Characteristic processing method of underwater acoustic signal, and underwater acoustic signal identification method and device

Abstract

The invention provides an underwater acoustic signal feature processing method and device and an underwater acoustic signal recognition method and device, and is applied to the field of underwater acoustic signal processing. The feature processing method of the underwater acoustic signal comprises the steps that a first spectrogram and a second spectrogram are acquired, the first spectrogram comprises global information of the underwater acoustic signal, and the second spectrogram comprises local information of the underwater acoustic signal; performing feature extraction on the first spectrogram through a first feature extraction module in a feature extraction network to obtain global feature representation of the underwater acoustic signal; performing feature extraction on the second spectrogram through a second feature extraction module in the feature extraction network to obtain local feature representation of the underwater acoustic signal; and performing feature fusion on the global feature representation and the local feature representation to obtain a target feature representation of the underwater acoustic signal, the target feature representation being at least used for identifying an underwater acoustic target. Therefore, the global features and the local features of the underwater acoustic signals are concerned in the feature extraction process, and the feature extraction accuracy of the underwater acoustic signals is improved.

Description

Technical Field

[0001] The present application is applied to the field of underwater acoustic signal processing, and in particular relates to a feature processing method for underwater acoustic signals, an underwater acoustic signal recognition method and a device. Background Art

[0002] Downstream tasks of underwater acoustic signals include target detection, sound synthesis, and anomaly detection based on underwater acoustic signals. The main research objects of these downstream tasks are the noise radiated by underwater targets, including mechanical noise, propeller noise, and hydrodynamic noise.

[0003] Taking underwater acoustic target detection as an example, in related technologies, the underwater acoustic signal is input into a deep learning network, and the features of the underwater acoustic signal are extracted through the deep learning network to obtain the characteristic information of the underwater acoustic signal. Then, based on the characteristic information of the underwater acoustic signal, the radiation source is identified.

[0004] However, the feature extraction accuracy of the above method is low, which affects the accuracy of downstream task execution. Summary of the Invention

[0005] In order to solve the above problems, the present application proposes a feature processing method of underwater acoustic signals, an underwater acoustic signal recognition and a device, which can improve the accuracy of feature extraction of underwater acoustic signals.

[0006] In a first aspect, the present application provides a feature processing method for an underwater acoustic signal, comprising: obtaining a first spectrum graph and a second spectrum graph, wherein the first spectrum graph contains global information of the underwater acoustic signal, and the second spectrum graph contains local information of the underwater acoustic signal; performing feature extraction on the first spectrum graph through a first feature extraction module in a feature extraction network to obtain a global feature representation of the underwater acoustic signal; performing feature extraction on the second spectrum graph through a second feature extraction module in the feature extraction network to obtain a local feature representation of the underwater acoustic signal; and performing feature fusion on the global feature representation and the local feature representation to obtain a target feature representation of the underwater acoustic signal, wherein the target feature representation is at least used to identify an underwater acoustic target.

[0007] In some embodiments, there are multiple second spectrum graphs, and different second spectrum graphs among the multiple second spectrum graphs contain frequency points with different energy distribution ranges; the second feature extraction module in the feature extraction network performs feature extraction on the second spectrum graph to obtain the local feature representation of the underwater acoustic signal, including: the second feature extraction module performs feature extraction on the multiple second spectrum graphs respectively to obtain the local feature representation, and the local feature representation includes the feature representations corresponding to the multiple second spectrum graphs respectively.

[0008] In some embodiments, the number of the second feature extraction modules is multiple. Among the multiple second feature extraction modules, different second feature extraction modules correspond to different feature extraction granularities. The obtaining of the local feature representation by performing feature extraction on the multiple second spectrograms respectively through the second feature extraction modules includes: determining the corresponding relationship between the multiple second spectrograms and the multiple second feature extraction modules according to the index information respectively corresponding to the multiple second spectrograms and the feature extraction granularities respectively corresponding to the multiple second feature extraction modules; inputting the multiple second spectrograms into the multiple second feature extraction modules according to the corresponding relationship, and performing feature extraction on the multiple second spectrograms respectively in the multiple second feature extraction modules to obtain the local feature representation; wherein, the index information respectively corresponding to the multiple second spectrograms is related to the energy values of the frequency points respectively included in the multiple second spectrograms.

[0009] In some embodiments, the multiple second feature extraction networks include multiple convolutional neural networks with different convolutional kernel sizes. In the corresponding relationship, the smaller the energy value of the frequency point included in the second spectrogram, the smaller the convolutional kernel size of the convolutional neural network corresponding to the second spectrogram.

[0010] In some embodiments, the generation process of the second spectrogram includes: dividing the first spectrogram to obtain the second spectrogram.

[0011] In some embodiments, the dividing the first spectrogram to obtain the second spectrogram includes: clustering the multiple frequency points in the first spectrogram according to the energy distribution of the multiple frequency points in the first spectrogram to obtain multiple clusters; and determining the second spectrogram according to the multiple clusters.

[0012] The second aspect of the present application provides an underwater acoustic signal recognition method, including: obtaining a target feature representation of an underwater acoustic signal, where the target feature representation is obtained by executing the feature processing method of the underwater acoustic signal according to the first aspect or any implementation manner of the first aspect; and recognizing a target radiation source corresponding to the underwater acoustic signal according to the target feature representation.

[0013] A third aspect of the present application provides a feature processing device for underwater acoustic signals, including: a spectrogram acquisition unit for acquiring a first spectrogram and a second spectrogram, where the first spectrogram contains the global information of the underwater acoustic signal, and the second spectrogram contains the local information of the underwater acoustic signal; a global feature extraction unit for extracting features from the first spectrogram through a first feature extraction module in a feature extraction network to obtain a global feature representation of the underwater acoustic signal; a local feature extraction unit for extracting features from the second spectrogram through a second feature extraction module in the feature extraction network to obtain a local feature representation of the underwater acoustic signal; and a feature fusion unit for performing feature fusion on the global feature representation and the local feature representation to obtain a target feature representation of the underwater acoustic signal, where the target feature representation is at least used to identify an underwater acoustic target.

[0014] A fourth aspect of the present application provides an underwater acoustic signal recognition device, including: an acquisition unit for acquiring a target feature representation of an underwater acoustic signal, where the target feature representation is obtained by executing the feature processing method for underwater acoustic signals according to the first aspect or any implementation manner of the first aspect; and an identification unit for identifying a target radiation source corresponding to the underwater acoustic signal according to the target feature representation.

