Pulse stream deinterleaving method, apparatus, device, and storage medium
Patent Information
- Application Number
- CN202411190163.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-28
- Publication Date
- 2026-09-04
- Estimated Expiration
- 2044-08-28
AI Technical Summary
[0003]然而，如今高密度脉冲数据环境下，脉冲重叠和脉冲丢失现象严重，常规序列信息利用的脉冲流去交织方法在搜索阶段时易失效、正确率低，严重影响电子侦查系统的效果
（1）本发明通过对重叠脉冲的处理恢复特定脉冲流中因脉冲重叠而丢失的特征，然后利用LSTM来挖掘和学习脉冲流中的长期时间模式，跟踪特定脉冲流的变化特性，从而实现该脉冲的有效分离，在高脉冲丢失、重叠概率下能取得较高的分选正确率。
Smart Images
Figure CN119202710B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of signal sorting technology, and particularly relates to pulse stream deinterleaving methods, apparatus, devices and storage media. Background Technology
[0002] The pulse stream deinterleaving method based on sequence information plays an important role in the main sorting process and can effectively sort out radar individuals.
[0003] However, in today's high-density pulse data environment, pulse overlap and pulse loss are serious problems. The pulse stream deinterleaving method used for conventional sequence information is prone to failure and has low accuracy during the search phase, which seriously affects the effectiveness of electronic reconnaissance systems. Summary of the Invention
[0004] The purpose of this invention is to overcome the shortcomings of the prior art by providing a pulse stream deinterleaving method, apparatus, device, and storage medium. By processing overlapping pulses, the invention recovers the features lost due to pulse overlap in a specific pulse stream, and then uses LSTM to mine and learn the long-term time pattern in the pulse stream to track the changing characteristics of the specific pulse stream, thereby achieving effective separation of the pulse.
[0005] The objective of this invention is achieved through the following technical solution: A pulse stream deinterleaving method, the method comprising: For each type of pulse flow, the LSTM prediction model is trained bidirectionally to obtain the forward LSTM prediction model and the backward LSTM prediction model for each type of pulse flow. The trained bidirectional LSTM model is used for deinterleaving. The possible features of overlapping pulses are calculated through stack operations. Then, the prediction and matching of each pulse is started from the initial prediction pulse to separate the same pulse before and after the initial prediction pulse.
[0006] Furthermore, the method also includes: Each type of radar pulse stream is represented by a continuous two-dimensional vector, which includes the PRI and PW characteristics of the pulse. Furthermore, the bidirectional training of the LSTM prediction model for each type of pulse flow specifically includes: The input pulse stream is segmented into segments of a preset length to obtain several samples. The feature of the next pulse is a label of the feature of the previous pulse. For each type of pulse data, train by traversing all samples in the training set to obtain the forward LSTM prediction model corresponding to each type of pulse data; For each type of pulse data, starting from the termination time, calculate each pulse in reverse order. PRI Features, used as the training set for inverse pulse data; The inverse LSTM prediction model is trained by iterating through all samples in the inverse pulse data training set to obtain the inverse LSTM prediction model corresponding to each type of pulse data.
[0007] Furthermore, the method for selecting the initial prediction pulse specifically includes: Calculate the potential characteristics of each pulse; The latent features are input into the trained bidirectional LSTM model to calculate the matching error between the real pulse and the predicted pulse. The forward matching error of each pulse is obtained by forward LSTM, and the forward sequence number of each pulse is obtained by sorting them from smallest to largest. The pulse with the smallest forward sequence number is taken as the starting pulse.
[0008] Furthermore, the method also includes: If a pulse is lost, the predicted value is used to replace the lost pulse feature, and an error threshold is set to determine whether the next pulse is lost. If all the matching errors calculated within the matching range are greater than the error threshold, then the next pulse is considered lost.
[0009] On the other hand, the present invention also provides a pulsed stream deinterleaving device, the device comprising: The model training module performs bidirectional training of the LSTM prediction model for each type of pulse flow to obtain the forward LSTM prediction model and the reverse LSTM prediction model for each type of pulse flow. The pulse separation module uses a trained bidirectional LSTM model for deinterleaving, calculates the possible features of overlapping pulses through stack operations, and selects a starting prediction pulse to predict and match pulses one by one, separating the same pulses before and after the starting prediction pulse.
[0010] On the other hand, the present invention also provides a computer device, which includes a processor and a memory, wherein the memory stores a computer program, and the computer program is loaded and executed by the processor to implement any of the pulse flow deinterleaving methods described above.
