Radar real-time data transmission processing classification identification method and system

Abstract

The application relates to the technical field of data processing, and discloses a radar real-time data transmission processing classification identification method and system.The method comprises the following steps: performing track mode identification on narrowband data of a target radar to obtain a real-time motion mode category of the narrowband data; performing directional region selection on synchronous wideband echo data of the target radar based on the real-time motion mode category to obtain a signal segment of the synchronous wideband echo data; performing sparse decomposition on the signal segment to obtain compact features of the signal segment; performing motion mode compatibility quantification on the compact features based on the real-time motion mode category to obtain category scores of the compact features; performing feedback adjustment on the real-time motion mode category based on the category scores and the compact features to obtain an optimized motion mode of the real-time motion mode category; and performing confidence fusion judgment on the optimized motion mode to obtain a classification identification of the real-time motion mode category; and the application can improve the efficiency of radar real-time data transmission processing classification identification.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for real-time data transmission processing, classification and identification of radar. Background Technology

[0002] In the field of radar real-time data transmission, processing, classification, and recognition, existing technologies generally suffer from a disconnect between narrowband and broadband data processing stages, making it difficult to balance real-time performance and accuracy in classification and recognition. Narrowband data track pattern identification often employs fixed filtering and clustering algorithms, which cannot adaptively match the dynamic motion characteristics of targets. It is also susceptible to clutter interference, generating false tracks, which in turn leads to a lack of precise guidance in subsequent broadband echo data processing. This results in indiscriminate, global analysis, increasing redundancy in data transmission and processing, and reducing the effectiveness of feature extraction due to interference from invalid signals, making it difficult to meet the real-time classification requirements in complex scenarios.

[0003] Existing technologies have significant shortcomings in the synergy between feature processing and classification decision-making. The extraction of signal features during sparse decomposition lacks specificity, making it difficult to form compact features that accurately characterize target motion patterns. Furthermore, motion pattern compatibility quantification relies solely on single-dimensional parameters, failing to fully incorporate multi-dimensional evaluation of the dynamic changes in target motion states. Simultaneously, the confidence fusion in the classification decision-making stage lacks an effective conflict resolution mechanism, and the feedback adjustment mechanism for preliminary classification results is imperfect, leading to poor stability of classification labels. This results in weak anti-interference capabilities and a high classification error rate when facing sudden changes in target motion patterns or complex electromagnetic environments. Therefore, improving the recognition efficiency of radar real-time data transmission processing and classification has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides a radar real-time data transmission processing, classification, and identification method and system to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a radar real-time data transmission processing classification and identification method, comprising: S01. Perform trajectory pattern identification on the narrowband data of the target radar to obtain the real-time motion pattern category of the narrowband data; S02. Based on the real-time motion mode category, perform directional region selection on the synchronous broadband echo data of the target radar to obtain the signal segment of the synchronous broadband echo data. S03. Perform sparse decomposition on the signal segment to obtain the compact features of the signal segment; S04. Based on the real-time motion mode category, perform motion mode compatibility quantification on the compact feature to obtain the category score of the compact feature; S05. Based on the category score and the compact feature, the real-time motion mode category is adjusted to obtain the optimized motion mode of the real-time motion mode category. S06. Perform confidence fusion judgment on the optimized motion mode to obtain the classification identifier of the real-time motion mode category.

[0006] In a preferred embodiment, the step of performing track pattern identification on the narrowband data of the target radar to obtain the real-time motion pattern category of the narrowband data includes: Adaptive spectral filtering is performed on the narrowband data of the target radar to obtain the cleaned data of the narrowband data; Density peak clustering is performed on the cleanup data to obtain candidate points in the cleanup data; Based on the candidate points, the cleanup data is probabilistically correlated to obtain the continuous motion trajectory of the cleanup data; Dynamic pattern matching is performed on the continuous motion trajectory to obtain the real-time motion pattern category of the narrowband data.

[0007] In a preferred embodiment, the step of performing guiding region selection on the synchronous broadband echo data of the target radar based on the real-time motion mode category to obtain signal segments of the synchronous broadband echo data includes: Multi-dimensional spatiotemporal clustering analysis is performed on the real-time motion pattern categories to obtain the predicted distribution area of ​​the real-time motion pattern categories; Based on the predicted distribution area, adaptive beamforming is performed on the synchronous broadband echo data of the target radar to obtain the focused energy spectrum of the synchronous broadband echo data. Based on the focused energy spectrum, coherent synthesis is performed on the predicted distribution region to obtain a coarsely selected signal segment of the predicted distribution region; The coarsely selected signal segment is iteratively optimized to obtain the signal segment of the synchronous broadband echo data.

[0008] In a preferred embodiment, the iterative optimization of the coarsely selected signal segment to obtain the signal segment of the synchronous broadband echo data includes: Perform time-frequency joint analysis on the coarsely selected signal segment to obtain the instantaneous frequency trajectory of the coarsely selected signal segment; Radial velocity fitting is performed on the motion state vector of the real-time motion mode category to obtain the instantaneous velocity of the real-time motion mode category; A nonlinear weighted transformation is performed on the probability density distribution of the predicted distribution region to obtain the time-domain weighting coefficients of the probability density distribution; Based on the instantaneous frequency trajectory and the real-time motion mode category, a consistency quantization measure is performed on the coarsely selected signal segment to obtain the instantaneous frequency matching degree of the coarsely selected signal segment. The formula for calculating the instantaneous frequency matching degree is as follows: ; In the formula, The instantaneous frequency matching degree, For time variables, The time support domain for the coarsely selected signal segment. The instantaneous frequency trajectory in time The value at that location, The carrier frequency of the transmitted signal of the target radar. The instantaneous velocity, The target radar operates at wavelength. The time-domain weighting coefficients are... This refers to the coarsely selected signal segment; Based on the instantaneous frequency matching degree, gradient-guided search is performed on the boundary parameters of the coarsely selected signal segment to obtain the parameter update amount of the coarsely selected signal segment; Based on the parameter update amount, the boundary of the coarsely selected signal segment is adjusted to obtain the signal segment of the synchronous broadband echo data.

[0009] In a preferred embodiment, the sparse decomposition of the signal segment to obtain the compact features of the signal segment includes: Perform multi-resolution time-frequency transformation on the signal segment to obtain the time-frequency representation of the signal segment; Local maxima detection is performed on the energy distribution in the time-frequency representation to obtain the dominant peak value of the energy distribution; Based on the dominant peak value, adaptive component extraction is performed on the time-frequency representation to obtain the local feature vector of the time-frequency representation; Based on the local feature vector, sparse projection is performed on the time-frequency representation to obtain the projection coefficients of the time-frequency representation; Based on the projection coefficients, the local feature vectors are linearly combined to obtain the compact features of the signal segment.