[0015] A fifth aspect of the present application provides an electronic device, including a memory and a processor; the memory is connected to the processor and is used for storing a program; the processor is used for implementing the feature processing method for underwater acoustic signals according to the first aspect or any implementation manner of the first aspect, or implementing the underwater acoustic signal recognition method according to the second aspect or any implementation manner of the second aspect by running the program in the memory.

[0016] A sixth aspect of the present application provides a chip, including a processor and a data interface, where the processor reads and runs a program stored on a memory through the data interface to execute the feature processing method for underwater acoustic signals according to the first aspect or any implementation manner of the first aspect, or execute the underwater acoustic signal recognition method according to the second aspect or any implementation manner of the second aspect.

[0017] A seventh aspect of the present application provides a computer program product, including a computer program, where when the computer program is executed by a processor, it implements the feature processing method for underwater acoustic signals according to the first aspect or any implementation manner of the first aspect, or implements the underwater acoustic signal recognition method according to the second aspect or any implementation manner of the second aspect.

[0018] The eighth aspect of the present application provides a storage medium, on which a computer program is stored. When the computer program is run by a processor, it implements the method for processing the characteristics of underwater acoustic signals as described in the first aspect or any embodiment of the first aspect, or implements the method for identifying underwater acoustic signals as described in the second aspect or any implementation manner of the second aspect.

[0019] According to the method for processing the characteristics of underwater acoustic signals, the method for identifying underwater acoustic signals and the device provided by the present application, from the perspective of data, a spectrogram containing the global information of the underwater acoustic signal and a spectrogram containing the local information of the underwater acoustic signal are used to more accurately extract the global characteristics and local characteristics of the underwater acoustic signal; from the perspective of the feature extraction network, separate feature extraction modules are provided for the spectrogram containing the global information and the spectrogram containing the local information, which can more accurately focus on the global characteristics and local characteristics of the underwater acoustic signal; thus, the accuracy of feature extraction of the underwater acoustic signal is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.

[0021] Figure 1 It is a schematic diagram of the implementation environment related to the embodiments of the present application;

[0022] Figure 2 It is a flowchart of the method for processing the characteristics of underwater acoustic signals provided by the embodiments of the present application Figure 1 ;

[0023] Figure 3 It is a flowchart of the method for processing the characteristics of underwater acoustic signals provided by the embodiments of the present application Figure 2 ;

[0024] Figure 4 It is a flowchart of the method for identifying underwater acoustic signals provided by the embodiments of the present application;

[0025] Figure 5 It is a structural example diagram of the target recognition network provided by the embodiments of the present application;

[0026] Figure 6 It is a structural schematic diagram of the device for processing the characteristics of underwater acoustic signals provided by the embodiments of the present application;

[0027] Figure 7 It is a structural schematic diagram of the device for identifying underwater acoustic signals provided by the embodiments of the present application;

[0028] Figure 8 It is a schematic structural diagram of an electronic device provided according to an embodiment of the present application. Detailed implementation manners

[0029] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the protection scope of the present application.

[0030] First, relevant terms in the field of underwater acoustic signal processing are explained.

[0031] Radiation source: The source that generates (which can be generated passively) and radiates an underwater acoustic signal outward in the water area, which can be an object, a device, or a living being. For example, a ship, a submarine, a dolphin.

[0032] Radiation noise: The noise signal generated by the radiation source and radiated and propagated in the water body. For example, mechanical noise, propeller noise, hydrodynamic noise.

[0033] Spectrum diagram of underwater acoustic signal: A graph showing the characteristics of the underwater acoustic signal in the frequency domain. There are several types of spectrum diagrams of underwater acoustic signals: low frequency analysis recording (Lofar) spectrum diagram (also known as Lofar feature), detection of envelope modulation on noise (DEMON) spectrum diagram, filter bank (FBank) feature map, wavelet transform feature map, Mel spectrogram, power spectral density (PSD) diagram, etc.

[0034] Secondly, the problems existing in the field of underwater acoustic signal processing are described.

[0035] On the one hand, the underwater acoustic signal contains target radiation noise (i.e., the radiation noise generated by the target radiation source to be identified) and a large amount of other noise (such as the sound signal generated by the sonar device itself for collecting the underwater acoustic signal, the radiation noise generated by the platform carrying the sonar device), resulting in a low signal-to-noise ratio of the underwater acoustic signal and bringing difficulties to the feature extraction of the underwater acoustic signal.

[0036] On the other hand, in the underwater acoustic signals collected in different acquisition scenarios, or in the underwater acoustic signals collected using different acquisition devices, the frequency-domain position distributions of the effective information are different. Taking the LOFAR spectrogram as an example, the LOFAR spectrogram is a spectrogram formed by processing the underwater acoustic signal in both the time domain and the frequency domain through short-time Fourier transform (STFT). When underwater acoustic signals are collected in two different scenarios to obtain two batches of acquisition data, the position distributions of the effective information in the LOFAR spectrograms corresponding to the underwater acoustic signals in different batches of acquisition data are different. This characteristic also brings difficulties to the feature extraction of underwater acoustic signals.

[0037] In related technologies, traditional underwater acoustic target recognition methods or deep learning-based underwater acoustic target recognition methods are used for underwater acoustic target recognition.

[0038] In traditional underwater acoustic target recognition methods, the time-frequency characteristics of underwater acoustic signals are used, such as time-frequency diagrams, or the auditory characteristics of underwater acoustic signals are used, such as Mel Frequency Cepstrum Coefficents (MFCC). The above characteristics of underwater acoustic signals are not considered, resulting in low accuracy of feature extraction.

[0039] In deep learning-based underwater acoustic target recognition methods, the deep learning network adopts a network architecture based on the attention mechanism. As one of the additional components in this network architecture, the attention module can distinguish different parts of the input data and automatically assign different weights, enabling the deep learning network to identify and process key information and improving the attention to important features related to target recognition. Taking the channel attention mechanism as an example, it automatically evaluates the importance of each feature channel through learning, models the interdependence between feature channels according to the importance of each feature channel, enabling the deep learning network to focus on key information and ignore irrelevant information.

[0040] However, in the deep learning network based on the attention mechanism, the overall information and local information are calculated through the reverse gradient for feature extraction. For underwater acoustic signals with low signal-to-noise ratio and uncertain frequency-domain distribution positions of effective information, the deep learning network based on the attention mechanism cannot accurately capture the relationship between the local features and global features of underwater acoustic signals and cannot accurately extract effective features, resulting in low accuracy of feature extraction of underwater acoustic signals.

