A method, apparatus, and electronic device for reconstructing and identifying individual fingerprints of self-calibrated radiation sources for distributed reception.

Abstract

This application provides a method, apparatus, and electronic device for reconstructing and identifying individual radiation source fingerprints for distributed reception. The individual radiation source identification method includes: acquiring observation signals from multiple distributed receiving nodes for the same radiation source, and estimating the ideal reference signal of the radiation source; estimating the equivalent frequency domain response of each receiving node based on the observation signals and the ideal reference signal, and separating initial fingerprint information from the equivalent frequency domain response; aggregating the initial fingerprint information from multiple receiving nodes to obtain an initial reconstructed fingerprint; constructing an amplitude residual to be estimated, obtaining the optimal amplitude residual by minimizing a joint optimization objective function within the effective frequency point set, and performing amplitude calibration on the initial reconstructed fingerprint based on the optimal amplitude residual to obtain a calibrated reconstructed fingerprint; extracting features from the calibrated reconstructed fingerprint, and obtaining the identification result of the individual radiation source based on the extracted identification features.

Description

Technical Field

[0001] This invention relates to the fields of signal processing technology, distributed multi-node collaborative observation, and individual identification of specific radiation sources, specifically to a method, apparatus, and electronic device for reconstructing and identifying individual fingerprints of self-calibrated radiation sources for distributed reception. Background Technology

[0002] Individual identification of specific radiation sources aims to distinguish between different radiation sources by utilizing subtle individual characteristic differences caused by hardware manufacturing errors, power amplifier nonlinearity, IQ imbalance, and frequency response differences within the same type of radiation source. In signal processing scenarios, these hardware non-ideal characteristics will form relatively stable transmitter signal distortion information in the amplitude, phase, frequency response, and local spectral shape of the received signal, which is the main source of individual radiation source fingerprints.

[0003] Existing methods for individual radiation source identification often employ IQ sequences, spectrograms, time-frequency plots, or statistical features to directly train classifiers. These methods typically prioritize identification accuracy while lacking explicit separation of the coupling between the transmitter fingerprint and the receiver channel, receiver node response, noise, multipath propagation, Doppler interference, and other factors. When receiving conditions change, the classifier may utilize channel or receiver features instead of the true transmitter fingerprint, leading to decreased generalization performance across nodes, scenarios, and under strong perturbation conditions.

[0004] In distributed multi-receiver node cooperative observation scenarios, multiple receiver nodes are typically deployed in different spatial locations, exhibiting differences in receiver gain, propagation path, channel response, synchronization error, local interference, and noise conditions. The observation signals of the same radiation source at different receiver nodes contain both relatively stable transmitter fingerprints and non-ideal perturbations related to the receiver nodes and propagation environment. If the consistency among multi-node observations cannot be effectively utilized, and the influence of abnormal nodes, low-reliability frequencies, and local interference frequencies cannot be suppressed, the reconstructed fingerprint is prone to incorporating receiver link features, leading to unstable subsequent radiation source identification results. Therefore, a cooperative fingerprint reconstruction method for distributed multi-receiver nodes is needed to extract stable transmitter fingerprints from multi-node observations without relying on the true fingerprint value.

[0005] To identify individual radiation sources using transmitter fingerprints, supervised learning-based methods require training with known ground truth fingerprints as supervisory labels. However, in real-world applications, transmitter ground truth fingerprints are difficult to obtain directly. If the reconstruction method relies on ground truth fingerprints during training, parameter selection, or inference, it becomes unsuitable for practical scenarios. Summary of the Invention

[0006] In view of this, the present invention proposes a method, apparatus, and electronic device for reconstructing and identifying individual fingerprints of self-calibrated radiation sources for distributed reception, to solve the problems of fingerprint reconstruction relying on ground truth fingerprints, inability to be directly deployed in real-world scenarios, and attenuation of fingerprint details in a single observation under complex scenarios. Specifically, the present invention is achieved through the following technical solutions:

[0007] According to a first aspect of the embodiments of this specification, a method for reconstructing and identifying individual fingerprints of self-calibrated radiation sources for distributed reception is provided, the method comprising the following steps:

[0008] Step S1: Obtain the observation signals of multiple receiving nodes in a distributed deployment for the same radiation source, and estimate the ideal reference signal of the radiation source based on the observation signals. The ideal reference signal is used to characterize the fingerprint information of the radiation source without superimposed and the ideal baseband waveform of external non-ideal link distortion.

[0009] Step S2: Based on the observed signals of each receiving node and the ideal reference signal, estimate the equivalent frequency domain response from the radiation source to each receiving node, separate the initial fingerprint information from the equivalent frequency domain response based on the effective frequency point set, and aggregate the initial fingerprint information of multiple receiving nodes to obtain the initial reconstructed fingerprint of the radiation source.

[0010] Step S3: Construct the amplitude residual to be estimated based on the preset orthogonal basis functions and the sparse local residual to be estimated. The amplitude residual reflects the amplitude difference between the initial reconstructed fingerprint and the true value of the radiation source fingerprint. Within the effective frequency point set, obtain the optimal amplitude residual by minimizing the joint optimization objective function. Then, perform amplitude calibration on the initial reconstructed fingerprint based on the optimal amplitude residual to obtain the calibrated reconstructed fingerprint. The joint optimization objective function includes at least:

[0011] The observation reconstruction consistency constraint is used to constrain the observation reconstruction residual of each receiving node to approach zero. The observation reconstruction residual is the difference between the reconstructed spectrum of each receiving node and the spectrum of the observed signal of that receiving node. The reconstructed spectrum of each receiving node is constructed using the equivalent frequency domain response of each receiving node, the ideal reference signal, and the calibrated reconstruction fingerprint.

[0012] Multi-node consistency constraint is used to constrain the initial fingerprint information separated by each receiving node to approximate the initial reconstructed fingerprint;

[0013] Multiple observation consistency constraints are used to ensure that the reconstructed fingerprints of the radiation source after calibration at different observation times approximate each other;

[0014] A priori deviation constraint is used to constrain the magnitude of the magnitude residual to approach zero;

[0015] Spectral smoothness constraint is used to construct a frequency domain differential signal by interpolation of the amplitude residual at adjacent frequency points, and to constrain the amplitude of the frequency domain differential signal to approach zero.

[0016] Local detail preservation constraints are used to perform frequency domain local difference processing on the calibrated reconstructed fingerprint and the initial reconstructed fingerprint respectively to obtain a first difference feature and a second difference feature, and to constrain the first difference feature to approximate the second difference feature;

[0017] Step S4: Extract features from the calibrated reconstructed fingerprint, and obtain the identification result of the radiation source based on the extracted identification features.

