Radar operating mode recognition method, apparatus, device, medium, and program product

Abstract

The application provides a radar working mode recognition method, device, equipment, medium and program product, and relates to the technical field of artificial intelligence. The method comprises the following steps: acquiring a pulse description word (PDW) sequence of a radar emitter; determining an intra-pulse feature of the radar emitter and an inter-pulse feature of the radar emitter based on the PDW sequence; inputting the intra-pulse feature and the inter-pulse feature into a working mode feature extraction model to obtain an intra-pulse feature vector and an inter-pulse feature vector output by the working mode feature extraction model; and inputting a feature fusion vector of the intra-pulse feature vector and the inter-pulse feature vector into a radar working mode recognition model to obtain a radar working mode recognition result output by the radar working mode recognition model. The intra-pulse feature and the inter-pulse feature are obtained first, and then the intra-pulse feature and the inter-pulse feature are subjected to feature extraction to obtain the intra-pulse feature vector and the inter-pulse feature vector with context information, so that high-precision recognition of the radar working mode can be realized for each scene.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a radar operating mode recognition method, apparatus, device, medium, and program product. Background Technology

[0002] Radar operating modes refer to the operational patterns adopted by radar by changing the style and parameters of its transmitted signals to adapt to different targets, interference environments, and missions. Radar operating modes can include, but are not limited to: velocity search, track-plus-search, range-while-track search, multi-target tracking, search-while-track, and single-target tracking, etc. In radar application scenarios, radar operating mode identification is typically required.

[0003] Currently, by statistically analyzing and storing pulse descriptors such as carrier frequency, pulse width, and pulse repetition interval of different radar radiation sources into a database, radar operating mode identification can be achieved by matching each descriptor with a template from the database. However, this existing technology is only suitable for scenarios with large differences in radar signal parameters and a simple electromagnetic environment. For other scenarios, the accuracy of this radar operating mode identification scheme is low. Summary of the Invention

[0004] This invention provides a radar operating mode recognition method, apparatus, device, medium, and program product to address the shortcomings of low accuracy in radar operating mode recognition in the prior art, and to achieve highly accurate radar operating mode recognition.

[0005] This invention provides a radar operating mode recognition method, comprising: Obtain the pulse descriptor word (PDW) sequence of the radar radiation source; the PDW sequence includes PDWs at multiple times. Based on the PDW sequence, the intra-pulse characteristics and inter-pulse characteristics of the radar radiation source are determined. The intrapulse features and interpulse features are input into the working mode feature extraction model to obtain the intrapulse feature vector and interpulse feature vector output by the working mode feature extraction model; the working mode feature extraction model is used to extract intrapulse feature vectors with contextual information based on the intrapulse features, and the working mode feature extraction model is also used to extract interpulse feature vectors with contextual information based on the interpulse features. The feature fusion vector of the intra-pulse feature vector and the inter-pulse feature vector is input into the radar operating mode recognition model to obtain the radar operating mode recognition result output by the radar operating mode recognition model; the radar operating mode recognition model is used to perform radar operating mode recognition on the feature fusion vector.

[0006] According to the radar operating mode identification method provided by the present invention, any of the PDWs includes a carrier frequency, a pulse width, and a time of arrival; The step of determining the intra-pulse characteristics and inter-pulse characteristics of the radar radiation source based on the PDW sequence includes: The carrier frequency, pulse width, and arrival time of the PDW sequence are extracted respectively to obtain the carrier frequency sequence, pulse width sequence, and arrival time sequence; The intra-pulse feature is obtained by concatenating the carrier frequency sequence and the pulse width sequence. The arrival time series is differentially processed to obtain the inter-pulse features.

[0007] According to a radar operating mode recognition method provided by the present invention, the step of inputting the intra-pulse features and the inter-pulse features into an operating mode feature extraction model to obtain the intra-pulse feature vector and the inter-pulse feature vector output by the operating mode feature extraction model includes: The intrapulse features are input into the first long short-term memory network layer in the working mode feature extraction model to obtain the intrapulse feature vector output by the first long short-term memory network layer. The inter-pulse features are input into the second long short-term memory network layer in the working mode feature extraction model to obtain the inter-pulse feature vector output by the second long short-term memory network layer. The working mode feature extraction model is obtained by unsupervised training based on sample PDW sequences.

[0008] According to the radar operating mode recognition method provided by the present invention, the operating mode feature extraction model is a bidirectional contrastive predictive coding model, and the operating mode feature extraction model is used to extract essential features.

[0009] According to a radar operating mode recognition method provided by the present invention, after acquiring the pulse descriptor word (PDW) sequence of the radar radiation source, the method further includes: The PDW sequence is input into the radar operating mode detection model to obtain the radar operating mode detection result output by the radar operating mode detection model; the radar operating mode detection model is used to detect radar operating modes based on the PDW sequence. After inputting the feature fusion vector of the intra-pulse feature vector and the inter-pulse feature vector into the radar operating mode recognition model to obtain the radar operating mode recognition result output by the radar operating mode recognition model, the method further includes: Based on the radar operating mode identification result and the radar operating mode detection result, the final radar operating mode identification result is determined.

[0010] According to a radar operating mode recognition method provided by the present invention, the step of inputting the PDW sequence into a radar operating mode detection model to obtain the radar operating mode detection result output by the radar operating mode detection model includes: Each PDW in the PDW sequence is input into the radar operating mode detection model to obtain the radar operating mode detection result output by the radar operating mode detection model. The radar operating mode detection model is used to detect radar operating modes based on a PDW, and the radar operating mode detection result includes the radar operating mode detection sub-results corresponding to each PDW.

[0011] The present invention also provides a radar operating mode identification device, comprising: A sequence acquisition module is used to acquire the pulse descriptor word (PDW) sequence of the radar radiation source; the PDW sequence includes PDWs at multiple times. The feature determination module is used to determine the intra-pulse features and inter-pulse features of the radar radiation source based on the PDW sequence. The feature extraction module is used to input the intrapulse features and the interpulse features into the working mode feature extraction model to obtain the intrapulse feature vector and the interpulse feature vector output by the working mode feature extraction model; the working mode feature extraction model is used to extract intrapulse feature vectors with contextual information based on the intrapulse features, and the working mode feature extraction model is also used to extract interpulse feature vectors with contextual information based on the interpulse features; The pattern recognition module is used to input the feature fusion vector of the intra-pulse feature vector and the inter-pulse feature vector into the radar operating mode recognition model to obtain the radar operating mode recognition result output by the radar operating mode recognition model; the radar operating mode recognition model is used to perform radar operating mode recognition on the feature fusion vector.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the radar operating mode recognition method as described above.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the radar operating mode recognition method as described above.

[0014] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the radar operating mode recognition method as described above.

