Cognitive radio spectrum sensing method based on mixed graph transformation, medium and device

Abstract

The invention provides a cognitive radio spectrum sensing method based on mixed graph transformation, a medium and equipment, and the method comprises the steps: firstly carrying out the processing of an observation signal in two different modes, then obtaining two simple undirected graphs through different normalization, quantization and graph construction, calculating the feature value of an A alpha spectrum of the obtained undirected graphs, and carrying out the calculation of the feature value of the A alpha spectrum of the obtained undirected graphs; and aggregating the obtained eigenvalue vectors to obtain a detection statistical magnitude, and comparing the detection statistical magnitude with a set threshold to complete a spectrum sensing task. According to the method, the characteristic value vectors of different graph structures are fused to perform inner accumulation combination operation to obtain the detection statistics, so that spectrum sensing of common communication signals is realized, the method has better graph topology resolution capability, the separability of graph characteristics under two hypotheses is improved, and the detection performance at a low signal-to-noise ratio is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of signal processing, and in particular relates to a cognitive radio spectrum sensing method, medium and device based on hybrid graph transformation. Background Art

[0002] In today's era, with the innovation of communication technology and the surge in the number of wireless devices, the scarcity of spectrum resources has become increasingly prominent. Traditional static spectrum allocation methods fail to make full use of limited spectrum resources and will produce spectrum holes. Dynamic Spectrum Access (DSA) has emerged as a new way to improve the efficiency of spectrum resource utilization and reduce the waste of spectrum resources. Under the DSA framework, cognitive radio (CR) technology allows secondary users (SU) to use idle licensed frequency bands without interfering with the communication of primary users (PU), thereby making full use of existing spectrum resources.

[0003] Spectrum sensing technology is the premise and foundation for the effective operation of cognitive radio systems. Its core task is to detect whether the PU channel is occupied, that is, to detect spectrum holes for SU use. Existing spectrum sensing algorithms are mainly divided into two categories: one is model-driven spectrum sensing algorithms, which mainly include methods based on energy detection, covariance matrix detection, cyclostationary feature detection and matched filter detection; the other is data-driven spectrum sensing algorithms, which mainly use machine learning technology to complete signal detection and spectrum sensing. These algorithms usually extract features from different angles such as the signal's time domain waveform, frequency domain spectrum, and transform domain function. In order to significantly improve the performance of the algorithm, the distinguishability of features is indispensable. However, this requirement often means that the scale of the observation sample must be expanded, which correspondingly leads to an increase in the complexity of the algorithm. Therefore, in the design of spectrum sensing algorithms, there is a trade-off between algorithm complexity and feature distinguishability.

[0004] To meet this challenge, the emerging graph-domain spectrum sensing algorithm in recent years provides a solution. The algorithm maps signal samples into a specific graph topology and further processes it to convert time domain or spectrum data into a corresponding graph structure, and then uses the inherent properties of the graph and related algorithms to perform spectrum sensing. This method helps to resolve the contradiction between algorithm complexity and feature distinctiveness by deeply exploring the intrinsic structural characteristics of the signal. By analyzing and processing graph signals, the graph-domain spectrum sensing algorithm can obtain more accurate and reliable spectrum sensing results, further improving the performance of cognitive radio systems. However, the existing methods either use the power spectrum or the autocorrelation function, and there is no feature fusion between the two. Their detection performance needs to be improved at low signal-to-noise ratios. In addition, the problem of spectrum sensing of commonly used communication signals under the background of Gaussian white noise is studied. Since it becomes difficult to perform spectrum sensing through traditional time domain and frequency domain under low signal-to-noise ratios, Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a cognitive radio spectrum sensing method, medium and device based on hybrid graph transformation.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] The cognitive radio spectrum sensing method based on hybrid graph transformation is characterized by comprising the following steps:

[0008] Obtain the observed signal, calculate the power spectrum of the observed signal, perform nonlinear normalization and uniform quantization on the power spectrum, and then map it to obtain the graph G a , for graph G a A a The spectrum is eigendecomposed to obtain all eigenvalues, and the eigenvalues ​​are sorted to obtain the eigenvalue vector a;

[0009] Calculate the autocorrelation function of the observed signal, perform linear normalization and uniform quantization on the autocorrelation function, and then map it to obtain the graph G b , for graph G b A α The spectrum is decomposed to obtain all eigenvalues, and the eigenvalues ​​are sorted to obtain the eigenvalue vector b;

[0010] Perform inner product aggregation operation on eigenvalue vector a and eigenvalue vector b to obtain the detection statistic F;

[0011] The detection statistic F is compared with the set threshold to determine whether there is a primary user signal in the observed signal.