[0018] According to a second aspect of the embodiments of this specification, a self-calibrating radiation source individual fingerprint reconstruction and identification device for distributed reception is provided, the device comprising:

[0019] The signal acquisition unit is used to acquire the observation signals of multiple receiving nodes in a distributed deployment for the same radiation source, and to estimate the ideal reference signal of the radiation source based on the observation signals. The ideal reference signal is used to characterize the fingerprint information of the radiation source without superimposed and the ideal baseband waveform of external non-ideal link distortion.

[0020] The fingerprint calculation unit is used to estimate the equivalent frequency domain response from the radiation source to each receiving node based on the observation signals of each receiving node and the ideal reference signal, separate the initial fingerprint information from the equivalent frequency domain response based on the effective frequency point set, and aggregate the initial fingerprint information of multiple receiving nodes to obtain the initial reconstructed fingerprint of the radiation source.

[0021] A fingerprint calibration unit is used to construct an amplitude residual to be estimated based on a preset orthogonal basis function and the sparse local residual to be estimated. The amplitude residual reflects the amplitude difference between the initial reconstructed fingerprint and the true value of the radiation source fingerprint. Within the effective frequency point set, an optimal amplitude residual is obtained by minimizing a joint optimization objective function. The initial reconstructed fingerprint is then calibrated based on the optimal amplitude residual to obtain a calibrated reconstructed fingerprint. The joint optimization objective function includes at least the following:

[0022] The observation reconstruction consistency constraint is used to constrain the observation reconstruction residual of each receiving node to approach zero. The observation reconstruction residual is the difference between the reconstructed spectrum of each receiving node and the spectrum of the observed signal of that receiving node. The reconstructed spectrum of each receiving node is constructed using the equivalent frequency domain response of each receiving node, the ideal reference signal, and the calibrated reconstruction fingerprint.

[0023] Multi-node consistency constraint is used to constrain the initial fingerprint information separated by each receiving node to approximate the initial reconstructed fingerprint;

[0024] Multiple observation consistency constraints are used to ensure that the reconstructed fingerprints of the radiation source after calibration at different observation times approximate each other;

[0025] A priori deviation constraint is used to constrain the magnitude of the magnitude residual to approach zero;

[0026] Spectral smoothness constraint is used to construct a frequency domain differential signal by interpolation of the amplitude residual at adjacent frequency points, and to constrain the amplitude of the frequency domain differential signal to approach zero.

[0027] Local detail preservation constraints are used to perform frequency domain local difference processing on the calibrated reconstructed fingerprint and the initial reconstructed fingerprint respectively to obtain a first difference feature and a second difference feature, and to constrain the first difference feature to approximate the second difference feature;

[0028] The radiation source individual identification unit is used to extract features from the calibrated reconstructed fingerprint and obtain the identification result of the radiation source based on the extracted identification features.

[0029] According to a third aspect of the embodiments of this specification, an electronic device is provided, comprising: a processor; and a computer-readable storage medium storing computer program instructions that, when executed by the processor, cause the processor to perform the method as described in the first aspect.

[0030] The embodiments of the present invention have at least the following technical effects:

[0031] (1) It does not rely on fingerprint truth labels and is applicable to real radiation source observation scenarios. This invention constructs a joint optimization objective function and uses multiple reconstruction errors calculated based on the observation signal, spatial redundancy of multiple receiving nodes, and temporal redundancy of multiple observations as self-supervised signal-driven amplitude residual calibration. No fingerprint truth labels are required to participate in training, optimization, or parameter selection during the fingerprint reconstruction process, which is more in line with the actual situation that the transmitter truth fingerprint cannot be directly obtained in real radiation source observation scenarios.

[0032] (2) Explicitly decouple and reconstruct the individual fingerprint of the transmitter to improve the generalization capability across scenarios. This invention reconstructs the individual fingerprint of the radiation source through initial reconstruction in the frequency domain and self-supervised amplitude residual calibration, which can reduce the impact of receiving channel, node response and random disturbance on the individual identification of radiation source, and make the reconstructed fingerprint closer to the real individual characteristics of radiation source. Attached Figure Description

[0033] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Some specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings in an exemplary and non-limiting manner. The same reference numerals in the drawings indicate the same or similar parts or components. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings:

[0034] Figure 1 This is a flowchart illustrating an exemplary embodiment of the present invention of a method for reconstructing and identifying individual fingerprints of a self-calibrated radiation source for distributed reception;

[0035] Figure 2 This is a schematic diagram illustrating an exemplary embodiment of the present invention of a self-calibrating radiation source individual fingerprint reconstruction and identification system for distributed reception;

[0036] Figure 3 This is a schematic diagram of a self-supervised radiation source individual fingerprint calibration model structure, as shown in an exemplary embodiment of the present invention.

[0037] Figure 4 This is a block diagram illustrating an electronic device according to an exemplary embodiment of the present invention;

[0038] Figure 5 This is a block diagram illustrating an exemplary embodiment of the present invention of a self-calibrated radiation source individual fingerprint reconstruction and identification device for distributed reception. Detailed Implementation

[0039] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0040] Unless otherwise specified, the observation signals from multiple receiving nodes mentioned in this specification refer to multi-node observation signals that have undergone spatiotemporal synchronization processing. This spatiotemporal synchronization processing includes unifying the time reference, sampling time, observation batch, and spatial coordinate reference of multiple receiving nodes, and correcting for propagation delay differences, sampling time deviations, and correspondences of observation events between nodes based on node positions, timestamps, and / or reference signals, so that the observation signals output by each receiving node correspond to the same transmission event from the same radiation source and a unified spatiotemporal reference. This spatiotemporal synchronization processing is not used to eliminate differences in propagation channels, node frequency domain responses, and local noise corresponding to each receiving node.

[0041] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular forms “a,” “the,” and “the” used in this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0042] To address the issues of existing fingerprint reconstruction methods relying on ground truth fingerprints, being unsuitable for direct deployment in real-world scenarios, and the attenuation of fingerprint details in single observations under complex conditions, this invention employs constraints such as initial reconstruction in the frequency domain, weighted reliably at reliable frequency points, multi-node consistency, consistency of multiple observations, and consistency of observation reconstruction to perform self-supervised amplitude residual calibration on the individual fingerprints of radiation sources. This improves the quality of reconstructed fingerprints without using ground truth fingerprints, thereby enhancing the accuracy of identifying individual radiation sources.

[0043] The following detailed explanation of each step is provided in conjunction with the accompanying drawings.