[0015] The radar operating mode recognition method, apparatus, device, medium, and program products provided by this invention acquire the PDW sequence of a radar radiation source. The PDW sequence includes PDWs at multiple time points, thus enabling accurate recognition even if some PDWs contain erroneous parameters, thereby improving the accuracy of radar operating mode recognition. Based on the PDW sequence, the intra-pulse and inter-pulse characteristics of the radar radiation source are determined, allowing for accurate recognition of the radar operating mode based on these characteristics, further improving the accuracy of radar operating mode recognition. The intra-pulse and inter-pulse characteristics are input into an operating mode feature extraction model to obtain intra-pulse and inter-pulse feature vectors output by the model. Based on these vectors, further feature extraction is performed on the intra-pulse and inter-pulse characteristics, allowing for the extraction of operating mode features for parameters within and between radar pulses, ensuring more comprehensive extraction of operating mode features and thus improving radar operating mode recognition. Accuracy; the working mode feature extraction model is used to extract intra-pulse feature vectors with contextual information based on intra-pulse features, and the working mode feature extraction model is also used to extract inter-pulse feature vectors with contextual information based on inter-pulse features, thereby fully extracting the implicit features and temporal correlation information within the PDW sequence, and thus improving the accuracy of radar working mode recognition; based on the input features and feature extraction methods determined above, the feature fusion vector of intra-pulse feature vector and inter-pulse feature vector is input into the radar working mode recognition model, which can obtain accurate radar working mode recognition results. At the same time, the above-mentioned method of first obtaining intra-pulse and inter-pulse features based on the PDW sequence, and then extracting working mode features from the intra-pulse and inter-pulse features to obtain intra-pulse feature vectors and inter-pulse feature vectors with contextual information, can achieve high-precision recognition of radar working modes even in complex electromagnetic environment scenarios such as incomplete radar signal parameters or overlapping radar signal parameters, and can achieve high-precision recognition in any scenario. In summary, this invention can achieve high-precision recognition of radar working modes in various scenarios, that is, this invention can improve the accuracy of radar working mode recognition and enhance the adaptability of radar working modes. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is one of the flowcharts of the radar operating mode recognition method provided by the present invention.

[0018] Figure 2This is the second flowchart of the radar operating mode recognition method provided by the present invention.

[0019] Figure 3 This is the third flowchart of the radar operating mode recognition method provided by the present invention.

[0020] Figure 4 This is the fourth flowchart of the radar operating mode recognition method provided by the present invention.

[0021] Figure 5 This is the fifth flowchart of the radar operating mode recognition method provided by the present invention.

[0022] Figure 6 This is the sixth flowchart of the radar operating mode recognition method provided by the present invention.

[0023] Figure 7 This is a schematic diagram of the radar operating mode recognition device provided by the present invention.

[0024] Figure 8 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0026] Current radar operating mode recognition solutions mostly rely on template matching based on expert experience. This means that each radar operating mode is matched against a template library. However, template matching is only suitable for scenarios with highly varied radar signal parameters and a uniform electromagnetic environment. For other scenarios, its accuracy is low.

[0027] Given the low accuracy of current radar operating mode recognition schemes, this invention conducted research and found that existing schemes mainly use several dimensional parameters of radar pulse descriptors, extract statistical features through mathematical statistical methods, and then perform template matching. However, these features only statistically analyze the distribution patterns of pulse descriptors, leading to low accuracy in template matching. Therefore, the initial approach of this invention was to use traditional machine learning-based pattern recognition methods, that is, to directly input the several dimensional parameters of radar pulse descriptors into a machine learning multi-classifier.

[0028] This invention, through research on the aforementioned approach, reveals that while it eliminates the need for template matching, it still primarily utilizes several dimensional parameters of radar pulse descriptors and extracts statistical features using mathematical statistical methods. However, it fails to extract deeper grammatical patterns and temporal relationships. Therefore, the feature description information provided by this approach is incomplete and insufficient. While it performs well in radar operating mode recognition under fixed scenarios, it exhibits poor adaptability to complex electromagnetic environments or scenarios with overlapping radar signal parameters. In other words, it is only suitable for scenarios with significant differences in radar signal parameters and a uniform electromagnetic environment; for other scenarios, the accuracy of radar operating mode recognition using this approach remains low.

[0029] To address the problems with the aforementioned approach, this invention continued its research, considering the creation of artificially designed physical features such as signal envelopes, amplitude features, and syntactic features. These features were constructed based on pulse descriptors and then input into a multi-classifier for radar operating mode recognition. However, while these artificially designed features possess good interpretability in physical meaning, their information expression capabilities are limited, and some implicit features contained within the radar descriptor sequence have not been fully explored.

[0030] To address the shortcomings of the aforementioned approaches, this invention further proposes a radar operating mode recognition method. This method extracts intra-pulse feature vectors with contextual information based on intra-pulse features and extracts inter-pulse feature vectors with contextual information based on inter-pulse features, thereby mining the implicit features and temporal correlation information within the radar descriptor sequence and improving the accuracy of radar operating mode recognition.

[0031] The radar operating mode recognition method provided by the present invention will be described below through the following embodiments. Figures 1-6 The radar operating mode recognition method of the present invention is described.

[0032] Figure 1 This is one of the flowcharts illustrating the radar operating mode recognition method provided by the present invention, such as... Figure 1 As shown, the radar operating mode recognition method includes the following steps 110, 120, 130 and 140.

[0033] Step 110: Obtain the Pulse Description Word (PDW) sequence of the radar radiation source.

[0034] Here, the radar radiation source is the device in the radar system responsible for generating, modulating, and transmitting electromagnetic wave signals, and this radar radiation source is the radiation source to be identified in the radar's operating mode. In one embodiment, the radar radiation source includes a transmitter, an antenna system, and a frequency synthesizer.

[0035] Here, the PDW (Pulse Description Word) sequence includes PDWs at multiple time points, and this PDW sequence is the sequence for radar operating mode identification. The PDW is a digital description of the key parameters of the radar pulse signal of a radar radiation source, which can be used to identify and classify radar radiation sources. In one embodiment, the PDW is obtained by receiving the radar pulse signal and then extracting parameters from the radar pulse signal.

[0036] The PDW may include, but is not limited to, at least one of the following: carrier frequency (CF), pulse width (PW), time of arrival (TOA), direction of arrival (DOA), pulse amplitude (PA), pulse repetition interval (PRI), etc.

[0037] It should be noted that the reason for obtaining PDW sequences is that PDW sequences include PDWs from multiple time points, allowing for the extraction of features with contextual information. This means mining the implicit features and temporal correlation information within the PDW sequence, thereby improving the accuracy of radar operating mode recognition. Furthermore, the short time intervals between PDWs in a PDW sequence result in higher accuracy for radar operating mode recognition compared to relying on a single PDW sequence, avoiding misidentification due to erroneous parameters in some PDWs.

[0038] Step 120: Based on the PDW sequence, determine the intra-pulse characteristics and inter-pulse characteristics of the radar radiation source.

[0039] Here, intra-pulse features are used to characterize the signal characteristics and parameters within a single radar pulse. These intra-pulse features can be determined by at least one parameter in the PDW sequence. The specific determination method can refer to existing technologies or the method studied in this invention. The determination method studied in this invention is as follows: extract the carrier frequency and pulse width of the PDW sequence separately to obtain a carrier frequency sequence and a pulse width sequence; then concatenate the carrier frequency sequence and the pulse width sequence to obtain the intra-pulse features.

