Cognitive radio spectrum sensing method and system based on horizontal visual graph and graph single-hop filter

Abstract

The invention provides a cognitive radio spectrum sensing method and system based on a horizontal visual graph and a graph single-hop filter. The method comprises the following steps: firstly, calculating a power spectrum of an observation signal; then carrying out grouping summation on the power spectrums to obtain a grouping summation sequence of the power spectrums; calculating an autocorrelation module value square sequence of the grouping summation sequence, and normalizing and quantifying the autocorrelation module value square sequence; performing horizontal visual graph conversion on the quantized sequence to obtain a horizontal visual graph corresponding to the quantized sequence, and further obtaining a graph unsigned Laplace matrix; taking the quantization sequence as a graph signal, and constructing a graph single-hop filter by the quantization sequence and the graph unsigned Laplace matrix; and finally, calculating the Euclidean norm of the graph single-hop filter as a detection statistic, setting a judgment threshold, and completing the detection of the existence of the main user signal. According to the method, the detection performance of the algorithm under the low signal-to-noise ratio is effectively improved while the calculation complexity is not increased.

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 and system based on a horizontal visibility graph and a graph single-hop filter. Background Art

[0002] With the development of B5G / 6G mobile communication systems, the scale of wireless communication equipment has further increased, exacerbating the shortage of spectrum resources. Cognitive Radio (CR) can make full use of the idle spectrum of the primary user by introducing an opportunistic spectrum access mechanism, providing an important way to improve the utilization of spectrum resources. For CR, spectrum sensing is the primary link to achieve spectrum sharing and resource management, and its performance directly affects the reliability and effectiveness of the entire cognitive signal processing system. Generally, the spectrum sensing problem can be modeled as a binary hypothesis testing model. The basic idea is to construct a detection statistic using limited observed signal samples, and design a judge based on specific criteria to realize the test of the existence of the primary user signal. However, how to select a detection statistic with good separability and controllable computational complexity under limited sample conditions has always been a research difficulty, which is specifically manifested as follows:

[0003] (1) Various noise or background interference. At present, as the electromagnetic environment becomes more and more complex, signals are easily interfered by various noises.

[0004] (2) Traditional detection algorithms have poor detection performance under low signal-to-noise ratios. Traditional detection algorithms such as energy detection and cyclostationary detection are easily affected by noise and have poor detection performance under low signal-to-noise ratios.

[0005] (3) The image domain detection algorithm needs to be optimized. Existing image domain detection algorithms, such as quantization graph and visibility graph, have poor performance under low signal-to-noise ratio and high computational complexity. Summary of the invention

[0006] Aiming at the problems that the existing graph-domain spectrum sensing algorithms have poor detection performance and high complexity under low signal-to-noise ratio, the present invention provides a cognitive radio spectrum sensing method and system based on a horizontal visibility graph and a graph single-hop filter.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] In a first aspect, the present invention provides a cognitive radio spectrum sensing method based on a horizontal visibility graph and a graph single-hop filter, comprising the following steps:

[0009] S1: Obtain the observed signal and calculate its power spectrum;

[0010] S2: performing group summation processing on the power spectrum to obtain a group summation sequence, then calculating the autocorrelation modulus square sequence of the group summation sequence, normalizing and quantizing the autocorrelation modulus square sequence to obtain a quantized sequence;

[0011] S3: Perform horizontal visible graph transformation on the quantized sequence to obtain an unsigned Laplace matrix. At the same time, define the quantized sequence as an image signal on the horizontal visible graph, and construct an image single-hop filter together with the unsigned Laplace matrix.

[0012] S4: Calculate the Euclidean norm of the graph single-hop filter as a detection statistic to determine whether the primary user signal exists in the observed signal.

[0013] Optionally, in step S1, the observation signal is a BPSK signal, which is defined as:

[0014]

[0015] Where r(n) represents the observed signal, s(n) represents the signal component, w(n) represents the additive Gaussian white noise, N is the number of observed signal samples, H0 represents that there is no primary user signal in the observed signal, and H1 represents that the observed signal consists of the primary user signal and noise;

[0016] The power spectrum of the observed signal is calculated by the following formula:

[0017]

[0018] Where R(k) represents the power spectrum of the observed signal r(n).