[0044] Figure 1 This is a flowchart illustrating an exemplary embodiment of the present invention of a method for reconstructing and identifying individual fingerprints of a self-calibrated radiation source for distributed reception, as shown below. Figure 1 As shown, the method for identifying individual radiation sources includes the following steps:

[0045] Step S1: Obtain the observation signals of multiple receiving nodes in a distributed deployment for the same radiation source, and estimate the ideal reference signal of the radiation source based on the observation signals. The ideal reference signal is used to characterize the fingerprint information of the radiation source without superimposed on it and the ideal baseband waveform of external non-ideal link distortion.

[0046] Step S2: Based on the observed signals of each receiving node and the ideal reference signal, estimate the equivalent frequency domain response from the radiation source to each receiving node, separate the initial fingerprint information from the equivalent frequency domain response based on the effective frequency point set, and aggregate the initial fingerprint information of multiple receiving nodes to obtain the initial reconstructed fingerprint of the radiation source.

[0047] Step S3: Construct the amplitude residual to be estimated based on the preset orthogonal basis functions and the sparse local residual to be estimated. The amplitude residual reflects the amplitude difference between the initial reconstructed fingerprint and the true value of the radiation source fingerprint. Within the effective frequency point set, obtain the optimal amplitude residual by minimizing the joint optimization objective function. Then, perform amplitude calibration on the initial reconstructed fingerprint based on the optimal amplitude residual to obtain the calibrated reconstructed fingerprint. The joint optimization objective function includes at least:

[0048] The observation reconstruction consistency constraint is used to constrain the observation reconstruction residual of each receiving node to approach zero. The observation reconstruction residual is the difference between the reconstructed spectrum of each receiving node and the spectrum of the observed signal of that receiving node. The reconstructed spectrum of each receiving node is constructed using the equivalent frequency domain response of each receiving node, the ideal reference signal, and the calibrated reconstruction fingerprint.

[0049] Multi-node consistency constraint is used to constrain the initial fingerprint information separated by each receiving node to approximate the initial reconstructed fingerprint;

[0050] Multiple observation consistency constraints are used to ensure that the reconstructed fingerprints of the radiation source after calibration at different observation times approximate each other;

[0051] A priori deviation constraint is used to constrain the magnitude of the magnitude residual to approach zero;

[0052] Spectral smoothness constraint is used to construct a frequency domain differential signal by interpolation of the amplitude residual at adjacent frequency points, and to constrain the amplitude of the frequency domain differential signal to approach zero.

[0053] Local detail preservation constraints are used to perform frequency domain local difference processing on the calibrated reconstructed fingerprint and the initial reconstructed fingerprint respectively to obtain a first difference feature and a second difference feature, and to constrain the first difference feature to approximate the second difference feature;

[0054] Step S4: Extract features from the calibrated reconstructed fingerprint, and obtain the identification result of the radiation source based on the extracted identification features.

[0055] Figure 1 The radiation source individual identification method shown above, through steps S1 to S4, realizes a self-calibrating fingerprint reconstruction and identification scheme that does not rely on ground truth fingerprint labels. This method first utilizes the frequency domain response relationship between multi-node observation signals and an ideal reference signal to initially separate and aggregate coarse-grained fingerprint information of the radiation source, providing physical priors for subsequent precise calibration. Then, through parameterized amplitude residual modeling and the construction of a joint optimization objective function, multiple constraints are used to replace traditional supervision signals, achieving fine-grained amplitude calibration of the coarse-grained fingerprint without the participation of any ground truth fingerprint labels. Finally, the individual radiation source identification is completed based on the calibrated reconstructed fingerprint, effectively suppressing the influence of external non-ideal link distortion and receiver node differences on the identification results. Therefore, this invention can stably reconstruct high-quality individual fingerprints and maintain high identification performance even in observation scenarios where ground truth fingerprint labels for radiation sources cannot be directly obtained.

[0056] Next, combined Figure 2 The present invention provides a detailed description of the radiation source individual identification scheme.

[0057] The radiation source individual identification method provided by this invention can be deployed in one or more signal processing devices in a radiation source individual identification system. Figure 2 An exemplary block diagram of a radiation source individual identification system is shown. The signal processing device includes, but is not limited to, a general-purpose computer, an embedded signal processing platform, a digital signal processor (DSP), a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), or a combination of the above devices. The signal processing device includes at least a processor and a memory, the memory storing computer program instructions that, when executed by the processor, cause the signal processing device to perform the method described in this invention.

[0058] The system first acquires the observation signals of multiple receiving nodes in a distributed manner for the same radiation source. The observation signals of the multiple nodes have all undergone spatiotemporal synchronization processing and an ideal reference signal has been constructed.

[0059] Assume this system has a total of The receiving node, at the _ ... In the second observation, the first The time-domain observation signals collected by each receiving node are ,in, , , Represents discrete-time sampling points. Indicates the number of observations.

[0060] The ideal reference signal is used to characterize the ideal waveform when the transmitter's individual fingerprint, receiver node response, and complex channel disturbances are not superimposed, such as a linear frequency modulated pulse, a frequency modulated continuous wave segment, or other known radiation source baseband waveforms.

[0061] The ideal reference signal can be obtained based on a known emission waveform template of the radiation source, prior waveform parameters, or conventional parameter estimation methods.

[0062] Specifically, when the waveform parameters of the radiation source, such as its operating mode, modulation method, pulse width, bandwidth, sweep slope, carrier frequency, or sampling rate, are known, an ideal reference signal can be directly generated based on a preset baseband signal model. For example, when the radiation source transmits a linear frequency modulated (LFM) signal, an ideal baseband waveform can be directly generated based on the known pulse width, bandwidth, carrier frequency, and sweep slope, according to the baseband analytical expression of the LFM signal.

[0063] When some waveform parameters are unknown, conventional pulse detection, time synchronization, frequency offset estimation, modulation parameter estimation, or matched filtering can be performed on the observed signal to obtain waveform parameter estimates. Then, a corresponding ideal reference signal can be generated based on these waveform parameter estimates. For example, for radiation source signals with unknown pulse width and bandwidth, pulse width estimates can be obtained through pulse detection, and bandwidth and sweep slope estimates can be obtained through time-frequency analysis or autocorrelation processing. Finally, an ideal reference signal can be reconstructed based on the estimated waveform parameters.

[0064] Through the above methods, the present invention can adapt to radiation source application scenarios with different degrees of known waveform parameters. It can directly generate high-precision reference signals when the waveform parameters are completely known, and can also obtain suitable ideal reference signals based on conventional estimation methods in scenarios where some parameters are unknown or not fully known.