[0041] Embodiments of the present application propose a method for processing characteristics of underwater acoustic signals, a method for identifying underwater acoustic signals, and a device. In view of the above characteristics of underwater acoustic signals, especially the uncertain frequency-domain position distribution of effective information in underwater acoustic signals, different feature extraction modules are used to extract the features of the spectrogram containing the global information of the underwater acoustic signal and the features of the spectrogram containing the local information of the underwater acoustic signal respectively, paying more attention to the global features and local features of the underwater acoustic signal more accurately, learning the relationship between the global features and local features in the underwater acoustic signal, extracting effective features from the underwater acoustic signal, and improving the accuracy of feature extraction. The feature information extracted from the underwater acoustic signal can be used at least for target recognition of the underwater acoustic signal, thereby improving the accuracy of target recognition of the underwater acoustic signal.

[0042] Among them, the implementation principle and technical effect of the device can be referred to the description of the method and will not be elaborated here.

[0043] Exemplary implementation environment

[0044] Please refer to Figure 1 , Figure 1 FIG. is a schematic diagram of the implementation environment involved in the embodiments of the present application. In the implementation environment involved in the present application, it includes an acquisition device 110 for underwater acoustic signals and a processing device 120 for underwater acoustic signals. The processing device 120 can obtain the underwater acoustic signal from the acquisition device 110 and generate a spectrogram containing the global information of the underwater acoustic signal and a spectrogram containing the local information of the underwater acoustic signal.

[0045] A feature extraction network 130 is deployed on the processing device 120. The feature extraction network 130 includes a feature extraction module 131 for extracting global features and a feature extraction module 132 for extracting local features; through the feature extraction module 131 for extracting global features, feature extraction is performed on the spectrogram containing the global information of the underwater acoustic signal to obtain a global feature representation of the underwater acoustic signal; through the feature extraction module 132 for extracting local features, feature extraction is performed on the spectrogram containing the local information of the underwater acoustic signal to obtain a local feature representation of the underwater acoustic signal; the global feature representation and the local feature representation are fused to obtain a target feature representation of the underwater acoustic signal.

[0046] Among them, the processing device 120 can be a terminal or a server. Figure 1 Taking the processing device 120 as a server as an example. Further, the implementation environment can be a target recognition scenario of underwater acoustic signals. At this time, the processing device 120 can complete the target recognition of the underwater acoustic signal based on the target feature representation of the underwater acoustic signal.

[0047] Exemplary method

[0048] Please refer to Figure 2, in an exemplary embodiment, a method for processing the characteristics of an underwater acoustic signal is provided. The method for processing the characteristics of the underwater acoustic signal includes the following steps:

[0049] S201, obtain a first spectrogram and a second spectrogram. The first spectrogram contains the global information of the underwater acoustic signal, and the second spectrogram contains the local information of the underwater acoustic signal.

[0050] Wherein, for the sake of simplicity of description and easy distinction, the spectrogram containing the global information of the underwater acoustic signal is called the first spectrogram, and the spectrogram containing the local information of the underwater acoustic signal is called the second spectrogram. The underwater acoustic signal may refer to an underwater acoustic time-domain signal.

[0051] Wherein, the first spectrogram and the second spectrogram may be at least one of the following: Lofar spectrogram, DEMON spectrogram, FBank feature map, wavelet transform feature map, Mel spectrogram, PSD map. In particular, the first spectrogram and the second spectrogram are LOFAR spectrograms, LOFAR spectrograms.

[0052] In one example, the underwater acoustic signal can be processed to obtain the first spectrogram and the second spectrogram. Further, when the first spectrogram is a Lofar spectrogram, the first spectrogram can be generated by performing a short-time Fourier transform on the underwater acoustic signal.

[0053] In another example, the first spectrogram and the second spectrogram can be obtained from a database, or the first spectrogram and the second spectrogram received from other devices are pre-processed from the underwater acoustic signal.

[0054] S202, through the first feature extraction module in the feature extraction network, perform feature extraction on the first spectrogram to obtain the global feature representation of the underwater acoustic signal.

[0055] Wherein, the feature extraction network is a pre-trained neural network model.

[0056] Wherein, the feature extraction network includes a first feature extraction module and a second feature extraction module. The first feature extraction module is used to extract the global features of the underwater acoustic signal, and the second feature extraction module is used to extract the local features of the underwater acoustic signal. The first feature extraction module and the second feature extraction module can respectively adopt network structures suitable for global feature extraction and local feature extraction to improve the accuracy of feature extraction.

[0057] In this embodiment, the first spectrogram is input into the first feature extraction module. In the first feature extraction module, through the feature extraction layer, the overall information of the first spectrogram is learned, and the feature representation of the first spectrogram, that is, the global feature representation of the underwater acoustic signal, is extracted.

[0058] S203. Use the second feature extraction module in the feature extraction network to extract features from the second spectrogram, obtaining a local feature representation of the underwater acoustic signal.

[0059] In this embodiment, input the second spectrogram into the second feature extraction module. In the second feature extraction module, through the feature extraction layer, learn the local information of the second spectrogram and extract the feature representation of the second spectrogram, which is the local feature representation of the underwater acoustic signal.

[0060] S204. Perform feature fusion on the global feature representation and the local feature representation to obtain a target feature representation of the underwater acoustic signal, where the target feature representation is at least used for identifying underwater acoustic targets.

[0061] In one example, the global feature representation and the local feature representation can be concatenated to obtain a concatenated feature representation, and the concatenated feature representation is the target feature representation. Thus, through the method of feature concatenation, the feature fusion of the global feature representation and the local feature representation is realized. Among them, the data formats of the global feature representation and the local feature representation can be vectors, and the concatenation between the global feature representation and the local feature representation can be vector concatenation.

[0062] In another example, the global feature representation and the local feature representation can be weighted to obtain the target feature representation. Thus, through the method of weighted operation, the feature fusion of the global feature representation and the local feature representation is realized. In this method, the weight parameters can be flexibly adjusted to improve the fusion effect of the global feature representation and the local feature representation, so as to obtain a more accurate target feature representation.

[0063] In this embodiment, the target feature representation is at least used for identifying underwater acoustic targets, that is, in the target recognition task of underwater acoustic signals, extract the target feature representation of the underwater acoustic signal, and then perform the target recognition of the underwater acoustic signal based on the target feature representation. Specifically, reference can be made to the embodiments of the subsequent underwater acoustic signal recognition method. It should be noted that the target feature representation can also be used for other downstream tasks of underwater acoustic signals, such as the synthesis of underwater acoustic signals, the anomaly detection of underwater acoustic signals, etc.

