPool(F)]))

[0072] Figure 8 This is a structural diagram of the expert mixture model of the present invention. The expert mixture model consists of a routing layer and an expert layer. The routing layer (Routing layer) is composed of a softmax function, and the expert layer (Exper) is composed of a multi-layer perceptron with two linear layers. Different expert layers focus on differentiated deep features, thereby enhancing the classification ability of the model. The structure of each expert is the same, but his parameters are independent, and each expert will only be activated when encountering suitable input. Such a structure significantly improves the performance of underwater acoustic target recognition by capturing the potential characteristics of high-level features and adaptively learning diverse data.

[0073] Set the batch size to n and the input spectrogram to {x i}, and its corresponding label is set to {y i}, (i=1,2...n), the number of expert layers (Exper) is set to m, and the expert layers are set to E1(·),…E m (·), input x i , after passing through the front-end backbone network, the deep features are obtained and the deep features r i Input into the routing layer S(·) to get the routing score Then the routing score is normalized to the routing probability p using the softmax function i , p i represents the probability of assigning the sample to each expert.

[0074] The model transforms the deep feature r i Input to the expert layer with the highest corresponding probability, the entire routing allocation is as described in the following formula, as shown in the UBM-MOER model of 4 experts, when p i=(0.1,0.1,0.2,0.6), the depth feature r i It will be input to E4(·), and finally the result (logits) is obtained, that is, the category of the ship is identified.

[0075] p i =softmax(S(r i )),i∈[1,n]

[0076]

[0077] Example 2

[0078] The present invention proposes an underwater acoustic target recognition system based on an expert mixture model, such as Fig. 9 Said, including:

[0079] A data preprocessing module is used to obtain underwater acoustic signal samples, preprocess the underwater acoustic signal samples, and obtain an unbalanced bandwidth Mel-spectrogram time-frequency diagram;

[0080] The method of obtaining a water acoustic signal sample and preprocessing the water acoustic signal sample to obtain an unbalanced bandwidth Mel-spectrogram time-frequency diagram is as follows:

[0081] The underwater acoustic signal samples are divided into multiple frequency bands by frames, each frame is windowed, and each frame group is Fourier transformed. Then the underwater acoustic signal samples are input into different filters for logarithmic operation to obtain sub-band Mel spectrograms. The sub-band Mel spectrograms are concatenated along the frequency direction to obtain Mel features, and the unbalanced bandwidth Mel spectrogram time-frequency diagram is obtained based on the Mel features.

[0082] A feature extraction module is used to convert the unbalanced bandwidth Mel-spectrogram time-frequency graph into a fixed-dimensional time-frequency graph using a front-end backbone network to obtain deep features;

[0083] The front-end backbone network structure is based on ResNet, uses LayerNorm as the normalization method, and GELU as the activation function. Specifically, the unbalanced bandwidth Mel-spectrogram time-frequency graph passes through the convolution layer and the LN layer, and the following four stages use the downsampling module and stacked ConvNeXt Blocks of different times to realize feature extraction, and finally passes through the global average pooling layer to output the deep features;

[0084] Among them, the downsampling module is composed of LN layer and convolution layer, and ConvNeXt Block is composed of convolution layer, LN layer, GELU layer and mixed domain attention mechanism module.

[0085] The mixed domain attention mechanism module combines the channel attention module and the spatial attention module;

[0086] The feature map input by the channel attention mechanism module is processed by global maximum pooling based on width and global average pooling based on height, and then operated through a shared fully connected layer. After summation and Sigmoid activation, the channel attention feature map is obtained;

[0087] The spatial attention mechanism takes the output of the channel attention module as input, performs global maximum pooling and global average pooling operations on the input feature map, then concatenates the results of the two, reduces the channel dimension to 1 through convolution operation, and generates a spatial attention feature map through Sigmoid.

[0088] The target recognition module is used to input deep features into the pre-built expert mixture model to obtain underwater acoustic target recognition results.

[0089] The expert mixture model consists of a routing layer and an expert layer. The routing layer is composed of a softmax function, and the expert layer is composed of a multi-layer perceptron with two linear layers.

[0090] The deep features are input into the pre-built expert mixture model to obtain the underwater acoustic target recognition result, specifically:

[0091] Set the batch size to n and the input unbalanced bandwidth Mel spectrogram to {x i}, and its corresponding label is set to {y i}, (i=1,2...n), the number of expert layers is set to m, and the expert level is set to E1(·),…E m (·), input unbalanced bandwidth Mel spectrogram time-frequency graph x i , after passing through the front-end backbone network, the deep feature r is obtained i , the deep feature r i Input into the routing layer S(·) to get the routing score The softmax function is used to normalize the routing score to the routing probability p i , routing probability p i represents the probability of assigning the sample to each expert; the deep feature r i Input it to the expert layer with the highest corresponding probability and get the category of the ship, which is the result of hydroacoustic target recognition.

[0092] Example 3

[0093] See also Fig.10As shown, the present invention also provides an electronic device 100 for an underwater acoustic target recognition method based on an expert mixture model; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.

[0094] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the method for underwater acoustic target recognition based on the expert mixture model described in Example 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data (such as audio data) created according to the use of the electronic device 100, etc. In addition, the memory 101 can include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0095] The at least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and uses various interfaces and lines to connect various parts of the entire electronic device 100.