[0065] In this embodiment, the ideal reference signal is used to compare with the observed signal in the frequency domain in subsequent steps to estimate the equivalent frequency domain response. Therefore, the waveform shape of the ideal reference signal should match the ideal waveform component in the observed signal to ensure the accuracy of the estimated equivalent frequency domain response.

[0066] After constructing the ideal reference signal, the observation signals from multiple receiving nodes that have completed spatiotemporal synchronization are subjected to complex gain normalization and frequency domain transformation to obtain the observation spectrum. and reference spectrum :

[0067] (1)

[0068] in, Indicates Fourier transform, Indicates frequency point index, Indicates the first The first observation The observed spectrum of each receiving node, Indicates the first Ideal time-domain reference signal under this observation This represents the reference spectrum.

[0069] Next, this system constructs the initial reconstructed fingerprint.

[0070] For the The first observation The equivalent frequency domain response of a receiving node can be obtained by the ratio between the observed spectrum and the reference spectrum. This response includes the receiving node response, propagation channel perturbations, and individual residual information from the transmitting end. The calculation formula is as follows:

[0071] (2)

[0072] in, Indicates complex conjugation. To prevent the denominator from being too small, a regularization constant is used.

[0073] This embodiment extracts the transmitter's individual residuals from the equivalent frequency domain response, performs amplitude normalization on the equivalent frequency domain response, and obtains the initial fingerprint information:

[0074] (3)

[0075] in, Indicates the first The set of effective frequency points corresponding to each observation Indicates the number of valid frequency points. This indicates the initial fingerprint information.

[0076] This embodiment reduces the impact of receiver gain, propagation loss, and global scale variations on individual fingerprints by removing the overall average amplitude response.

[0077] The set of effective frequency points can be determined jointly based on the frequency energy of the ideal reference signal, the frequency energy of the observed signal, the consistency between nodes, and the observation reconstruction residual. For example, when the reference signal energy of a certain frequency point is higher than a preset energy threshold, the response difference of multiple receiving nodes at that frequency point is lower than a preset consistency threshold, and the observation reconstruction residual corresponding to that frequency point is lower than a preset residual threshold, that frequency point is determined as an effective frequency point; otherwise, low-energy frequency points, strong interference frequency points, frequency points with excessively large differences between nodes, or frequency points with excessively large observation residuals are removed from the set of effective frequency points, or assigned a lower reliability weight.

[0078] In some embodiments, the effective frequency point set is expressed as:

[0079] (4)

[0080] in, Indicates the first The set of effective frequency points corresponding to each observation The ideal reference signal is represented in the first... Energy at each frequency point This indicates that the observed signals from multiple receiving nodes are at the [number]th [time]. Amplitude inconsistency measurement at each frequency point Indicates that multiple receiving nodes are in the first stage. The spectral residuals between the observed signal and the ideal reference signal at each frequency point , , These represent the preset energy threshold, consistency threshold, and residual threshold, respectively.

[0081] Furthermore, to reduce the impact of individual node anomalies, local interference, and low-reliability frequency points, this embodiment uses the median, truncated mean, or quality-weighted mean to aggregate the initial fingerprint information of multiple receiving nodes.

[0082] Specifically, the initial fingerprint information of multiple receiving nodes is aggregated through the following steps:

[0083] Obtain the observation reconstruction residuals of each receiving node Inconsistency measurement quality factor Frequency point missing ratio .

[0084] The observation reconstructed residual To reconstruct the spectrum of the observed signal of the receiving node using the equivalent frequency domain response of each receiving node, the ideal reference signal, and the calibrated reconstructed fingerprint, the difference between the reconstructed spectrum and the spectrum of the observed signal of the receiving node is measured.

[0085] The inconsistency measure To determine the metric center of multiple receiving nodes based on equivalent frequency domain response, initial fingerprint information and / or observation reconstruction residual at the same acquisition time, the degree of deviation of each receiving node from the metric center is used to measure the relative consistency of each receiving node in terms of frequency domain response, fingerprint information or reconstruction residual.

[0086] The quality factor The quality factor is determined based on the signal-to-noise ratio (SNR) estimate of the observed signal at each receiving node and is used to reflect the quality of the observed signal at that receiving node. The higher the SNR, the larger the quality factor.

[0087] The frequency point missing ratio It is the ratio of the number of invalid frequency points in the observed signal of each receiving node to the total number of frequency points in the set of valid frequency points; the invalid frequency points are those whose signal energy is lower than a preset energy threshold, whose interference intensity is higher than a preset interference threshold, whose data is missing, or whose observation reconstruction residual is greater than a preset residual threshold.

[0088] The node quality weight of each receiving node is obtained based on the observation reconstruction residual, inconsistency metric, quality factor, and frequency point missing ratio. The node quality weight measures the reliability of different receiving nodes in the current observation; the larger the observation reconstruction residual, the smaller its node quality weight; the more stable the node observation, the larger its quality weight. Specifically, the node quality weight is calculated based on the following formula:

[0089] (5)

[0090] in, Indicates the first The node quality weight of each receiving node. , This indicates the preset adjustment coefficient. .

[0091] Based on the normalized node quality weights of each receiving node, the initial fingerprint information of the multiple receiving nodes is weighted and aggregated to obtain the initial reconstructed fingerprint. :

[0092] (6)

[0093] in, This represents the normalized node quality weight, used to adjust the contribution of different receiving nodes to the joint optimization objective function.

[0094] refer to Figure 2 In this embodiment, the normalized node quality weights can be calculated by the weight construction module. In some embodiments, this weight construction module is also used to calculate the normalized reliable frequency point weights. The normalized reliable frequency point weights are determined based on one or more of the following: the ideal reference signal energy of each frequency point in the effective frequency point set, the equivalent frequency domain response consistency among multiple receiving nodes, and the observation reconstruction residuals. They are used to adjust the contribution of different frequency points to the joint optimization objective function, and their expression is as follows:

[0095] (7)

[0096] in, Indicates the weight of reliable frequency points. Indicates the first The first observation The energy weight of the frequency point can be determined according to the first frequency point. The reference signal energy at each frequency point is calculated relative to the average energy within the effective frequency point set, and is used to reduce the impact of low-energy frequencies on self-supervised optimization. Indicates the first The equivalent frequency domain response consistency weight among multiple receiving nodes under the first observation can be used to calculate the first observation of each receiving node. The mean of the equivalent frequency domain response at each frequency point is used to calculate the ratio of the equivalent frequency domain response of each receiving node to this mean. ; Indicates the first The weights of the observation reconstruction residuals under the nth observation can be based on the nth observation. The residual of observation reconstruction at each frequency point is calculated relative to the average residual of observation reconstruction within the effective frequency point set, and is used to reduce the impact of frequencies with large reconstruction errors. Residual index reconstructed from frequency-level observations Sure.

