Segmented Fourier transform feature-based mixed interference type identification method and device

Abstract

The invention discloses a mixed interference type identification method and device based on segmented Fourier transform features, and the method comprises the steps: extracting the basic features and segmented Fourier features of a to-be-identified signal, and enabling the segmented Fourier features to be obtained through the Fourier transform of the to-be-identified signal after the to-be-identified signal is segmented; and inputting the basic features and the segmented Fourier features into a trained identification network to obtain a classification result of the to-be-identified signal. Due to the fact that the segmented Fourier features can more accurately represent the difference between different interference signals, the method extracts the segmented Fourier features of the to-be-recognized signals, recognizes the to-be-recognized signals according to the segmented Fourier features and the basic features, and can improve the recognition precision of the interference signals.

Description

Technical Field

[0001] The present invention belongs to the technical field of radar signal processing, and in particular relates to a method and device for identifying mixed interference types based on segmented Fourier transform features. Background Art

[0002] Traditional radar active jammer detection involves manually extracting jammer features in feature domains such as the time and frequency domains after receiving a jammer signal using mathematical statistics. These features are then input into a trained classifier to classify the jammer. In this process, the quality of the feature parameters extracted manually through expert knowledge directly impacts the classifier's training and recognition results. Therefore, the extracted features must exhibit stable discriminability, meaning they must be free of aliasing within the feature space. Furthermore, the classifiers used in traditional jammer detection are mostly derived using machine learning algorithms, primarily including decision trees, support vector machines, K-nearest neighbor algorithms, and BP neural networks. The advantages of these machine learning algorithms include fast computational speed, low memory usage, relatively simple algorithms, ease of understanding, and strong interpretability. These classifiers also require a sufficient number of valid samples as a training set to optimize the classifier parameters.

[0003] With the increasing complexity of modern electromagnetic environments, various active interference signals frequently appear, significantly reducing radar detection performance. Effective countermeasures are essential to ensure the proper functioning of radar systems. Interference identification, as the primary step in countering interference, plays a crucial role in accurately identifying interference types and guiding the implementation of effective countermeasures. The development of this technology is of great practical significance and urgency. However, traditional interference identification methods suffer from poor interference identification performance due to the lack of representativeness of the extracted features. Summary of the Invention

[0004] The embodiments of the present invention provide a method and device for identifying mixed interference types based on segmented Fourier transform features, which can solve the above technical problems.

[0005] In a first aspect, an embodiment of the present invention provides a method for identifying mixed interference types based on segmented Fourier transform features, the method comprising:

[0006] Extracting basic features and segmented Fourier features of the signal to be identified, wherein the segmented Fourier features are obtained by performing Fourier transform on the segmented signal to be identified;

[0007] The basic features and the segmented Fourier features are input into a trained recognition network to obtain a classification result of the signal to be recognized.

[0008] In a second aspect, an embodiment of the present invention provides a mixed interference type identification device based on segmented Fourier transform features, the device including a feature extraction unit and an identification unit;

[0009] The feature extraction unit is used to extract the basic features and segmented Fourier features of the signal to be identified, wherein the segmented Fourier features are obtained by performing Fourier transform on the segmented signal to be identified;

[0010] The recognition unit is used to input the basic features and the segmented Fourier features into a trained recognition network to obtain a classification result of the signal to be recognized.

[0011] The advantageous effect of the embodiments of the present invention compared with the prior art is that: since the segmented Fourier features can more accurately represent the differences between different interference signals, the present invention can improve the recognition accuracy of interference signals by extracting the segmented Fourier features of the signal to be identified and identifying the signal to be identified based on the segmented Fourier features and the basic features as a benchmark. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A schematic diagram of the structure of an identification network provided by an embodiment of the present invention;

[0013] Figure 2 A flowchart of a method for identifying mixed interference types based on segmented Fourier transform features provided by an embodiment of the present invention;

[0014] Figure 3 A schematic structural diagram of a mixed interference type identification device based on segmented Fourier transform features provided by an embodiment of the present invention;

[0015] Figures 4-10 Schematic diagram of the change of different segmented Fourier characteristics of the suppressed interference and its composite interference with JNR provided by an embodiment of the present invention;

[0016] Figure 11 A schematic diagram of a recognition rate curve of composite interference provided by an embodiment of the present invention;

[0017] Figure 12 A comparison chart of the effects of recognition based on different features provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In the following description, specific details such as particular system structures and techniques are provided for purposes of illustration, not limitation, to facilitate a thorough understanding of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the present invention may be practiced in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present invention with unnecessary detail.

