Active sonar detection waveform recognition method and device based on generative adversarial network

Abstract

The present application provides a method and device for active sonar detection waveform recognition based on a generative adversarial network. The method includes: obtaining a sonar signal received by a sonar detector after sending a detection signal; when the discriminator in the GAN network is used to determine that the sonar signal contains an echo signal corresponding to the detection signal, the generator in the GAN network is used to perform distortion compensation on the sonar signal to obtain a compensated sonar signal; and the compensated sonar signal is matched and detected by a matched filter. The method uses a pre-trained generator to perform distortion compensation on the sonar signal, which can solve the problem of the performance degradation of the matched filter when the matched filter is used to detect the sonar signal due to the spatiotemporal frequency variation of the underwater acoustic channel.

Description

Technical Field

[0001] The present application relates to the field of underwater detection technology, and in particular to a method and device for active sonar detection waveform recognition based on a generative adversarial network. Background Art

[0002] Active sonar is a method of underwater detection. Specifically, the sonar detector actively sends a detection signal into the water, then receives the echo signal reflected by the target in the water, and identifies the target based on the echo signal. When identifying, it is usually necessary to perform waveform detection on the echo signal to detect the target waveform signal from the echo signal.

[0003] Since the matched filter can guarantee the maximum signal-to-noise ratio when detecting a known signal under a white noise background, the traditional detection method usually uses a matched filter to match the echo signal. However, when performing underwater detection, the interference of factors such as the platform on which the sonar detector is installed and the movement of the underwater target, as well as multiple reflections on the water surface and the bottom of the water, will cause the underwater acoustic channel to be distorted, which seriously reduces the detection performance of the matched filter. Summary of the invention

[0004] The present application provides a method and device for active sonar detection waveform recognition based on a generative adversarial network, which uses a generator to perform distortion compensation on a received echo signal to improve matched filtering performance.

[0005] In a first aspect, the present application provides an active sonar detection waveform recognition method, which is applied to a sonar detector.

[0006] The method comprises: obtaining a sonar signal; the sonar signal comprises a signal received by a sonar detector after sending a detection signal; in the case where a discriminator in a GAN network is used to determine that the sonar signal contains an echo signal corresponding to the detection signal, a generator in the GAN network is used to perform distortion compensation on the sonar signal to obtain a compensated sonar signal; the GAN network is obtained by training according to ideal signal samples, distorted signal samples and reverberation signal samples; and a matched filter is used to perform matched correlation detection on the compensated sonar signal.

[0007] In the above scheme, waveform recognition is performed through a pre-trained discriminator, and distortion compensation and matching correlation detection are performed when it is determined that the sonar signal contains an echo signal, which can improve the signal processing efficiency. In addition, using a pre-trained generator for distortion compensation can solve the distortion problem of the sonar signal caused by the spatiotemporal frequency variation of the underwater acoustic channel, improve the detection performance of the matched filter, and obtain a more accurate waveform signal.

[0008] In a possible implementation, the training process of the GAN network includes: inputting the reverberation signal sample into the generator to obtain the output of the generator; respectively inputting the waveform signal sample and the output of the generator into the discriminator to obtain the output of the discriminator; the waveform signal sample includes the ideal signal sample and the distorted signal sample.

[0009] In a possible implementation, the ideal signal sample is obtained by convolving the detection signal sample with a transfer function of an ideal channel.

[0010] In a possible implementation, the distorted signal sample is obtained by the following steps: convolving the detection signal sample with the transfer function of the distorted channel corresponding to the ideal channel to obtain an initial distorted signal sample; and adding the initial distorted signal sample and the reverberation signal sample according to a preset signal mixing ratio to obtain the distorted signal sample.

[0011] In a possible implementation, the reverberation signal sample is obtained by convolving the detection signal sample with a reverberation response function.

[0012] In a possible implementation manner, the distorted channel includes: one or more of a flat fading channel, a time selective fading channel, a frequency selective fading channel, and a time-frequency two-dimensional selective fading channel.

[0013] In a possible implementation, using a discriminator in a GAN network to determine that the sonar signal contains an echo signal corresponding to the detection signal includes: inputting the sonar signal into the discriminator to obtain a probability value output by the discriminator; and determining, based on the probability value, that the sonar signal contains an echo signal corresponding to the detection signal.