[0012] To optimize the above technical solutions, the specific measures taken also include:

[0013] Furthermore, the power spectrum of the observed signal is calculated by performing an N-point Fourier transform on the observed signal x(n) and taking the modulus, and the power spectrum R(k) is obtained as:

[0014]

[0015] Where N is the number of signal samples and k is the serial number of the power spectrum.

[0016] Furthermore, after the power spectrum is subjected to nonlinear normalization and uniform quantization, the graph mapping obtains the graph G a , specifically:

[0017] 1) After the power spectrum R(k) passes through the sigmoid function, it is mapped to the interval [0, 1] one by one to obtain the sequence Y(k):

[0018] 2) Uniformly quantize the sequence Y(k) to obtain the sequence Q(k):

[0019]

[0020] In the formula, μ represents the quantization level, and N0 represents the maximum quantization level;

[0021] 3) Map Q(k) to an undirected unweighted graph structure G a ={E a , V a}, E a and V a Represent the edge set and vertex set of the graph respectively; the vertex v of the graph μ The mapping rule with the quantization level in the quantization interval [0, 1] is:

[0022]

[0023] Establish the vertex set of the graph according to the mapping rule Create edge set E based on the relationship between vertices a ={e α，β |v α ∈V,v β ∈V}, where the vertex corresponding to Q(k) is v α , the vertex v corresponding to Q(k+h) β , h is the step length, and 1≤h≤N0-1, when v α ≠v β When , it means that there is an edge e between these two vertices. α，β , recorded as 1; otherwise, recorded as 0.

[0024] Furthermore, the pair G a A αThe spectrum is decomposed to obtain all eigenvalues, and the eigenvalues ​​are sorted to obtain the eigenvalue vector a, which is:

[0025] Computational graph G a The adjacency matrix A(G a ):

[0026]

[0027] Computational graph G a The degree matrix D(G a ):

[0028]

[0029] In the formula, the vertex v δ The degree is defined as

[0030] According to A α (G a )=αA(G a )+(1-α)D(G a ) to obtain graph G a A α The spectrum is:

[0031]

[0032] Where α is a coefficient between (0, 1);

[0033] A α (G a ) matrix is ​​decomposed to obtain all eigenvalues, and arranged in order from small to large to obtain the eigenvalue vector Eigenvalue

[0034] Furthermore, the autocorrelation function of the observed signal is calculated by first removing the mean of the observed signal x(n) to obtain an observed signal r(n) with a zero mean, and then calculating the autocorrelation function X(m) of r(n):

[0035]

[0036] Where m is the independent variable of the autocorrelation function.

[0037] Furthermore, after the autocorrelation function is subjected to linear normalization and uniform quantization, the graph mapping obtains the graph G b , specifically:

[0038] 1) Perform maximum and minimum linear normalization on the autocorrelation function X(m) to obtain the normalized sequence U(m);

[0039] 2) Uniformly quantize the sequence U(m) to obtain the sequence ζ(m):

[0040]

[0041] In the formula, μ represents the quantization level, and N0 represents the maximum quantization level;

[0042] 3) Map ζ(m) to an undirected unweighted graph structure G b ={E b , V b}, E b and V b Represent the edge set and vertex set of the graph respectively; the vertex v of the graph μ The mapping rule with the quantization level in the quantization interval [0, 1] is:

[0043]

[0044] Establish the vertex set of the graph according to the mapping rule Create edge set E based on the relationship between vertices b ={e α，β |v α ∈V,v β ∈V}, where the vertex corresponding to ζ(m) is v a , the vertex v corresponding to ζ(m+h) β , h is the step length, and 1≤h≤N0-1, when v α ≠v β When , it means that there is an edge e between these two vertices. α，β , recorded as 1; otherwise, recorded as 0.

[0045] Furthermore, the pair G b A α The spectrum is decomposed to obtain all eigenvalues, and the eigenvalues ​​are sorted to obtain the eigenvalue vector b, which is:

[0046] Computational graph G b The adjacency matrix A(G b ):

[0047]

[0048] Computational graph G b The degree matrix D(G b ):

[0049]

[0050] In the formula, the vertex v δ The degree is defined as

[0051] According to Aα (G b )=αA(G b )+(1-α)D(G b ) to obtain graph G b A α The spectrum is:

[0052]

[0053] Where α is a coefficient between (0, 1);

[0054] A α (G b ) matrix is ​​decomposed to obtain all eigenvalues, and arranged in order from small to large to obtain the eigenvalue vector Eigenvalue

[0055] Furthermore, the set threshold is obtained through the ROC curve. When the false alarm rate is set to 0.001, the corresponding threshold value is the set threshold η; if F≥η, there is a primary user signal in the observed signal; otherwise, there is no primary user signal in the observed signal.