[0064] In the embodiments of this application, different feature extraction modules are used to extract the features of the spectrogram containing the global information of the underwater acoustic signal and the spectrogram containing the local information of the underwater acoustic signal respectively, which can more accurately focus on the global features and local features of the underwater acoustic signal, learn the relationship between the global features and local features in the underwater acoustic signal, extract effective feature representations from the underwater acoustic signal with uncertain frequency domain position distribution of effective information, and improve the accuracy of feature extraction of the underwater acoustic signal.

[0065] Next, for some steps in the above embodiments, more implementation methods are provided.

[0066] In a possible implementation, S201 includes: obtaining a first spectrogram; partitioning the first spectrogram to obtain a second spectrogram. The first spectrogram contains the global information of the underwater acoustic signal, and the second spectrogram contains the local information in the first spectrogram, that is, contains the local information of the underwater acoustic signal.

[0067] In this implementation, the underwater acoustic signal can be subjected to time-frequency domain transformation to obtain the first spectrogram, or the first spectrogram can be obtained from a database or other devices, where the first spectrogram is obtained by previously performing time-frequency domain transformation on the underwater acoustic signal. Then, the information contained in the first spectrogram is partitioned to obtain a second spectrogram, and the number of second spectrograms can be multiple.

[0068] It can be seen that the second spectrogram is a part of the first spectrogram, and there is an association relationship between the second spectrogram and the first spectrogram. By respectively performing feature extraction on the first spectrogram and the second spectrogram through the first feature extraction network and the second feature extraction network, the global feature representation and the local feature representation with an association relationship can be extracted, and the dependence relationship between the global feature and the local feature of the underwater acoustic signal can be learned, improving the accuracy of feature extraction of the underwater acoustic signal.

[0069] Optionally, partitioning the first spectrogram to obtain a second spectrogram includes: clustering a plurality of frequency points in the first spectrogram according to the energy distribution of the plurality of frequency points in the first spectrogram to obtain a plurality of clusters; determining the second spectrogram according to the plurality of clusters.

[0070] In this optional method, the first spectrogram contains a plurality of frequency points. Based on the first spectrogram, the plurality of frequency points can be respectively integrated in the time domain to obtain the energy values corresponding to the plurality of frequency points, and the energy values corresponding to the plurality of frequency points reflect the energy distribution of the plurality of frequency points. Then, according to the energy values corresponding to the plurality of frequency points in the first spectrogram, the plurality of frequency points can be clustered by a clustering algorithm, so that the frequency points with similar energy values are clustered into the same cluster, and the frequency points with far-different energy values are clustered into different clusters, finally obtaining a plurality of clusters. For each cluster, a corresponding spectrogram is formed by the frequency points in the cluster, and finally the spectrograms corresponding to the plurality of clusters are obtained. The spectrograms corresponding to the plurality of clusters are the second spectrograms. Therefore, a plurality of second spectrograms can be finally obtained, and different second spectrograms contain frequency points with different energy distribution ranges, enabling the second feature extraction module to extract local features with different energy distribution situations of the underwater acoustic signal in the frequency domain.

[0071] Among them, no specific limitation is imposed on the above-mentioned clustering algorithm.

[0072] In a possible implementation, the number of second spectrograms is multiple. Among the multiple second spectrograms, different second spectrograms include frequency points with different energy distribution ranges. At this time, S203 may include: through the second feature extraction module, performing feature extraction on each of the multiple second spectrograms respectively to obtain a local feature representation of the underwater acoustic signal, where the local feature representation includes the feature representations corresponding to each of the multiple second spectrograms respectively. Thus, the feature extraction of the multiple second spectrograms is performed independently, and the feature representations corresponding to each of the multiple second spectrograms are obtained. The feature representations corresponding to each of the multiple second spectrograms correspond to the local features with different energy distribution conditions of the underwater acoustic signal in the frequency domain, so that the local features with different energy distribution conditions of the underwater acoustic signal in the frequency domain are concerned during the feature extraction process.

[0073] Optionally, the number of second feature extraction modules is multiple. The multiple second spectrograms can be respectively subjected to feature extraction through the multiple second feature extraction modules, so that the feature extraction of the multiple second spectrograms can be performed independently and in parallel, improving the accuracy and efficiency of local feature extraction.

[0074] Optionally, the number of second feature extraction modules is the same as the number of second spectrograms.

[0075] Optionally, among the multiple second feature extraction modules, different second feature extraction modules correspond to different feature extraction granularities to extract local features of different granularities of the underwater acoustic signal.

[0076] Please refer to Figure 3 , in another exemplary embodiment, a method for processing features of an underwater acoustic signal is provided. The method for processing features of an underwater acoustic signal includes the following steps:

[0077] S301, obtaining a first spectrogram and multiple second spectrograms.

[0078] Among them, among the multiple second spectrograms, different second spectrograms include frequency points with different energy distribution ranges. The generation process of the second spectrogram can refer to the foregoing embodiments and will not be elaborated here.

[0079] S302, through the first feature extraction module in the feature extraction network, performing feature extraction on the first spectrogram to obtain a global feature representation of the underwater acoustic signal.

[0080] Among them, the implementation principles and technical effects of S301 to S302 can refer to the foregoing embodiments and will not be elaborated here.

[0081] Optionally, the first feature extraction module is a feature extraction module based on an attention mechanism. During the process of performing feature extraction on the first spectrogram through the first feature extraction module, feature extraction is performed based on the attention mechanism to improve the accuracy of global feature extraction.

[0082] Further, the first feature extraction module is an attention mechanism Transformer.

[0083] S303. Determine the correspondence between the multiple second spectrograms and the multiple second feature extraction modules according to the index information respectively corresponding to the multiple second spectrograms and the feature extraction granularities respectively corresponding to the multiple second feature extraction modules in the feature extraction network.

[0084] Among them, the feature extraction network includes multiple second feature extraction modules, and different second feature extraction modules have different feature extraction granularities. The correspondence between the multiple second spectrograms and the multiple second feature extraction modules can be a one-to-one correspondence. Different second spectrograms are input into different second feature extraction modules for feature extraction to extract local features of different granularities of the underwater acoustic signal.