[0014] In a second aspect, the present application provides an active sonar detection waveform recognition device, which is applied to a sonar detector. The device includes: an acquisition module, a compensation module and a detection module.

[0015] The acquisition module is used to acquire a sonar signal; the sonar signal includes an echo signal received by the sonar detector after the sonar detector sends a detection signal;

[0016] The compensation module is used to use the generator in the GAN network to perform distortion compensation on the sonar signal when the discriminator in the GAN network determines that the sonar signal contains an echo signal corresponding to the detection signal, so as to obtain a compensated sonar signal; the GAN network is trained based on ideal signal samples, distorted signal samples and reverberation signal samples;

[0017] The detection module is used to perform matching correlation detection on the compensated sonar signal using a matched filter.

[0018] In a possible implementation, the compensation module is further used to: input the reverberation signal sample into the generator to obtain the output of the generator; input the waveform signal sample and the output of the generator into the discriminator respectively to obtain the output of the discriminator; the waveform signal sample includes the ideal signal sample and the distorted signal sample.

[0019] In a possible implementation, the compensation module is further configured to obtain an ideal signal sample by convolving the detection signal sample with a transfer function of an ideal channel.

[0020] In a possible implementation, the compensation module is further used to obtain a distorted signal sample through the following steps: convolving the detection signal sample with the transfer function of the distorted channel corresponding to the ideal channel to obtain an initial distorted signal sample; and adding the initial distorted signal sample and the reverberation signal sample according to a preset signal mixing ratio to obtain the distorted signal sample.

[0021] In a possible implementation, the reverberation signal sample is obtained by convolving the detection signal sample with a reverberation response function.

[0022] In a possible implementation manner, the distorted channel includes: one or more of a flat fading channel, a time selective fading channel, a frequency selective fading channel, and a time-frequency two-dimensional selective fading channel.

[0023] In a possible implementation, the detection module is further used to: input the sonar signal into the discriminator to obtain a probability value output by the discriminator; and determine, based on the probability value, whether the sonar signal contains an echo signal corresponding to the detection signal.

[0024] In a third aspect, the present application further provides a sonar detector, which includes a processor and a memory, wherein the processor is used to execute computer instructions stored in the memory to implement the active sonar detection waveform recognition method described in the first aspect and its optional embodiments.

[0025] In a fourth aspect, the present application further provides a computer-readable storage medium, comprising instructions, which, when executed on a sonar detector, enable the sonar detector to execute the active sonar detection waveform recognition method described in the first aspect and its optional embodiments.

[0026] In a fifth aspect, the present application further provides a computer program product, which includes program code, and when the sonar detector runs the computer program product, the sonar detector executes the active sonar detection waveform recognition method described in the first aspect and its optional embodiments.

[0027] Any of the above-mentioned devices, computer storage media or computer program products are used to execute the method provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects of the corresponding schemes in the corresponding methods provided above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a schematic diagram of an application scenario provided by an embodiment of the present application;

[0029] Figure 2 It is a flow chart of an active sonar detection waveform recognition method based on a generative adversarial network provided in an embodiment of the present application;

[0030] Figure 3 It is a flow chart of a training method of a generative adversarial network provided in an embodiment of the present application;

[0031] Figure 4 It is a data flow diagram of a generative adversarial network provided in an embodiment of the present application;

[0032] Figure 5 It is a structural schematic diagram of an active sonar detection waveform recognition device provided in an embodiment of the present application;

[0033] Figure 6 It is a structural schematic diagram of a training device for a generative adversarial network provided in an embodiment of the present application;

[0034] Figure 7 It is a structural diagram of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0036] In the description of the embodiments of the present application, words such as "exemplary", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "for example" or "for example" is intended to present related concepts in a concrete way.

[0037] In the description of the embodiments of the present application, the term "and / or" is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, B exists alone, and A and B exist at the same time. In addition, unless otherwise specified, the term "multiple" means two or more. For example, multiple systems refers to two or more systems, and multiple screen terminals refers to two or more screen terminals.