[0019] Optionally, in step S2, the power spectrum is subjected to group summation processing to obtain a group summation sequence, specifically:

[0020] The power spectrum R(k) is grouped and summed, with each δ samples forming a group, with a total of K = [N / δ] groups. The samples in each group are summed to obtain a grouped sum sequence:

[0021]

[0022] Where Z(m) represents the group summation sequence of the power spectrum R(k).

[0023] Optionally, in step S2, the autocorrelation modulus square sequence of the grouped summation sequence is calculated by the following formula:

[0024]

[0025] Wherein, X(l) represents the square sequence of the autocorrelation modulus values ​​of the group summation sequence Z(m).

[0026] Optionally, in step S2, the autocorrelation modulus square sequence is normalized and quantized to obtain a quantized sequence, specifically:

[0027] The autocorrelation modulus square sequence X(l) is transformed to the interval [0, 1] according to the maximum minimization normalization principle:

[0028]

[0029] Where U(l) is the normalized sequence of the autocorrelation modulus square sequence X(l), θ min and θ max are the minimum and maximum values ​​of the autocorrelation modulus square sequence X(l) respectively;

[0030] Perform uniform quantization on the normalized sequence U(l):

[0031]

[0032] Where P(l) is the quantization sequence, μ∈{1, 2, 3, ..., M}, and M is the quantization level.

[0033] Optionally, in step S3, the quantized sequence is subjected to a horizontal visual transformation to obtain an unsigned Laplace matrix, specifically:

[0034] Convert the quantized sequence P(l) into a horizontal visibility graph G(V,E), where V is the node set of the graph, E is the edge set of the graph, and the number of nodes is given by the number of samples K of the quantized sequence;

[0035] Based on the horizontal visibility graph G(V, E), the adjacency matrix A and degree matrix D of the quantized sequence are calculated:

[0036]

[0037] In the formula, a ij represents the element value of the i-th row and j-th column in the adjacency matrix A, d ii Represents the element value in the degree matrix D;

[0038] The unsigned Laplace matrix of the graph is calculated by the following formula:

[0039] Q = D + A;

[0040] Where Q represents the unsigned Laplace matrix of the graph.

[0041] Optionally, in step S3, a graph single-hop filter is constructed by the following formula:

[0042]

[0043] In the formula, Represents a single-hop filter, Represents the graph signal on the horizontal visible graph.

[0044] Optionally, in step S4, the Euclidean norm of the graph single-hop filter is calculated as the detection statistic by the following formula:

[0045]

[0046] In the formula, A single-hop filter The Euclidean norm of the graph is the unsigned Laplace matrix Q = (q ij ) K×K , quantized sequence As a graph signal,

[0047] Set the threshold ξ, if the Euclidean norm If it is less than or equal to the threshold value ξ, it is considered to meet H1, indicating that there is a primary user signal in the observed signal; otherwise, it is considered to meet H0, indicating that there is no primary user signal in the observed signal.

[0048] Optionally, in step S4, the threshold value ξ is calculated by the following formula:

[0049]

[0050] Where P fa is the false alarm probability, α0, c0, and v0 are the scale parameter, the first shape parameter, and the second shape parameter obtained by fitting the detection statistic of H0 with Burr distribution, respectively.

[0051] In a first aspect, the present invention provides a cognitive radio spectrum sensing system based on a horizontal visibility graph and a graph single-hop filter, comprising:

[0052] An acquisition unit, used for acquiring an observation signal;

[0053] A calculation unit is used to calculate the power spectrum of the observed signal; perform group summation processing on the power spectrum to obtain a group summation sequence, then calculate the autocorrelation modulus square sequence of the group summation sequence, normalize and quantize the autocorrelation modulus square sequence to obtain a quantized sequence; perform horizontal visible graph transformation on the quantized sequence to obtain a graph unsigned Laplace matrix, and define the quantized sequence as a graph signal on the horizontal visible graph, and construct a graph single-hop filter together with the graph unsigned Laplace matrix;

[0054] The decision unit is used to calculate the Euclidean norm of the graph single-hop filter as a detection statistic to determine whether the primary user signal exists in the observed signal.