[0056] Accordingly, the present invention proposes a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the cognitive radio spectrum sensing method based on hybrid graph transformation as described above.

[0057] Accordingly, the present invention proposes an electronic device, characterized in that it includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the cognitive radio spectrum sensing method based on hybrid graph transformation as described above is implemented.

[0058] The beneficial effects of the present invention are as follows: the present invention obtains detection statistics by fusing eigenvalue vectors of different graph structures to perform inner product aggregation operations, and compares it with the decision threshold to achieve spectrum perception of commonly used communication signals. From the perspective of graph features, the present invention performs different graph transformations on the autocorrelation function and power spectrum of the observed signal, which can further explore the topological features of the graph; the graph conversion is performed after the power spectrum is nonlinearly transformed, and the graphs obtained under the two assumptions are both non-connected graphs, and the graph transformation is performed after the autocorrelation function is normalized to the maximum and minimum, and the graphs obtained under the two assumptions are respectively a non-connected graph and an approximately complete graph. The graphs obtained by these two processing methods are fused, which can further increase the difference between the graphs under the two assumptions. The present invention defines Graph A α The inner product aggregation result is used as the detection statistic. αWhen the number of vertices of the spectrum is about 10, compared with the existing Laplace matrix, it has better graph topology resolution ability, which increases the separability of graph features under the two assumptions, and is conducive to improving the detection performance of the algorithm at low signal-to-noise ratio. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flow chart of the cognitive radio spectrum sensing method based on hybrid graph transformation.

[0060] Figure 2a and Figure 2b They are the graph structures obtained by processing the power spectrum of the observed signal under the H0 and H1 assumptions when the signal-to-noise ratio is 5dB; Figure 2c and Figure 2d They are the graph structures obtained by processing the observed signals through the autocorrelation function under the H0 and H1 assumptions when the signal-to-noise ratio is 5dB.

[0061] Figure 3a and Figure 3b Figure A is obtained by processing the power spectrum of the observed signal under the H0 and H1 assumptions. α Spectral eigenvalue vector; Figure 3c and Figure 3d Figure A is obtained by processing the observed signal through the autocorrelation function under the H0 and H1 assumptions. α The eigenvalue vector of the spectrum.

[0062] Figure 4 It is a schematic diagram comparing the mean values ​​of the test statistics under two hypotheses.

[0063] Figure 5a and Figure 5b They are comparative schematic diagrams of receiver operating characteristic curves (ROC) of the method of the present invention, the autocorrelation function nonlinear transformation method, the power spectrum grouping range method, and the power spectrum ungrouping method when SNR=-5dB and -10dB respectively. DETAILED DESCRIPTION

[0064] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0065] In one embodiment, the present invention proposes a cognitive radio spectrum sensing method based on hybrid graph transformation, such as Figure 1 As shown, the specific steps include:

[0066] Step 1: Obtain the observed signal and calculate its power spectrum.

[0067] In this embodiment, the primary user signal uses a phase coded modulation signal. When the observed signal contains the primary user signal, its expression is:

[0068] x(n)=s(n)+w(n)=κexp(j[2πf0nΔt+πd(n)+θ])+w(n), 0≤n≤N-1;

[0069] In the formula, x(n) is the observed signal, s(n) is the signal component, κ is the signal amplitude, N is the number of signal samples, f0 is the signal carrier frequency, Δt is the sampling interval, θ is the initial phase of the signal, and d(n) is the phase function; w(n) is zero-mean Gaussian white noise, whose real and imaginary parts are independent and whose variance is

[0070] When there is no primary user signal in the observed signal, it is only Gaussian white noise.