[0096] The memory 101 in the electronic device 100 stores a plurality of instructions to implement an underwater acoustic target recognition method based on an expert mixture model, and the processor 102 can execute the plurality of instructions to implement:

[0097] Acquire an underwater acoustic signal sample, pre-process the underwater acoustic signal sample, and acquire an unbalanced bandwidth Mel-spectrogram time-frequency diagram;

[0098] The front-end backbone network is used to convert the unbalanced bandwidth Mel-spectrogram into a fixed-dimensional time-frequency graph to obtain deep features;

[0099] The deep features are input into the pre-built expert mixture model to obtain the underwater acoustic target recognition results.

[0100] Example 4

[0101] If the module / unit integrated in the electronic device 100 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory and read-only memory (ROM, Read-Only Memory).

[0102] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0103] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0104] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0105] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for underwater acoustic target recognition based on expert mixture model, characterized in that: include: Acquire an underwater acoustic signal sample, pre-process the underwater acoustic signal sample, and acquire an unbalanced bandwidth Mel-spectrogram time-frequency diagram; The front-end backbone network is used to convert the unbalanced bandwidth Mel-spectrogram into a fixed-dimensional time-frequency graph to obtain deep features; The deep features are input into the pre-built expert mixture model to obtain the underwater acoustic target recognition results.

2. The underwater acoustic target recognition method based on expert mixture model according to claim 1 is characterized in that: The method of obtaining a water acoustic signal sample and preprocessing the water acoustic signal sample to obtain an unbalanced bandwidth Mel-spectrogram time-frequency diagram is as follows: The underwater acoustic signal samples are divided into multiple frequency bands by frames, each frame is windowed, and each frame group is Fourier transformed. Then the underwater acoustic signal samples are input into different filters for logarithmic operation to obtain sub-band Mel spectrograms. The sub-band Mel spectrograms are concatenated along the frequency direction to obtain Mel features, and the unbalanced bandwidth Mel spectrogram time-frequency diagram is obtained based on the Mel features.

3. The underwater acoustic target recognition method based on expert mixture model according to claim 1 is characterized in that: The front-end backbone network structure is based on ResNet, uses LayerNorm as the normalization method, and GELU as the activation function. Specifically, the unbalanced bandwidth Mel-spectrogram time-frequency graph passes through the convolution layer and the LN layer, and the following four stages use the downsampling module and stacked ConvNeXt Blocks of different times to realize feature extraction, and finally passes through the global average pooling layer to output the deep features; Among them, the downsampling module is composed of LN layer and convolution layer, and ConvNeXt Block is composed of convolution layer, LN layer, GELU layer and mixed domain attention mechanism module.

4. The underwater acoustic target recognition method based on expert mixture model according to claim 3 is characterized in that: The mixed domain attention mechanism module combines the channel attention module and the spatial attention module; The feature map input by the channel attention mechanism module is processed by global maximum pooling based on width and global average pooling based on height, and then operated through a shared fully connected layer. After summation and Sigmoid activation, the channel attention feature map is obtained; The spatial attention mechanism takes the output of the channel attention module as input, performs global maximum pooling and global average pooling operations on the input feature map, then concatenates the results of the two, reduces the channel dimension to 1 through convolution operation, and generates a spatial attention feature map through Sigmoid.

5. The underwater acoustic target recognition method based on expert mixture model according to claim 1 is characterized in that: The expert mixture model consists of a routing layer and an expert layer. The routing layer is composed of a softmax function, and the expert layer is composed of a multi-layer perceptron with two linear layers.

6. The underwater acoustic target recognition method based on expert mixture model according to claim 5 is characterized in that: The deep features are input into the pre-built expert mixture model to obtain the underwater acoustic target recognition result, specifically: Set the batch size to n and the input unbalanced bandwidth Mel spectrogram to {x i }, and its corresponding label is set to {y i }, (i=1,2...n), the number of expert layers is set to m, and the expert level is set to E1(·),…E m (·), input unbalanced bandwidth Mel spectrogram time-frequency graph x i , after passing through the front-end backbone network, the deep feature r is obtained i , the deep feature r i Input into the routing layer S(·) to get the routing score The softmax function is used to normalize the routing score to the routing probability p i , routing probability p i represents the probability of assigning the sample to each expert; the deep feature r i Input it to the expert layer with the highest corresponding probability and get the category of the ship, which is the result of hydroacoustic target recognition.

7. An underwater acoustic target recognition system based on an expert mixture model, characterized in that: include: A data preprocessing module is used to obtain underwater acoustic signal samples, preprocess the underwater acoustic signal samples, and obtain an unbalanced bandwidth Mel-spectrogram time-frequency diagram; A feature extraction module is used to convert the unbalanced bandwidth Mel-spectrogram time-frequency graph into a fixed-dimensional time-frequency graph using a front-end backbone network to obtain deep features; The target recognition module is used to input deep features into the pre-built expert mixture model to obtain underwater acoustic target recognition results.

8. The underwater acoustic target recognition system based on expert mixture model according to claim 7 is characterized in that: The method of obtaining a water acoustic signal sample and preprocessing the water acoustic signal sample to obtain an unbalanced bandwidth Mel-spectrogram time-frequency diagram is as follows: The underwater acoustic signal samples are divided into multiple frequency bands by frames, each frame is windowed, and each frame group is Fourier transformed. Then the underwater acoustic signal samples are input into different filters for logarithmic operation to obtain sub-band Mel spectrograms. The sub-band Mel spectrograms are concatenated along the frequency direction to obtain Mel features, and the unbalanced bandwidth Mel spectrogram time-frequency diagram is obtained based on the Mel features.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the underwater acoustic target recognition method based on the expert mixture model as described in any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the underwater acoustic target recognition method based on the expert mixture model as claimed in any one of claims 1 to 6 are implemented.

Images