[0010] In a preferred embodiment, the step of quantifying the motion mode compatibility of the compact feature based on the real-time motion mode category to obtain a category score for the compact feature includes: The motion elements of the real-time motion mode category are analyzed to obtain the motion parameters of the real-time motion mode category. Based on the motion parameters, spatial registration is performed on the compact feature to obtain the calibrated feature representation of the compact feature; Based on the motion parameters, parameter mapping derivation is performed on the motion pattern of the narrowband data to obtain the candidate categories of the narrowband data; Based on the calibration feature representation, a similarity measure is performed on the candidate categories to obtain the association strength distribution of the candidate categories; A competitive evaluation is performed on the association strength distribution to obtain the category score of the compact feature.

[0011] In a preferred embodiment, the step of adjusting the real-time motion mode category based on the category score and the compact feature to obtain an optimized motion modality for the real-time motion mode category includes: The confidence scores of the dominant category and competing categories in the category ratings are evaluated to obtain the confidence scores of the dominant category and competing categories. Based on the confidence level of the competing categories, multi-class feature residual inference is performed on the compact features to obtain the feature difference distribution of the compact features; Based on the feature difference distribution, backpropagation of error is performed on the real-time motion pattern category to obtain the error propagation path of the real-time motion pattern category; Based on the error propagation path and the confidence level of the dominant category, the real-time motion mode category is iteratively weighted and corrected to obtain the optimized motion mode of the real-time motion mode category.

[0012] In a preferred embodiment, the step of performing multi-class feature residual deduction on the compact feature based on the competitive class confidence level to obtain the feature difference distribution of the compact feature includes: The confidence scores of the competing categories are normalized, and the Gini impurity is derived from the normalized confidence scores to obtain the confidence impurity of the competing category confidence scores. Principal component energy is extracted from the compact feature to obtain the first principal component energy and the second principal component energy of the compact feature. Based on the confidence impurity, the energy of the first principal component, and the energy of the second principal component, the feature difference distribution of the compact feature is calculated, wherein the formula for calculating the feature difference distribution is: ; In the formula, The feature difference distribution, The confidence level is impurity. The energy of the second principal component. The energy of the first principal component. For the preset smallest positive number, The preset power-law adjustment coefficient, It is an exponential function. This is the preset exponential decay adjustment coefficient.

[0013] In a preferred embodiment, the step of performing confidence fusion judgment on the optimized motion modality to obtain the classification identifier of the real-time motion mode category includes: Spatiotemporal consistency analysis is performed on the optimized motion mode to obtain the consistency evaluation index of the optimized motion mode; Based on the aforementioned consistency evaluation index, the optimized motion modes are dynamically weighted and integrated to obtain a weighted evidence body of the optimized motion modes; The weighted evidence body is conflict resolved, and based on the resolved evidence body, a category decision is made on the optimized motion mode to obtain the classification label of the real-time motion mode category.

[0014] To address the above problems, the present invention also provides a radar real-time data transmission processing classification and recognition system, the system comprising: The narrowband track pattern identification module is used to identify the track pattern of the narrowband data of the target radar and obtain the real-time motion pattern category of the narrowband data. A broadband guiding region selection module is used to select a guiding region for the synchronous broadband echo data of the target radar based on the real-time motion mode category, so as to obtain a signal segment of the synchronous broadband echo data. The signal sparse feature extraction module is used to perform sparse decomposition on the signal segment to obtain the compact features of the signal segment; The motion mode compatibility quantification module is used to quantify the motion mode compatibility of the compact feature based on the real-time motion mode category, and obtain the category score of the compact feature. A real-time hypothesis feedback adjustment module is used to adjust the real-time motion mode category based on the category score and the compact feature to obtain the optimized motion mode of the real-time motion mode category. The confidence fusion decision module is used to perform confidence fusion judgment on the optimized motion mode to obtain the classification label of the real-time motion mode category.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention effectively improves the targeting and efficiency of radar data processing by synergistically linking narrowband data track pattern identification with broadband echo data guidance region selection. Based on real-time motion pattern categories, it accurately locks the broadband data processing region, significantly reducing the transmission and processing overhead of invalid signals. Combined with sparse decomposition technology to extract compact signal features, it enhances the feature representation ability of target motion patterns, ensuring the accuracy and effectiveness of feature extraction. This provides high-quality data support for subsequent classification and recognition, significantly improving overall processing efficiency and feature extraction reliability.

[0016] 2. This invention achieves dynamic optimization and accurate decision-making in motion pattern classification by constructing a motion pattern compatibility quantification and feedback adjustment mechanism. Relying on category scoring and compact feature iterative correction of motion modes, combined with confidence fusion and conflict resolution, it realizes classification identification, improving the stability and accuracy of classification results. It can accurately capture the dynamic changes of target motion patterns, enhance adaptability to complex scenes, effectively reduce classification errors, and ensure the accuracy and anti-interference capability of real-time radar data classification and recognition. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a radar real-time data transmission processing classification and identification method according to an embodiment of the present invention. Figure 2 This is a functional block diagram of a radar real-time data transmission processing classification and recognition system provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] This application provides a radar real-time data transmission processing classification and identification method. The executing entity of the radar real-time data transmission processing classification and identification method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the radar real-time data transmission processing classification and identification method can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0020] Reference Figure 1 The diagram shown is a flowchart illustrating a radar real-time data transmission processing, classification, and identification method according to an embodiment of the present invention. In this embodiment, the radar real-time data transmission processing, classification, and identification method includes: S01. Perform trajectory pattern identification on the narrowband data of the target radar to obtain the real-time motion pattern category of the narrowband data; In this embodiment of the invention, the step of performing track pattern identification on the narrowband data of the target radar to obtain the real-time motion pattern category of the narrowband data includes: Adaptive spectral filtering is performed on the narrowband data of the target radar to obtain the cleaned data of the narrowband data; Density peak clustering is performed on the cleanup data to obtain candidate points in the cleanup data; Based on the candidate points, the cleanup data is probabilistically correlated to obtain the continuous motion trajectory of the cleanup data; Dynamic pattern matching is performed on the continuous motion trajectory to obtain the real-time motion pattern category of the narrowband data.

[0021] Adaptive spectral filtering is applied to the narrowband data of the target radar. By capturing the differences in spectral characteristics between the signal and clutter in the data, the filtering frequency range is dynamically adjusted to retain only the part that matches the spectral characteristics of the target signal, while completely filtering out clutter signals and interference components, ultimately obtaining cleaned narrowband data after removing various interferences.

[0022] Density peak clustering is performed on the purified data. All data points in the purified data are traversed, and the distribution of other data points within a certain range around each data point is counted one by one. Based on this, the density of each data point is determined, and core data points with a density that meets the set conditions are selected. Then, data points surrounding the core data points and whose distance meets the requirements are grouped into the same set to form multiple data point clusters. Each cluster corresponds to a set of candidate points of purified data.

[0023] Based on the candidate points, the cleaned data is probabilistically correlated. Combining the distribution patterns of the candidate points in the time and space dimensions, the candidate points generated at different times are compared one by one to determine the inherent correlation between the candidate points at different times. Candidate points with continuity in the time and space dimensions are retained, and isolated candidate points without correlation are eliminated. The continuous candidate points are connected sequentially in time order to form a continuous motion track of the cleaned data that can reflect the target's motion trajectory.