[0097] Thus, under different observations, different receiver node states, or different interference conditions, the reliable frequency point weights corresponding to each frequency point can be updated as the observed signal changes. Therefore, the reliable frequency point weights can adaptively suppress the influence of low-energy frequencies, strong interference frequencies, and inconsistent frequencies between nodes on the self-supervised calibration process.

[0098] To avoid instability caused by excessively large or small individual weights, this embodiment also normalizes the weights of reliable frequency points:

[0099] (8)

[0100] Next, combined Figure 3 The self-supervised radiation source individual fingerprint calibration model of the present invention is described in general.

[0101] Figure 3 This is a schematic diagram illustrating the structure of a self-supervised radiation source individual fingerprint calibration model, as shown in an exemplary embodiment of the present invention. Figure 3 As shown, firstly, the multi-node observation spectrum formed by receiving nodes 1 to n for the same radiation source is obtained. Then, based on the frequency energy of the ideal reference spectrum, the consistency between nodes, and the observation reconstruction residual, low-energy frequency points, strong interference frequency points, and multi-node inconsistent frequency points are masked or weighted to obtain reliable frequency points and their corresponding reliable frequency point weights.

[0102] Subsequently, by combining the reliable frequency point weights and node quality weights, the frequency domain information corresponding to the effective frequency points is weighted and organized to form weighted spectral features, which are then input into the amplitude residual estimation module. The amplitude residual is composed of a low-dimensional smoothing basis function term and a sparse local residual term. The low-dimensional smoothing basis function term is used to correct the overall smoothing amplitude deviation, while the sparse local residual term is used to compensate for amplitude detail deviations at a small number of local frequency points.

[0103] The amplitude residual and the logarithmic amplitude spectrum of the initial reconstructed fingerprint are in Figure 3 The residual fusion nodes shown are superimposed and input into the fingerprint calibration module for amplitude calibration; the phase of the initial reconstructed fingerprint is input into the fingerprint calibration module via a bypass and remains unchanged during the calibration process. The calibrated amplitude spectrum is then recombined with the initial phase to obtain the current calibrated reconstructed fingerprint.

[0104] The current calibration and reconstructed fingerprint, along with the ideal reference spectrum and the equivalent frequency domain response of each receiving node, are used to reconstruct the observed spectrum of each receiving node. A self-supervised joint loss is constructed based on observation reconstruction consistency, multi-node consistency, multiple observation consistency, prior deviation, spectral smoothness, and preservation of local details. The amplitude residual parameters are updated according to the self-supervised joint loss until a preset convergence condition is met. The calibrated reconstructed fingerprint is then output, and feature extraction and individual radiation source identification are performed on it. The fingerprint calibration process does not use the ground truth fingerprint or radiation source category label; the radiation source category label can be used for subsequent classifier training but does not participate in the amplitude residual estimation and self-supervised joint optimization process.

[0105] After constructing the initial reconstructed fingerprint and related weights through the above embodiments, this system performs self-supervised amplitude residual calibration on the initial reconstructed fingerprint without using the truth fingerprint, and obtains the calibrated reconstructed fingerprint.

[0106] In practical applications, the reconstructed fingerprint is usually a complex frequency domain fingerprint. Therefore, its phase can be kept unchanged, and only the amplitude part can be calibrated to obtain the calibrated complex reconstructed fingerprint. :

[0107] (9)

[0108] in, Indicates the initial reconstructed fingerprint phase, Reconstructed fingerprint after calibration The logarithmic amplitude spectrum.

[0109] Based on formula (9), it can be seen that calibrating the complex initial reconstructed fingerprint is equivalent to calibrating its logarithmic amplitude spectrum:

[0110] (10)

[0111] in, Indicates the initial reconstructed fingerprint The logarithmic amplitude spectrum, This represents the magnitude residual to be estimated.

[0112] In other words, for the initial reconstructed fingerprint The calibration is equivalent to the calibration of the amplitude residual. The estimation is used to calibrate the global and local biases in the logarithmic amplitude spectrum of the initial reconstructed fingerprint without altering the phase structure of the initial reconstructed fingerprint.

[0113] In this embodiment, to reduce the risk of overfitting, the magnitude residual to be estimated is constructed using low-dimensional basis function terms and sparse local terms. Specifically, the expression for the magnitude residual to be estimated is as follows:

[0114] (11)

[0115] in, This represents the magnitude residual to be estimated. Indicates the number of basis functions. Indicates the first A preset basis function, which can be a DCT basis, a PCA basis, or other preset smoothing basis; This represents the low-dimensional coefficients to be estimated. This represents the sparse local residual to be estimated.

[0116] By modeling using formula (11), the amplitude residual can simultaneously possess both overall smoothing correction capability and a small amount of local detail correction capability.

[0117] In some embodiments, the joint optimization objective function uses normalized reliable frequency point weights and normalized node quality weights to weight each constraint; the normalized reliable frequency point weights are determined based on one or more of the ideal reference signal energy of each frequency point in the effective frequency point set, the equivalent frequency domain response consistency among multiple receiving nodes, and the observation reconstruction residuals, and are used to adjust the contribution of different frequency points in the joint optimization objective function; the normalized node quality weights are used to adjust the contribution of different receiving nodes in the joint optimization objective function.

[0118] The calculation process of the normalized reliable frequency point weight and the normalized node quality weight in this embodiment can be referred to the previous embodiment on the weight construction module, and will not be repeated here.

[0119] The joint optimization objective function in this embodiment includes at least the observation reconstruction consistency constraint. Multi-node consistency constraints Consistency constraint of multiple observations Prior deviation constraint Spectral smoothness constraint Maintain constraints on local details .