[0085] Among them, the index information respectively corresponding to the multiple second spectrograms is related to the energy values of the frequency points respectively included in the multiple second spectrograms. For example, the index information corresponding to the second spectrogram includes the maximum energy value and the minimum energy value of the frequency points included in the second spectrogram; another example is that the index information corresponding to the second spectrogram includes the average energy value of the frequency points included in the second spectrogram. Thus, in the case of referring to the energy values of the spectra included in the second spectrogram, a second feature extraction network with a corresponding feature extraction granularity is assigned to the second spectrogram, so that the assignment of the second feature extraction network conforms to the energy distribution of the target radiation signal in the underwater acoustic signal. Among the multiple second feature extraction networks, through the second feature extraction network with a smaller feature extraction granularity, more detailed local features are extracted from the second spectrogram that is more likely to contain relevant information of the target radiation signal, improving the extraction accuracy of the effective features of the underwater acoustic signal.

[0086] In one example, the multiple second spectrograms can be sorted according to the index information respectively corresponding to the multiple second spectrograms; according to the sorting order of the multiple second spectrograms and the size order of the feature extraction granularities respectively corresponding to the multiple second feature extraction modules, determine the correspondence between the multiple second spectrograms and the multiple second feature extraction modules.

[0087] In this example, since the index information corresponding to multiple second spectrograms reflects the frequency point energy values in the multiple second spectrograms, referring to the index information corresponding to the multiple second spectrograms, the multiple second spectrograms are sorted. The sorting order of the multiple second spectrograms reflects the order of the energy values of the frequency points included in the multiple second spectrograms (from high to low or from low to high). In the process of determining the correspondence between the multiple second spectrograms and the multiple second feature extraction modules according to the sorting order of the multiple second spectrograms and the order of the feature extraction granularities corresponding to the multiple second feature extraction modules, a one-to-one correspondence can be established between the sorted multiple second spectrograms and the multiple second feature extraction modules sorted according to the size of the feature extraction granularity. Thus, by sorting the multiple second spectrograms with reference to the index information reflecting the frequency point energy values, and determining the correspondence according to the sorting order of the multiple second spectrograms and the order of the feature extraction granularities of the multiple second feature extraction modules, the accuracy of allocating corresponding second feature extraction modules to the multiple second spectrograms is improved.

[0088] Optionally, the multiple second feature extraction networks include multiple convolutional neural networks with different convolutional kernel sizes. Thus, local feature extraction with different granularities is achieved through convolutional kernels of different sizes.

[0089] Optionally, in the case where the multiple second feature extraction networks include multiple convolutional neural networks with different convolutional kernel sizes, in the correspondence between the multiple second spectrograms and the multiple second feature extraction modules, the smaller the energy value of the frequency points included in the second spectrogram, the smaller the convolutional kernel size of the convolutional neural network corresponding to the second spectrogram.

[0090] In this optional method, in underwater acoustic signals, the energy of the target radiated noise is weak after attenuation. Therefore, the smaller the energy value of the frequency points included in the second spectrogram, the higher the probability that the frequency domain information of the target radiated noise is included in the second spectrogram. The smaller the energy value of the frequency points included in the second spectrogram, the smaller the convolutional kernel size of the convolutional neural network corresponding to the second spectrogram. Thus, through a convolutional neural network with a smaller convolutional kernel size (i.e., through a second feature extraction network with a smaller feature extraction granularity), feature extraction is performed on the second spectrogram that is more likely to contain valid information, so as to extract more valid features and improve the accuracy of local feature extraction of underwater acoustic signals.

[0091] Optionally, the multiple second feature extraction networks are multiple residual networks with different convolutional kernel sizes. Through multiple residual networks with different convolutional kernel sizes, local features with different granularities are extracted, improving the richness of local features and extracting the features of valid information in underwater acoustic signals from the second spectrogram.

[0092] S304. According to the correspondence between multiple second spectrograms and multiple second feature extraction modules, input the multiple second spectrograms into the multiple second feature extraction modules. In the multiple second feature extraction modules, perform feature extraction on the multiple second spectrograms respectively to obtain local feature representations.

[0093] In this embodiment, according to the correspondence between multiple second spectrograms and multiple second feature extraction modules, input the multiple second feature spectrograms into the corresponding second feature extraction modules respectively. In the second feature extraction module, perform feature extraction on the input second spectrogram to obtain the feature representation corresponding to the input second spectrogram. Finally, obtain the feature representations corresponding to the multiple second spectrograms respectively, that is, obtain the local feature representation of the underwater acoustic signal.

[0094] S305. Perform feature fusion on the global feature representation and the local feature representation to obtain the target feature representation of the underwater acoustic signal.

[0095] Among them, the implementation principle and technical effect of S305 can refer to the foregoing embodiments and will not be elaborated here.

[0096] In the embodiment of the present application, according to the index information corresponding to the multiple second spectrograms respectively and the feature extraction granularity corresponding to the multiple second feature extraction modules in the feature extraction network, arrange the second feature extraction modules with appropriate feature extraction granularity for the multiple second spectrograms. While realizing the multi-granularity local feature extraction of the underwater acoustic signal, the accuracy of the local feature extraction of the underwater acoustic signal is improved. Furthermore, the accuracy of the feature extraction of the underwater acoustic signal is improved.

[0097] Please refer to Figure 4 , in an exemplary embodiment, an underwater acoustic signal recognition method is provided. The underwater acoustic signal recognition method may include:

[0098] S401. Obtain the target feature representation of the underwater acoustic signal.

[0099] Among them, the target feature representation of the underwater acoustic signal is obtained by performing the feature processing method of the underwater acoustic signal provided in any of the foregoing embodiments, and will not be elaborated here.

[0100] In this embodiment, the target recognition network of the underwater acoustic signal includes the feature extraction network in any of the foregoing embodiments. Through this feature extraction network, according to the feature processing method of the underwater acoustic signal provided in any of the foregoing embodiments, the target feature representation of the underwater acoustic signal can be extracted.

[0101] S402. Identify the target radiation source corresponding to the underwater acoustic signal according to the target feature representation.

[0102] In this embodiment, in the target recognition network of the underwater acoustic signal, the target feature representation can be input into other network modules (such as the classification head) after the feature extraction network. In other network modules, based on this target feature representation, the target radiation source corresponding to the underwater acoustic signal is recognized.

[0103] In the embodiments of the present application, through the feature processing method of the underwater acoustic signal provided in any of the above embodiments, the accuracy of the target feature representation of the underwater acoustic signal is improved. Based on this target feature representation, the target recognition of the underwater acoustic signal is performed, and the accuracy of the target recognition of the underwater acoustic signal is improved.