[0040] Here, inter-pulse features are used to characterize the parameter variation patterns between different radar pulses. These inter-pulse features can be determined by at least one parameter in the PDW sequence. The specific determination method can refer to existing technologies or the method studied in this invention. The determination method studied in this invention is as follows: extract the arrival time of the PDW sequence to obtain an arrival time series, and perform differential processing on the arrival time series to obtain the inter-pulse features.

[0041] It should be understood that extracting intra-pulse features can improve the accuracy of radar operating mode identification, and extracting inter-pulse features can also improve the accuracy of radar operating mode identification.

[0042] Step 130: Input the intrapulse features and interpulse features into the working mode feature extraction model to obtain the intrapulse feature vector and interpulse feature vector output by the working mode feature extraction model.

[0043] Here, the operating mode feature extraction model is used to extract a feature that can effectively characterize the radar operating mode attributes, thereby facilitating the accurate identification of the radar operating mode in the future.

[0044] The working mode feature extraction model is a deep learning model. For example, the working mode feature extraction model may include, but is not limited to, at least one of the following: Recurrent Neural Network (RNN), Long Short-Term Memory Network (LSTM), Convolutional Neural Network (CNN), Residual Neural Network (ResNet), Generative Adversarial Network (GAN), or Transformer network, etc., thereby utilizing the superior expressive power of deep learning networks to extract deep patterns, that is, to extract the hidden features inside the PDW sequence, thereby improving the accuracy of radar working mode recognition.

[0045] In one embodiment, the working mode feature extraction model is obtained through supervised training based on sample PDW sequences, and the specific training method can employ existing techniques. In another embodiment, the working mode feature extraction model is obtained through unsupervised training based on sample PDW sequences, and this training method can be a contrastive learning approach.

[0046] Specifically, the working mode feature extraction model is used to extract intra-pulse feature vectors with contextual information based on the intra-pulse features, and the working mode feature extraction model is also used to extract inter-pulse feature vectors with contextual information based on the inter-pulse features. That is, the working mode feature extraction model is used to further extract working mode features with contextual information.

[0047] In one embodiment, the working mode feature extraction model includes a long short-term memory network layer, which is used to extract working mode features with contextual information.

[0048] It should be understood that extracting inter-pulse feature vectors and intra-pulse feature vectors with contextual information is necessary to fully extract (mined) the implicit features and temporal correlation information within the PDW sequence, that is, to extract deeper grammatical rules and temporal correlations, that is, to extract more comprehensive and complete features, thereby improving the accuracy of radar working mode recognition.

[0049] In one specific embodiment, the working mode feature extraction model includes two branches, such as a first working mode feature extraction layer and a second working mode feature extraction layer. Specifically, intra-pulse features are input to the first working mode feature extraction layer to obtain an intra-pulse feature vector output by the first working mode feature extraction layer; inter-pulse features are input to the second working mode feature extraction layer to obtain an inter-pulse feature vector output by the second working mode feature extraction layer.

[0050] It should be understood that further feature extraction of intra-pulse and inter-pulse features allows for the extraction of contextual information for features within and between radar pulses, ensuring more comprehensive extraction of operating mode features and thus improving the accuracy of radar operating mode recognition. It should be noted that this invention, through continuous research, has found that further feature extraction of intra-pulse and inter-pulse features significantly improves the accuracy of radar operating mode recognition.

[0051] Step 140: Input the feature fusion vector of the intra-pulse feature vector and the inter-pulse feature vector into the radar operating mode recognition model to obtain the radar operating mode recognition result output by the radar operating mode recognition model.

[0052] Here, the feature fusion vector is obtained by fusing intra-pulse feature vectors and inter-pulse feature vectors. This feature fusion method can include, but is not limited to: concatenation, element-wise operations (adding, multiplying, or averaging vectors of the same dimension), weighted fusion, attention mechanism fusion, etc.

[0053] The radar operating mode recognition model is used to identify the radar operating mode of the feature fusion vector.

[0054] In one specific embodiment, the radar operating mode recognition model is a classification model, thus eliminating the need for template matching and improving the accuracy of radar operating mode recognition. Furthermore, the radar operating mode recognition model includes a deep residual network (ResNet), thereby increasing the network model depth through residual learning without causing network degradation, effectively improving the accuracy of radar operating mode recognition.

[0055] In one embodiment, the radar operating mode recognition model is obtained through supervised training based on sample PDW sequences, and the specific training method can employ existing techniques. In another embodiment, the radar operating mode recognition model is obtained through unsupervised training based on sample PDW sequences, and this training method can also employ existing techniques.

[0056] Here, the radar operating mode identification result is used to indicate the radar operating mode of the radar radiation source. The radar operating mode may include, but is not limited to: velocity search, track plus search, range-while-track search, multi-target tracking, search-while-track, and single-target tracking, etc.

[0057] The radar operating mode recognition method provided in this invention acquires the PDW sequence of the radar radiation source, and the PDW sequence includes PDWs at multiple times. Therefore, even if the PDW at some times contains erroneous parameters, accurate recognition can still be achieved based on PDWs at multiple times, thereby improving the accuracy of radar operating mode recognition. Based on the PDW sequence, the intra-pulse features and inter-pulse features of the radar radiation source are determined, thereby accurately recognizing the radar operating mode based on the intra-pulse and inter-pulse features, further improving the accuracy of radar operating mode recognition. The intra-pulse and inter-pulse features are input into the operating mode feature extraction model to obtain the intra-pulse feature vector and inter-pulse feature vector output by the operating mode feature extraction model. Based on this, further feature extraction is performed on the intra-pulse and inter-pulse features, allowing for the extraction of operating mode features for parameters within and between radar pulses, ensuring more comprehensive extraction of operating mode features, thereby improving the accuracy of radar operating mode recognition. The working mode feature extraction model is used to extract intra-pulse feature vectors with contextual information based on intra-pulse features. The working mode feature extraction model is also used to extract inter-pulse feature vectors with contextual information based on inter-pulse features, thereby fully extracting the implicit features and temporal correlation information within the PDW sequence, thus improving the accuracy of radar working mode recognition. Based on the input features and feature extraction methods determined above, the feature fusion vector of intra-pulse and inter-pulse feature vectors is input into the radar working mode recognition model, which can obtain accurate radar working mode recognition results. Furthermore, by first obtaining intra-pulse and inter-pulse features based on the PDW sequence, and then extracting working mode features from the intra-pulse and inter-pulse features to obtain intra-pulse and inter-pulse feature vectors with contextual information, high-precision recognition of radar working modes can be achieved even in complex electromagnetic environments such as incomplete or overlapping radar signal parameters, and high-precision recognition can be achieved regardless of the scenario. In summary, this invention can achieve high-precision recognition of radar working modes for various scenarios, that is, this invention can improve the accuracy of radar working mode recognition and enhance the adaptability of radar working modes.