[0019] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0020] It will also be understood that the term "and / or" used in the present description and appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0021] As used in the present specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0022] In addition, in the description of the present specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0023] References to "one embodiment" or "some embodiments" in the present specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present invention. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in yet other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0024] The present invention will be further described in detail below with reference to specific examples, but the embodiments of the present invention are not limited thereto.

[0025] Example 1

[0026] Figure 1 The figure shows a schematic diagram of the structure of an identification network provided by an embodiment of the present invention.

[0027] In one possible implementation, see Figure 1 The identification network 100 may include a primary identification network and multiple secondary identification networks. A secondary identification network may correspond one-to-one to each classification of the primary identification network. The primary identification network may be used to identify the interference category of the signal to be identified, and input the identified interference category into the corresponding secondary identification network. The secondary identification network then identifies the interference subcategory and outputs the interference subcategory as the classification result.

[0028] For example, if the primary recognition network can identify five categories, there can be five secondary recognition networks. When the primary recognition network identifies a signal as interference of suppression interference and its combined interference, the signal to be identified can be input into the secondary recognition network corresponding to that category for further identification.

[0029] For example, the recognition network can be a neural network model, whose structure is developed by simulating the information exchange between neurons. Multiple neurons can form a neural network. Complex neural networks composed of neurons are robust to noise and adaptable to different environments. With sufficient data and training time, they can effectively approximate complex linear relationships.

[0030] For example, the recognition network can be trained by minimizing the loss function using a back-propagation algorithm, and an optimization algorithm such as gradient descent or conjugate gradient is used to adjust the weights and biases of each neuron.

[0031] Specifically, the loss function can be expressed as the mean squared error or absolute error between the predicted and actual results. When the first sample enters the network, it initially follows the forward propagation method to obtain the output of each layer, and then calculates the loss based on the output of each layer based on the loss function. This loss is back-propagated through the network to calculate the error of each neuron output; this iterative process continues until the output loss is minimized.

[0032] In one example, the major categories of interference that the first-level identification network can identify may include: "suppression interference and its composite interference", "deception interference and its composite interference", "noise amplitude modulation interference + deception interference", "noise frequency modulation interference + deception interference", and "noise sweep frequency interference + deception interference".

[0033] Exemplarily, the jamming may include noise amplitude modulation jamming (AMJ), noise frequency modulation jamming (FMJ), and noise sweep-frequency jamming (NSJ).

[0034] Specifically, amplitude modulation refers to the use of a modulating signal to influence the amplitude of a carrier signal. That is, the amplitude of the carrier signal is a function of the modulating signal, but the carrier frequency is not affected by the modulating signal. Noise AM interference refers to the amplitude of the RF signal being affected by noise and being a function of the modulating noise. Its mathematical model is expressed as:

[0035]

[0036] Among them, J(t) is the suppressed interference signal, A is the amplitude of the carrier signal, u(t) is the modulation noise, ω c is the carrier frequency, is the initial phase.

[0037] Specifically, frequency modulation (FM) involves changing the frequency of a carrier signal according to a modulating signal, while the amplitude of the carrier signal is not affected by the modulating signal. In noisy FM interference, the frequency of the RF signal is affected by the noise and becomes a function of the modulating noise. Its mathematical model can be expressed as:

[0038]

[0039] Among them, K FM is the frequency modulation index, which represents the change in signal frequency per unit amplitude of the modulation signal.

[0040] Specifically, the characteristics of noise scanning interference include a modulated noise spectrum that changes periodically over time. The mathematical expression of this interference can be described as:

[0041]

[0042] Where μ″ represents the frequency modulation slope, and the noise sweep interference signal appears as noise distributed around zero in the time domain.

[0043] Exemplarily, deception jamming may include deception jamming for distance (DDJ), interrupted sampling and repeater jamming (ISRJ), and spectrum dispersion jamming (SDJ).