[0104] Taking the first feature extraction module as the attention mechanism transformer and the second feature extraction module as the residual network as an example, Figure 5 It is a structural example diagram of the target recognition network provided by the embodiments of the present application.

[0105] As Figure 5 shown, the target recognition network includes an attention mechanism transformer, a plurality of residual networks (residual network_1, residual network_2,..., residual network_n, where n is the total number of residual networks), a fully connected layer (FC layer), and a softmax layer. Among them, the fully connected layer and the softmax layer constitute the classification head in the target recognition network.

[0106] As Figure 5 shown, first, the underwater acoustic signal is converted into a spectrogram (i.e., the first spectrogram, such as the Lofar spectrogram); then, the frequency points in the spectrogram are clustered according to the frequency point energy distribution to form n clusters, and according to these n clusters, n sub-spectrograms (sub-spectrogram_1, sub-spectrogram_2,..., sub-spectrogram_n) can be obtained; then, the n sub-spectrograms are respectively subjected to feature extraction through n residual networks (feature extraction is performed on sub-spectrogram_1 through residual network_1, feature extraction is performed on sub-spectrogram_2 through residual network_2,..., feature extraction is performed on sub-spectrogram_n through residual network_n), and n local feature representations (local feature representation_1, local feature representation_2,..., local feature representation_n) are obtained. In addition, the spectrogram is subjected to feature extraction through the attention mechanism transformer to obtain a global feature representation. The global feature representation is concatenated with the n local feature representations to obtain a target feature representation. Then, the target feature representation can be input into the fully connected layer, and through the fully connected layer and the softmax layer, the target recognition result of the underwater acoustic signal is obtained.

[0107] Exemplary device

[0108] Correspondingly, the embodiments of the present application also provide a feature processing device for underwater acoustic signals.

[0109] Please refer to Figure 6, in an exemplary embodiment, a feature processing device 600 for underwater acoustic signals is provided. The feature processing device 600 for underwater acoustic signals includes: a spectrogram acquisition unit 601, a global feature extraction unit 602, a local feature extraction unit 603, and a feature fusion unit 604. Among them:

[0110] The spectrogram acquisition unit 601 is configured to acquire a first spectrogram and a second spectrogram. The first spectrogram contains the global information of the underwater acoustic signal, and the second spectrogram contains the local information of the underwater acoustic signal. The global feature extraction unit 602 is configured to perform feature extraction on the first spectrogram through a first feature extraction module in the feature extraction network to obtain a global feature representation of the underwater acoustic signal. The local feature extraction unit 603 is configured to perform feature extraction on the second spectrogram through a second feature extraction module in the feature extraction network to obtain a local feature representation of the underwater acoustic signal. The feature fusion unit 604 is configured to perform feature fusion on the global feature representation and the local feature representation to obtain a target feature representation of the underwater acoustic signal, and the target feature representation is at least used to identify underwater acoustic targets.

[0111] In some embodiments, the number of the second spectrograms is multiple, and different second spectrograms among the multiple second spectrograms contain frequency points with different energy distribution ranges. The local feature extraction unit 603 is specifically configured to: perform feature extraction on the multiple second spectrograms respectively through the second feature extraction module to obtain local feature representations, and the local feature representations include the feature representations corresponding to the multiple second spectrograms respectively.

[0112] In some embodiments, the number of the second feature extraction modules is multiple, and among the multiple second feature extraction modules, different second feature extraction modules correspond to different feature extraction granularities. The local feature extraction unit 603 is specifically configured to: determine the correspondence between the multiple second spectrograms and the multiple second feature extraction modules according to the index information corresponding to the multiple second spectrograms respectively and the feature extraction granularities corresponding to the multiple second feature extraction modules respectively; input the multiple second spectrograms into the multiple second feature extraction modules according to the correspondence, and in the multiple second feature extraction modules, perform feature extraction on the multiple second spectrograms respectively to obtain local feature representations; wherein, the index information corresponding to the multiple second spectrograms respectively is related to the energy values of the frequency points included in the multiple second spectrograms respectively.

[0113] In some embodiments, the multiple second feature extraction networks include multiple convolutional neural networks with different convolutional kernel sizes. In the correspondence, the smaller the energy value of the frequency point included in the second spectrogram, the smaller the convolutional kernel size of the convolutional neural network corresponding to the second spectrogram.

[0114] In some embodiments, the generation process of the second spectrogram includes: dividing the first spectrogram to obtain the second spectrogram.

[0115] In some embodiments, the first spectrogram is divided to obtain a second spectrogram, including: clustering a plurality of frequency points in the first spectrogram according to the energy distribution of the plurality of frequency points in the first spectrogram to obtain a plurality of clusters; and determining the second spectrogram according to the plurality of clusters.

[0116] The underwater acoustic signal feature processing device 600 provided in this embodiment belongs to the same inventive concept as the underwater acoustic signal feature processing method provided in the foregoing embodiments of the present application, and can execute the underwater acoustic signal feature processing method provided in any of the foregoing embodiments of the present application, and has corresponding functional modules and beneficial effects for executing the underwater acoustic signal feature processing method. For technical details not described in detail in this embodiment, reference may be made to the specific processing content of the underwater acoustic signal feature processing method provided in the foregoing embodiments of the present application, which will not be elaborated herein.

[0117] Correspondingly, an embodiment of the present application further provides an underwater acoustic signal recognition device.

[0118] Please refer to Figure 7 , in an exemplary embodiment, an underwater acoustic signal recognition device 700 is provided. The underwater acoustic signal recognition device 700 includes: an acquisition unit 701 and a recognition unit 702. Among them:

[0119] The acquisition unit 701 is configured to acquire a target feature representation of the underwater acoustic signal, and the target feature representation is obtained by executing the underwater acoustic signal feature processing method provided in the foregoing embodiments of the present application; the recognition unit 702 is configured to recognize a target radiation source corresponding to the underwater acoustic signal according to the target feature representation.

[0120] The underwater acoustic signal recognition device provided in this embodiment belongs to the same inventive concept as the underwater acoustic signal recognition method provided in the foregoing embodiments of the present application, and can execute the underwater acoustic signal recognition method provided in any of the foregoing embodiments of the present application, and has corresponding functional modules and beneficial effects for executing the underwater acoustic signal recognition method. For technical details not described in detail in this embodiment, reference may be made to the specific processing content of the underwater acoustic signal recognition method provided in the foregoing embodiments of the present application, which will not be elaborated herein.

[0121] The functions implemented by each unit in the above device can be realized by the same or different processors respectively, which is not limited in the embodiments of the present application.