[0058] Based on any of the above embodiments, a specific embodiment of the radar operating mode recognition method is given below. Figure 2 This is the second flowchart of the radar operating mode recognition method provided by the present invention, as shown below. Figure 2 As shown, the radar operating mode identification method includes steps 110, 121, 122, 123, 130, and 140.

[0059] Step 110: Obtain the Pulse Description Word (PDW) sequence of the radar radiation source.

[0060] Here, the PDW sequence includes PDW at multiple time points, and the PDW at any given time point includes the carrier frequency, pulse width, and arrival time.

[0061] Step 121: Extract the carrier frequency, pulse width, and arrival time of the PDW sequence to obtain the carrier frequency sequence, pulse width sequence, and arrival time sequence.

[0062] Here, the carrier frequency sequence includes carrier frequencies at multiple times. The pulse width sequence includes pulse widths at multiple times. The arrival time sequence includes arrival times at multiple times.

[0063] Step 122: Concatenate the carrier frequency sequence and the pulse width sequence to obtain the intra-pulse features.

[0064] Specifically, the carrier frequency sequence and the pulse width sequence are concatenated together to obtain intra-pulse features. Concatenating the carrier frequency sequence and the pulse width sequence yields intra-pulse features that characterize the signal features and parameters within a single radar pulse. In one specific embodiment, the intra-pulse feature obtained by this concatenation is a single feature, which, for the operating mode feature extraction model, can still extract an intra-pulse feature vector with contextual information.

[0065] It should be understood that by splicing the carrier frequency sequence and the pulse width sequence, more accurate intra-pulse features can be obtained, thereby improving the accuracy of radar operating mode identification.

[0066] Step 123: Perform differential processing on the arrival time series to obtain the inter-pulse features.

[0067] Here, differential processing can include, but is not limited to, first-order differential and second-order differential, etc. For example, second-order differential is performed on the arrival time series to obtain inter-pulse features; or, first-order differential is performed on the arrival time series, and then second-order differential is performed on the resulting sequence to obtain inter-pulse features. Differential processing of the arrival time series can yield inter-pulse features used to characterize the parameter variation patterns between different radar pulses.

[0068] It should be understood that differential processing of the arrival time series can yield more accurate inter-pulse features, thereby improving the accuracy of radar operating mode identification.

[0069] Step 130: Input the intrapulse features and interpulse features into the working mode feature extraction model to obtain the intrapulse feature vector and interpulse feature vector output by the working mode feature extraction model.

[0070] It should be noted that by inputting intra-pulse and inter-pulse features into the working mode feature extraction model, potential semantic features can be further extracted, that is, inter-pulse feature vectors and intra-pulse feature vectors with contextual information can be extracted. This fully extracts (mines) the implicit features and temporal correlation information within the PDW sequence, that is, extracts deeper grammatical rules and temporal correlations. In other words, more comprehensive and complete features are extracted, thereby improving the accuracy of radar working mode recognition.

[0071] In one specific embodiment, the working mode feature extraction model includes two branches to extract working mode features of intra-pulse features and inter-pulse features, respectively.

[0072] Step 140: Input the feature fusion vector of the intra-pulse feature vector and the inter-pulse feature vector into the radar operating mode recognition model to obtain the radar operating mode recognition result output by the radar operating mode recognition model.

[0073] The radar operating mode recognition method provided in this invention can obtain more accurate intra-pulse features by splicing carrier frequency sequences and pulse width sequences, thereby further improving the recognition accuracy of radar operating modes; and can obtain more accurate inter-pulse features by performing differential processing on arrival time sequences, thereby further improving the recognition accuracy of radar operating modes.

[0074] Based on any of the above embodiments, a specific embodiment of the radar operating mode recognition method is given below. Figure 3 This is the third flowchart of the radar operating mode recognition method provided by the present invention, as shown below. Figure 3 As shown, the radar operating mode identification method includes steps 110, 120, 131, 132 and 140.

[0075] Step 110: Obtain the Pulse Description Word (PDW) sequence of the radar radiation source.

[0076] Here, the PDW sequence includes PDWs at multiple times, and this PDW sequence is the sequence to be identified by radar operating mode.

[0077] Step 120: Based on the PDW sequence, determine the intra-pulse characteristics and inter-pulse characteristics of the radar radiation source.

[0078] It should be noted that the PDW sequence includes temporal information, so the intra-pulse features and inter-pulse features still implicitly contain the temporal information of the PDW sequence. For the working mode feature extraction model, it can still extract intra-pulse feature vectors and inter-pulse feature vectors with contextual information.

[0079] Step 131: Input the intrapulse features into the first long short-term memory network layer in the working mode feature extraction model to obtain the intrapulse feature vector output by the first long short-term memory network layer.

[0080] Here, the first Long Short-Term Memory (LSTM) network layer utilizes the superior expressive power of deep learning networks to extract deep-seated patterns, namely, the implicit features within the PDW sequence, thereby improving the accuracy of radar operating mode recognition. Furthermore, the first LSM network layer is used to extract intra-pulse feature vectors with contextual information based on intra-pulse features, that is, it is used to further extract operating mode features with contextual information, thereby fully extracting (mining) the implicit features and temporal correlation information within the PDW sequence, that is, extracting deeper-level grammatical rules and temporal correlations. In other words, it extracts more comprehensive and complete features, thereby improving the accuracy of radar operating mode recognition.

[0081] Furthermore, the intrapulse features are input into the encoding layer of the working mode feature extraction model to obtain the first encoded vector output by the encoding layer. The first encoded vector is then input into the first long short-term memory network layer of the working mode feature extraction model to obtain the intrapulse feature vector output by the first long short-term memory network layer. This encoding layer can be set according to actual needs; for example, this encoding layer includes 5 one-dimensional convolutional layers.

[0082] Step 132: Input the inter-pulse features into the second long short-term memory network layer in the working mode feature extraction model to obtain the inter-pulse feature vector output by the second long short-term memory network layer.

[0083] Here, the second long short-term memory network layer utilizes the superior expressive power of deep learning networks to extract deep patterns, namely, to extract the implicit features within the PDW sequence, thereby improving the accuracy of radar operating mode recognition. Furthermore, the second long short-term memory network layer is used to extract inter-pulse feature vectors with contextual information based on inter-pulse features, that is, it is used to further extract operating mode features with contextual information, thereby fully extracting (mining) the implicit features and temporal correlation information within the PDW sequence, that is, extracting deeper grammatical rules and temporal correlation relationships. In other words, it extracts more comprehensive and complete features, thereby improving the accuracy of radar operating mode recognition.

[0084] Furthermore, the inter-pulse features are input into the encoding layer of the working mode feature extraction model to obtain the second encoding vector output by the encoding layer. The second encoding vector is then input into the second long short-term memory network layer of the working mode feature extraction model to obtain the inter-pulse feature vector output by the second long short-term memory network layer. This encoding layer can be set according to actual needs; for example, this encoding layer includes 5 one-dimensional convolutional layers.

[0085] The working mode feature extraction model is obtained through unsupervised training based on sample PDW sequences. These sample PDW sequences serve as training samples and can be referenced from the aforementioned PDW sequences. Unsupervised training methods may include, but are not limited to, contrastive learning, generative adversarial learning, deep clustering, etc.