[0044] Specifically, range deception jamming typically occurs when a reconnaissance receiver intercepts a radar signal, delays it by a certain amount, and retransmits it. This creates a false target echo near the actual target, interfering with the radar's ability to accurately detect the real target. The specific mathematical expression for the range deception jamming signal is:

[0045]

[0046] Where S(t) is the deceptive jamming signal, t0 is the target echo delay, and τ is the jamming delay.

[0047] Specifically, Interrupted Sampling Repeater Jamming (ISRJ) is a type of radar active jamming that, unlike traditional jamming methods, captures intercepted signal fragments in a short period of time and quickly retransmits them. Furthermore, the signal can be retransmitted multiple times, and its power can be amplified to interfere with the normal operation of the radar. This makes ISRJ a very effective jamming suppression technology. The mathematical expression for ISRJ is:

[0048]

[0049] Where n is the number of samples, T' is the width of each intermittent sampling, T S is the sampling period, T P is the radar pulse width.

[0050] Specifically, spectral dispersion interference is composed of multiple sub-pulses, generated by time-domain compression of captured radar signals and repeated retransmission. This process can generate multiple false targets, thereby interfering with the radar system. Its specific mathematical expression is:

[0051]

[0052] Among them, μ' is the frequency modulation slope, A i is the amplitude of the i-th false target.

[0053] Therefore, the interference subcategories under the general category of suppression interference and its combined interference may include: AMJ+FMJ+NSJ, AMJ+FMJ, FMJ+NSJ, and AMJ+NSJ. The interference subcategories under the general category of deceptive interference and its combined interference may include: DDJ+ISRJ+SSJ, DDJ+ISRJ, ISRJ+SSJ, and DDJ+SSJ. The interference subcategories under the general category of noise amplitude modulation interference + deceptive interference may include: AMJ+DDJ+SSJ, AMJ+DDJ, AMJ+SSJ, etc. The interference subcategories under the general category of noise frequency modulation interference + deceptive interference may include: FMJ+SSJ+ISRJ, FMJ+SSJ, FMJ+ISRJ, etc. The interference subcategories under the general category of noise sweep frequency interference + deceptive interference may include: NSJ+DDJ+ISRJ, NSJ+DDJ, NSJ+ISRJ, etc.

[0054] The present invention identifies interference signals through a two-stage cascaded identification network, can accurately locate each interference component of the composite interference signal, and improve the identification accuracy of the interference signal.

[0055] Example 2

[0056] The mixed interference type identification method based on segmented Fourier transform characteristics provided by the embodiment of the present invention can be applied to electronic devices such as mobile terminals, personal laptops, supercomputers, etc. The embodiment of the present invention does not impose any restrictions on the specific type of electronic devices.

[0057] Figure 2 The flowchart shown is an implementation flow of a method for identifying mixed interference types based on segmented Fourier transform features, provided by an embodiment of the present invention. By way of example and not limitation, the method can be applied to the aforementioned electronic device. The method may include steps S201-S202, each of which is described below.

[0058] S201, extracting basic features and segmented Fourier features of the signal to be identified.

[0059] In a possible implementation, the segmented Fourier features are obtained by performing Fourier transform on the signal to be identified after segmenting it.

[0060] In one example, the signal to be identified can be segmented first, and each segmented sub-signal can be Fourier transformed to obtain a segmented frequency domain signal; then, the Fourier features of the segmented frequency domain signal are extracted, and the mean of the Fourier features of all segmented frequency domain signals is used as the segmented Fourier feature of the signal to be identified.

[0061] For example, the piecewise Fourier transform (SFT) is a method for analyzing signals in both the time and frequency domains. It involves breaking the signal into shorter segments, each of which can be approximated as a stationary signal within a local time frame, and then performing a Fourier transform on each segment. The advantage of the SFT is that it provides local information about the signal in both the time and frequency domains, allowing for more precise extraction of characteristic parameters and facilitating accurate signal recognition.

[0062] S202: Input the basic features and the segmented Fourier features into a trained recognition network to obtain a classification result of the signal to be recognized.

[0063] Exemplarily, the identification network may be the identification network 100 described above.

[0064] Since the segmented Fourier features can more accurately represent the differences between different interference signals, the present invention can improve the recognition accuracy of interference signals by extracting the segmented Fourier features of the signal to be identified and identifying the signal to be identified based on the segmented Fourier features and basic features.