[0038] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprises", "has" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0039] When detecting sonar detectors, matched filters are commonly used filters. DO North established the matched filter theory. He pointed out that if the signal input to the linear time-invariant filter is a known signal and the noise is additive stationary Gaussian white noise, then under the condition of a certain input power signal-to-noise ratio, the filter that maximizes the output power signal-to-noise ratio is the matched filter. In the marine environment, the best likelihood ratio detector for an ideal channel is a matched filter, and its equivalent form is a replica correlator. However, since the underwater acoustic channel in the water environment is prone to spatiotemporal frequency variation (see the introduction in the background technology for specific reasons), and it is impossible to know the expansion time and expansion frequency of the specific underwater acoustic channel, the performance of the matched filter is reduced.

[0040] When performing waveform recognition on echo signals, if the echo signals are searched and matched in the two-dimensional space of time and frequency, it will take a lot of time and real-time performance is difficult to guarantee.

[0041] Specifically, the channel that causes Doppler frequency spread distortion of the echo signal can be modeled as a time selective fading channel (or frequency spread channel). The best likelihood ratio detector for the frequency spread channel is a segment replica correlator (SRC). The essence of SRC is to segment the echo signal according to the spread frequency (or coherence time) of the channel, and then perform replica correlation with the transmitted signal and accumulate the sum. The echo signal in each segment can be considered as a channel without distortion, and a matched filtering operation is performed with the transmitted signal. When the coherence time corresponding to the spread frequency of the actual channel matches the coherence time used by the SRC detector, the detection performance of SRC is optimal, otherwise, the detection performance of SRC is seriously reduced. However, in actual applications, since the coherence time of the actual channel is difficult to predict, it is difficult to match the coherence time of the actual channel with the coherence time used by the SRC detector.

[0042] Specifically, the channel that causes the time expansion of the echo signal can be modeled as a frequency selective fading channel (or referred to as a time expansion channel). The best likelihood ratio detector for the time expansion channel is a replica correlation integrator (RCI). The RCI test statistic can be expressed as the integral of the replica correlator over the echo expansion time. When the expansion time (coherence bandwidth) of the actual channel matches the expansion time adopted by the RCI detector, the detection performance of RCI is optimal. Otherwise, when the actual channel expansion time does not match the expansion integration time adopted by the SRC detector, the detection performance of RCI is seriously degraded. However, in actual applications, since the expansion time of the actual channel is difficult to predict, it is difficult to match the actual channel expansion time with the expansion time adopted by the RCI detector.

[0043] In this regard, an embodiment of the present application provides an active sonar detection waveform recognition method based on a generative adversarial network. Specifically, a generative adversarial network (GAN) is pre-trained using signal samples. Figure 1 As shown in Figure 2, during detection, the generator in GAN is used to compensate for the distortion of the echo signal of the sonar detector. Figure 1 Give a detailed introduction.

[0044] Figure 2 1 is a flow chart of an active sonar detection waveform recognition method based on a generative adversarial network provided in an embodiment of the present application. Figure 2 As shown, the method includes the following steps S201-S203.

[0045] In step S201, a sonar signal received by a sonar detector is obtained.

[0046] When conducting underwater detection, a sonar detector is used to send out a detection signal, and the detector receives the signal in the water. The detection signal propagates in the water environment, and after being reflected by various objects in the water, the sonar detector receives the reflected echo signal. In other words, when the detector receives the signal, it may receive the aforementioned echo signal, and it may also receive signals from other objects. Signals emitted by other physical objects can be understood as signals unrelated to the detection signal, such as sound signals emitted by dolphins.

[0047] The detection signal and the sonar signal are both acoustic signals. Specifically, the transducer included in the sonar detector can convert the pre-modulated electrical signal into the detection signal and transmit the detection signal.

[0048] In step S202, when the discriminator in the GAN is used to determine that the sonar signal contains an echo signal corresponding to the detection signal, the generator in the GAN network is used to perform distortion compensation on the sonar signal to obtain a compensated sonar signal.

[0049] Specifically, the step includes inputting the sonar signal into the discriminator to obtain the probability value output by the discriminator. When the output probability value is greater than a preset value, it can be determined that the sonar signal contains an echo signal corresponding to the detection signal, and then the sonar signal is input into the generator for distortion compensation, and the compensated sonar signal is obtained according to the output of the generator.

[0050] Before distortion compensation, a discriminator is used to detect whether an echo signal corresponding to the detection signal is included, so that distortion compensation and subsequent detection can be avoided when the sonar detector does not receive an echo signal, thereby improving signal processing efficiency. That is, when it is determined that the sonar signal does not contain an echo signal corresponding to the detection signal, distortion compensation and detection are not performed on it.