[0055] The beneficial effects of the present invention are as follows: the present invention integrates the topological structure features with the statistical characteristics by constructing a graph single-hop filter, defines the Euclidean norm of the graph single-hop filter as the detection statistic, and effectively improves the anti-interference ability and detection performance of the spectrum sensing algorithm. The present invention also verifies through data exploration that the proposed detection statistic approximately obeys the Burr distribution under the null hypothesis, and derives an approximate analytical expression for the detection threshold, which facilitates the calculation of the detection threshold and the calculation of the detection probability under low signal-to-noise ratio; compared with the existing graph domain spectrum sensing algorithm, the present invention effectively improves the detection performance of the algorithm under low signal-to-noise ratio without increasing the computational complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a flow chart of a cognitive radio spectrum sensing method based on a horizontal visibility graph and a graph single-hop filter;

[0057] Figure 2a is the power spectrum sequence under the H0 hypothesis;

[0058] Figure 2b is the power spectrum sequence under the H1 hypothesis;

[0059] Figure 3a It is the group sum sequence under the H0 hypothesis;

[0060] Figure 3b is the group summation sequence under the H1 hypothesis;

[0061] Figure 4a It is the normalized sequence of the square sequence of the autocorrelation modulus value under the H0 assumption;

[0062] Figure 4b It is the normalized sequence of the square sequence of the autocorrelation modulus value under the H1 assumption;

[0063] Figure 5a is a quantitative sequence under the H0 hypothesis;

[0064] Figure 5b It is the quantitative sequence under the H1 hypothesis;

[0065] Figure 6a It is the horizontal visualization graph corresponding to the quantized sequence under the H0 hypothesis;

[0066] Figure 6b It is the horizontal visualization graph corresponding to the quantized sequence under the H1 hypothesis;

[0067] Figure 7 It is the Burr distribution fitting diagram of the detection statistic at some signal-to-noise ratios under two assumptions;

[0068] Figure 8 It is a schematic diagram comparing the theoretical detection probability and the actual detection probability;

[0069] Figure 9a It is a schematic diagram comparing receiver working curves of the present invention and the existing spectrum sensing algorithm under Gaussian channel;

[0070] Figure 9b It is a schematic diagram comparing receiver working curves of the present invention and the existing spectrum sensing algorithm under a flat fading channel;

[0071] Fig.9c It is a schematic diagram comparing receiver operating curves of the present invention and the existing spectrum sensing algorithm under a selective fading channel. DETAILED DESCRIPTION

[0072] The present invention will now be described in further detail with reference to the accompanying drawings.

[0073] In one embodiment, the present invention proposes a cognitive radio spectrum sensing method based on a horizontal visibility graph and a graph single-hop filter, such as Figure 1 The method comprises the following steps:

[0074] Step 1: Obtain the observed signal and calculate the power spectrum of the observed signal.

[0075] In this embodiment, the observation signal used is a binary phase shift keying (BPSK) signal. BPSK signal is widely used in wireless communication, radar communication and other scenarios due to its advantages such as simple implementation, mature technology and strong noise resistance.

[0076] The question of whether there is a primary user signal in the received signal sequence is described as a binary hypothesis test. Define H0: There is no primary user signal in the observed signal; H1: The observed signal consists of the primary user signal and noise; N is the number of observed signal samples. Therefore, the received signal is:

[0077]

[0078] Where s(n) represents the signal component and w(n) represents the additive white Gaussian noise.

[0079] Calculate the power spectrum R(k) of the observed signal r(n), that is:

[0080]

[0081] Step 2: Perform group summation processing on the power spectrum to obtain a group summation sequence; calculate the autocorrelation modulus square sequence of the group summation sequence; normalize and quantize the autocorrelation modulus square sequence to obtain a quantized sequence.

[0082] Step 2.1: Calculate the group summation sequence of the power spectrum.

[0083] The power spectrum R(k) is grouped and summed, with each δ samples forming a group, with a total of K = [N / δ] groups. The samples in each group are summed to obtain the grouped sum sequence Z(m):

[0084]

[0085] Step 2.2: Calculate the autocorrelation modulus square sequence of the grouped summation sequence.