[0122] It should be understood that the units in the above device can be implemented in the form of a processor calling software. For example, the device includes a processor, the processor is connected to a memory, instructions are stored in the memory, and the processor calls the instructions stored in the memory to implement any of the above methods or the functions of each unit of the device. The processor can be a general-purpose processor, such as a CPU or a microprocessor, etc., and the memory can be a memory inside the device or a memory outside the device. Alternatively, the units in the device can be implemented in the form of hardware circuits. By designing the hardware circuits, the functions of some or all of the units can be implemented. The hardware circuits can be understood as one or more processors. For example, in one implementation, the hardware circuit is an ASIC, and by designing the logical relationship of the components in the circuit, the functions of some or all of the above units are implemented. Again, for example, in another implementation, the hardware circuit can be implemented by a PLD. Taking an FPGA as an example, it can include a large number of logic gate circuits, and the connection relationship between the logic gate circuits is configured through a configuration file, so as to implement the functions of some or all of the above units. All units of the above device can be all implemented in the form of a processor calling software, or all implemented in the form of hardware circuits, or some implemented in the form of a processor calling software, and the remaining part implemented in the form of hardware circuits.

[0123] In the embodiments of the present application, a processor is a circuit with the ability to process signals. In one implementation, the processor can be a circuit with the ability to read and run instructions, such as a CPU, a microprocessor, a GPU, or a DSP, etc. In another implementation, the processor can implement certain functions through the logical relationship of hardware circuits, and the logical relationship of the hardware circuits is fixed or can be reconfigured. For example, the processor is a hardware circuit implemented by an ASIC or a PLD, such as an FPGA, etc. In a reconfigurable hardware circuit, the process of the processor loading a configuration document to implement the configuration of the hardware circuit can be understood as the process of the processor loading instructions to implement the functions of some or all of the above units. In addition, it can also be a hardware circuit designed for artificial intelligence, which can be understood as a kind of ASIC, such as an NPU, a TPU, a DPU, etc.

[0124] It can be seen that each unit in the above device can be one or more processors (or processing circuits) configured to implement the above methods. For example: CPU, GPU, NPU, TPU, DPU, microprocessor, DSP, ASIC, FPGA, or a combination of at least two of these processor forms.

[0125] In addition, each unit in the above device can be integrated in whole or in part, or can be implemented independently. In one implementation, these units are integrated together and implemented in the form of an SOC. The SOC may include at least one processor for implementing any of the above methods or the functions of each unit of the device. The types of the at least one processor may be different, for example, including a CPU and an FPGA, a CPU and an artificial intelligence processor, a CPU and a GPU, etc.

[0126] Exemplary electronic device

[0127] Another embodiment of the present application further provides an electronic device. Refer to Figure 8 As shown, the electronic device may include: a memory 800 and a processor 810; wherein, the memory 800 is connected to the processor 810 for storing programs; the processor 810 is configured to implement the feature processing method of the underwater acoustic signal disclosed in any of the above embodiments, or implement the underwater acoustic signal recognition method disclosed in any of the above embodiments by running the programs stored in the memory 800.

[0128] Specifically, the above electronic device may further include: a bus, a communication interface 820, an input device 830, and an output device 840.

[0129] The processor 810, the memory 800, the communication interface 820, the input device 830, and the output device 840 are interconnected via the bus. Among them:

[0130] The bus may include a path for transmitting information between various components of the computer system.

[0131] The processor 810 may be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or may be an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the solution of the present application. It may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0132] The processor 810 may include a main processor, and may also include a baseband chip, a modem, etc.

[0133] The program for implementing the technical solution of this application is stored in the memory 800. The operating system and other key services may also be stored therein. Specifically, the program may include program code, and the program code includes computer operation instructions. More specifically, the memory 800 may include a read-only memory (ROM), other types of static storage devices that can store static information and instructions, a random access memory (RAM), other types of dynamic storage devices that can store information and instructions, a disk memory, a flash memory, etc.

[0134] The input device 830 may include devices for receiving data and information input by a user, such as a keyboard, a mouse, a camera, a scanner, a light pen, a voice input device, a touch screen, a pedometer, or a gravity sensor, etc.

[0135] The output device 840 may include devices for allowing information to be output to a user, such as a display screen, a printer, a speaker, etc.

[0136] The communication interface 820 may include devices of any transceiver type for communicating with other devices or communication networks, such as an Ethernet, a radio access network (RAN), a wireless local area network (WLAN), etc.

[0137] The processor 810 executes the program stored in the memory 800 and calls other devices, and can be used to implement each step of any one of the underwater acoustic signal feature processing methods provided in the above embodiments of this application, or implement each step of any one of the underwater acoustic signal recognition methods provided in the above embodiments of this application.

[0138] An embodiment of this application also proposes a chip, which includes a processor and a data interface. The processor reads and runs the program stored on the memory through the data interface to execute any one of the underwater acoustic signal feature processing methods provided in the above embodiments or any one of the underwater acoustic signal recognition methods provided in the above embodiments. For the specific processing process and its beneficial effects, reference can be made to the embodiment introduction of the above underwater acoustic signal feature processing method or the embodiment introduction of the above underwater acoustic signal recognition method.

[0139] Exemplary computer program product and storage medium

[0140] In addition to the above methods and devices, an embodiment of this application may also be a computer program product, which includes computer program instructions. When the computer program instructions are run by a processor, the processor is caused to execute the steps in the underwater acoustic signal feature processing method according to various embodiments of this application described in any of the above embodiments of this specification, or execute the steps in the underwater acoustic signal recognition method according to various embodiments of this application described in any of the above embodiments of this specification.

[0141] The computer program product can be written in any combination of one or more programming languages for executing the program code of the operations of the embodiments of the present application. The programming languages include object-oriented programming languages such as Java, C++, etc., and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, executed as an independent software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0142] In addition, an embodiment of the present application can also be a storage medium on which a computer program is stored. The computer program is executed by a processor to perform the steps in the method for processing the characteristics of underwater acoustic signals according to various embodiments of the present application described in any of the above embodiments of this specification, or to perform the steps in the method for identifying underwater acoustic signals according to various embodiments of the present application described in any of the above embodiments of this specification.

[0143] For the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present application.

[0144] It should be noted that the various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other. For the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0145] The steps in the methods of the various embodiments of the present application can be adjusted, combined, and deleted according to actual needs. The technical features recorded in the various embodiments can be replaced or combined.