[0051] Among them, GAN is trained based on ideal signal samples, distorted signal samples and reverberation signal samples. The specific training process will be introduced later and will not be repeated here.

[0052] In step S203, matched correlation detection is performed on the compensated sonar signal using a matched filter.

[0053] After the distortion compensation in step S202, the distortion (time expansion and frequency expansion) of the sonar signal is compensated. However, since the GAN network lacks comparison with the standard signal during discrimination, a matched filter is still needed to perform matched correlation detection on the waveform signal. The matched filter can be obtained based on the aforementioned ideal signal sample and distorted signal sample.

[0054] Figure 3 A training method for a GAN network is provided in an embodiment of the present application. The method can be applied to a training device, wherein the training device may include the aforementioned sonar detector. The method includes the following steps S301 and S302.

[0055] In step S301, training samples are obtained, where the training samples include ideal signal samples, distorted signal samples, and reverberation signal samples.

[0056] Specifically, the ideal signal sample can be obtained by convolving the detection signal sample with the transfer function of the ideal channel.

[0057] Specifically, the distorted signal sample can be obtained by the following steps: convolving the detection signal sample with the transfer function of the distorted channel corresponding to the ideal channel to obtain the initial distorted signal sample; adding the initial distorted signal sample and the reverberation signal sample according to a preset signal mixing ratio to obtain the distorted signal sample. The distorted channel includes one or more of a flat fading channel, a time selective fading channel, a frequency selective fading channel, and a time-frequency two-dimensional selective fading channel.

[0058] Specifically, the reverberation signal sample may be obtained by convolving the reverberation response function with the detection signal sample.

[0059] In step S302, the parameters of the GAN network are updated according to the training samples.

[0060] The entire training process of GAN is a process in which the discriminator D and the generator G in the network compete with each other and update alternately.

[0061] After the structure of the GAN network is determined, the reverberation signal sample z is input into the generator in the GAN network to obtain the forged waveform signal sample G(z) output by the generator. The waveform signal sample x and the output of the generator G are respectively input into the discriminator D to obtain the output of the discriminator D(x) and D(G(z)). The waveform signal sample includes the ideal signal sample and the distorted signal sample.

[0062] The mechanism of GAN is an optimization problem, and the optimization function is shown in formula (4). The generator will output a forged waveform signal sample based on the input to deceive the discriminator D. The discriminator D makes a true or false judgment on the waveform signal, with the goal of judging that the waveform signal sample x is true and judging that the forged waveform signal sample G(z) output by the generator G is false. During the training process, D should maximize the difference between the real data and the generated data, that is, D(x) is close to 1, and D(G(z)) is close to 0. G should make the generated signal closer to the waveform signal sample, that is, D(G(z)) is close to 1.

[0063]

[0064] In formula (4), Represents the average value of the logarithm of the probability value corresponding to each waveform signal sample output by the discriminator, It represents the average value of the logarithm of the difference between the probability value corresponding to the forged waveform signal output by the discriminator and the first label value, and the first label value is 1; P(x) represents the probability distribution of x, and P(z) represents the probability distribution of z.

[0065] When training the network, the loss value is calculated based on the output value of D, the label value of the sample and the loss function. According to the loss value, the optimizer is used to update the parameters of the generator and the discriminator. When the loss value is 0, the best training effect is achieved, and the training of the generative adversarial network is completed. Among them, the loss function can be one of the L1 loss function, the L2 loss function, the SSIM loss function, and the MS-SSIM loss function. The optimizer can be one of the batch gradient descent algorithm, the stochastic gradient descent algorithm, the Adagrad optimization algorithm, and the Adam optimization algorithm. The Adam optimization algorithm is an extension of the stochastic gradient descent algorithm.