[0086] It should be understood that the working mode feature extraction model is obtained through unsupervised training, thus eliminating the need for labeling and reducing labeling costs. Furthermore, it can fully utilize massive amounts of unlabeled sample PDW sequences for model training, thereby effectively improving model performance and ultimately enhancing the accuracy of radar working mode recognition.

[0087] Step 140: Input the feature fusion vector of the intra-pulse feature vector and the inter-pulse feature vector into the radar operating mode recognition model to obtain the radar operating mode recognition result output by the radar operating mode recognition model.

[0088] The radar operating mode recognition method provided in this invention further extracts features from intra-pulse and inter-pulse features through a first long short-term memory (LSTM) network layer and a second LSM network layer, respectively. This allows for the extraction of contextual information from features within and between radar pulses, ensuring more comprehensive extraction of operating mode features and thus improving the accuracy of radar operating mode recognition. Furthermore, the operating mode feature extraction model is obtained through unsupervised training based on sample PDW sequences, thereby reducing labeling costs and fully utilizing a large number of unlabeled sample PDW sequences for model training, thus improving the robustness of the operating mode feature extraction model and ultimately enhancing the accuracy of radar operating mode recognition.

[0089] Based on any of the above embodiments, a specific embodiment of the radar operating mode recognition method is given below. The operating mode feature extraction model is a bidirectional contrastive predictive coding model, which is used to extract essential features.

[0090] Here, the bidirectional contrastive predictive coding model is constructed based on the BiCPC (Bidirectional Contrastive Predictive Coding) method to extract the latent semantic features of the PDW sequence.

[0091] The BiCPC method is used to construct the working mode feature extraction model because it is a self-supervised learning method that combines bidirectional context modeling and contrastive predictive coding (CPC). It aims to learn richer sequence data representations by simultaneously utilizing past and future contextual information. Moreover, it is a high-level representation extraction technique based on unsupervised learning, which has achieved excellent performance in multiple fields such as text, image, and speech, thereby improving the accuracy of radar working mode recognition.

[0092] For example, the working mode feature extraction model is trained in the following manner: The first step is to determine the intra-pulse features and inter-pulse features of the sample at time t based on the sample PDW sequence at time t. The second step is to input the intra-pulse features and inter-pulse features of the sample at time t into the encoding layer of the working mode feature extraction model to obtain the first encoding vector and the second encoding vector of the sample at time t output by the encoding layer. The third step is to input the first and second encoding vectors of the sample at time t into the LSTM layer of the working mode feature extraction model to obtain the intra-pulse feature vector and inter-pulse feature vector of the sample at time t output by the LSTM layer. The LSTM layer determines the intra-pulse feature vector and inter-pulse feature vector of the sample at time t based on the encoding vector at time t and the state output vector of the previous time t-1. The fourth step is to predict the encoding vectors for the next n times based on the intra-pulse feature vectors and inter-pulse feature vectors of the samples at time t, that is, to obtain the first encoding vector and the second encoding vector of the samples from time t+1 to time t+n; the prediction can be performed through a linear layer. The fifth step is to calculate the similarity between the predicted first and second encoding vectors of the samples and the actual first and second encoding vectors of the samples. The sixth step is to determine the loss value based on the similarity calculation results, and then train the working mode feature extraction model based on the loss value.

[0093] In other words, let the set of sample PDW sequences in a single training batch be {X(i)}, and the number of samples in this set of sample PDW sequences be M. The training objective (training criterion) of contrastive learning is to predict the similarity score of the encoding vector of sample PDW sequence X(i) in the next n time steps by using the intra-pulse feature vector and inter-pulse feature vector of sample PDW sequence X(i) to get higher and higher, while predicting the similarity score of the encoding vector of other sample PDW sequences in the next n time steps to get lower and lower.

[0094] For example, the loss function of the working pattern feature extraction model is as follows: ; In the formula, This indicates the number of future moments (the predicted number of future time steps). It is a positive integer greater than 0; Represents a similarity calculation function (such as a cosine similarity calculation function); Represents the future time of the sample PDW sequence X(i). The predicted encoding vector, Represents the future time of the sample PDW sequence X(i). The actual encoded vector; This represents the set of sample PDW sequences for a single training batch.

[0095] It should be noted that the bidirectional contrastive predictive coding model uses the coding vectors from the current and past times to predict the coding vectors from the future. This effectively extracts a stable, essential feature (baseline feature) from the PDW sequence that exists throughout the entire signal phase, thereby improving the accuracy of radar operating mode recognition. Furthermore, through the training criterion of the aforementioned contrastive learning—that the current sample can only predict the coding vector of its future time, but not the coding vectors of other samples in the same training batch—it can extract a discriminative information from the PDW sequence. Combined with the extraction of the essential feature (baseline feature) of the PDW sequence data, a highly discriminative operating mode feature is extracted, thus improving the accuracy of radar operating mode recognition.

[0096] The radar operating mode recognition method provided in this invention uses a bidirectional contrastive predictive coding model for operating mode feature extraction. This model uses the coding vectors from the current and past times to predict the coding vectors from the future times. This effectively extracts a stable and essential feature from the PDW sequence that exists throughout the entire signal phase, thereby improving the feature extraction capability of the operating mode feature extraction model and thus improving the accuracy of radar operating mode recognition. Furthermore, the bidirectional contrastive predictive coding model can extract a highly discriminative operating mode feature from the PDW sequence, further enhancing the accuracy of radar operating mode recognition.

[0097] Based on any of the above embodiments, another embodiment of the radar operating mode recognition method is given below. Figure 4 This is the fourth flowchart of the radar operating mode recognition method provided by the present invention, as shown below. Figure 4 As shown, the radar operating mode recognition method includes steps 110, 120, 130, 140, 150, and 160. Step 150 follows step 110, and step 160 follows step 140.

[0098] Step 150: Input the PDW sequence into the radar operating mode detection model to obtain the radar operating mode detection result output by the radar operating mode detection model.

[0099] The radar operating mode detection model is used to detect radar operating modes based on PDW sequences.

[0100] In one specific embodiment, the radar operating mode detection model includes a machine learning multi-classifier, thereby eliminating the need for template matching and improving the accuracy of radar operating mode recognition. This radar operating mode detection model can be constructed based on algorithms such as Support Vector Machine (SVM), Random Forest, and Decision Tree.

[0101] It should be noted that traditional machine learning pattern recognition methods are based on manually defined physical features, which have good interpretability. Compared with deep learning methods (i.e., the way radar operating mode recognition results are obtained as described above), machine learning can also be effective for operating modes with a small number of samples, so it is not completely unusable. It is very suitable as a fallback model for radar operating mode recognition. Therefore, a radar operating mode detection model is also used for radar operating mode detection.

[0102] In one specific embodiment, the radar operating mode detection model is constructed based on the random forest algorithm. The training method of the radar operating mode detection model is as follows: multiple training sets are randomly extracted from the prior sample set, multiple decision tree classifiers are constructed using the training sets, and finally, the operating mode detection result is obtained by identifying and voting on the feature vector (i.e., the parameters in the PDW sequence) through multiple decision tree classifiers.