[0024] Dynamic pattern matching is performed on the continuous motion track. Feature templates of multiple typical motion modes are preset, and motion features of the continuous motion track are extracted, including the direction of the motion trajectory, the speed change trend, the turning pattern, etc. The extracted motion features are compared with the preset feature templates one by one to find the feature template that perfectly matches the motion features of the continuous motion track. The motion mode corresponding to the feature template is the real-time motion mode category of the narrowband data.

[0025] The beneficial effects include accurately capturing the spectral characteristics differences between signals and clutter through adaptive spectral filtering, dynamically adjusting the filtering frequency band range, effectively filtering out clutter signals and interference components, and obtaining purified data after removing various interferences, providing a high-quality data foundation for subsequent processing. Density peak clustering is used to traverse all data points, statistically analyzing the distribution around each data point to determine its density, selecting core data points, and grouping data points that meet distance requirements into clusters to obtain accurate candidate tracks. Based on the spatiotemporal distribution patterns of the candidate tracks, tracks at different times are compared, retaining continuous tracks and removing isolated tracks, connecting them in chronological order to form continuous motion tracks, fully reflecting the target's motion trajectory. Multiple motion features of the continuous motion tracks are extracted and compared one by one with preset typical motion pattern feature templates. A perfectly matching template is found to determine the real-time motion pattern category, improving the accuracy and stability of motion pattern identification, ensuring a reliable initial motion pattern basis for subsequent radar data processing, and enhancing the effectiveness and coherence of overall data processing.

[0026] S02. Based on the real-time motion mode category, perform directional region selection on the synchronous broadband echo data of the target radar to obtain the signal segment of the synchronous broadband echo data. In this embodiment of the invention, the step of performing guiding region selection on the synchronous broadband echo data of the target radar based on the real-time motion mode category to obtain signal segments of the synchronous broadband echo data includes: Multi-dimensional spatiotemporal clustering analysis is performed on the real-time motion pattern categories to obtain the predicted distribution area of ​​the real-time motion pattern categories; Based on the predicted distribution area, adaptive beamforming is performed on the synchronous broadband echo data of the target radar to obtain the focused energy spectrum of the synchronous broadband echo data. Based on the focused energy spectrum, coherent synthesis is performed on the predicted distribution region to obtain a coarsely selected signal segment of the predicted distribution region; The coarsely selected signal segment is iteratively optimized to obtain the signal segment of the synchronous broadband echo data.

[0027] The iterative optimization of the coarsely selected signal segment to obtain the signal segment of the synchronous broadband echo data includes: Perform time-frequency joint analysis on the coarsely selected signal segment to obtain the instantaneous frequency trajectory of the coarsely selected signal segment; Radial velocity fitting is performed on the motion state vector of the real-time motion mode category to obtain the instantaneous velocity of the real-time motion mode category; A nonlinear weighted transformation is performed on the probability density distribution of the predicted distribution region to obtain the time-domain weighting coefficients of the probability density distribution; Based on the instantaneous frequency trajectory and the real-time motion mode category, a consistency quantization measure is performed on the coarsely selected signal segment to obtain the instantaneous frequency matching degree of the coarsely selected signal segment. The formula for calculating the instantaneous frequency matching degree is as follows: ; In the formula, The instantaneous frequency matching degree, For time variables, The time support domain for the coarsely selected signal segment. The instantaneous frequency trajectory in time The value at that location, The carrier frequency of the transmitted signal of the target radar. The instantaneous velocity, The target radar operates at wavelength. The time-domain weighting coefficients are... This refers to the coarsely selected signal segment; Based on the instantaneous frequency matching degree, gradient-guided search is performed on the boundary parameters of the coarsely selected signal segment to obtain the parameter update amount of the coarsely selected signal segment; Based on the parameter update amount, the boundary of the coarsely selected signal segment is adjusted to obtain the signal segment of the synchronous broadband echo data.

[0028] Multi-dimensional spatiotemporal clustering analysis is performed on the real-time motion pattern category. From the time dimension, the motion sequence pattern of the target under the category is sorted out, and the position change characteristics of the target in different time periods are clarified. From the spatial dimension, the motion spatial range of the target under the category is divided, and the spatial boundary of the target motion is defined. Combining the feature information of the two dimensions of time and space, a specific area that perfectly matches the target motion state is delineated, and the predicted distribution area of ​​the real-time motion pattern category is obtained.

[0029] Based on the predicted distribution area, adaptive beamforming is performed on the synchronous broadband echo data of the target radar to adjust the pointing range of signal reception, concentrate the signal reception energy within the predicted distribution area, filter out irrelevant signals and interference components outside the area, and retain only the effective broadband echo data within the area to obtain the focused energy spectrum of the synchronous broadband echo data.

[0030] Based on the focused energy spectrum, coherent synthesis is performed on the predicted distribution area. Signal components with concentrated energy and consistent with the target motion characteristics in the focused energy spectrum are selected. These signal components are integrated and superimposed according to their temporal order and spatial position relationship. The effective signal part in the predicted distribution area is aggregated to obtain the coarsely selected signal segment of the predicted distribution area.

[0031] The coarsely selected signal segments are iteratively optimized, and the degree of fit between each part of the signal within the coarsely selected signal segments and the target motion characteristics is continuously verified. Invalid signal parts with insufficient fit are gradually eliminated, and the core signal components that conform to the target motion characteristics are retained, and finally the signal segments of the synchronous broadband echo data are obtained.

[0032] The coarse-selected signal segment is subjected to time-frequency joint analysis. The continuous change process of the signal within the coarse-selected signal segment is tracked from the time dimension, and the waveform characteristics of the signal at different time points are recorded. The characteristic frequency band distribution of the signal within the coarse-selected signal segment is identified from the frequency dimension, and the frequency composition of the signal is clarified. Combining the analysis results of the two dimensions of time and frequency, the complete trajectory of the signal frequency change over time is depicted, and the instantaneous frequency trajectory of the coarse-selected signal segment is obtained.

[0033] Radial velocity fitting is performed on the motion state vector of the real-time motion mode category. Combining the characteristics of the target's motion direction and position change under this category, the relationship between the target's position and time is analyzed, and the continuous change of the target's velocity over time is derived to obtain the instantaneous velocity of the real-time motion mode category.

[0034] A nonlinear weighted transformation is performed on the probability density distribution of the predicted distribution area. Based on the density of data points in the predicted distribution area, the weight values ​​corresponding to different times are determined. Higher weights are assigned to times with dense distribution, and lower weights are assigned to times with sparse distribution. The allocation and transformation of weight values ​​are completed to obtain the time-domain weighting coefficient of the probability density distribution.

[0035] Based on the instantaneous frequency trajectory and the real-time motion mode category, the consistency quantification measurement of the coarsely selected signal segment is performed. By comparing the frequency change characteristics corresponding to the real-time motion mode category, the fit between the instantaneous frequency trajectory and the characteristic is compared moment by moment. The degree of fit between the instantaneous frequency trajectory and the frequency characteristics of the real-time motion mode category is determined, and the instantaneous frequency matching degree of the coarsely selected signal segment is obtained.