[0065] In some embodiments, the basic features may include the time domain moment skewness, frequency domain moment skewness, time domain moment kurtosis, frequency domain moment kurtosis, time domain envelope fluctuation, frequency domain envelope fluctuation, normalized instantaneous amplitude spectrum maximum value, standard deviation of the absolute value of the central nonlinear component of the instantaneous phase, standard deviation of the direct value of the central nonlinear component of the instantaneous phase, distribution bandwidth, distribution variance, correlation ratio parameter, fast intrapulse modulation mode identification parameter, carrier factor, probability within phase threshold, and normalized spectrum 3dB bandwidth.

[0066] In one possible implementation, the moment skewness coefficient can reflect the degree of symmetry of the statistical distribution of the signal to be identified X. The time domain moment skewness and the frequency domain moment skewness can be calculated using the following formula:

[0067]

[0068] Where a3 is the time domain moment skewness / frequency domain moment skewness, μ represents the mean of the signal to be identified X in the time domain / frequency domain, and σ represents the standard deviation of the signal to be identified in the time domain / frequency domain.

[0069] Generally, the moment skewness of the Gaussian distribution can be used as a standard to measure the degree of symmetry of the signal to be identified X. The moment skewness of the Gaussian distribution is 0. When the moment skewness is less than 0, it is considered that the left tail of the statistical distribution of the signal to be identified is longer than the right tail. If the moment skewness is greater than 0, it is considered that the right tail of the statistical distribution of the signal to be identified is longer than the left tail.

[0070] In one possible implementation, the kurtosis coefficient can reflect the sharpness of the statistical distribution of the signal to be identified X. The time domain moment kurtosis and the frequency domain moment kurtosis can both be calculated using the following formula:

[0071]

[0072] Where a4 is the time domain moment kurtosis / frequency domain moment kurtosis.

[0073] Similarly, the kurtosis of a Gaussian distribution can be used as a criterion to measure the symmetry of a signal to be identified. The kurtosis of a Gaussian distribution is 3. When the kurtosis is less than 3, the statistical distribution of the signal to be identified is considered relatively flat. If the kurtosis is greater than 3, the statistical distribution of the signal to be identified is considered relatively sharp.

[0074] In a possible implementation, the envelope fluctuation reflects the degree of change of the signal envelope. Assume that the received signal to be identified is Its sampling discrete sequence is expressed as x(n). By performing fast Fourier transform on its discrete sampling sequence, we can obtain the discrete Fourier transform sequence X(m). Then the square of its instantaneous envelope can be expressed as: |x(n)| 2 =Re 2 [x(n)]+Im 2 [x(n)], the square of its frequency domain envelope can be expressed as |X(m)| 2 =Re 2 [X(m)]+Im 2 [X(m)], the envelope fluctuation of the signal to be identified can be calculated from this. The envelope fluctuation of the time domain and frequency domain can be calculated by the following formula:

[0075]

[0076] Here, R is the time domain / frequency domain envelope fluctuation of the signal to be identified, σ is the standard deviation of the instantaneous envelope square of the signal or the signal frequency domain envelope square of the signal, and μ is the mean of the instantaneous envelope square of the signal or the signal frequency domain envelope square of the signal.

[0077] In one possible implementation, the Hilbert transform of the discrete sequence of samples of the signal to be identified can be used to obtain the analytical expression of x(n): Thus, the transformed signal to be identified can be separated, and the maximum value of its normalized instantaneous amplitude spectrum can be extracted according to its sampling discrete sequence and the transformed signal to be identified.

[0078] Exemplarily, the maximum value of the normalized instantaneous amplitude spectrum of the signal to be identified can be calculated using the following formula:

[0079]

[0080] Among them, γ max is the maximum value of the normalized instantaneous amplitude spectrum of the signal to be identified, N s is the number of sampling points.

[0081] in:

[0082] a cn (i) = a(i) / E[a(i)] - 1

[0083]

[0084] in, is the Hilbert transform of x(n), i.e. the signal to be identified after the transformation, and a(i) is the instantaneous amplitude of the signal to be identified.

[0085] Specifically, normalizing the instantaneous amplitude by the average value can eliminate the influence of the signal gain. Therefore, the maximum value of the instantaneous amplitude spectrum can be used to distinguish between signals containing amplitude fluctuation information and signals without amplitude fluctuation information.