[0071] Assume that the hypothesis test is divided into: (1) H0 hypothesis: there is no primary user signal in the observed signal, or only noise signal exists; (2) H1 hypothesis: there is a primary user signal in the observed signal. The hypothesis test can be represented by a binary detection model, that is,

[0072]

[0073] Perform N-point Fourier transform on the observed signal x(n) and take the modulus to obtain the power spectrum:

[0074]

[0075] Step 2: After sigmoid nonlinear normalization of the power spectrum, uniform quantization and graph mapping are performed to obtain graph G a , computational graph G a The adjacency matrix A(G a ) and degree matrix D(G a ), and then find the graph G a A a Spectrum is A α (G a )=αA(G a )+(1-α)D(G a ), by finding A α (G a ) to obtain the eigenvalue vector a. Specifically, it includes the following sub-steps:

[0076] Step 2.1: The nonlinear transformation uses the sigmoid function, which is defined as:

[0077]

[0078] After the power spectrum R(k) passes through the sigmoid function, it is mapped to the interval [0, 1] one by one, recorded as the sequence Y(k), that is,

[0079] Y(k)=f(R(k)).

[0080] Step 2.2: Set the maximum quantization level N0 (i.e., the number of vertices in the graph), and uniformly quantize the sequence Y(k) to obtain the sequence Q(k), which is expressed as follows:

[0081]

[0082] Step 2.3: Map Q(k) to an undirected unweighted graph structure (referred to as graph structure) G a ={E a , V a}, E a and V a denote the edge set and vertex set of a simple undirected graph, respectively, where the vertex v μ The mapping rule with a specific sub-interval (i.e., quantization level) within the quantization interval [0, 1] is:

[0083]

[0084] Thus, we can get graph G a ={E a , V a} Create edge set E based on the relationship between vertices a , if the vertex corresponding to Q(k) is v α , the vertex v corresponding to Q(k+h) β , h is the step length, and 1≤h≤N0-1, when v a ≠v β When , it means there is an edge between the two vertices, which is recorded as 1; otherwise, it is recorded as 0. a It is expressed as follows:

[0085] E a ={e α，β |v α ∈V,v β ∈V};

[0086] Therefore, Figure G a The adjacency matrix A(G a ) can be expressed as:

[0087]

[0088] Vertex v δ The degree is defined as: Then graph G a The degree matrix D(G a) can be expressed as:

[0089]

[0090] Then we get graph G a A α Spectrum is A α (G a )=αA(G a )+(1-α)D(G a ):

[0091]

[0092] Among them, the value range of α is [0, 1]. When α is greater than 0.5, calculate A α When the adjacency matrix is ​​used, the weight is larger; when the value of α is less than 0.5, the calculation of A a When the spectrum is calculated, the degree matrix has a larger weight, so in this embodiment

[0093] Step 2.4: Compute graph G a A α Spectrum A α (G a ) and arrange them in order from small to large, that is, Get the eigenvalue vector

[0094] Step 3: Calculate the autocorrelation function of the observed signal, perform maximum and minimum linear normalization on the autocorrelation function, and then perform uniform quantization and graph mapping to obtain the graph G b , similarly, find graph G b A α Spectrum is A α (G b )=αA(G b )+(1-α)D(G b ), by finding A α (G b ) to obtain the eigenvalue vector b. Specifically, it includes the following sub-steps:

[0095] Step 3.1: Remove the mean of the observed signal x(n) to obtain an observed signal with zero mean, as follows:

[0096]

[0097] In the formula, Represents the mean of the observed signal x(n);

[0098] The autocorrelation function of r(n) is calculated as:

[0099]

[0100] Where m is the independent variable of the autocorrelation function.

[0101] Step 3.2: The minimum value of the sequence X(m) is ρ min =min(X(m)), the maximum value is ρ max =max(X(m)), then the normalized sequence U(m) is calculated as follows:

[0102]

[0103] Step 3.3: Set the quantization level N0 (i.e., the number of vertices in the graph), and uniformly quantize the sequence U(m) to obtain the sequence ζ(m), which is expressed as follows:

[0104]

[0105] Step 3.4: Map ζ(m) to graph G b ={E b , V b}, and calculate the graph G b A α Spectrum A α (G b ), the calculation process is exactly the same as step 2.3.

[0106] Step 3.5: Compute graph G b A α Spectrum A α (G b ) and arrange them in order from small to large, that is, Get the eigenvalue vector

[0107] Step 4: Perform inner product aggregation operation on the eigenvalue vectors a and b to obtain the detection statistic F. Set the threshold η. If F ≥ η, the primary user signal exists in the observed signal; otherwise, the primary user signal does not exist in the observed signal. The threshold is obtained through the ROC curve. Generally, when the false alarm rate is set to 0.001, the corresponding threshold value is the set threshold η.