[0011] On the other hand, the present invention also provides a computer-readable storage medium storing a computer program, which is loaded and executed by a processor to implement any of the pulse stream deinterleaving methods described above.
[0012] The beneficial effects of this invention are as follows: (1) This invention recovers the features lost due to pulse overlap in a specific pulse stream by processing overlapping pulses, and then uses LSTM to mine and learn the long-term time pattern in the pulse stream to track the changing characteristics of the specific pulse stream, thereby achieving effective separation of the pulse. It can achieve a high sorting accuracy under high pulse loss and overlap probability.
[0013] (2) Compared with traditional methods, the present invention can reduce the probability of pulse loss by retaining overlapping pulses and performing relevant processing to restore the characteristics of the target pulse lost due to pulse overlap, thereby effectively reducing the probability of pulse loss.
[0014] (3) This invention utilizes LSTM to mine and learn long-term time patterns in the target pulse stream, enabling it to adaptively approach the target pulse stream pattern during the sorting process and gradually complete self-correction when misselection occurs due to pulse loss. Attached Figure Description
[0015] Figure 1 This is a flowchart of the pulse stream deinterleaving method according to an embodiment of the present invention; Figure 2 This is a schematic diagram representing the pulse flow according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the LSTM prediction network training framework according to an embodiment of the present invention; Figure 4 This is a stack diagram of TOA and TOD in an embodiment of the present invention; Figure 5 This is a schematic diagram of the starting pulse selection in an embodiment of the present invention; Figure 6 This is a schematic diagram of forward and reverse search matching according to an embodiment of the present invention; Figure 7 This is a structural block diagram of the pulse flow deinterleaving device according to an embodiment of the present invention. Detailed Implementation
[0016] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.
[0017] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.
[0018] In high-density pulse data environments, pulse overlap and pulse loss are severe. The pulse stream deinterleaving method used for conventional sequence information is prone to failure and has low accuracy during the search phase, which seriously affects the effectiveness of electronic reconnaissance systems.
[0019] To address the aforementioned technical problems, the following embodiments of the pulse stream deinterleaving method, apparatus, device, and storage medium of the present invention are proposed.
[0020] Example 1 This embodiment provides a pulse stream deinterleaving method based on a bidirectional LSTM prediction model. In this embodiment, each pulse is first processed into a two-dimensional vector composed of PRI features and PW features. In this characterization, each type of radar pulse flow can be represented by a series of continuous two-dimensional vectors. (Refer to...) Figure 2 ,like Figure 2 The diagram shown is a schematic representation of the pulse flow in this embodiment.
[0021] Reference Figure 1 ,like Figure 1 The diagram shown is a flowchart of the pulse flow deinterleaving method in this embodiment. The method specifically includes the following steps: Step 1: Train the LSTM prediction model bidirectionally for each type of pulse flow. (Refer to...) Figure 3 ,like Figure 3 The diagram shown is a schematic of the LSTM prediction network training framework in this embodiment. Figure 3 The image shows a typical many-to-many prediction scenario using LSTM. The specific training method is as follows: Step 1.1: Assume the original pure pulse feature sequence is as follows: , the elements The data was standardized using the classic z-score standardization method, and the standardized result is X={X1,X2,…,X...} N},in The corresponding formula is as follows: ,
[0022] ,
[0023] Step 1.2: Divide the input pulse stream into segments of length L to obtain several samples. The first sample is... The label corresponding to this sample is That is, the feature of the next pulse is a label of the feature of the previous pulse, which reflects the mechanism of single-step prediction.
[0024] Step 1.3: For each type of pulse data, traverse all samples in the training set to train fully and obtain the forward LSTM prediction model corresponding to each type of pulse data.
[0025] Step 1.4: For each type of pulse data, starting from the end time, calculate the pulse value in reverse order for each pulse. PRIThe features are used as the training set for the inverse pulse data. Following the same steps as above, the inverse LSTM prediction model corresponding to each type of pulse data can be obtained.
[0026] Step 2: Perform impulse deinterleaving using the trained bidirectional LSTM model. This includes the following sub-steps: Step 2.1: Stacking of overlapping pulses.
[0027] by The received aliased and interleaved pulse stream sequence is described. Stacking operations are performed on overlapping pulses during the testing phase to calculate the possible characteristics of the overlapping pulses, referring to... Figure 4 ,like Figure 4 The diagram shown is a stacked schematic of TOA and TOD in this embodiment. The representation vector of the overlapping pulse consists of four parts:
[0028] Step 2.2: Selection of the starting pulse.