[0036] Based on the instantaneous frequency matching degree, a gradient-guided search is performed on the boundary parameters of the coarsely selected signal segment. According to the changing trend of the instantaneous frequency matching degree, the direction of the boundary adjustment of the coarsely selected signal segment is determined. The parameters corresponding to the start and end boundaries of the signal segment are gradually adjusted, and the matching degree changes after each adjustment are recorded to obtain the parameter update amount of the coarsely selected signal segment.

[0037] Based on the parameter update amount, the boundary of the coarsely selected signal segment is adjusted. The boundary range of the coarsely selected signal segment is modified according to the parameter update amount. The signal part with the instantaneous frequency matching degree at the boundary is removed, and the core signal component with the instantaneous frequency matching degree within the boundary range is retained to obtain the signal segment of the synchronous broadband echo data.

[0038] The instantaneous frequency trajectory is derived from the joint time-frequency analysis of the coarse-selected signal segments, tracking the frequency changes of the segments at different time points to form a continuous trajectory. The transmitted signal carrier frequency is a fixed transmission frequency set by the target radar itself, extracted directly from the radar's configuration information. The instantaneous velocity is derived from radial velocity fitting of the motion state vector of the real-time motion mode category, obtaining continuous velocity values ​​after analyzing the changes in the target's motion direction and position. The operating wavelength is an inherent attribute of the target radar, determined by the correspondence between the transmitted signal carrier frequency and the electromagnetic wave propagation speed, extracted directly from the radar's technical parameters. The time-domain weighting coefficients are derived from a nonlinear weighted transformation of the probability density distribution of the predicted distribution area, assigning corresponding weight values ​​to different times based on the density of data points within the area. The time support domain is the time range of the coarse-selected signal segments, determined directly from the start and end points of the segments.

[0039] This calculation compares the instantaneous frequency trajectory with the theoretical frequency changes corresponding to the real-time motion pattern, and combines time-domain weighting coefficients to quantify the degree of fit between the frequency characteristics of the coarse-selected signal segment and the target motion pattern. The results are then used to adjust the boundaries of the coarse-selected signal segment.

[0040] When the instantaneous frequency trajectory perfectly matches the theoretical frequency change, the calculation result is zero. When the instantaneous frequency trajectory deviates from the theoretical frequency change, the calculation result increases with the degree of deviation. The time-domain weighting coefficient amplifies the impact of deviations at times with densely distributed data points on the calculation result, while the impact of deviations at times with sparsely distributed data points is weaker.

[0041] The beneficial effects are as follows: by conducting multi-dimensional spatiotemporal clustering analysis on real-time motion pattern categories, the predicted distribution area is accurately delineated, providing clear direction for broadband echo data processing. Adaptive beamforming based on the predicted distribution area can concentrate received energy, filter out irrelevant interference, and obtain a focused energy spectrum. Coherent synthesis of the predicted distribution area is then performed based on the focused energy spectrum, aggregating effective signal components to form a coarsely selected signal segment. Iterative optimization of the coarsely selected signal segment is then performed, combining time-frequency joint analysis to depict the instantaneous frequency trajectory. Instantaneous velocity is determined through radial velocity fitting, and time-domain weighting coefficients are obtained through nonlinear weighted transformation. Consistency quantization is then performed to obtain the instantaneous frequency matching degree. Gradient-guided search is conducted based on the instantaneous frequency matching degree to obtain parameter update amounts. Adjusting the boundaries of the coarsely selected signal segment according to the parameter update amounts accurately eliminates invalid signal components and retains core signals that highly match the target motion characteristics. Finally, high-quality synchronous broadband echo data signal segments are obtained, providing reliable data support for subsequent sparse decomposition and motion pattern compatibility quantization, ensuring the smooth progress and accurate implementation of the overall radar data classification and identification process.

[0042] All inputs relied upon by this calculation come from clearly defined results already generated within the process, eliminating the need for additional redundant operations and ensuring the consistency and reliability of data sources. By comparing the instantaneous frequency trajectory with the theoretical frequency changes corresponding to the real-time motion pattern, and combining time-domain weighting coefficients to quantify the degree of fit between the frequency characteristics of the coarsely selected signal segments and the target motion pattern, a clear basis is provided for the subsequent boundary adjustment of the coarsely selected signal segments, ensuring the accuracy of the adjustment process. The calculation results can directly reflect the fit between the signal and the motion pattern; the result is zero when they fit perfectly, and increases with the degree of deviation when they deviate. Furthermore, the time-domain weighting coefficients can amplify the influence of the moment when data points are densely distributed, improving the targeting and effectiveness of signal optimization. Ultimately, high-quality synchronous broadband echo data signal segments are obtained, providing reliable support for subsequent sparse decomposition and motion pattern compatibility quantification.

[0043] S03. Perform sparse decomposition on the signal segment to obtain the compact features of the signal segment; In this embodiment of the invention, the step of sparsely decomposing the signal segment to obtain the compact features of the signal segment includes: Perform multi-resolution time-frequency transformation on the signal segment to obtain the time-frequency representation of the signal segment; Local maxima detection is performed on the energy distribution in the time-frequency representation to obtain the dominant peak value of the energy distribution; Based on the dominant peak value, adaptive component extraction is performed on the time-frequency representation to obtain the local feature vector of the time-frequency representation; Based on the local feature vector, sparse projection is performed on the time-frequency representation to obtain the projection coefficients of the time-frequency representation; Based on the projection coefficients, the local feature vectors are linearly combined to obtain the compact features of the signal segment.

[0044] The signal segment is subjected to multi-resolution time-frequency transformation, which decomposes the signal segment into layers according to different time scales, and processes it sequentially from the shortest time scale to the longest time scale. Each time scale corresponds to a different frequency coverage range. Within each time scale, the frequency components of the signal change with time are tracked, and the frequency components corresponding to each time point are presented in the form of a two-dimensional distribution, which completely preserves the correlation information of the signal in the time and frequency dimensions, and finally obtains the time-frequency representation of the signal segment.

[0045] Local maxima detection is performed on the energy distribution in the time-frequency representation. All energy points in the time-frequency representation are traversed. For each energy point, the energy points in the time and frequency directions adjacent to it are checked. If the energy value of the energy point is higher than the energy values ​​of all adjacent points, the point is determined to be a local maximum point. After collecting all local maximum points, the points with the highest energy values ​​are selected in descending order of energy value. These points are the dominant peaks of the energy distribution.

[0046] Based on the dominant peak, adaptive component extraction is performed on the time-frequency representation. Taking each dominant peak as the center, the extraction time and frequency boundaries are determined according to the energy range of the peak. Local regions containing the dominant peak and its surrounding associated energy are segmented in the time-frequency representation. The energy distribution in each local region is arranged in chronological order to form a one-dimensional vector. Each dominant peak corresponds to a one-dimensional vector. These vectors are the local feature vectors of the time-frequency representation.

