Wideband spectrum sensing method and system based on Hankel matrix Prony-Kung
Through the Prony-Kung method based on Hankel matrix, the problems of large time consumption and heavy equipment load in drone broadband spectrum perception are solved, and the ability to accurately detect frequency points in a short time is realized, complexity is reduced, and suitable for dynamic communication is suitable for Scene.
Patent Information
- Application Number
- CN202210868400.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-07-21
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2042-07-21
AI Technical Summary
The existing broadband spectrum perception methods have problems such as high time consumption, heavy equipment load and conditional limitations in drone applications, making it difficult to accurately detect frequency points in a short period of time, especially in scenarios with complex electromagnetic spectrum environments.
Using the Prony-Kung method based on the Hankel matrix, the Hankel matrix is sampled, constructed, singular value decomposition and threshold settings to separate the signal and noise space, and using the Prony-Kung method to estimate the frequency point position, which improves the accuracy of spectrum detection and efficiency.
Accurately detecting the number and location of the main user occupied by the broadband spectrum in a short time, reducing the complexity of spectrum perception, improving the accuracy of spectrum detection and the load capacity of the equipment, and is suitable for dynamic communication scenarios of drones.
Smart Images

Figure CN115173977B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of wireless communication for target detection, such as unmanned aerial vehicles, automatic driving and other technical fields, and specifically relates to a Prony-Kung broadband spectrum sensing method and system based on a Hankel matrix. Background Art
[0002] Drones are highly maneuverable, have low equipment costs, and are less risky. As the performance of drones gradually improves, they will gradually become tasks for communication, interference, reconnaissance, and security. On the other hand, the electromagnetic spectrum environment is complex and changeable when drones are flying, and is interfered by multiple factors. Therefore, being able to accurately find the frequency point for mutual communication in a short period of time, that is, spectrum sensing, is one of the foundations for ensuring the completion of communication tasks, and is also the basis for completing this application.
[0003] Drones have their own characteristics. First, the load capacity of drones is limited. Second, they keep high-speed movement most of the time. Considering these characteristics, the existing spectrum sensing methods have obvious shortcomings when applied to drones. So far, there are many broadband spectrum sensing methods. The main methods are serial or parallel scanning sensing, frequency detection based on compressed sensing, etc. Serial scanning sensing uses a narrowband bandpass filter with adjustable center frequency, and uses narrowband spectrum sensing technology to detect one channel at a time, and scan the entire frequency band in a serial manner. In this way, it takes a lot of time to sense a broadband frequency band, which is obviously not suitable for scenes that change instantly. Parallel scanning sensing is to recognize that a group of narrowband bandpass filters are closely followed by the front end of the wireless device, and the sub-band division is completed using the filter group using the existing sub-band information, and multiple sub-bands are detected simultaneously in a parallel manner. Obviously, parallel scanning sensing saves detection time than serial scanning detection, but it requires much more time on the device than serial scanning detection. This requirement is not very applicable in drones because the load of drones is limited. In addition, the detection method based on compressed sensing is also a spectrum detection method that has been proposed and rapidly developed in recent years. However, this method requires that the occupied frequency band is sparse, which greatly limits the application scope of such methods.
[0004] In view of the problems existing in the above prior art, the present invention proposes a new broadband spectrum sensing technical solution, namely, a broadband spectrum sensing method and system of Prony-Kung based on Hankel matrix. The Prony method involved here is an ancient and effective method for solving the exponential sum problem, but the stability of the Prony method is weak and depends on the properties of the Hankel matrix. Therefore, the present invention generates an improved Prony method, namely, the Prony-Kung method. Summary of the invention
[0005] In view of the above problems existing in the prior art, the present invention proposes a Prony-Kung broadband spectrum sensing method and system based on the Hankel matrix, so that under the action of a single sensing user (a drone), the entire broadband spectrum can be retrieved, and the number of frequency points occupied by the main user can be accurately detected in a short time, and the specific positions of the occupied frequencies can be determined at the same time, thereby improving the accuracy of spectrum detection within the broadband range and reducing the complexity of broadband spectrum sensing.
[0006] The present invention adopts the following technical solutions:
[0007] The wideband spectrum sensing method of Prony-Kung based on Hankel matrix includes the following steps:
[0008] S1. Use a filter to sample the wideband spectrum to obtain the collected received signal.
[0009] S2. Construct a Hankel matrix according to the received signal and the set reference information.
[0010] S3. Perform singular value decomposition on the Hankel matrix, separate the signal space from the noise space using the thresholds set under high and low signal-to-noise ratios, and determine the number of primary user signals occupying the frequency band.
[0011] S4. Use the Prony-Kung method to estimate the specific frequency location occupied by the primary user.