[0108] Figure 2a , 2b They are the graph structures obtained by processing the power spectrum under the assumptions of H0 and H1 respectively. The figure assumes that the observed signal is a Gaussian white noise signal under the H0 condition and the observed signal is an interfered BPSK signal under the H1 condition. The number of sample points is 200 points, the sampling frequency of the BPSK signal is 2MHz, the oversampling factor is 10, the quantization level is 10, and the signal-to-noise ratio is 5dB. The subsequent simulations are the same unless otherwise specified. It can be seen from the figure that the connectivity of the graph under different assumptions is relatively small, and the connectivity of the graph under the H1 assumption is slightly higher. Figure 2c , 2d The graph structures obtained by processing the autocorrelation function under the H0 and H1 assumptions are shown in Figure 2. As can be seen from the figure, under the E0 assumption, the connectivity of the graph is weak, while under the H1 assumption, the connectivity of the graph is strong.

[0109] Figure 3a , 3b Figure A is obtained after processing the power spectrum under the assumptions H0 and H1. α The eigenvalue vector of the spectrum. The simulation conditions are the same as above. It can be seen from the figure that under the H1 assumption, Figure A α The eigenvalues ​​of the spectrum are relatively high. Figure 3c , 3d Figure A shows the autocorrelation function after processing 2 under the assumptions H0 and H1. α The eigenvalue vector of the spectrum. As can be seen from the figure, under the H0 hypothesis, Figure A α The number of zero eigenvalues ​​in the spectrum is relatively large.

[0110] Figure 4 This is a schematic diagram comparing the mean values ​​of the detection statistic F obtained under the two assumptions when the signal-to-noise ratio is -8dB to 2dB. The mean value of the detection statistic F is obtained by 1000 simulations at each signal-to-noise ratio. It can be seen from the figure that under the H0 assumption, the mean value of F is constant at around 8, while under the H1 assumption, as the signal-to-noise ratio increases, the mean value of F also increases accordingly.

[0111] Figure 5a , 5b The following are the ROC curve comparison diagrams of the receiver operating curves of the method of the present invention, the autocorrelation function nonlinear transformation method, the power spectrum grouping extreme difference method and the power spectrum ungrouping method when SNR=-5dB and -10dB respectively. The other simulation conditions are the same as those mentioned above. It can be seen from the figure that under low signal-to-noise ratio conditions, the AUC value of the method of the present invention is better than that of other comparison algorithms. Therefore, when the signal-to-noise ratio is low, the detection performance of the method of the present invention is better.

[0112] In another embodiment, the present invention proposes a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the cognitive radio spectrum sensing method based on hybrid graph transformation as described in the first embodiment.

[0113] In another embodiment, the present invention proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the cognitive radio spectrum sensing method based on hybrid graph transformation as described above is implemented.

[0114] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the above. More specific examples of computer storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0115] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application 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. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0116] The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A cognitive radio spectrum sensing method based on hybrid graph transformation, characterized in that: The steps include: Obtain the observed signal, calculate the power spectrum of the observed signal, perform nonlinear normalization and uniform quantization on the power spectrum, and then map it to obtain the graph G a , for graph G a A α The spectrum is eigendecomposed to obtain all eigenvalues, and the eigenvalues ​​are sorted to obtain the eigenvalue vector a; Calculate the autocorrelation function of the observed signal, perform linear normalization and uniform quantization on the autocorrelation function, and then map it to obtain the graph G b , for graph G b A α The spectrum is decomposed to obtain all eigenvalues, and the eigenvalues ​​are sorted to obtain the eigenvalue vector b; Perform inner product aggregation operation on eigenvalue vector a and eigenvalue vector b to obtain the detection statistic F; The detection statistic F is compared with the set threshold to determine whether there is a primary user signal in the observed signal.

2. The cognitive radio spectrum sensing method based on hybrid graph transformation according to claim 1, characterized in that: The power spectrum of the observed signal is calculated by performing N-point Fourier transform on the observed signal x(n) and taking the modulus, and the power spectrum R(k) is obtained as follows: Where N is the number of signal samples and k is the serial number of the power spectrum.