[0047] Based on the local feature vectors, sparse projection is performed on the time-frequency representation. Each energy point of the time-frequency representation is compared with all local feature vectors one by one to determine the degree of correlation between each energy point and each local feature vector. Only the correlation value corresponding to the local feature vector with the highest correlation is retained. These correlation values ​​are arranged in the order of the local feature vectors to form a set of values, which are the projection coefficients of the time-frequency representation.

[0048] Based on the projection coefficients, the local feature vectors are linearly combined. Each local feature vector is multiplied by the corresponding projection coefficient to obtain a weighted local feature vector. All weighted local feature vectors are superimposed and integrated in the order of time and frequency association to form a vector that can completely characterize the core features of the signal segment. This vector is the compact feature of the signal segment.

[0049] The beneficial effects are as follows: By decomposing signal segments through multi-resolution time-frequency transformation, the correlation information of the signal in the time and frequency dimensions is fully preserved, resulting in an accurate time-frequency representation and providing comprehensive basic data for subsequent feature extraction. Local maximum detection traverses all energy points and selects the local maxima points with the highest energy values, locating the dominant peak of the energy distribution and clarifying the core direction for subsequent feature extraction. Adaptive component extraction segments local regions with the dominant peak as the center, generating targeted local feature vectors to ensure that the feature vectors highly match the core energy components of the signal. Sparse projection compares the energy points in the time-frequency representation with the local feature vectors one by one, selecting the feature corresponding values ​​with the highest correlation to obtain concise projection coefficients, achieving initial compression of the feature dimension. Finally, by linearly combining and weighting the local feature vectors, a compact feature that can fully represent the core features of the signal segment is formed, preserving the core information of the signal while further compressing the feature dimension, providing high-quality feature support for subsequent motion mode compatibility quantification, and ensuring the smooth progress and accurate implementation of the overall radar data classification and recognition process.

[0050] S04. Based on the real-time motion mode category, perform motion mode compatibility quantification on the compact feature to obtain the category score of the compact feature; In this embodiment of the invention, the step of quantifying the motion mode compatibility of the compact feature based on the real-time motion mode category to obtain a category score for the compact feature includes: The motion elements of the real-time motion mode category are analyzed to obtain the motion parameters of the real-time motion mode category. Based on the motion parameters, spatial registration is performed on the compact feature to obtain the calibrated feature representation of the compact feature; Based on the motion parameters, parameter mapping derivation is performed on the motion pattern of the narrowband data to obtain the candidate categories of the narrowband data; Based on the calibration feature representation, a similarity measure is performed on the candidate categories to obtain the association strength distribution of the candidate categories; A competitive evaluation is performed on the association strength distribution to obtain the category score of the compact feature.

[0051] The motion elements of the real-time motion mode category are analyzed, and the core features of the target motion under this category are sorted out, including the changing pattern of motion direction, the continuity of motion trajectory, the migration characteristics of position over time, and the stability attributes of motion state. These sorted core features are integrated and summarized to form an information set that can comprehensively characterize the motion state of this category, and the motion parameters of the real-time motion mode category are obtained.

[0052] Based on the motion parameters, spatial registration is performed on the compact features. Referring to the target motion spatial distribution range and position migration rules clearly defined in the motion parameters, the spatial dimension coordinates of the compact features are adjusted to ensure that the spatial distribution features of the compact features are consistent with the motion spatial features of the real-time motion mode category. The spatial feature parts of the compact features that do not match the motion parameters are corrected so that the compact features can accurately correspond to the actual motion spatial state of the target, thus obtaining the calibrated feature representation of the compact features.

[0053] Based on the motion parameters, parameter mapping is performed on the motion patterns of the narrowband data. Feature information of various motion patterns recorded in the narrowband data is retrieved. The core features in the motion parameters are matched one by one with the feature information of various motion patterns in the narrowband data. Narrowband data motion patterns that match the motion parameter features are selected. These selected motion patterns are determined as candidate objects associated with the real-time motion pattern category, thus obtaining the candidate category of the narrowband data.

[0054] Based on the calibration feature representation, a similarity measure is performed on the candidate categories, and a typical feature template for each candidate category is extracted. Each feature of the calibration feature representation is compared with the typical feature template of the candidate category dimension by dimension to determine the fit between the calibration feature representation and the typical feature template of each candidate category. The fit is converted into a value that can characterize the degree of association. The association strength values ​​of all candidate categories are arranged by category to obtain the association strength distribution of the candidate categories.

[0055] A competitive evaluation is performed on the association strength distribution. The association strength values ​​of each candidate category in the association strength distribution are compared to clarify the differences in association strength between different candidate categories. A corresponding score is assigned to each candidate category according to the association strength. Candidate categories with high association strength receive high scores, and candidate categories with low association strength receive low scores. The scores of all candidate categories are integrated and summarized to form a score set that can quantify the compatibility of compact features with real-time motion mode categories, and the category score of the compact feature is obtained.

[0056] The beneficial effects are as follows: By analyzing motion elements of real-time motion pattern categories, the core features of target motion are comprehensively identified and integrated into motion parameters, providing a precise reference for subsequent compatibility quantification. Spatial registration of compact features based on motion parameters corrects mismatches between compact features and motion states, ensuring consistency between calibrated feature representations and actual target motion spatial features, thus improving the fit between features and motion patterns. Parameter mapping derivation of narrowband data motion patterns based on motion parameters accurately filters out associated candidate categories, narrowing the scope of similarity measurement. Similarity measurement of candidate categories is performed based on calibrated feature representations, comparing feature fit dimension by dimension and forming a correlation strength distribution, clarifying the degree of correlation between different candidate categories and calibrated feature representations. Competitive evaluation of the correlation strength distribution is conducted, assigning corresponding scores based on correlation strength and integrating them to form a category score. This achieves precise quantification of the compatibility between compact features and real-time motion pattern categories, providing reliable numerical support for subsequent feedback adjustments to real-time motion pattern categories, and ensuring the accuracy and stability of overall radar data classification and recognition.

[0057] S05. Based on the category score and the compact feature, the real-time motion mode category is adjusted to obtain the optimized motion mode of the real-time motion mode category. In this embodiment of the invention, the step of adjusting the real-time motion mode category based on the category score and the compact feature to obtain the optimized motion modality of the real-time motion mode category includes: The confidence scores of the dominant category and competing categories in the category ratings are evaluated to obtain the confidence scores of the dominant category and competing categories. Based on the confidence level of the competing categories, multi-class feature residual inference is performed on the compact features to obtain the feature difference distribution of the compact features; Based on the feature difference distribution, backpropagation of error is performed on the real-time motion pattern category to obtain the error propagation path of the real-time motion pattern category; Based on the error propagation path and the confidence level of the dominant category, the real-time motion mode category is iteratively weighted and corrected to obtain the optimized motion mode of the real-time motion mode category.