[0066] Specifically, the following steps can be used to train GAN:

[0067] 1) Fix the generator G and train the discriminator D. During the first training, the generator G has not been trained yet and its weight parameters are random values. Therefore, after the reverberation signal sample passes through the generator G, a random signal is obtained with a false label. The waveform signal sample and the signal generated by the generator G are input into the discriminator D to train the discriminator D;

[0068] 2) Fix the discriminator D and train the generator G. After the training in the previous step, the discriminator D has the initial ability to judge whether the waveform is true or false. At this time, the generator G is improved and updated in the direction of continuously decreasing the second term of the optimization function (1) until the waveform signal generated by the generator G can deceive the discriminator D;

[0069] 3) Repeat 1) and 2), and the generator G and the discriminator D are continuously optimized and trained until the output of the discriminator D converges to 1 / 2, and the training is completed. The trained generator G and discriminator D are obtained. The former can judge whether the waveform signal is true or false, and the latter can generate a forged waveform signal with the same distribution as the real waveform signal sample.

[0070] In the above training process, the waveform signal samples input to the discriminator include distorted signal samples, so that the discriminator can be used to judge the signal received by the detector to determine whether it contains an echo signal corresponding to the detection signal. In addition, the generator in the GAN network is used to compensate for the distortion of the signal when applied. Therefore, in order to improve the distortion compensation effect, in the training samples, the number of ideal signal samples input to the discriminator should be higher than the number of distorted signal samples.

[0071] based on Figure 2 The above-mentioned identification method, the embodiment of the present application also provides an identification device, which is applied to the above-mentioned sonar detector.

[0072] Figure 5 FIG. 5 is a schematic diagram of a structure of an identification device 500 provided in an embodiment of the present application. Figure 5As shown, the identification device 500 includes: an acquisition module 501 , a compensation module 502 and a detection module 503 .

[0073] The acquisition module 501 is used to acquire a sonar signal; the sonar signal includes an echo signal received by the sonar detector after sending a detection signal.

[0074] Among them, the compensation module 502 is used to use the generator in the GAN network to perform distortion compensation on the sonar signal when the discriminator in the GAN network determines that the sonar signal contains an echo signal corresponding to the detection signal, so as to obtain a compensated sonar signal; the GAN network is trained based on ideal signal samples, distorted signal samples and reverberation signal samples.

[0075] The detection module 503 is used to perform matching correlation detection on the compensated sonar signal using a matched filter.

[0076] The specific implementation process of each module in the above-mentioned identification device 500 can be found in the introduction of the above-mentioned invention content or the above-mentioned method embodiment, and will not be repeated here.

[0077] based on Figure 2 The training method described above, the embodiment of the present application also provides a training device, which is applied to the above-mentioned training equipment.

[0078] Figure 6 Schematic diagram of the structure of a training device 600 provided in an embodiment of the present application. Figure 6 As shown, the training device 600 includes: an acquisition module 601 and an update module 602.

[0079] The acquisition module 601 is used to acquire training samples, and the training samples include ideal signal samples, distorted signal samples and reverberation signal samples.

[0080] Among them, the updating module 602 is used to update the parameters of the GAN network according to the training samples.

[0081] The specific implementation process of each module in the above-mentioned training device 600 can be found in the introduction of the above-mentioned method embodiment, which will not be repeated here.

[0082] Figure 7 It is a schematic diagram of the hardware structure of a computing device 700 provided in an embodiment of the present application.

[0083] The computing device 700 may be the sonar detector or the training device. Figure 7The computing device 700 includes a processor 701, a memory 702, a communication interface 703, and a bus 704. The processor 701, the memory 702, and the communication interface 703 are connected to each other via the bus 704. The processor 701, the memory 702, and the communication interface 703 may also be connected in other connection modes besides the bus 704.

[0084] Among them, the memory 702 can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), flash memory, optical storage, hard disk, etc.

[0085] The processor 701 may be a general-purpose processor, which may be a processor that performs specific steps and / or operations by reading and executing the contents stored in a memory (e.g., the memory 702). For example, the general-purpose processor may be a central processing unit (CPU). The processor 701 may include at least one circuit to execute Figure 2 All or part of the steps of the identification method provided in the embodiment shown, or Figure 3 All or part of the steps of the training method shown.

[0086] The communication interface 703 includes an input / output (I / O) interface, a physical interface, and a logical interface, etc., which are used to interconnect devices within the computing device 700, and an interface for interconnecting the computing device 700 with other devices (such as other computing devices or user devices). The physical interface can be an Ethernet interface, a fiber optic interface, an ATM interface, etc.

[0087] The bus 704 may be any type of communication bus for interconnecting the processor 701 , the memory 702 , and the communication interface 703 , such as a system bus.