[0086] In one possible implementation, according to The instantaneous phase of the signal to be identified can be obtained However, since the phase calculated by the inverse tangent function is defined in the interval Therefore, phase aliasing will occur. Therefore, it is necessary to perform phase aliasing on Φ(n) to transform the signal into the interval [0,2π]. After obtaining the instantaneous phase sequence Φ after aliasing, NL (n), we can then NL (n) Calculate the standard deviation of the absolute value of the central nonlinear component of the instantaneous phase of the signal to be identified.

[0087] Exemplarily, the phase aliasing process can be expressed as:

[0088]

[0089] Exemplarily, the standard deviation of the absolute value of the central nonlinear component of the instantaneous phase of the signal to be identified can be calculated by the following formula:

[0090]

[0091] Among them, σ dp is the standard deviation of the absolute value of the central nonlinear component of the instantaneous phase of the signal to be identified, a t It is an amplitude judgment threshold for judging weak signals. If it is lower than this threshold, the estimation of instantaneous phase is very sensitive to noise. c is Φ NL(i) Satisfies condition a n (i)>a t The number of signal points.

[0092] In one possible implementation, the standard deviation σ of the direct value of the central nonlinear component of the instantaneous phase of the signal to be identified is dp It can be calculated by the following formula:

[0093]

[0094] In a possible implementation, the correlation parameter ratio of the signal to be identified can be calculated using the following formula:

[0095]

[0096] Where S is the correlation parameter ratio of the signal to be identified, N and P are the number of positive and negative data points of B(t,τ) respectively. * (t-τ) is the autocorrelation function of the signal to be identified.

[0097] In a possible implementation, the fast intra-pulse modulation mode identification parameter C of the signal to be identified can be calculated using the following formula:

[0098]

[0099] Wherein, R(n) is the autocorrelation function of the signal, n1, n2, and n3 are the positions of the sampling points of the autocorrelation function, and the distances between them are equal.

[0100] In a possible implementation, the probability P of the phase of the signal to be identified being within the threshold is pth It can be calculated by the following formula:

[0101]

[0102] Specifically, P pth It can be defined as the ratio of the number of points within a certain threshold above and below the phase mean to the total number of points.

[0103] In a possible implementation, the normalized spectrum 3dB bandwidth B of the signal to be identified ω It can be calculated by the following formula:

[0104]

[0105] in:

[0106]

[0107] Wherein, S(n)=|FFT(x(n))| is the Fast Fourier Transform modulus of the signal to be identified, and N′ is the sampling length. The threshold is generally 0.707.

[0108] In an example, the Fourier features may include frequency domain moment skewness, frequency domain moment kurtosis, carrier frequency factor, additive white Gaussian noise factor A1, additive white Gaussian noise factor A2, normalized spectrum 3dB bandwidth, and frequency domain envelope fluctuation of the segmented frequency domain signal.

[0109] For example, the frequency domain moment skewness, frequency domain moment kurtosis, carrier frequency factor, normalized spectrum 3dB bandwidth, and frequency domain envelope fluctuation of the segmented frequency domain signal can be calculated by the above formulas (1.1), (1.2), (1.10), and (1.3), respectively.

[0110] Example 3

[0111] Figure 3 The diagram shows a schematic diagram of the structure of a mixed interference type identification device based on segmented Fourier transform features provided by an embodiment of the present invention. As an example and not a limitation, the device 300 may include a feature extraction unit and an identification unit.

[0112] Exemplarily, the feature extraction unit is used to extract the basic features and segmented Fourier features of the signal to be identified, wherein the segmented Fourier features are obtained by segmenting the signal to be identified and performing Fourier transform; the recognition unit is used to input the basic features and segmented Fourier features into the trained recognition network to obtain the classification results of the signal to be identified.

[0113] In one example, the basic features include the time domain moment skewness, frequency domain moment skewness, time domain moment kurtosis, frequency domain moment kurtosis, time domain envelope fluctuation, frequency domain envelope fluctuation, normalized instantaneous amplitude spectrum maximum value, standard deviation of the absolute value of the central nonlinear component of the instantaneous phase, standard deviation of the direct value of the central nonlinear component of the instantaneous phase, distribution bandwidth, distribution variance, correlation ratio parameter, fast intrapulse modulation mode identification parameter, carrier factor, probability within phase threshold, and normalized spectrum 3dB bandwidth.