[0058] The step of performing multi-class feature residual deduction on the compact feature based on the confidence level of the competing categories to obtain the feature difference distribution of the compact feature includes: The confidence scores of the competing categories are normalized, and the Gini impurity is derived from the normalized confidence scores to obtain the confidence impurity of the competing category confidence scores. Principal component energy is extracted from the compact feature to obtain the first principal component energy and the second principal component energy of the compact feature. Based on the confidence impurity, the energy of the first principal component, and the energy of the second principal component, the feature difference distribution of the compact feature is calculated, wherein the formula for calculating the feature difference distribution is: ; In the formula, The feature difference distribution, The confidence level is impurity. The energy of the second principal component. The energy of the first principal component. For the preset smallest positive number, The preset power-law adjustment coefficient, It is an exponential function. This is the preset exponential decay adjustment coefficient.

[0059] The confidence levels of the dominant and competing categories in the category scores are assessed. The category with the highest score is extracted as the dominant category, and several categories with scores second only to the dominant category are extracted as competing categories. The proportion of the scores of the dominant category and each competing category to the total category score is calculated. Based on this proportion and the matching of the corresponding motion features of the categories, the confidence level of the dominant category and each competing category is determined, thus obtaining the confidence levels of the dominant category and the competing categories in the category scores.

[0060] Based on the confidence level of the competing categories, multi-class feature residual inference is performed on the compact feature, the correlation information between the confidence level of the competing categories and the compact feature is integrated, the feature performance differences of the compact feature under different competing categories are analyzed, and the feature difference distribution of the compact feature is obtained.

[0061] The confidence scores of the competing categories are normalized to adjust the confidence scores of all competing categories to a uniform range, so that the adjusted confidence scores can directly reflect the relative strength between different competing categories. The Gini impurity is then derived from the normalized confidence scores to calculate the uniformity of the confidence score distribution of the competing categories. The higher the uniformity, the higher the corresponding impurity value, and the lower the uniformity, the lower the corresponding impurity value, thus obtaining the confidence impurity of the competing category confidence scores.

[0062] Principal component energy extraction is performed on the compact feature. All components of the compact feature are sorted out, the energy percentage of each component is calculated, the component with the highest energy percentage is extracted as the first principal component energy, and the component with the second highest energy percentage is extracted as the second principal component energy, thus obtaining the first principal component energy and the second principal component energy of the compact feature.

[0063] Based on the confidence impurity, the energy of the first principal component, and the energy of the second principal component, the feature difference distribution of the compact feature is calculated. The confidence impurity is used as the basis for measuring the difference. Combined with the energy ratio relationship between the energy of the first principal component and the energy of the second principal component, the feature difference of the compact feature under different competitive categories is comprehensively determined, forming distribution information that can comprehensively characterize the differences of the compact feature.

[0064] Based on the feature difference distribution, backpropagation of errors is performed on the real-time motion mode category. The processing steps of the real-time motion mode category corresponding to each difference part in the feature difference distribution are traced back, the deviation points that may cause differences in each processing step are identified, and the correlation and transmission path between the deviation points are sorted out to obtain the error propagation path of the real-time motion mode category.

[0065] Based on the error propagation path and the confidence level of the dominant category, the real-time motion mode category is iteratively weighted and corrected. The processing steps and adjustment directions that need to be adjusted are determined according to the error propagation path. Adjustment weights are assigned to different processing steps according to the confidence level of the dominant category. Steps with higher confidence levels of the dominant category correspond to higher adjustment weights, and steps with lower confidence levels of the dominant category correspond to lower adjustment weights. The feature parameters of the real-time motion mode category are adjusted according to the adjustment direction and adjustment weights. The adjustment process is repeated until the degree of fit between the feature parameters and the target motion mode reaches a stable state, thereby obtaining the optimized motion mode of the real-time motion mode category.

[0066] The confidence impurity is derived from the normalization of the confidence scores of competing categories and the Gini impurity extrapolation. First, the confidence scores of all competing categories are adjusted to a uniform range. Then, the uniformity of the confidence score distribution is calculated; higher uniformity corresponds to higher impurity values, and lower uniformity corresponds to lower impurity values. The second principal component energy is derived from the principal component energy extraction of compact features. All components of the compact feature are analyzed, the energy proportion of each component is determined, and the component with the second-highest energy proportion after the first principal component is extracted. The first principal component energy is derived from the principal component energy extraction of compact features. All components of the compact feature are analyzed, the energy proportion of each component is determined, and the component with the highest energy proportion is extracted. The minimum positive constant is a pre-set fixed value used to avoid zero denominators in the calculation. The power-law adjustment coefficient is a pre-set fixed value used to adjust the influence of the ratio of the second principal component energy to the first principal component energy. The exponential decay adjustment coefficient is a pre-set fixed value used to adjust the influence of the proportion of the first principal component energy in the total energy.

[0067] This calculation quantifies the performance differences of compact features under different competitive categories by combining the confidence level impurity and the ratio and proportion of principal component energy, forming distribution information that can characterize feature differences. This provides a clear basis for subsequent error backpropagation and motion pattern correction, enabling the feedback adjustment process to accurately locate the source and extent of feature deviation.

[0068] The calculation result increases with increasing confidence level impurity. The calculation result increases with increasing ratio of the energy of the second principal component to that of the first principal component. The calculation result decreases with increasing proportion of the energy of the first principal component in the total energy. The power-law adjustment coefficient amplifies or diminishes the effect of the energy ratio of the second principal component. The exponential decay adjustment coefficient amplifies or diminishes the effect of the proportion of the energy of the first principal component.

[0069] The beneficial effects include: assessing the confidence levels of the dominant and competing categories in the category scoring to clarify the credibility of different categories and provide a precise reference for subsequent feedback adjustments; performing multi-category feature residual extrapolation based on the confidence level of the competing categories, combining normalization processing and Gini impurity extrapolation to obtain the confidence impurity, simultaneously extracting the principal component energy of compact features, and forming a feature difference distribution by integrating multiple information sources to accurately capture the performance differences of compact features under different categories; conducting error backpropagation based on the feature difference distribution to trace the deviation points and propagation paths of the differences, clarifying the adjustment direction of the real-time motion mode category processing stage; combining the error propagation path and the confidence level of the dominant category for iterative weighted correction, assigning corresponding adjustment weights to different processing stages, and achieving a stable state of fit between feature parameters and target motion modes through repeated adjustments, ultimately obtaining optimized motion modes, effectively improving the accuracy and stability of motion mode categories, and providing high-quality modal basis for subsequent confidence fusion judgment.

[0070] All inputs relied upon in this calculation are derived from explicitly generated results within the workflow. Confidence impurity is obtained through normalization and Gini impurity derivation, while principal component energy is obtained through compact feature extraction. Preset values ​​ensure computational stability, eliminating the need for additional redundant operations and guaranteeing the consistency and reliability of data sources. By combining the ratio and proportion of confidence impurity to principal component energy, the performance differences of compact features under different competing categories are quantified, forming a clear distribution of feature differences. This provides a clear basis for subsequent error backpropagation and motion pattern correction, enabling the feedback adjustment stage to accurately pinpoint the source and extent of feature deviations. The calculation results increase with the ratio of confidence impurity to principal component energy and decrease with the increase of the proportion of the first principal component energy. Preset coefficients can adjust the degree of influence of each factor, accurately reflecting changes in feature differences, improving the targeting and effectiveness of feedback adjustments, and ultimately obtaining accurate optimized motion modes, ensuring the stable progress of the overall radar data classification and recognition process.