[0120] Observation Reconstruction Consistency Constraint This is used to constrain the observation reconstruction residuals of each receiving node to approach zero, and its calculation formula is as follows:

[0121] (12)

[0122] Multi-node consistency constraints The initial fingerprint information separated by each receiving node is used to constrain the approximation of the initial reconstructed fingerprint, and its calculation formula is as follows:

[0123] (13)

[0124] Consistency constraint of multiple observations The formula is used to constrain the reconstructed fingerprints of the radiation source after calibration at different observation times to approximate each other, and its calculation formula is as follows:

[0125] (14)

[0126] Prior deviation constraint The magnitude residual is used to constrain its magnitude to approach zero, and its calculation formula is as follows:

[0127] (15)

[0128] Spectral smoothness constraint The amplitude of the frequency domain differential signal is used to constrain it to approach zero, and its calculation formula is as follows:

[0129] (16)

[0130] Local detail preservation constraints This is used to perform frequency domain local difference processing on the calibrated reconstructed fingerprint and the initial reconstructed fingerprint respectively to obtain a first difference feature and a second difference feature, and to constrain the first difference feature to approximate the second difference feature. The calculation formula is as follows:

[0131] (17)

[0132] in, and For observation index, This refers to the set of observed signals from the radiation source within the same time period. For receiving node index, For frequency point index, To preserve the weights of pre-defined details used to represent local spectral peaks or valleys, or high-discrimination frequencies, This is a frequency domain local difference operator. Those skilled in the art can construct this frequency domain local difference operator using relevant technical solutions. This embodiment does not impose any special limitations on this.

[0133] After constructing various constraints through the above embodiments, the joint optimization objective function of this embodiment can be obtained. :

[0134] (18)

[0135] in, , , , , and These represent the weighting coefficients of the corresponding loss terms.

[0136] After obtaining the joint optimization objective function Then, iterative optimization is performed alternately.

[0137] In a specific implementation, first let , and with As the initial reconstructed fingerprint.

[0138] The above initialization means that no additional magnitude correction is introduced at the beginning of the iteration. The initial reconstructed fingerprint is used as the baseline for self-supervised calibration, so that subsequent optimization only learns the necessary magnitude residuals relative to the initial reconstruction results, thereby reducing the risk of overcalibration under truth-less conditions.

[0139] Then, fix the current reconstructed fingerprint. Estimating the equivalent frequency domain response Then fix the equivalent frequency domain response Update the parameters in the magnitude residual.

[0140] set up This represents the set of parameters in the magnitude residual to be estimated. Including low-dimensional coefficients and sparse local residuals Then the first In each iteration, updates can be performed as follows:

[0141] (19)

[0142] in, This indicates the step size for optimization.

[0143] This embodiment can employ gradient descent, Adam, LBFGS, alternating minimization, or other numerical optimization methods to achieve the update. After each iteration, the update can be performed on... Amplitude clipping, smoothing constraints, and sparse gating are applied to avoid excessive fingerprint correction caused by individual abnormal frequency points.

[0144] When the decrease in the total loss function is less than a preset threshold, or when the preset number of iterations is reached, the optimization stops, thus obtaining the optimal magnitude residual. Calculate the final logarithmic amplitude spectrum:

[0145] (20)

[0146] Based on the final logarithmic amplitude spectrum, the calibrated reconstructed fingerprint is obtained. :

[0147] (twenty one)

[0148] Because this embodiment performs residual calibration on the logarithmic amplitude spectrum of the initial reconstructed fingerprint, the initial phase remains unchanged when constructing the final complex reconstructed fingerprint. No real fingerprint tags are used throughout the optimization process, making it suitable for real-world radiation source observation scenarios.

[0149] After obtaining the calibrated reconstructed fingerprint through the above embodiments, this system completes the individual identification of radiation sources based on the calibrated reconstructed fingerprint.

[0150] This system first extracts reconstructed fingerprint features. Based on the calibrated reconstructed fingerprint, one or more of the following can be extracted as identification features: amplitude spectrum, phase spectrum, group delay, real part of complex spectrum, imaginary part of complex spectrum, and amplitude residual.

[0151] For example, the vector representation of the identified features is as follows:

[0152] (twenty two)

[0153] in, Indicates the first The reconstructed fingerprint recognition feature vector under the second observation. Indicates phase, Indicates group delay, , These represent the real and imaginary parts of the complex number spectrum, respectively.

[0154] The identification features are input into a classifier to obtain the individual identification results of the radiation source:

[0155] (twenty three)

[0156] in, Indicates the first The predicted category corresponding to the next observation. This represents the class probability output by the classifier.

[0157] The individual identification results of the radiation source under multiple observations are obtained, and the individual identification results under multiple observations are aggregated to obtain the final identification result of the radiation source.

[0158] Specifically, multiple observations were conducted on the radiation source, and based on... Figure 1 The method shown calculates the corresponding individual identification result for each observation, and aggregates the classification probabilities of multiple identifications by averaging, median, truncated mean, or quality-weighted aggregation to obtain the final identification category. , This represents the weighted result of class probabilities after multiple identifications.

[0159] Figure 4This is a schematic diagram of an electronic device illustrated in this specification according to an exemplary embodiment. Please refer to... Figure 4 At the hardware level, the device includes a processor 410, an internal bus 420, a network interface 430, memory 440, a hardware acceleration device 450, and non-volatile memory 460, and may also include other hardware required for its functions. One or more embodiments of the present invention can be implemented in software, for example, the processor 410 reads the corresponding computer program from the non-volatile memory 460 into the memory 440 and then runs it. Of course, in addition to software implementation, one or more embodiments of the present invention do not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution subject of the above processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0160] Figure 5 This is a structural block diagram of a self-calibrating radiation source individual fingerprint reconstruction and identification device for distributed reception, as illustrated in an exemplary embodiment of the present invention. The radiation source individual identification device can be applied to, for example... Figure 4 The electronic device shown implements the technical solution of the present invention. The radiation source individual identification device includes: a signal acquisition unit 510, a fingerprint calculation unit 520, a fingerprint calibration unit 530, and a radiation source individual identification unit 540, wherein:

[0161] The signal acquisition unit 510 is used to acquire the observation signals of multiple receiving nodes in a distributed deployment for the same radiation source, and estimate the ideal reference signal of the radiation source based on the observation signals. The ideal reference signal is used to characterize the fingerprint information of the radiation source without superimposed and the ideal baseband waveform of external non-ideal link distortion.

[0162] The fingerprint calculation unit 520 is used to estimate the equivalent frequency domain response from the radiation source to each receiving node based on the observation signals of each receiving node and the ideal reference signal, separate the initial fingerprint information from the equivalent frequency domain response based on the effective frequency point set, and aggregate the initial fingerprint information of multiple receiving nodes to obtain the initial reconstructed fingerprint of the radiation source.

[0163] The fingerprint calibration unit 530 is used to construct an amplitude residual to be estimated based on a preset orthogonal basis function and the sparse local residual to be estimated. The amplitude residual reflects the amplitude difference between the initial reconstructed fingerprint and the true value of the radiation source fingerprint. Within the effective frequency point set, the optimal amplitude residual is obtained by minimizing a joint optimization objective function. The initial reconstructed fingerprint is then calibrated based on the optimal amplitude residual to obtain a calibrated reconstructed fingerprint. The joint optimization objective function includes at least the following:

[0164] The observation reconstruction consistency constraint is used to constrain the observation reconstruction residual of each receiving node to approach zero. The observation reconstruction residual is the difference between the reconstructed spectrum of each receiving node and the spectrum of the observed signal of that receiving node. The reconstructed spectrum of each receiving node is constructed using the equivalent frequency domain response of each receiving node, the ideal reference signal, and the calibrated reconstruction fingerprint.