[0086] The autocorrelation function can reflect the correlation between a signal and its different delayed signals, and is usually symmetrical. Therefore, for the group summation sequence Z(m), the modulus square sequence X(l) of its half-length autocorrelation function is calculated. In the following text, unless otherwise specified, X(l) is referred to as the modulus square sequence. That is:

[0087]

[0088] Step 2.3: Normalize and quantize the autocorrelation modulus square sequence.

[0089] Through normalized quantization processing, the algorithm's anti-interference ability can be effectively improved.

[0090] (1) Normalization

[0091] Transform the sequence X(l) to the interval [0, 1] according to the maximum minimization normalization principle:

[0092]

[0093] Among them, θ min and θ max are the minimum and maximum values ​​of the sequence X(l) respectively.

[0094] (2) Uniform quantization

[0095] Perform uniform quantization on the normalized sequence U(l):

[0096]

[0097] Wherein, M is the quantization level, μ∈{1, 2, 3, .., M}, thereby obtaining a uniform quantization sequence P(l).

[0098] Step 3: Perform horizontal visual transformation on the quantized sequence to obtain a graph unsigned Laplace matrix; define the quantized sequence as a graph signal, and construct a graph single-hop filter with the graph unsigned Laplace matrix.

[0099] Perform horizontal visualization transformation on the quantized sequence P(l) to obtain the graph adjacency matrix A. The adjacency matrix is ​​used to describe the connection relationship matrix between nodes in the graph, that is, in an undirected graph, if there is a connection between node i and node , then the element value of the i-th row, i-th column and the i-th row, i-th column in the adjacency matrix is ​​1, otherwise it is 0. If the graph is a weighted graph, the elements of the adjacency matrix can be expressed as the strength of the connection, usually the weight value of the connection.

[0100] The horizontal visibility graph is an undirected and unweighted graph, so the horizontal visibility graph adjacency matrix A of the quantized sequence P(l) is defined as:

[0101]

[0102] The degree matrix is ​​a diagonal matrix whose diagonal elements represent the degree of each node, that is, the number of edges connected to the node. Therefore, in the visual graph, the degree of a node is the number of edges connected to it.

[0103] Defining the degree matrix D ,Right now:

[0104]

[0105] Step 3.1: Based on the adjacency matrix A and the degree matrix D, construct the graph unsigned Laplace matrix Q:

[0106] Q=D+A

[0107] Step 3.2: Define the quantization sequence P(l) as the image signal on the horizontal visual graph Based on the graph unsigned Laplace matrix, a graph single-hop filter is constructed Right now:

[0108]

[0109] Step 4: Calculate the Euclidean norm of the graph single-hop filter as the detection statistic, set the detection probability adjustment threshold, and determine whether the observed signal is H0 or H1.

[0110] Step 4.1: Calculate the Euclidean norm of the graph single-hop filter, that is:

[0111]

[0112] Among them, the unsigned Laplace matrix Q = (q ij ) K×K , As a graph signal,

[0113] In addition, through a large number of simulation experiments, it is found that the probability distribution of the detection statistic defined in this embodiment can be fitted by Burr distribution under the H0 assumption, and the goodness of fit is good; under low signal-to-noise ratio (SNR≤-6dB), the probability distribution of the detection statistic under the H1 assumption can still be well fitted by Burr distribution. Therefore, this embodiment obtains the analytical expression of the detection threshold ξ by fitting the probability distribution function of Burr distribution under the H0 assumption, and gives the detection probability P under low signal-to-noise ratio. d The analytical expression of .

[0114] Therefore, a certain amount of observed signals under the H0 hypothesis are simulated to obtain some detection statistic samples under the hypothesis, and a Burr fitting is performed on them. Assuming that the false alarm probability is P fa =0.1, then:

[0115]

[0116] Among them, f0(x; α0, c0, v0) and F0(ξ; α0, c0, v0) are the probability density function and cumulative distribution function of the H0 test statistic, respectively. α0, c0, v0 are the scale parameter, first shape parameter, and second shape parameter obtained by fitting the H0 test statistic with Burr distribution, respectively. Therefore, by solving formula (Δ), the approximate threshold ξ can be obtained as:

[0117]

[0118] Therefore, at low signal-to-noise ratio (SNR≤-6dB), the detection probability can be expressed as:

[0119]

[0120] Among them, f1(x; α1, c1, v1) and F1(ξ; α1, c1, v1) are the probability density function and cumulative distribution function of the H1 detection statistic, respectively. α1, c1, v1 are the scale parameter, first shape parameter, and second shape parameter obtained by fitting the Burr distribution of the H1 detection statistic, respectively.