[0071] S06. Perform confidence fusion judgment on the optimized motion mode to obtain the classification identifier of the real-time motion mode category; In this embodiment of the invention, the step of performing confidence fusion judgment on the optimized motion modality to obtain the classification identifier of the real-time motion mode category includes: Spatiotemporal consistency analysis is performed on the optimized motion mode to obtain the consistency evaluation index of the optimized motion mode; Based on the aforementioned consistency evaluation index, the optimized motion modes are dynamically weighted and integrated to obtain a weighted evidence body of the optimized motion modes; The weighted evidence body is conflict resolved, and based on the resolved evidence body, a category decision is made on the optimized motion mode to obtain the classification label of the real-time motion mode category.

[0072] Spatiotemporal consistency analysis is performed on the optimized motion mode. From the time dimension, the trajectory of the motion direction and position changes of the optimized motion mode at different times is sorted out, and the motion characteristics of adjacent times are checked to see if they are continuous without abrupt jumps. From the spatial dimension, the distribution range of the motion trajectory of the optimized motion mode is sorted out, and the range is checked to see if it conforms to the typical spatial boundary of the corresponding motion mode. The continuity of the time dimension and the boundary conformity of the spatial dimension are integrated to form an information set that can quantify the degree of spatiotemporal fit, and the consistency evaluation index of the optimized motion mode is obtained.

[0073] Based on the consistency evaluation index, the optimized motion mode is dynamically weighted and integrated. The optimized motion mode is decomposed into two parts: temporal features and spatial features. The temporal features are assigned corresponding weights according to the temporal continuity in the consistency evaluation index. If the temporal continuity meets the standard, the corresponding weight is increased. Similarly, the spatial features are assigned corresponding weights according to the spatial continuity in the consistency evaluation index. If the spatial continuity meets the standard, the corresponding weight is increased. The weighted temporal features and spatial features are integrated according to the motion sequence to form a set containing weighted feature information, thus obtaining the weighted evidence body of the optimized motion mode.

[0074] The weighted evidence body undergoes conflict resolution by comparing the temporal and spatial features of each element. Inconsistent information points are identified, and their credibility is assessed using a consistency evaluation index. High-credibility information points are retained, while low-credibility points are removed. The remaining consistent information points are then reassembled to form a set of evidence free of conflict, resulting in the resolved evidence body. Based on this resolved evidence body, its feature information is compared one by one with preset feature templates for various motion patterns. A template that perfectly matches the feature information of the resolved evidence body is found, and the motion pattern corresponding to this template becomes the final category result, thus obtaining the classification identifier for the real-time motion pattern category.

[0075] The beneficial effects are as follows: By conducting spatiotemporal consistency analysis on optimized motion modes, the continuity of motion features is examined from a temporal dimension, and the boundary compliance of motion trajectories is examined from a spatial dimension. This generates a consistency evaluation index that can quantify the degree of spatiotemporal fit, providing a precise reference for subsequent weighted integration. Dynamic weighted integration is then performed based on the consistency evaluation index, assigning corresponding weights to temporal and spatial features, strengthening the role of coherent and compliant features, and integrating them to form a weighted evidence body containing weighted feature information, thus improving the reliability of the evidence body. Conflict resolution is performed on the weighted evidence body, identifying and eliminating contradictory information points, integrating consistent information to obtain the resolved evidence body, and then obtaining the real-time motion mode category classification label through precise matching with a preset motion mode feature template. This effectively improves the accuracy and stability of the classification results, providing a clear and reliable final category basis for the overall radar data classification and identification process.

[0076] like Figure 2 The diagram shown is a functional block diagram of a radar real-time data transmission processing classification and recognition system provided in an embodiment of the present invention.

[0077] The radar real-time data transmission processing classification and recognition system 10 of this invention can be installed in an electronic device. Depending on the functions implemented, the radar real-time data transmission processing classification and recognition system 10 may include a narrowband track pattern identification module 11, a broadband guiding region selection module 12, a signal sparse feature extraction module 13, a motion pattern compatibility quantization module 14, a real-time hypothesis feedback adjustment module 15, and a confidence fusion decision module 16. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0078] In this embodiment, the functions of each module / unit are as follows: The narrowband track pattern identification module 11 is used to identify the track pattern of the narrowband data of the target radar to obtain the real-time motion pattern category of the narrowband data. The broadband guiding region selection module 12 is used to select a guiding region for the synchronous broadband echo data of the target radar based on the real-time motion mode category, so as to obtain a signal segment of the synchronous broadband echo data. The signal sparse feature extraction module 13 is used to perform sparse decomposition on the signal segment to obtain the compact features of the signal segment; The motion mode compatibility quantification module 14 is used to perform motion mode compatibility quantification on the compact feature based on the real-time motion mode category, and obtain the category score of the compact feature. The real-time hypothesis feedback adjustment module 15 is used to adjust the real-time motion mode category based on the category score and the compact feature to obtain the optimized motion mode of the real-time motion mode category. The confidence fusion decision module 16 is used to perform confidence fusion judgment on the optimized motion mode to obtain the classification identifier of the real-time motion mode category.

[0079] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0080] The modules described as separate components may or may not be physically separate. The components shown as modules 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.

[0081] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0082] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0083] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A radar real-time data transmission processing, classification, and identification method, characterized in that, The method includes: S01. Perform trajectory pattern identification on the narrowband data of the target radar to obtain the real-time motion pattern category of the narrowband data; S02. Based on the real-time motion mode category, perform directional region selection on the synchronous broadband echo data of the target radar to obtain the signal segment of the synchronous broadband echo data. S03. Perform sparse decomposition on the signal segment to obtain the compact features of the signal segment; S04. Based on the real-time motion mode category, perform motion mode compatibility quantification on the compact feature to obtain the category score of the compact feature; S05. Based on the category score and the compact feature, the real-time motion mode category is adjusted to obtain the optimized motion mode of the real-time motion mode category. S06. Perform confidence fusion judgment on the optimized motion mode to obtain the classification identifier of the real-time motion mode category.

2. The radar real-time data transmission processing classification and identification method as described in claim 1, characterized in that, The process of identifying the trajectory pattern of the narrowband data from the target radar to obtain the real-time motion pattern category of the narrowband data includes: Adaptive spectral filtering is performed on the narrowband data of the target radar to obtain the cleaned data of the narrowband data; Density peak clustering is performed on the cleanup data to obtain candidate points in the cleanup data; Based on the candidate points, the cleanup data is probabilistically correlated to obtain the continuous motion trajectory of the cleanup data; Dynamic pattern matching is performed on the continuous motion trajectory to obtain the real-time motion pattern category of the narrowband data.