[0029] Reference Figure 5 ,like Figure 5 The diagram shown is a schematic diagram of the starting pulse selection in this embodiment. Figure 5 This demonstrates the selection of the starting pulse in a segment of aliased and interleaved pulse streams. It is assumed that the forward LSTM and backward LSTM are trained using the pulse stream corresponding to the first pulse in the diagram (hereinafter referred to as pulse stream A). The specific steps for selecting the starting pulse are as follows: Step 2.2.1: Calculate the potential characteristics of each pulse.
[0030] First, calculate the range of the previous or next pulse of the same type. Kmin~Kmax :
[0031]
[0032]
[0033] In the formula, This indicates the maximum TOA value in the received pulse stream. This represents the number of pulses in the test set. This indicates the density of the pulse current. It is the floor function. It is the floor function. and These represent the minimum and maximum PRI values of pulse current A, respectively.
[0034] On Kmin~Kmax Within a pulse range, calculate the potential PRI characteristics of the current pulse, in order to Figure 4For example, the latent features of the current pulse can be represented as a matrix:
[0035] If above Kmin~Kmax If overlapping pulses exist within a given pulse range, the possible PRIs are calculated based on all TOAs in the overlapping pulse TOA stack to represent the potential characteristics of the current pulse. This process is repeated for all pulses in the pulse stream to obtain the potential characteristics of each pulse.
[0036] It should be noted that this embodiment will begin with... Kmax One pulse was discarded because the potential PRI characteristics could not be calculated.
[0037] Step 2.2.2: Calculate the matching error of the predicted pulse.
[0038] Inputting all possible features of the current pulse into the trained LSTM network yields the possible predicted pulse matrix:
[0039] In the next Kmin~Kmax Within each pulse range, the matching error between the actual pulse and the possible predicted pulse matrix is calculated. The formula for calculating the matching error is as follows:
[0040] In the formula and The coefficients can represent feature accuracy coefficients, which represent the prediction accuracy of PRI and PW features respectively. They can be adjusted according to the actual situation, and are set to 0.5 here.
[0041] If in the next Kmin~Kmax If overlapping pulses exist within a given pulse range, the matching error is calculated based on all possible characteristics of the overlapping pulses. The minimum matching error among all matching errors for the current pulse is taken as the matching error standard for that pulse. Then, the matching error criteria of all pulses are sorted.
[0042] Step 2.2.3: Take the pulse with the smallest sequence number as the starting pulse.
[0043] The forward matching error standard for each pulse is obtained through forward LSTM. After sorting from smallest to largest, the forward sequence number of each pulse is obtained. Similarly, the inverse matching error standard for each pulse is obtained through inverse LSTM. After sorting, the reverse sequence number of each pulse is obtained. Therefore, the sequence number of each pulse is represented as... The pulse with the smallest sequence number I is taken as the starting pulse, which is the pulse of pulse stream A. This pulse is used as a springboard to perform prediction matching forward and backward.
[0044] Step 2.3: Pulse-by-pulse prediction and matching.
[0045] Reference Figure 6 ,like Figure 6 The diagram illustrates the forward and reverse search matching in this embodiment. The features of the initial pulse are input into the forward LSTM network to obtain the predicted features of the next pulse. The next pulse is then calculated according to the formula in step 2.2.2. Kmin~ Kmax The matching error of each pulse is calculated, and the pulse corresponding to the smallest matching error value is the next A pulse. Then, all selected A pulse features are continuously input into the forward LSTM network to obtain subsequent prediction features. This process is repeated to select the A pulses after the initial pulse in the received pulse stream. Similarly, the features of the initial pulse are input into the reverse LSTM network for reverse prediction and matching to select all A pulses before the initial pulse. In this way, the A pulse stream is separated from the aliased and interleaved pulse stream.
[0046] When encountering overlapping pulses, all possible features of the overlapping pulses are treated as individual pulses for matching and updating. If a pulse is lost due to other reasons, the predicted value is used to replace the lost pulse features. An error threshold is set. To determine whether the next pulse is lost, all matching errors calculated within the matching range are checked; if all of them exceed the error threshold... If the next pulse is lost, the predicted value is used to replace the lost value.
[0047] This embodiment recovers features lost due to pulse overlap in a specific pulse stream by processing overlapping pulses. Then, it utilizes LSTM to mine and learn long-term temporal patterns in the pulse stream, tracking the changing characteristics of the specific pulse stream, thereby achieving effective pulse separation. This method has two advantages over traditional methods: First, it reduces the probability of pulse loss by preserving overlapping pulses and processing them to recover features lost due to pulse overlap. Second, it leverages the powerful adaptive properties of deep learning. Using LSTM to mine and learn long-term temporal patterns in the target pulse stream allows the sorting process to adaptively align with the target pulse stream pattern, gradually self-correcting when misselection occurs due to pulse loss. Therefore, this method achieves a high sorting accuracy even with high pulse loss and overlap probabilities.