[0121] Next, the effectiveness of the cognitive radio spectrum sensing method based on the horizontal visibility graph and the graph single-hop filter proposed in this embodiment is further explained with reference to specific examples.

[0122] The signal to be observed is a BPSK modulated signal, the SNR range is -13 to 0 dB, the number of samples NN = 1024, the quantization level M = 8, the packet length δ = 16, the channel is additive white Gaussian noise (AWGN), and the number of simulations is 1000.

[0123] Figure 2a , 2b The following are schematic diagrams of power spectrum sequences under two assumptions. As can be seen from the figure, there are obvious differences between the power spectrum sequences under the two assumptions.

[0124] Figure 3a , 3b The following are schematic diagrams of the group sum sequence under the two assumptions. As can be seen from the figure, the fluctuation range of the group sum sequence under the two assumptions is also quite different.

[0125] Figure 4a , 4b The following are schematic diagrams of the normalized autocorrelation modulus square sequence of the group summation sequence under two assumptions. It can be seen from the figure that the fluctuation frequency and fluctuation amplitude of the normalized sequence under the two assumptions are relatively large.

[0126] From the above three figures, it can be seen that the sequence that has not been quantized has a large fluctuation amplitude and frequency under the H1 hypothesis, and is easily affected by interference such as noise, which reduces the quality of the observed signal and reduces the performance of the spectrum sensing algorithm.

[0127] Figure 5a , 5b The following are schematic diagrams of quantization sequences under two assumptions. As can be seen from the figure, after quantization, the difference between the quantization sequences under the two assumptions is quite obvious. After quantization, it is found that the fluctuation frequency of the observed signal under the H1 assumption is significantly reduced compared with the normalized sequence, indicating that the quantization process improves the anti-interference ability of the algorithm and still maintains the difference with the H0 assumption.

[0128] Figure 6a , 6b They are the horizontal visualization graphs corresponding to the quantization sequences under the two assumptions. As can be seen from the figure, the quantization sequences under the two assumptions have obvious differences, so their graph topological structures are also quite different.

[0129] The unsigned Laplace matrices of the horizontal visible graphs under the two assumptions correspond to their graph topological structure characteristics. By defining the quantized sequence as a graph signal, the original statistical characteristics of the observed signal are reused. This embodiment integrates the topological structure characteristics with the statistical characteristics by constructing a graph single-hop filter, defines the Euclidean norm of the graph single-hop filter as the detection statistic, and improves the detection performance of the spectrum sensing algorithm.

[0130] Figure 7The following is a schematic diagram of the Burr distribution fitting of the detection statistic under the two hypotheses. As can be seen from the figure, the probability distribution of the detection statistic under the H0 hypothesis can be fitted by the Burr distribution, and the goodness of fit is good. Under low signal-to-noise ratio (SNR≤-6dB), the detection statistic under the H1 hypothesis can also be well fitted by the Burr distribution. This facilitates the calculation of the detection threshold and the detection probability under low signal-to-noise ratio.

[0131] Figure 8 It is a comparison between the theoretical value of the detection probability of the present invention and the actual value. As can be seen from the figure, the theoretical value of the detection probability calculated by the formula is well consistent with the detection probability simulation value obtained by the experiment. The reason is that the probability distribution of the detection statistic under the H0 hypothesis has a high goodness of fit when the Burr distribution is fitted, and at a low signal-to-noise ratio (SNR≤-6dB), the detection statistic of the H1 hypothesis can also be well fitted by the Burr distribution. Therefore, the threshold calculated by the probability density function under the H0 hypothesis can effectively distinguish the two hypotheses; when the signal-to-noise ratio is high (SNR≥-5dB), although the complexity of the detection statistic under the H1 hypothesis is high, it is difficult to find a suitable distribution for fitting, but at this time, the distance between the detection statistics under the two hypotheses is large, and the detection probability of the algorithm is close to 100%, so even if the fitting result is poor, it will not affect the overall performance of the algorithm.