3. The radar real-time data transmission processing classification and identification method as described in claim 1, characterized in that, The step of performing directional region selection on the synchronous broadband echo data of the target radar based on the real-time motion mode category to obtain signal segments of the synchronous broadband echo data includes: Multi-dimensional spatiotemporal clustering analysis is performed on the real-time motion pattern categories to obtain the predicted distribution area of ​​the real-time motion pattern categories; Based on the predicted distribution area, adaptive beamforming is performed on the synchronous broadband echo data of the target radar to obtain the focused energy spectrum of the synchronous broadband echo data. Based on the focused energy spectrum, coherent synthesis is performed on the predicted distribution region to obtain a coarsely selected signal segment of the predicted distribution region; The coarsely selected signal segment is iteratively optimized to obtain the signal segment of the synchronous broadband echo data.

4. The radar real-time data transmission processing classification and identification method as described in claim 3, characterized in that, The iterative optimization of the coarsely selected signal segment to obtain the signal segment of the synchronous broadband echo data includes: Perform time-frequency joint analysis on the coarsely selected signal segment to obtain the instantaneous frequency trajectory of the coarsely selected signal segment; Radial velocity fitting is performed on the motion state vector of the real-time motion mode category to obtain the instantaneous velocity of the real-time motion mode category; A nonlinear weighted transformation is performed on the probability density distribution of the predicted distribution region to obtain the time-domain weighting coefficients of the probability density distribution; Based on the instantaneous frequency trajectory and the real-time motion mode category, a consistency quantization measure is performed on the coarsely selected signal segment to obtain the instantaneous frequency matching degree of the coarsely selected signal segment. The formula for calculating the instantaneous frequency matching degree is as follows: ； In the formula, The instantaneous frequency matching degree, For time variables, The time support domain for the coarsely selected signal segment. The instantaneous frequency trajectory in time The value at that location, The carrier frequency of the transmitted signal of the target radar. The instantaneous velocity, The target radar operates at wavelength. The time-domain weighting coefficients are... This refers to the coarsely selected signal segment; Based on the instantaneous frequency matching degree, gradient-guided search is performed on the boundary parameters of the coarsely selected signal segment to obtain the parameter update amount of the coarsely selected signal segment; Based on the parameter update amount, the boundary of the coarsely selected signal segment is adjusted to obtain the signal segment of the synchronous broadband echo data.

5. The radar real-time data transmission processing classification and identification method as described in claim 1, characterized in that, The sparse decomposition of the signal segment to obtain its compact features includes: Perform multi-resolution time-frequency transformation on the signal segment to obtain the time-frequency representation of the signal segment; Local maxima detection is performed on the energy distribution in the time-frequency representation to obtain the dominant peak value of the energy distribution; Based on the dominant peak value, adaptive component extraction is performed on the time-frequency representation to obtain the local feature vector of the time-frequency representation; Based on the local feature vector, sparse projection is performed on the time-frequency representation to obtain the projection coefficients of the time-frequency representation; Based on the projection coefficients, the local feature vectors are linearly combined to obtain the compact features of the signal segment.

6. The radar real-time data transmission processing classification and identification method as described in claim 1, characterized in that, The step of quantifying the motion mode compatibility of the compact feature based on the real-time motion mode category to obtain a category score for the compact feature includes: The motion elements of the real-time motion mode category are analyzed to obtain the motion parameters of the real-time motion mode category. Based on the motion parameters, spatial registration is performed on the compact feature to obtain the calibrated feature representation of the compact feature; Based on the motion parameters, parameter mapping derivation is performed on the motion pattern of the narrowband data to obtain the candidate categories of the narrowband data; Based on the calibration feature representation, a similarity measure is performed on the candidate categories to obtain the association strength distribution of the candidate categories; A competitive evaluation is performed on the association strength distribution to obtain the category score of the compact feature.

7. The radar real-time data transmission processing classification and identification method as described in claim 1, characterized in that, The step of adjusting the real-time motion mode category based on the category score and the compact feature to obtain the optimized motion mode of the real-time motion mode category includes: The confidence scores of the dominant category and competing categories in the category ratings are evaluated to obtain the confidence scores of the dominant category and competing categories. Based on the confidence level of the competing categories, multi-class feature residual inference is performed on the compact features to obtain the feature difference distribution of the compact features; Based on the feature difference distribution, backpropagation of error is performed on the real-time motion pattern category to obtain the error propagation path of the real-time motion pattern category; Based on the error propagation path and the confidence level of the dominant category, the real-time motion mode category is iteratively weighted and corrected to obtain the optimized motion mode of the real-time motion mode category.

8. The radar real-time data transmission processing classification and identification method as described in claim 7, characterized in that, The step of performing multi-class feature residual deduction on the compact feature based on the confidence level of the competing categories to obtain the feature difference distribution of the compact feature includes: The confidence scores of the competing categories are normalized, and the Gini impurity is derived from the normalized confidence scores to obtain the confidence impurity of the competing category confidence scores. Principal component energy is extracted from the compact feature to obtain the first principal component energy and the second principal component energy of the compact feature. Based on the confidence impurity, the energy of the first principal component, and the energy of the second principal component, the feature difference distribution of the compact feature is calculated, wherein the formula for calculating the feature difference distribution is: ； In the formula, The feature difference distribution, The confidence level is impurity. The energy of the second principal component. The energy of the first principal component. For the preset smallest positive number, The preset power-law adjustment coefficient, It is an exponential function. This is the preset exponential decay adjustment coefficient.

9. The radar real-time data transmission processing classification and identification method as described in claim 1, characterized in that, The step of performing confidence fusion judgment on the optimized motion modality to obtain the classification identifier of the real-time motion mode category includes: Spatiotemporal consistency analysis is performed on the optimized motion mode to obtain the consistency evaluation index of the optimized motion mode; Based on the aforementioned consistency evaluation index, the optimized motion modes are dynamically weighted and integrated to obtain a weighted evidence body of the optimized motion modes; The weighted evidence body is conflict resolved, and based on the resolved evidence body, a category decision is made on the optimized motion mode to obtain the classification label of the real-time motion mode category.

10. A radar real-time data transmission processing classification and recognition system, characterized in that, The system for implementing the radar real-time data transmission processing classification and identification method according to claim 1 includes: The narrowband track pattern identification module is used to identify the track pattern of the narrowband data of the target radar and obtain the real-time motion pattern category of the narrowband data. A broadband guiding region selection module is used to select a guiding region for the synchronous broadband echo data of the target radar based on the real-time motion mode category, so as to obtain a signal segment of the synchronous broadband echo data. The signal sparse feature extraction module is used to perform sparse decomposition on the signal segment to obtain the compact features of the signal segment; The motion mode compatibility quantification module is used to quantify the motion mode compatibility of the compact feature based on the real-time motion mode category, and obtain the category score of the compact feature. A real-time hypothesis feedback adjustment module is used to adjust the real-time motion mode category based on the category score and the compact feature to obtain the optimized motion mode of the real-time motion mode category. The confidence fusion decision module is used to perform confidence fusion judgment on the optimized motion mode to obtain the classification label of the real-time motion mode category.

Images