[0114] In one example, the feature extraction unit can be specifically used to segment the signal to be identified, and perform Fourier transform on each segmented sub-signal to obtain a segmented frequency domain signal; extract the Fourier features of the segmented frequency domain signal, and take the mean of the Fourier features of all segmented frequency domain signals as the segmented Fourier features of the signal to be identified.

[0115] In one example, the Fourier features include frequency domain moment skewness, frequency domain moment kurtosis, carrier frequency factor, additive white Gaussian noise factor A1, additive white Gaussian noise factor A2, normalized spectrum 3dB bandwidth, and frequency domain envelope fluctuation of the segmented frequency domain signal.

[0116] In order to better demonstrate the beneficial effects of the present invention, the following simulation experiments were conducted:

[0117] Simulation Experiment 1

[0118] For example, simulation experiment 1 illustrates the effectiveness of the proposed method by comparing the changing trends of composite interference amplitude with signal-to-interference ratio under characteristic parameters such as segmented Fourier transform frequency domain moment skewness, segmented Fourier transform frequency domain moment kurtosis, segmented Fourier transform carrier frequency factor, segmented Fourier transform additive white Gaussian noise factor A1, segmented Fourier transform additive white Gaussian noise factor A2, segmented Fourier transform normalized spectrum 3dB bandwidth, and segmented Fourier transform frequency domain envelope fluctuation of five major types of interference composite signals.

[0119] Specifically, the dataset used in Experiment 1 can be simulated data. It simulates the six aforementioned interference signals and their composite interference signals within the interference-to-noise ratio (JNR) range of 0 to 20 dB. For each interference signal, 1000 random samples are generated, and their time domain, frequency domain, and segmented Fourier transform domain feature parameters are extracted as a training dataset. Furthermore, for each interference signal, 500 samples are generated with integer JNR values ​​within the JNR range of 10 to 20 dB as a validation dataset. The radar bandwidth is 10 MHz, the sampling frequency is 200 MHz, and the pulse width is 10 μs.

[0120] Specifically, in Experiment 1, three suppression jamming and three deceptive jamming signals were combined to generate a composite jamming signal. The radar signal waveform used in the radar suppression jamming simulation was a linear frequency modulation (LFM) signal. The cascaded inverse neural network was implemented using the built-in neural network toolbox in Matlab. All networks were configured as four-layer networks, with each layer containing 16 neurons. The activation function for each layer was set to the tanh function. Seven feature parameters were obtained using the segmented Fourier transform feature extraction method.

[0121] The characteristic parameters of the piecewise Fourier transform of the suppressed interference and its composite interference signal under 7 characteristic parameters vary with JNR. Figure 4-10 As shown. Figure 4-10 As can be seen, the overall trend of the characteristic parameters obtained through the piecewise Fourier transform is consistent with the frequency domain characteristic parameters. However, a key difference is that the distribution size of the piecewise Fourier transform characteristic parameters is larger and relatively more stable. This enhanced stability makes it more suitable for classification and recognition in neural networks.

[0122] Simulation Experiment 2

[0123] For example, in simulation experiment 2, the interference signal is first identified using the basic features, and then the segmented Fourier features of the signal to be identified are extracted and input into the recognition network together with the basic features, and the following is obtained: Figure 11 The recognition rate curves of the five composite interferences shown in Figure 12The comparison diagram before and after adding 7 segmented Fourier features is shown.

[0124] from Figure 11 As can be seen from the results, both for suppressive interference and its combined forms, and for deceptive interference and its combined forms, the recognition results are excellent. At a level greater than 3dB, the recognition accuracy is above 90%. For the remaining combined interference signals, the recognition accuracy reaches over 90% at a level greater than 5dB. This demonstrates the effectiveness of the present invention in identifying combined interference.

[0125] Afterwards, simulation experiment 2 also compared the recognition rate under interference-to-signal ratio conditions, ranging from 0 to 20 dB, using only 16 time and frequency domain features and using 23 features, including time, frequency and segmented Fourier transform domains. The comparison results are shown in Figure 12 By comparing the recognition rates using 16 features and 23 features, it is clear that the recognition rate using 23 features (including STFT features) is always higher than the recognition rate using only 16 features. The overall recognition rate is improved by 8%, which proves that the present invention can improve the recognition accuracy.