[0103] In one embodiment, based on the PDW sequence, a carrier frequency sequence, a pulse width sequence, and a pulse repetition period sequence are determined. These sequences are then input into a radar operating mode detection model to obtain the radar operating mode detection result output by the model. It should be understood that using the carrier frequency sequence, pulse width sequence, and pulse repetition period sequence as the identification feature vectors of the radar operating mode detection model can improve the accuracy of radar operating mode detection.

[0104] Here, the radar operating mode detection result is used to indicate the radar operating mode of the radar radiation source. The radar operating mode may include, but is not limited to: velocity search, track plus search, range-while-track search, multi-target tracking, search-while-track, and single-target tracking, etc.

[0105] Step 160: Based on the radar operating mode identification result and the radar operating mode detection result, determine the final radar operating mode identification result.

[0106] Specifically, the radar operating mode identification result and the radar operating mode detection result are combined to obtain the final radar operating mode identification result.

[0107] For example, assuming the total number of radar operating modes is M, the first confidence level of the M radar operating modes is determined based on the radar operating mode identification results, the second confidence level of the M radar operating modes is determined based on the radar operating mode detection results, and the comprehensive confidence level of the M radar operating modes is determined based on the M first confidence levels and the M second confidence levels (the comprehensive confidence level of any radar operating mode is obtained by fusing the first and second confidence levels of that radar operating mode, and the fusing method can be average or weighted average, etc.). Based on the comprehensive confidence level of the M radar operating modes, the final radar operating mode identification result is determined, that is, the radar operating mode with the highest comprehensive confidence level is determined as the final radar operating mode identification result.

[0108] Here, the final radar operating mode identification result is used to indicate the radar operating mode of the radar radiation source. This radar operating mode may include, but is not limited to: velocity search, track plus search, range-while-track search, multi-target tracking, search-while-track, and single-target tracking, etc.

[0109] It should be noted that, through experimental verification, using both the radar operating mode recognition model and the radar operating mode detection model for identification significantly improves the radar operating mode recognition performance compared to using only the radar operating mode recognition model.

[0110] The radar operating mode recognition method provided in this embodiment of the invention also inputs the PDW sequence into the radar operating mode detection model to obtain the radar operating mode detection result output by the radar operating mode detection model. Based on the radar operating mode recognition result and the radar operating mode detection result, the final radar operating mode recognition result is determined comprehensively. Thus, the radar operating mode is recognized in two ways, thereby improving the recognition accuracy of the radar operating mode.

[0111] Based on any of the above embodiments, another embodiment of the radar operating mode recognition method is given below. Figure 5 This is the fifth flowchart of the radar operating mode recognition method provided by the present invention, as shown below. Figure 5 As shown, the radar operating mode recognition method includes steps 110, 120, 130, 140, 151, and 160. Step 151 follows step 110, and step 160 follows step 140.

[0112] Step 151: Input each PDW in the PDW sequence into the radar operating mode detection model to obtain the radar operating mode detection result output by the radar operating mode detection model.

[0113] The radar operating mode detection model is used to detect radar operating modes based on a PDW.

[0114] In one embodiment, any PDW includes a carrier frequency, pulse width, and pulse repetition period. Based on this, the carrier frequency, pulse width, and pulse repetition period are input into the radar operating mode detection model to obtain the radar operating mode detection sub-result output by the radar operating mode detection model. It should be understood that using the carrier frequency, pulse width, and pulse repetition period as the identification feature vector of the radar operating mode detection model can improve the accuracy of radar operating mode detection. Of course, other parameters can also be input into the radar operating mode detection model, for example, the carrier frequency, pulse width, time of arrival, and direction of arrival can be input into the radar operating mode detection model to obtain the radar operating mode detection sub-result output by the radar operating mode detection model.

[0115] The radar operating mode detection result includes the radar operating mode detection sub-result corresponding to each PDW. That is, a PDW is input into the radar operating mode detection model, and a radar operating mode detection sub-result is output by the radar operating mode detection model.

[0116] Here, any radar operating mode detection sub-result can be used to indicate the radar operating mode of the radar radiation source. The radar operating mode may include, but is not limited to: velocity search, track plus search, range-while-track search, multi-target tracking, search-while-track, and single-target tracking, etc.

[0117] It should be noted that by simply inputting a single PDW into the radar operating mode detection model, the radar operating mode detection model still has a strong ability to distinguish operating modes with large differences in radar signal parameters, and it does not require the extraction of time-series correlation information.

[0118] Step 160: Based on the radar operating mode identification result and the radar operating mode detection result, determine the final radar operating mode identification result.

[0119] Specifically, based on the radar operating mode identification results and the detection sub-results of each radar operating mode, a comprehensive decision is made to obtain the final radar operating mode identification result.

[0120] In one specific embodiment, the detection result of each radar operating mode is determined by comprehensively considering the detection results of each radar operating mode, and the final identification result of the radar operating mode is obtained by comprehensively considering the radar operating mode identification result and the detection result of the operating mode.

[0121] For example, assuming the total number of radar operating modes is M and the number of radar operating mode detection sub-results is N, based on each radar operating mode detection sub-result, N*M confidence levels are determined. Based on the N*M confidence levels, the second confidence levels of M radar operating modes are determined. Based on the radar operating mode identification results, the first confidence levels of M radar operating modes are determined. Based on the M first confidence levels and the M second confidence levels, the comprehensive confidence level of the M radar operating modes is determined (the comprehensive confidence level of any radar operating mode is obtained by fusing the first and second confidence levels of that radar operating mode, and the fusing method can be average or weighted average, etc.). Based on the comprehensive confidence level of the M radar operating modes, the final radar operating mode identification result is determined, that is, the radar operating mode with the highest comprehensive confidence level is determined as the final radar operating mode identification result.

[0122] The radar operating mode recognition method provided in this embodiment of the invention only requires inputting a single PDW into the radar operating mode detection model. This radar operating mode detection model still has strong distinguishability for operating modes with large differences in radar signal parameters. Therefore, the radar operating mode detection model can be used as a fallback model for radar operating mode recognition, thereby improving the accuracy of radar operating mode recognition.

[0123] To facilitate understanding of the above embodiments, a specific embodiment will be described here. For example... Figure 6 As shown, in the first branch, based on the PDW sequence, the carrier frequency sequence, pulse width sequence, and arrival time sequence are extracted. Based on the carrier frequency sequence and pulse width sequence, intra-pulse features are determined, and based on the arrival time sequence, inter-pulse features are determined. Then, the intra-pulse features are input into the coding layer and the first long short-term memory network layer in the operating mode feature extraction model to obtain the intra-pulse feature vector output by the first long short-term memory network layer. The inter-pulse features are input into the coding layer and the second long short-term memory network layer in the operating mode feature extraction model to obtain the inter-pulse feature vector output by the second long short-term memory network layer. Then, the feature fusion vector of the intra-pulse feature vector and the inter-pulse feature vector is input into the radar operating mode recognition model to obtain the radar operating mode recognition result output by the radar operating mode recognition model. In the second branch, each PDW in the PDW sequence is input into the radar operating mode detection model to obtain the radar operating mode detection sub-results output by the radar operating mode detection model. Finally, based on the radar operating mode recognition result and the radar operating mode detection sub-results, the final radar operating mode recognition result is determined.