[0165] Multi-node consistency constraint is used to constrain the initial fingerprint information separated by each receiving node to approximate the initial reconstructed fingerprint;

[0166] Multiple observation consistency constraints are used to ensure that the reconstructed fingerprints of the radiation source after calibration at different observation times approximate each other;

[0167] A priori deviation constraint is used to constrain the magnitude of the magnitude residual to approach zero;

[0168] Spectral smoothness constraint is used to construct a frequency domain differential signal by interpolation of the amplitude residual at adjacent frequency points, and to constrain the amplitude of the frequency domain differential signal to approach zero.

[0169] Local detail preservation constraints are used to perform frequency domain local difference processing on the calibrated reconstructed fingerprint and the initial reconstructed fingerprint respectively to obtain a first difference feature and a second difference feature, and to constrain the first difference feature to approximate the second difference feature;

[0170] The radiation source individual identification unit 540 is used to extract features from the calibrated reconstructed fingerprint and obtain the identification result of the radiation source based on the extracted identification features.

[0171] In some embodiments, the fingerprint calculation unit 520 is used to acquire the observation reconstruction residual, inconsistency metric, quality factor, and frequency point missing ratio of each receiving node; wherein, the inconsistency metric is the degree of deviation of each receiving node from the metric center determined by multiple receiving nodes based on the equivalent frequency domain response, initial fingerprint information, and / or observation reconstruction residual at the same acquisition time; the quality factor is determined based on the signal-to-noise ratio estimate of the observed signal of each receiving node; the frequency point missing ratio is the ratio of the number of invalid frequency points in the observed signal of each receiving node to the total number of frequency points in the set of valid frequency points; the node quality weight of each receiving node is obtained based on the observation reconstruction residual, inconsistency metric, quality factor, and frequency point missing ratio; and the initial fingerprint information of the multiple receiving nodes is weighted and aggregated based on the normalized node quality weight of each receiving node to obtain the initial reconstructed fingerprint.

[0172] In some embodiments, the identification features include one or more of amplitude spectrum, phase spectrum, group delay, real part of complex spectrum, imaginary part of complex spectrum, and amplitude residual. The radiation source individual identification unit 540 is used to input the identification features into a classifier to obtain the individual identification result of the radiation source; obtain the individual identification results of the radiation source under multiple observations; and aggregate the individual identification results under multiple observations to obtain the final identification result of the radiation source.

[0173] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the present invention according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0174] Accordingly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any of the above embodiments.

[0175] Accordingly, embodiments of the present invention also provide a computer program product configured to perform the methods described in any of the above embodiments.

[0176] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer, which can take the form of a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email sending and receiving device, game console, tablet computer, wearable device, or any combination of these devices.

[0177] In a typical configuration, a computer includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0178] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0179] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random-access memory (SRAM), dynamic random-access memory (DRAM), other types of random-access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage, quantum memory, graphene-based storage media or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0180] While this specification contains numerous specific implementation details, these should not be construed as limiting the scope of any invention or the scope of the claims, but rather are primarily intended to describe features of specific embodiments of a particular invention. Certain features described in the various embodiments herein may also be implemented in combination in a single embodiment. Conversely, various features described in a single embodiment may also be implemented separately in various embodiments or in any suitable sub-combination. Furthermore, while features may function in certain combinations as described above and even initially claimed in this way, one or more features from a claimed combination may be removed from that combination in some cases, and a claimed combination may refer to a sub-combination or a variation thereof.

[0181] Similarly, although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0182] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims may be performed in a different order and still achieve the desired result. Furthermore, the processes depicted in the drawings are not necessarily shown in a specific order or sequence to achieve the desired result. In some implementations, multitasking and parallel processing may be advantageous.

[0183] It should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0184] 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, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A self-calibration radiation source individual fingerprint reconstruction and identification method for distributed reception, characterized in that, Includes the following steps: Step S1: Obtain the observation signals of multiple receiving nodes in a distributed deployment for the same radiation source, and estimate the ideal reference signal of the radiation source based on the observation signals. The ideal reference signal is used to characterize the fingerprint information of the radiation source without superimposed and the ideal baseband waveform of external non-ideal link distortion. Step S2: Based on the observed signals of each receiving node and the ideal reference signal, estimate the equivalent frequency domain response from the radiation source to each receiving node, separate the initial fingerprint information from the equivalent frequency domain response based on the effective frequency point set, and aggregate the initial fingerprint information of multiple receiving nodes to obtain the initial reconstructed fingerprint of the radiation source. Step S3: Construct the amplitude residual to be estimated based on the preset orthogonal basis functions and the sparse local residual to be estimated. The amplitude residual reflects the amplitude difference between the initial reconstructed fingerprint and the true value of the radiation source fingerprint. Within the effective frequency point set, obtain the optimal amplitude residual by minimizing the joint optimization objective function. Then, perform amplitude calibration on the initial reconstructed fingerprint based on the optimal amplitude residual to obtain the calibrated reconstructed fingerprint. The joint optimization objective function includes at least: The observation reconstruction consistency constraint is used to constrain the observation reconstruction residual of each receiving node to approach zero. The observation reconstruction residual is the difference between the reconstructed spectrum of each receiving node and the spectrum of the observed signal of that receiving node. The reconstructed spectrum of each receiving node is constructed using the equivalent frequency domain response of each receiving node, the ideal reference signal, and the calibrated reconstruction fingerprint. Multi-node consistency constraint is used to constrain the initial fingerprint information separated by each receiving node to approximate the initial reconstructed fingerprint; Multiple observation consistency constraints are used to ensure that the reconstructed fingerprints of the radiation source after calibration at different observation times approximate each other; A priori deviation constraint is used to constrain the magnitude of the magnitude residual to approach zero; Spectral smoothness constraint is used to construct a frequency domain differential signal by interpolation of the amplitude residual at adjacent frequency points, and to constrain the amplitude of the frequency domain differential signal to approach zero. Local detail preservation constraints are used to perform frequency domain local difference processing on the calibrated reconstructed fingerprint and the initial reconstructed fingerprint respectively to obtain a first difference feature and a second difference feature, and to constrain the first difference feature to approximate the second difference feature; Step S4: Extract features from the calibrated reconstructed fingerprint, and obtain the identification result of the radiation source based on the extracted identification features.