[0126] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0127] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

Claims

1. A mixed interference type identification method based on segmented Fourier transform features, characterized in that: include: Extracting basic features and segmented Fourier features of the signal to be identified, wherein the segmented Fourier features are obtained by performing Fourier transform on the segmented signal to be identified; The basic features and the segmented Fourier features are input into a trained recognition network to obtain a classification result of the signal to be recognized.

2. The method according to claim 1, characterized in that The basic features include the time domain moment skewness, frequency domain moment skewness, time domain moment kurtosis, frequency domain moment kurtosis, time domain envelope fluctuation, frequency domain envelope fluctuation, normalized instantaneous amplitude spectrum maximum value, standard deviation of the absolute value of the central nonlinear component of the instantaneous phase, standard deviation of the direct value of the central nonlinear component of the instantaneous phase, distribution bandwidth, distribution variance, correlation ratio parameter, fast pulse modulation mode identification parameter, carrier factor, probability within phase threshold, and normalized spectrum 3dB bandwidth of the signal to be identified.

3. The method according to claim 1, characterized in that Extracting the segmented Fourier features of the signal to be identified includes: Segmenting the signal to be identified, and performing Fourier transform on each segmented sub-signal to obtain a segmented frequency domain signal; The Fourier features of the segmented frequency domain signals are extracted, and the mean of the Fourier features of all segmented frequency domain signals is used as the segmented Fourier features of the signal to be identified.

4. The method according to claim 3, characterized in that The Fourier features include the frequency domain moment skewness, frequency domain moment kurtosis, carrier frequency factor, additive white Gaussian noise factor A1, additive white Gaussian noise factor A2, normalized spectrum 3dB bandwidth, and frequency domain envelope fluctuation of the segmented frequency domain signal.

5. The method according to claim 1, wherein The recognition network includes a primary recognition network and multiple secondary recognition networks, and the number of the secondary recognition networks is the same as the total number of categories of the primary recognition network; The first-level identification network is used to identify the interference category of the signal to be identified; The secondary recognition network is used to identify the interference subcategory of the signal to be identified, and output the interference subcategory as the classification result.

6. A mixed interference type identification device based on segmented Fourier transform features, characterized in that: including a feature extraction unit and a recognition unit; The feature extraction unit is used to extract the basic features and segmented Fourier features of the signal to be identified, wherein the segmented Fourier features are obtained by performing Fourier transform on the segmented signal to be identified; The recognition unit is used to input the basic features and the segmented Fourier features into a trained recognition network to obtain a classification result of the signal to be recognized.

7. The device according to claim 6, characterized in that The basic features include the time domain moment skewness, frequency domain moment skewness, time domain moment kurtosis, frequency domain moment kurtosis, time domain envelope fluctuation, frequency domain envelope fluctuation, normalized instantaneous amplitude spectrum maximum value, standard deviation of the absolute value of the central nonlinear component of the instantaneous phase, standard deviation of the direct value of the central nonlinear component of the instantaneous phase, distribution bandwidth, distribution variance, correlation ratio parameter, fast pulse modulation mode identification parameter, carrier factor, probability within phase threshold, and normalized spectrum 3dB bandwidth of the signal to be identified.

8. The device according to claim 6, characterized in that The feature extraction unit is specifically used for: Segmenting the signal to be identified, and performing Fourier transform on each segmented sub-signal to obtain a segmented frequency domain signal; The Fourier features of the segmented frequency domain signals are extracted, and the mean of the Fourier features of all segmented frequency domain signals is used as the segmented Fourier features of the signal to be identified.

9. The device according to claim 8, characterized in that The Fourier features include the frequency domain moment skewness, frequency domain moment kurtosis, carrier frequency factor, additive white Gaussian noise factor A1, additive white Gaussian noise factor A2, normalized spectrum 3dB bandwidth, and frequency domain envelope fluctuation of the segmented frequency domain signal.

10. The device according to claim 6, characterized in that The recognition network includes a primary recognition network and multiple secondary recognition networks, and the number of the secondary recognition networks is the same as the total number of categories of the primary recognition network; The first-level identification network is used to identify the interference category of the signal to be identified; The secondary recognition network is used to identify the interference subcategory of the signal to be identified, and output the interference subcategory as the classification result.

Images