[0124] The intrapulse and interpulse features are input into the working mode feature extraction model to obtain the intrapulse feature vector and interpulse feature vector output by the working mode feature extraction model. This invention addresses the complex scenarios that may occur in electromagnetic environments, namely, the significant impact of a multi-functional radar on pulse signals under a certain operating mode, which exacerbates the problem of parameter overlap, and the low accuracy of radar operating mode recognition due to low signal-to-noise ratio. By employing the above embodiments, the invention effectively improves the accuracy and robustness of radar operating mode recognition in complex electromagnetic environments.

[0125] The radar operating mode recognition device provided by the present invention is described below. The radar operating mode recognition device described below and the radar operating mode recognition method described above can be referred to in correspondence.

[0126] Figure 7 This is a schematic diagram of the radar operating mode recognition device provided by the present invention, as shown below. Figure 7 As shown, the radar operating mode recognition device includes: a sequence acquisition module 710, a feature determination module 720, a feature extraction module 730, and a pattern recognition module 740.

[0127] The sequence acquisition module 710 is used to acquire the pulse descriptor (PDW) sequence of the radar radiation source; the PDW sequence includes PDWs at multiple times.

[0128] The feature determination module 720 is used to determine the intra-pulse features and inter-pulse features of the radar radiation source based on the PDW sequence.

[0129] The feature extraction module 730 is used to input the intrapulse features and the interpulse features into the working mode feature extraction model to obtain the intrapulse feature vector and the interpulse feature vector output by the working mode feature extraction model; the working mode feature extraction model is used to extract intrapulse feature vectors with context information based on the intrapulse features, and the working mode feature extraction model is also used to extract interpulse feature vectors with context information based on the interpulse features.

[0130] The pattern recognition module 740 is used to input the feature fusion vector of the intra-pulse feature vector and the inter-pulse feature vector into the radar operating mode recognition model to obtain the radar operating mode recognition result output by the radar operating mode recognition model; the radar operating mode recognition model is used to perform radar operating mode recognition on the feature fusion vector.

[0131] The radar operating mode recognition device provided in this embodiment of the invention acquires the PDW sequence of the radar radiation source, and the PDW sequence includes PDWs at multiple times. Therefore, even if the PDW at some times contains erroneous parameters, accurate recognition can still be achieved based on the PDWs at multiple times, thereby improving the accuracy of radar operating mode recognition. Based on the PDW sequence, the intra-pulse features and inter-pulse features of the radar radiation source are determined, thereby accurately recognizing the radar operating mode based on the intra-pulse and inter-pulse features, further improving the accuracy of radar operating mode recognition. The intra-pulse and inter-pulse features are input into the operating mode feature extraction model to obtain the intra-pulse feature vector and inter-pulse feature vector output by the operating mode feature extraction model. Based on this, further feature extraction is performed on the intra-pulse and inter-pulse features, respectively. Operating mode feature extraction can be performed on parameters within and between radar pulses, respectively, ensuring more comprehensive extraction of operating mode features, thereby improving the accuracy of radar operating mode recognition. The working mode feature extraction model is used to extract intra-pulse feature vectors with contextual information based on intra-pulse features. The working mode feature extraction model is also used to extract inter-pulse feature vectors with contextual information based on inter-pulse features, thereby fully extracting the implicit features and temporal correlation information within the PDW sequence, thus improving the accuracy of radar working mode recognition. Based on the input features and feature extraction methods determined above, the feature fusion vector of intra-pulse and inter-pulse feature vectors is input into the radar working mode recognition model, which can obtain accurate radar working mode recognition results. Furthermore, by first obtaining intra-pulse and inter-pulse features based on the PDW sequence, and then extracting working mode features from the intra-pulse and inter-pulse features to obtain intra-pulse and inter-pulse feature vectors with contextual information, high-precision recognition of radar working modes can be achieved even in complex electromagnetic environments such as incomplete or overlapping radar signal parameters, and high-precision recognition can be achieved regardless of the scenario. In summary, this invention can achieve high-precision recognition of radar working modes for various scenarios, that is, this invention can improve the accuracy of radar working mode recognition and enhance the adaptability of radar working modes.

[0132] Based on any of the above embodiments, any PDW includes a carrier frequency, pulse width, and arrival time; the feature determination module 720 is specifically used for: The carrier frequency, pulse width, and arrival time of the PDW sequence are extracted respectively to obtain the carrier frequency sequence, pulse width sequence, and arrival time sequence; The intra-pulse feature is obtained by concatenating the carrier frequency sequence and the pulse width sequence. The arrival time series is differentially processed to obtain the inter-pulse features.

[0133] Based on any of the above embodiments, the feature extraction module 730 is specifically used for: The intrapulse features are input into the first long short-term memory network layer in the working mode feature extraction model to obtain the intrapulse feature vector output by the first long short-term memory network layer. The inter-pulse features are input into the second long short-term memory network layer in the working mode feature extraction model to obtain the inter-pulse feature vector output by the second long short-term memory network layer. The working mode feature extraction model is obtained by unsupervised training based on sample PDW sequences.

[0134] Based on any of the above embodiments, the working mode feature extraction model is a bidirectional contrastive predictive coding model, and the working mode feature extraction model is used to extract essential features.

[0135] Based on any of the above embodiments, the device further includes: a pattern detection module and a result determination module.

[0136] The mode detection module is used to input the PDW sequence into the radar operating mode detection model to obtain the radar operating mode detection result output by the radar operating mode detection model; the radar operating mode detection model is used to perform radar operating mode detection based on the PDW sequence.

[0137] The result determination module is used to determine the final radar operating mode identification result based on the radar operating mode identification result and the radar operating mode detection result.

[0138] Based on any of the above embodiments, the pattern detection module is specifically used for: Each PDW in the PDW sequence is input into the radar operating mode detection model to obtain the radar operating mode detection result output by the radar operating mode detection model.

[0139] The radar operating mode detection model is used to detect radar operating modes based on a PDW, and the radar operating mode detection result includes the radar operating mode detection sub-results corresponding to each PDW.

[0140] Figure 8 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 8As shown, the electronic device may include: a processor 810, a communications interface 820, a memory 830, and a communications bus 840, wherein the processor 810, the communications interface 820, and the memory 830 communicate with each other through the communications bus 840. The processor 810 can call logical instructions in the memory 830 to execute a radar operating mode recognition method. This method includes: acquiring a pulse descriptor word (PDW) sequence of a radar radiation source; the PDW sequence includes PDWs at multiple time points; determining intra-pulse features and inter-pulse features of the radar radiation source based on the PDW sequence; inputting the intra-pulse features and inter-pulse features into an operating mode feature extraction model to obtain intra-pulse feature vectors and inter-pulse feature vectors output by the operating mode feature extraction model; the operating mode feature extraction model is used to extract intra-pulse feature vectors with contextual information based on the intra-pulse features, and is also used to extract inter-pulse feature vectors with contextual information based on the inter-pulse features; inputting a feature fusion vector of the intra-pulse feature vectors and the inter-pulse feature vectors into a radar operating mode recognition model to obtain a radar operating mode recognition result output by the radar operating mode recognition model; the radar operating mode recognition model is used to perform radar operating mode recognition on the feature fusion vector.