2. The method of claim 1, wherein, The expression for the set of effective frequency points in step S2 is: ； wherein, denotes the set of valid frequency points corresponding to the -th observation, denotes the energy of the ideal reference signal at the -th frequency point, denotes the amplitude inconsistency measure of the observation signals of the plurality of receiving nodes at the -th frequency point, denotes the spectral residual between the observation signals and the ideal reference signal at the , , denote the preset energy threshold, consistency threshold and residual threshold, respectively.

3. The method according to claim 1, characterized in that, The initial fingerprint information in step S2 is obtained using the following formula: ； in, Indicates the first The first observation The equivalent frequency domain response of each receiving node Indicates the first The set of effective frequency points corresponding to each observation This indicates the initial fingerprint information.

4. The method according to claim 1, characterized in that, Step S2 involves aggregating the initial fingerprint information of multiple receiving nodes, including: The observation reconstruction residual, inconsistency metric, quality factor, and frequency point missing ratio of each receiving node are obtained. The inconsistency metric is the deviation of each receiving node from the metric center determined by the equivalent frequency domain response, initial fingerprint information, and / or observation reconstruction residual at the same acquisition time. The quality factor is determined based on the estimated signal-to-noise ratio of the observed signals of each receiving node. The frequency point missing ratio is the ratio of the number of invalid frequency points in the observed signals of each receiving node to the total number of frequency points in the set of valid frequency points. The node quality weight of each receiving node is obtained based on the observation reconstruction residual, inconsistency measure, quality factor and frequency point missing ratio. Based on the normalized node quality weights of each receiving node, the initial fingerprint information of the multiple receiving nodes is weighted and aggregated to obtain the initial reconstructed fingerprint.

5. The method according to claim 1, characterized in that, The expression for the magnitude residual to be estimated in step S3 is as follows: ； in, This represents the magnitude residual to be estimated. Indicates the number of basis functions. Indicates the first A pre-defined basis function, This represents the low-dimensional coefficients to be estimated. This represents the sparse local residual to be estimated.

6. The method according to claim 4, characterized in that, The joint optimization objective function in step S3 uses normalized reliable frequency point weights and normalized node quality weights to weight each constraint; where: The normalized reliable frequency point weights are determined based on one or more of the ideal reference signal energy of each frequency point in the effective frequency point set, the consistency of the equivalent frequency domain response among multiple receiving nodes, and the observation reconstruction residuals, and are used to adjust the contribution of different frequency points to the joint optimization objective function. The normalized node quality weights are used to adjust the contribution of different receiving nodes to the joint optimization objective function.

7. The method according to claim 6, characterized in that: The observation reconstruction consistency constraint ; The multi-node consistency constraint ; The Consistency Consistency of Multiple Observations ; The prior deviation constraint ; The spectral smoothness constraint ; The local detail preservation constraint ;in, and For observation index, This refers to the set of observed signals from the radiation source within the same time period. For receiving node index, For frequency point index, To normalize the reliable frequency point weights, To normalize node quality weights, To observe the reconstructed residuals, The amplitude spectrum of the initial fingerprint information. To initially reconstruct the amplitude spectrum of the fingerprint, For amplitude residuals, The amplitude spectrum of the initial reconstructed fingerprint.

8. The method according to claim 1, characterized in that, The identification features include one or more of the following: amplitude spectrum, phase spectrum, group delay, real part of complex spectrum, imaginary part of complex spectrum, and amplitude residual. Step S4 includes: The identification features are input into a classifier to obtain the individual identification results of the radiation source; The individual identification results of the radiation source under multiple observations are obtained, and the individual identification results under multiple observations are aggregated to obtain the final identification result of the radiation source.

9. A self-calibrating radiation source individual fingerprint reconstruction and identification device for distributed reception, characterized in that, The device includes: The signal acquisition unit is used to acquire the observation signals of multiple receiving nodes in a distributed deployment for the same radiation source, and to estimate the ideal reference signal of the radiation source based on the observation signals. The ideal reference signal is used to characterize the fingerprint information of the radiation source without superimposed and the ideal baseband waveform of external non-ideal link distortion. The fingerprint calculation unit is used to estimate the equivalent frequency domain response from the radiation source to each receiving node based on the observation signals of each receiving node and the ideal reference signal, separate the initial fingerprint information from the equivalent frequency domain response based on the effective frequency point set, and aggregate the initial fingerprint information of multiple receiving nodes to obtain the initial reconstructed fingerprint of the radiation source. A fingerprint calibration unit is used to construct an amplitude residual to be estimated based on a preset orthogonal basis function and the sparse local residual to be estimated. The amplitude residual reflects the amplitude difference between the initial reconstructed fingerprint and the true value of the radiation source fingerprint. Within the effective frequency point set, an optimal amplitude residual is obtained by minimizing a joint optimization objective function. The initial reconstructed fingerprint is then calibrated based on the optimal amplitude residual to obtain a calibrated reconstructed fingerprint. The joint optimization objective function includes at least the following: The observation reconstruction consistency constraint is used to constrain the observation reconstruction residual of each receiving node to approach zero. The observation reconstruction residual is the difference between the reconstructed spectrum of each receiving node and the spectrum of the observed signal of that receiving node. The reconstructed spectrum of each receiving node is constructed using the equivalent frequency domain response of each receiving node, the ideal reference signal, and the calibrated reconstruction fingerprint. Multi-node consistency constraint is used to constrain the initial fingerprint information separated by each receiving node to approximate the initial reconstructed fingerprint; Multiple observation consistency constraints are used to ensure that the reconstructed fingerprints of the radiation source after calibration at different observation times approximate each other; A priori deviation constraint is used to constrain the magnitude of the magnitude residual to approach zero; Spectral smoothness constraint is used to construct a frequency domain differential signal by interpolation of the amplitude residual at adjacent frequency points, and to constrain the amplitude of the frequency domain differential signal to approach zero. Local detail preservation constraints are used to perform frequency domain local difference processing on the calibrated reconstructed fingerprint and the initial reconstructed fingerprint respectively to obtain a first difference feature and a second difference feature, and to constrain the first difference feature to approximate the second difference feature; The radiation source individual identification unit is used to extract features from the calibrated reconstructed fingerprint and obtain the identification result of the radiation source based on the extracted identification features.

10. An electronic device, characterized in that, include: processor; as well as A computer-readable storage medium storing computer program instructions that, when executed by the processor, cause the processor to perform the method as described in any one of claims 1 to 8.

Images