[0132] Figures 9a to 9c The figure is a schematic diagram comparing the receiver working curves of the present invention and the existing algorithm under different fading channels. In the simulation, SNR=-12dB is simulated 5000 times. The channels are set as Gaussian channel, flat fading channel and selective fading channel respectively (the maximum Doppler shift of the channel is 130Hz, the flat fading channel has only one path and no delay, and the single path gain is -6dB; the selective fading channel has four paths, the delays are 0, 5 / fs, 10 / fs, 15 / fs, and the multipath gains are 0dB, -3dB, -6dB, and -9dB, respectively, and fs is the sampling frequency. Under the same conditions, the ROC of the algorithm of the present invention, the original algorithm, the ungrouped algorithm, the energy detection, the cyclostationary detection method, the quantized graph method, the NVG+adjacency matrix method, the convolutional neural network method and other algorithms are compared, wherein the original algorithm uses the original sequence that has not been normalized and quantized as the graph transformation input and the graph signal, and the other steps are consistent with the present invention; the ungrouped algorithm does not perform group summation processing on the power spectrum, and the other steps are consistent with the present invention. Through comparison, it is found that the detection performance of the present invention is better than the existing typical spectrum sensing algorithm under various conditions, which reflects the superiority of the present invention.

[0133] In another embodiment, the present invention proposes a cognitive radio spectrum sensing system based on a horizontal visibility graph and a graph single-hop filter, which corresponds to the cognitive radio spectrum sensing method based on a horizontal visibility graph and a graph single-hop filter proposed in the above embodiment, including:

[0134] An acquisition unit, used for acquiring an observation signal;

[0135] A calculation unit is used to calculate the power spectrum of the observed signal; perform group summation processing on the power spectrum to obtain a group summation sequence, then calculate the autocorrelation modulus square sequence of the group summation sequence, normalize and quantize the autocorrelation modulus square sequence to obtain a quantized sequence; perform horizontal visible graph transformation on the quantized sequence to obtain a graph unsigned Laplace matrix, and define the quantized sequence as a graph signal on the horizontal visible graph, and construct a graph single-hop filter together with the graph unsigned Laplace matrix;

[0136] The decision unit is used to calculate the Euclidean norm of the graph single-hop filter as a detection statistic to determine whether the primary user signal exists in the observed signal.

[0137] In this system, the functions and execution steps of each unit are the same as the steps of the method in the aforementioned embodiment, so they are not repeated here.

[0138] The above are only preferred embodiments of the present invention, and 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 horizontal visibility graph and graph single-hop filter, characterized in that: The steps include: S1: Obtain the observed signal and calculate its power spectrum; S2: performing group summation processing on the power spectrum to obtain a group summation sequence, then calculating the autocorrelation modulus square sequence of the group summation sequence, normalizing and quantizing the autocorrelation modulus square sequence to obtain a quantized sequence; S3: Perform horizontal visible graph transformation on the quantized sequence to obtain an unsigned Laplace matrix. At the same time, define the quantized sequence as an image signal on the horizontal visible graph, and construct an image single-hop filter together with the unsigned Laplace matrix. S4: Calculate the Euclidean norm of the graph single-hop filter as a detection statistic to determine whether the primary user signal exists in the observed signal.

2. The cognitive radio spectrum sensing method based on horizontal visibility graph and graph single-hop filter as claimed in claim 1, characterized in that: In step S1, the observed signal is a BPSK signal, which is defined as: Where r(n) represents the observed signal, s(n) represents the signal component, w(n) represents the additive Gaussian white noise, N is the number of observed signal samples, H0 represents that there is no primary user signal in the observed signal, and H1 represents that the observed signal consists of the primary user signal and noise; The power spectrum of the observed signal is calculated by the following formula: Where R(k) represents the power spectrum of the observed signal r(n).

3. The cognitive radio spectrum sensing method based on horizontal visibility graph and graph single-hop filter as claimed in claim 1, characterized in that: In step S2, the power spectrum is grouped and summed to obtain a grouped summation sequence, which is specifically: The power spectrum R(k) is grouped and summed, with each δ samples forming a group, with a total of K = [N / δ] groups. The samples in each group are summed to obtain a grouped summation sequence: Where Z(m) represents the group summation sequence of the power spectrum R(k).