[0048] Example 2 Reference Figure 7 ,like Figure 7The diagram shown is a structural block diagram of the pulse flow deinterleaving device in this embodiment. The device specifically includes the following structures: The model training module performs bidirectional training of the LSTM prediction model for each type of pulse flow, resulting in a forward LSTM prediction model and a backward LSTM prediction model for each type of pulse flow. The pulse separation module uses a trained bidirectional LSTM model for deinterleaving, calculates the possible features of overlapping pulses through stack operations, and selects the starting prediction pulse to predict and match each pulse one by one, separating the same pulse before and after the starting prediction pulse.
[0049] Example 3 This preferred embodiment provides a computer device that can implement the steps of any embodiment of the pulse stream deinterleaving method provided in this application. Therefore, it can achieve the beneficial effects of the pulse stream deinterleaving method provided in this application. For details, please refer to the previous embodiments, which will not be repeated here.
[0050] Example 4 Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. Therefore, embodiments of the present invention provide a storage medium storing multiple instructions that can be loaded by a processor to execute the steps of any embodiment of the pulse stream deinterleaving method provided by the present invention.
[0051] The storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.
[0052] Since the instructions stored in the storage medium can execute the steps in any of the pulse stream deinterleaving method embodiments provided in this invention, the beneficial effects that any of the pulse stream deinterleaving methods provided in this invention can achieve can be realized, as detailed in the preceding embodiments, and will not be repeated here.
[0053] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A pulse stream deinterleaving method, characterized in that, The method includes: For each type of pulse flow, the LSTM prediction model is trained bidirectionally to obtain the forward LSTM prediction model and the backward LSTM prediction model for each type of pulse flow. The trained bidirectional LSTM model is used for deinterleaving. The possible features of overlapping pulses are calculated through stack operations. The prediction and matching of each pulse is started from the initial prediction pulse. The same pulse before and after the initial prediction pulse is separated. The bidirectional training of the LSTM prediction model for each type of pulse flow specifically includes: The input pulse stream is segmented into segments of a preset length to obtain several samples. The feature of the next pulse is a label of the feature of the previous pulse. For each type of pulse data, train by traversing all samples in the training set to obtain the forward LSTM prediction model corresponding to each type of pulse data; For each type of pulse data, starting from the termination time, calculate each pulse in reverse order. PRI Features, used as the training set for inverse pulse data; Train all samples in the inverse pulse data training set to obtain the inverse LSTM prediction model corresponding to each type of pulse data. The method for selecting the initial prediction pulse specifically includes: Calculate the potential characteristics of each pulse; The latent features are input into the trained bidirectional LSTM model to calculate the matching error between the real pulse and the predicted pulse. The forward matching error of each pulse is obtained by forward LSTM, and the forward sequence number of each pulse is obtained by sorting them from smallest to largest. The pulse with the smallest forward sequence number is taken as the starting pulse.
2. The pulse stream deinterleaving method as described in claim 1, characterized in that, The method further includes: Each type of radar pulse stream is represented by a continuous two-dimensional vector, which includes the PRI and PW characteristics of the pulse.
3. The pulse stream deinterleaving method as described in claim 1, characterized in that, The method further includes: If a pulse is lost, the predicted value is used to replace the lost pulse feature, and an error threshold is set to determine whether the next pulse is lost. If all the matching errors calculated within the matching range are greater than the error threshold, then the next pulse is considered lost.
4. A pulse stream deinterleaving device for implementing the pulse stream deinterleaving method according to any one of claims 1 to 3, characterized in that, The device includes: The model training module performs bidirectional training of the LSTM prediction model for each type of pulse flow to obtain the forward LSTM prediction model and the reverse LSTM prediction model for each type of pulse flow. The pulse separation module uses a trained bidirectional LSTM model for deinterleaving, calculates the possible features of overlapping pulses through stack operations, and selects a starting prediction pulse to predict and match pulses one by one, separating the same pulses before and after the starting prediction pulse.
5. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing a computer program, which is loaded and executed by the processor to implement the pulse stream deinterleaving method as described in any one of claims 1-3.
6. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which is loaded and executed by a processor to implement the pulse stream deinterleaving method as described in any one of claims 1-3.