[0141] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0142] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the radar operating mode recognition method provided by the above methods. The method includes: acquiring a pulse descriptor word (PDW) sequence of a radar radiation source; the PDW sequence includes PDWs at multiple times; determining intra-pulse features and inter-pulse features of the radar radiation source based on the PDW sequence; inputting the intra-pulse features and the inter-pulse features into an operating mode feature extraction model to obtain intra-pulse feature vectors and inter-pulse feature vectors output by the operating mode feature extraction model; the operating mode feature extraction model is used to extract intra-pulse feature vectors with contextual information based on the intra-pulse features, and the operating mode feature extraction model is also used to extract inter-pulse feature vectors with contextual information based on the inter-pulse features; inputting a feature fusion vector of the intra-pulse feature vectors and the inter-pulse feature vectors into a radar operating mode recognition model to obtain a radar operating mode recognition result output by the radar operating mode recognition model; the radar operating mode recognition model is used to perform radar operating mode recognition on the feature fusion vectors.

[0143] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the radar operating mode recognition method provided by the above methods. The method includes: acquiring a pulse descriptor word (PDW) sequence of a radar radiation source; the PDW sequence including multiple PDWs at different times; determining intra-pulse features and inter-pulse features of the radar radiation source based on the PDW sequence; inputting the intra-pulse features and the inter-pulse features into an operating mode feature extraction model to obtain intra-pulse feature vectors and inter-pulse feature vectors output by the operating mode feature extraction model; the operating mode feature extraction model is used to extract intra-pulse feature vectors with contextual information based on the intra-pulse features, and the operating mode feature extraction model is also used to extract inter-pulse feature vectors with contextual information based on the inter-pulse features; inputting a feature fusion vector of the intra-pulse feature vectors and the inter-pulse feature vectors into a radar operating mode recognition model to obtain a radar operating mode recognition result output by the radar operating mode recognition model; the radar operating mode recognition model is used to perform radar operating mode recognition on the feature fusion vector.

[0144] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0145] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A radar operating mode recognition method, characterized in that, include: Obtain the pulse descriptor word (PDW) sequence of the radar radiation source; the PDW sequence includes PDWs at multiple times. Based on the PDW sequence, the intra-pulse characteristics and inter-pulse characteristics of the radar radiation source are determined. The intrapulse features and interpulse features are input into the working mode feature extraction model to obtain the intrapulse feature vector and interpulse feature vector output by the working mode feature extraction model; the working mode feature extraction model is used to extract intrapulse feature vectors with contextual information based on the intrapulse features, and the working mode feature extraction model is also used to extract interpulse feature vectors with contextual information based on the interpulse features; The feature fusion vector of the intra-pulse feature vector and the inter-pulse feature vector is input into the radar operating mode recognition model to obtain the radar operating mode recognition result output by the radar operating mode recognition model; the radar operating mode recognition model is used to perform radar operating mode recognition on the feature fusion vector.

2. The radar operating mode recognition method according to claim 1, characterized in that, Any of the PDWs described includes a carrier frequency, pulse width, and time of arrival; The step of determining the intra-pulse characteristics and inter-pulse characteristics of the radar radiation source based on the PDW sequence includes: The carrier frequency, pulse width, and arrival time of the PDW sequence are extracted respectively to obtain the carrier frequency sequence, pulse width sequence, and arrival time sequence; The intra-pulse feature is obtained by concatenating the carrier frequency sequence and the pulse width sequence. The arrival time series is differentially processed to obtain the inter-pulse features.

3. The radar operating mode recognition method according to claim 1, characterized in that, The step of inputting the intrapulse features and the interpulse features into the working mode feature extraction model to obtain the intrapulse feature vector and interpulse feature vector output by the working mode feature extraction model includes: The intrapulse features are input into the first long short-term memory network layer in the working mode feature extraction model to obtain the intrapulse feature vector output by the first long short-term memory network layer. The inter-pulse features are input into the second long short-term memory network layer in the working mode feature extraction model to obtain the inter-pulse feature vector output by the second long short-term memory network layer. The working mode feature extraction model is obtained by unsupervised training based on sample PDW sequences.

4. The radar operating mode recognition method according to claim 3, characterized in that, The working mode feature extraction model is a bidirectional contrastive predictive coding model, which is used to extract essential features.

5. The radar operating mode recognition method according to any one of claims 1 to 4, characterized in that, After acquiring the pulse descriptor word (PDW) sequence of the radar radiation source, the method further includes: The PDW sequence is input into the radar operating mode detection model to obtain the radar operating mode detection result output by the radar operating mode detection model; the radar operating mode detection model is used to detect radar operating modes based on the PDW sequence. After inputting the feature fusion vector of the intra-pulse feature vector and the inter-pulse feature vector into the radar operating mode recognition model to obtain the radar operating mode recognition result output by the radar operating mode recognition model, the method further includes: Based on the radar operating mode identification result and the radar operating mode detection result, the final radar operating mode identification result is determined.

6. The radar operating mode recognition method according to claim 5, characterized in that, The step of inputting the PDW sequence into the radar operating mode detection model to obtain the radar operating mode detection result output by the radar operating mode detection model includes: Each PDW in the PDW sequence is input into the radar operating mode detection model to obtain the radar operating mode detection result output by the radar operating mode detection model. The radar operating mode detection model is used to detect radar operating modes based on a PDW, and the radar operating mode detection result includes the radar operating mode detection sub-results corresponding to each PDW.

7. A radar operating mode recognition device, characterized in that, include: A sequence acquisition module is used to acquire the pulse descriptor word (PDW) sequence of the radar radiation source; the PDW sequence includes PDWs at multiple times. The feature determination module is used to determine the intra-pulse features and inter-pulse features of the radar radiation source based on the PDW sequence. The feature extraction module is used to input the intrapulse features and the interpulse features into the working mode feature extraction model to obtain the intrapulse feature vector and the interpulse feature vector output by the working mode feature extraction model; the working mode feature extraction model is used to extract intrapulse feature vectors with context information based on the intrapulse features, and the working mode feature extraction model is also used to extract interpulse feature vectors with context information based on the interpulse features; The pattern recognition module is used to input the feature fusion vector of the intra-pulse feature vector and the inter-pulse feature vector into the radar operating mode recognition model to obtain the radar operating mode recognition result output by the radar operating mode recognition model; the radar operating mode recognition model is used to perform radar operating mode recognition on the feature fusion vector.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the radar operating mode recognition method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the radar operating mode recognition method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the radar operating mode recognition method as described in any one of claims 1 to 6.

Images