[0088] The above devices may be arranged on independent chips, or at least partially or completely on the same chip. Whether to arrange each device independently on different chips or to integrate them on one or more chips often depends on the needs of product design. The embodiments of the present application do not limit the specific implementation form of the above devices.

[0089] Figure 7 The computing device 700 shown is merely exemplary. During implementation, the computing device 700 may further include other components, which are not listed one by one herein.

[0090] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state disk (SSD)), etc.

[0091] It is understood that the various numerical numbers involved in the embodiments of the present application are only for the convenience of description and are not used to limit the scope of the embodiments of the present application. It should be understood that in the embodiments of the present application, the size of the sequence number of the above-mentioned processes does not mean the order of execution, and the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0092] The specific implementation methods described above further illustrate the purpose, technical solutions and beneficial effects of the present application in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present application. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solutions of the present application should be included in the scope of protection of the present application.

Claims

1. A method for identifying active sonar detection waveforms, characterized in that: include: Acquiring a sonar signal; the sonar signal includes a signal received by the sonar detector after the sonar detector sends a detection signal; When it is determined by using a discriminator in a GAN network that the sonar signal contains an echo signal corresponding to the detection signal, a generator in the GAN network is used to perform distortion compensation on the sonar signal to obtain a compensated sonar signal; the GAN network is obtained by training based on ideal signal samples, distorted signal samples, and reverberation signal samples; Using a matched filter to perform matched correlation detection on the compensated sonar signal; The training process of the GAN network includes: Inputting the reverberation signal sample into the generator to obtain the output of the generator; Inputting waveform signal samples and the output of the generator into the discriminator respectively to obtain the output of the discriminator; the waveform signal samples include the ideal signal samples and the distorted signal samples; According to the output of the discriminator and the loss function, the parameters in the generator and the discriminator are updated.

2. The method according to claim 1, characterized in that: The ideal signal samples are obtained by convolving the detection signal samples with the transfer function of the ideal channel.

3. The method according to claim 1, characterized in that: The distorted signal sample is obtained by the following steps: Convolve the detection signal sample with the transfer function of the distorted channel corresponding to the ideal channel to obtain an initial distorted signal sample; The initial distortion signal sample and the reverberation signal sample are added according to a preset signal mixing ratio to obtain the distortion signal sample.

4. The method according to claim 1, characterized in that: The reverberation signal samples are obtained by convolving the detection signal samples with the reverberation response function.

5. The method according to claim 3, characterized in that: The distorted channel includes one or more of a flat fading channel, a time selective fading channel, a frequency selective fading channel and a time-frequency two-dimensional selective fading channel.

6. The method according to claim 1, characterized in that: The method of using a discriminator in a GAN network to determine that the sonar signal contains an echo signal corresponding to the detection signal comprises: Inputting the sonar signal into the discriminator to obtain a probability value output by the discriminator; According to the probability value, it is determined that the sonar signal contains an echo signal corresponding to the detection signal.

7. An active sonar detection waveform recognition device, characterized in that: include: An acquisition module, used for acquiring a sonar signal; the sonar signal includes a signal received by the sonar detector after the sonar detector sends a detection signal; A compensation module, configured to, when determining by using a discriminator in a GAN network that the sonar signal contains an echo signal corresponding to the detection signal, use a generator in the GAN network to perform distortion compensation on the sonar signal to obtain a compensated sonar signal; the GAN network is obtained by training according to ideal signal samples, distorted signal samples, and reverberation signal samples; A detection module, used for performing matching correlation detection on the compensated sonar signal using a matched filter; The training process of the GAN network includes: Inputting the reverberation signal sample into the generator to obtain the output of the generator; Inputting waveform signal samples and the output of the generator into the discriminator respectively to obtain the output of the discriminator; the waveform signal samples include the ideal signal samples and the distorted signal samples; According to the output of the discriminator and the loss function, the parameters in the generator and the discriminator are updated.

8. A sonar detector, characterized in that: include: A processor and a memory, wherein the processor is used to execute computer instructions stored in the memory to implement the active sonar detection waveform recognition method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The invention comprises instructions, which, when executed on a sonar detector, enable the sonar detector to execute the active sonar detection waveform recognition method as described in any one of claims 1 to 6.

Images