4. The cognitive radio spectrum sensing method based on horizontal visibility graph and graph single-hop filter as claimed in claim 3, characterized in that: In step S2, the autocorrelation modulus square sequence of the group summation sequence is calculated by the following formula: Wherein, X(l) represents the square sequence of the autocorrelation modulus value of the group summation sequence Z(m).

5. The cognitive radio spectrum sensing method based on horizontal visibility graph and graph single-hop filter as claimed in claim 4, characterized in that: In step S2, the autocorrelation modulus square sequence is normalized and quantized to obtain a quantized sequence, specifically: The autocorrelation modulus square sequence X(l) is transformed to the interval [0, 1] according to the maximum minimization normalization principle: Where U(l) is the normalized sequence of the autocorrelation modulus square sequence X(l), θ min and θ max are the minimum and maximum values ​​of the autocorrelation modulus square sequence X(l) respectively; Perform uniform quantization on the normalized sequence U(l): Where P(l) is the quantization sequence, μ∈{1, 2, 3, ..., M}, and M is the quantization level.

6. The cognitive radio spectrum sensing method based on horizontal visibility graph and graph single-hop filter as claimed in claim 1, characterized in that: In step S3, the quantized sequence is subjected to a horizontal visual transformation to obtain an unsigned Laplace matrix, which is specifically: Convert the quantized sequence P(l) into a horizontal visibility graph G(V,E), where V is the node set of the graph, E is the edge set of the graph, and the number of nodes is given by the number of samples K of the quantized sequence; Based on the horizontal visibility graph G(V, E), the adjacency matrix A and degree matrix D of the quantized sequence are calculated: In the formula, a ij represents the element value of the i-th row and j-th column in the adjacency matrix A, d ii Represents the element value in the degree matrix D; The unsigned Laplace matrix of the graph is calculated by the following formula: Q = D + A; Where Q represents the unsigned Laplace matrix of the graph.

7. The cognitive radio spectrum sensing method based on horizontal visibility graph and graph single-hop filter as claimed in claim 6, characterized in that: In step S3, a graph single-hop filter is constructed by the following formula: In the formula, Represents a single-hop filter, Represents the graph signal on the horizontal visible graph.

8. The cognitive radio spectrum sensing method based on horizontal visibility graph and graph single-hop filter as claimed in claim 7, characterized in that: In step S4, the Euclidean norm of the graph single-hop filter is calculated as the detection statistic by the following formula: In the formula, A single-hop filter The Euclidean norm of the graph is the unsigned Laplace matrix Q = (q ij ) K×K , quantized sequence As a graph signal, Set the threshold value ξ, if the Euclidean norm If it is less than or equal to the threshold value ξ, it is considered to meet H1, indicating that there is a primary user signal in the observed signal; otherwise, it is considered to meet H0, indicating that there is no primary user signal in the observed signal.

9. The cognitive radio spectrum sensing method based on horizontal visibility graph and graph single-hop filter as claimed in claim 8, characterized in that: In step S4, the threshold value ξ is calculated by the following formula: Where P fa is the false alarm probability, α0, c0, and v0 are the scale parameter, the first shape parameter, and the second shape parameter obtained by fitting the detection statistic of H0 with Burr distribution, respectively.

10. A cognitive radio spectrum sensing system based on horizontal visibility graph and graph single-hop filter, characterized in that: include: An acquisition unit, used for acquiring an observation signal; A calculation unit, used for calculating the power spectrum of the observation signal; Performing group summation processing on the power spectrum to obtain a group summation sequence, then calculating an autocorrelation modulus square sequence of the group summation sequence, normalizing and quantizing the autocorrelation modulus square sequence to obtain a quantized sequence; Perform horizontal visible graph transformation on the quantized sequence to obtain an unsigned Laplace matrix. At the same time, define the quantized sequence as an image signal on the horizontal visible graph, and construct an image single-hop filter together with the unsigned Laplace matrix. The decision unit is used to calculate the Euclidean norm of the graph single-hop filter as a detection statistic to determine whether the primary user signal exists in the observed signal.

Images