3. The cognitive radio spectrum sensing method based on hybrid graph transformation according to claim 1, characterized in that: After the power spectrum is subjected to nonlinear normalization and uniform quantization, the graph Ga is obtained by graph mapping, which is specifically: 1) After the power spectrum R(k) passes through the sigmoid function, it is mapped to the interval [0, 1] one by one to obtain the sequence Y(k): 2) Uniformly quantize the sequence Y(k) to obtain the sequence Q(k): In the formula, μ represents the quantization level, and N0 represents the maximum quantization level; 3) Map Q(k) to an undirected unweighted graph structure G a ={E a , V a }, E a and V a Represent the edge set and vertex set of the graph respectively; the vertex v of the graph μ The mapping rule with the quantization level in the quantization interval [0, 1] is: Establish the vertex set of the graph according to the mapping rule Create edge set E based on the relationship between vertices a ={e α，β |v α ∈V,v β ∈V}, where the vertex corresponding to Q(k) is v α , the vertex v corresponding to Q(k+h) β , h is the step length, and 1≤h≤N0-1, when v α ≠v β When , it means that there is an edge e between these two vertices. α，β , recorded as 1; otherwise, recorded as 0.

4. The cognitive radio spectrum sensing method based on hybrid graph transformation according to claim 3, characterized in that: The pair G a A α The spectrum is decomposed to obtain all eigenvalues, and the eigenvalues ​​are sorted to obtain the eigenvalue vector a, which is: Computational graph G a The adjacency matrix A(G a ): Computational graph G a The degree matrix D(G a ): In the formula, the vertex v δ The degree is defined as According to A α (G a )=αA(G a )+(1-α)D(G a ) to obtain graph G a A α The spectrum is: Where α is a coefficient between (0, 1); A α (G a ) matrix is ​​decomposed to obtain all eigenvalues, and arranged in order from small to large to obtain the eigenvalue vector Eigenvalue 5. The cognitive radio spectrum sensing method based on hybrid graph transformation according to claim 1, characterized in that: The autocorrelation function of the observed signal is calculated by first removing the mean of the observed signal x(n) to obtain the observed signal r(n) with a zero mean, and then calculating the autocorrelation function X(m) of r(n): Where m is the independent variable of the autocorrelation function.

6. The cognitive radio spectrum sensing method based on hybrid graph transformation according to claim 1, characterized in that: After the autocorrelation function is linearly normalized and uniformly quantized, the graph G is obtained by graph mapping. b , specifically: 1) Perform maximum and minimum linear normalization on the autocorrelation function X(m) to obtain the normalized sequence U(m); 2) Uniformly quantize the sequence U(m) to obtain the sequence ζ(m): In the formula, μ represents the quantization level, and N0 represents the maximum quantization level; 3) Map ζ(m) to an undirected unweighted graph structure G b ={E b , V b }, H0 and V b Represent the edge set and vertex set of the graph respectively; the vertex v of the graph μ The mapping rule with the quantization level in the quantization interval [0, 1] is: Establish the vertex set of the graph according to the mapping rule Create edge set E based on the relationship between vertices b ={e α，β |v α ∈V,v β ∈V}, where the vertex corresponding to ζ(m) is v α , the vertex v corresponding to ζ(m+h) β , h is the step length, and 1≤h≤N0-1, when v α ≠v β When , it means that there is an edge e between these two vertices. α，β , recorded as 1; otherwise, recorded as 0.

7. The cognitive radio spectrum sensing method based on hybrid graph transformation according to claim 6, characterized in that: The pair G b A α The spectrum is decomposed to obtain all eigenvalues, and the eigenvalues ​​are sorted to obtain the eigenvalue vector b, which is: Computational graph G b The adjacency matrix A(G b ): Computational graph G b The degree matrix D(G b ): In the formula, the vertex v δ The degree is defined as According to A α (G b ))=αA(G b ))+(1-α)D(G b ))Get the graph G b A α The spectrum is: Where α is a coefficient between (0, 1); A α (G b ) matrix is ​​decomposed to obtain all eigenvalues, and arranged in order from small to large to obtain the eigenvalue vector Eigenvalue 8. The cognitive radio spectrum sensing method based on hybrid graph transformation according to claim 1, characterized in that: The set threshold is obtained through the ROC curve. When the false alarm rate is set to 0.001, the corresponding threshold value is the set threshold η; if F≥η, there is a primary user signal in the observed signal; Otherwise, there is no primary user signal in the observed signal.

9. A computer-readable storage medium storing a computer program, characterized in that: The computer program enables a computer to execute the cognitive radio spectrum sensing method based on hybrid graph transformation as described in any one of claims 1 to 8.

10. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the cognitive radio spectrum sensing method based on hybrid graph transformation as described in any one of claims 1 to 8 is implemented.

Images