[0146] The modules and sub-modules in the devices and terminals in the various embodiments of the present application can be combined, divided, and deleted according to actual needs.

[0147] In several embodiments provided by the present application, it should be understood that the disclosed terminals, devices, and methods can be implemented in other ways. For example, the terminal embodiments described above are merely illustrative. For example, the division of modules or sub-modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple sub-modules or modules can be combined or integrated into another module, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling, direct coupling, or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of devices or modules can be in electrical, mechanical, or other forms.

[0148] The modules or sub-modules described as separate components may or may not be physically separated. The components as modules or sub-modules may or may not be physical modules or sub-modules, that is, they can be located in one place, or can be distributed to multiple network modules or sub-modules. Some or all of the modules or sub-modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0149] In addition, in each embodiment of the present application, each functional module or sub-module can be integrated in a processing module, or each module or sub-module can exist physically alone, or two or more modules or sub-modules can be integrated in one module. The above-mentioned integrated modules or sub-modules can be implemented in the form of hardware or in the form of software functional modules or sub-modules.

[0150] Those skilled in the art can further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0151] The steps of the methods or algorithms described in combination with the embodiments disclosed in this article can be directly implemented by hardware, software units executed by a processor, or a combination of the two. The software units can be placed in a random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium well-known in the technical field.

[0152] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0153] The above description of the disclosed embodiments will enable those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is to be construed in the widest manner consistent with the principles and novel features disclosed herein.

Claims

1. A method for processing the characteristics of underwater acoustic signals, characterized in that, Comprising: Obtain a first spectrogram and a second spectrogram, where the first spectrogram contains global information of the underwater acoustic signal, and the second spectrogram contains local information of the underwater acoustic signal; Through a first feature extraction module in the feature extraction network, perform feature extraction on the first spectrogram to obtain a global feature representation of the underwater acoustic signal; Through a second feature extraction module in the feature extraction network, perform feature extraction on the second spectrogram to obtain a local feature representation of the underwater acoustic signal; Perform feature fusion on the global feature representation and the local feature representation to obtain a target feature representation of the underwater acoustic signal, and the target feature representation is at least used to identify an underwater acoustic target.

2. The method for processing the characteristics of the underwater acoustic signal according to claim 1, characterized in that, The number of the second spectrograms is multiple, and different second spectrograms among the multiple second spectrograms contain frequency points with different energy distribution ranges; The step of, through a second feature extraction module in the feature extraction network, performing feature extraction on the second spectrogram to obtain a local feature representation of the underwater acoustic signal includes: Through the second feature extraction module, perform feature extraction on the multiple second spectrograms respectively to obtain the local feature representation, and the local feature representation includes feature representations corresponding to the multiple second spectrograms respectively.

3. The method for processing the characteristics of the underwater acoustic signal according to claim 2, wherein The number of the second feature extraction modules is multiple, and among the multiple second feature extraction modules, different second feature extraction modules correspond to different feature extraction granularities; The step of, through the second feature extraction module, performing feature extraction on the multiple second spectrograms respectively to obtain the local feature representation includes: According to the index information corresponding to the multiple second spectrograms respectively and the feature extraction granularities corresponding to the multiple second feature extraction modules respectively, determine the corresponding relationship between the multiple second spectrograms and the multiple second feature extraction modules; According to the corresponding relationship, input the multiple second spectrograms into the multiple second feature extraction modules, and in the multiple second feature extraction modules, perform feature extraction on the multiple second spectrograms respectively to obtain the local feature representation; Wherein, the index information corresponding to the multiple second spectrograms respectively is related to the energy values of the frequency points included in the multiple second spectrograms respectively.

4. The method for processing the characteristics of the underwater acoustic signal according to claim 3, wherein The multiple second feature extraction networks include multiple convolutional neural networks with different convolutional kernel sizes. In the corresponding relationship, the smaller the energy value of the frequency points included in the second spectrogram, the smaller the convolutional kernel size of the convolutional neural network corresponding to the second spectrogram.

5. The method for processing the characteristics of the underwater acoustic signal according to any one of claims 1 to 4, characterized in that, The generation process of the second spectrogram includes: Divide the first spectrogram to obtain the second spectrogram.

6. The method for processing the characteristics of the underwater acoustic signal according to claim 5, wherein, The step of dividing the first spectrogram to obtain the second spectrogram includes: According to the energy distribution of multiple frequency points in the first spectrogram, cluster the multiple frequency points in the first spectrogram to obtain multiple clusters; According to the multiple clusters, determine the second spectrogram.

7. An underwater acoustic signal recognition method, characterized in that, Comprising: Obtain a target feature representation of an underwater acoustic signal, where the target feature representation is obtained by executing the feature processing method of the underwater acoustic signal according to any one of claims 1 to 6; According to the target feature representation, identify the target radiation source corresponding to the underwater acoustic signal.

8. An apparatus for processing the characteristics of an underwater acoustic signal, characterized in that, Comprising: A spectrogram acquisition unit, configured to acquire a first spectrogram and a second spectrogram, where the first spectrogram contains global information of an underwater acoustic signal, and the second spectrogram contains local information of the underwater acoustic signal; A global feature extraction unit, configured to perform feature extraction on the first spectrogram through a first feature extraction module in a feature extraction network to obtain a global feature representation of the underwater acoustic signal; A local feature extraction unit, configured to perform feature extraction on the second spectrogram through a second feature extraction module in the feature extraction network to obtain a local feature representation of the underwater acoustic signal; A feature fusion unit, configured to perform feature fusion on the global feature representation and the local feature representation to obtain a target feature representation of the underwater acoustic signal, where the target feature representation is at least used to identify an underwater acoustic target.

9. An underwater acoustic signal recognition device, characterized in that, Comprising: An acquisition unit, configured to acquire a target feature representation of an underwater acoustic signal, where the target feature representation is obtained by executing the feature processing method of the underwater acoustic signal according to any one of claims 1 to 6; An identification unit, configured to identify a target radiation source corresponding to the underwater acoustic signal according to the target feature representation.

10. An electronic device, characterized in that, Comprising a memory and a processor; The memory is connected to the processor and is configured to store a program; The processor is configured to implement the feature processing method of the underwater acoustic signal according to any one of claims 1 to 6 or the underwater acoustic signal identification method according to claim 7 by running the program in the memory.

11. A computer program product, characterized in that, Comprising a computer program, where when the computer program is executed by a processor, it implements the feature processing method of the underwater acoustic signal according to any one of claims 1 to 6 or the underwater acoustic signal identification method according to claim 7.

Images