[0012] As a preferred solution, in S1:
[0013] Assume that the bandwidth to be detected is B, and there are P users sending signals in this bandwidth. The center frequency of the Pth user is f P , the signal bandwidth is B P (Random), use a filter to sample the signal.
[0014] The exponential and mathematical expression of the acquired signal are:
[0015]
[0016] Among them, ρ represents the received signal-to-noise ratio, y N is the received signal after acquisition (including the main user signal and noise), h p represents the channel from the pth user to the detector, s p represents the pth random transmission signal, f p represents the frequency point position of the pth primary user in the broadband, Δt is the sampling interval of the collected signal data (equally spaced sampling data), N represents the number of sampled samples, and w N is a complex Gaussian white noise, obeying CN(0,δ 2 ).
[0017] As a preferred solution, the specific steps of S2 are as follows:
[0018] The data collected within N·Δt time is {y1,y2,…,y N}, can be written in the form of the following vector:
[0019]
[0020] Write the exponents and mathematical expressions in step S1 into the following matrix expression:
[0021]
[0022] Among them, c P =h P ·s P , represents the set of all complex numbers with N rows and 1 column, and j represents the imaginary unit. From the above formula, we can see that if we ignore the noise for the time being, the sample vector Y is in the support vector The generated space is completely composed of occupied frequency points f1, f2, ..., f P Therefore, the spectrum sensing problem is actually the problem of determining the subspace.
[0023] Next order
[0024]
[0025] Get the final model expression:
[0026]
[0027] Next, we will transform the one-dimensional time series, namely Y N×1 Converted to a matrix with multi-dimensional space, written in Hankel matrix form as follows:
[0028]
[0029] Where Q and L are two positive integers, satisfying the conditions Q+L-1=N, Q≥K, L≥K, and Q≥L+1; when N is an even number, then take When N is an odd number, then An integer that satisfies L=N-Q+1.
[0030] As a preferred solution, the specific steps of S3 are as follows:
[0031] S3.1, perform singular value decomposition (SVD) on the Hankel matrix in S2, which has the following form:
[0032] H Q×L =UDV H
[0033] Among them, U and V are unitary matrices of size Q×L and L×L respectively, D is a L×L diagonal matrix, D=diag(d1,d2,…,d L ), diag means taking diagonal elements, d1, d2, ..., d L All are diagonal elements, and d1≥d2≥…≥d P ≥d P+1 ≥d P+2 ≥…≥d L ≥0.
[0034] S3.2. Determine the threshold to separate signal singular values from noise singular values:
[0035] S3.2.1. Determining the threshold λ in the case of low signal-to-noise ratio w1
[0036] Using the binary hypothesis model, when only noise exists, the false alarm probability P f To set the threshold λ w1 .
[0037] The binary hypothesis model is as follows:
[0038]
[0039] Among them, ρ is the signal-to-noise ratio (the ratio of the primary user signal energy to the noise energy) generated in the channel from the transmitter to the receiver, N0 indicates when only noise exists, N1 indicates when both signal and noise exist, S is the primary user's random transmission signal, and W1 and W2 are both complex Gaussian noises with a mean of 0 and a variance of 1. The false alarm probability is used when only noise exists and the misjudged signal exists. The Hankel matrix is constructed for the noise signal, singular value decomposition is performed, and a Monte Carlo simulation experiment is performed to determine the threshold. The false alarm probability needs to be controlled between (0-0.1) (too high a false alarm probability will lead to an artificially high detection probability).
[0040] When the threshold is determined as λ w1 After that, the singular values obtained after singular value decomposition are arranged from large to small and compared with the threshold λ w1 Compare them respectively, set the singular values smaller than the threshold to zero, and retain those larger than the threshold λ w1 The singular value of is greater than the threshold λ w1 The number of singular values of is the number of primary user signals, denoted as
[0041] S3.2.2. Determining the threshold λ under high signal-to-noise ratio w2
[0042] Construct a Hankel matrix for the received signal of the one-dimensional time series, perform singular value decomposition on the Hankel matrix, and obtain the maximum singular value, denoted as d1, satisfying (δ W is the noise standard deviation), and Obeying the Tracy-Widom distribution, here Setting P f After that, the threshold λ under high signal-to-noise ratio can be obtained w2 , Estimated value of the number of users P The first p-value that satisfies the inequality: d 2·p ≤d1 / λ w2 .
[0043] As a preferred solution, in S4:
[0044] Frequency f p Estimation of (p=1,2,…,P): Prony-Kung method can be used, the specific method is as follows:
[0045] S4.1. Take the first row to the Q-1 row and the first column to the Q-1 column of the U matrix. Column, denoted as U1, take the 2nd to Qth row of the U matrix, the 1st to the Denoted as U2:
[0046]
[0047] S4.2, calculation matrix (U1 H U1) -1 U1 H The eigenvalue of U2 is:
[0048] eig{(U1 H U1) -1 U1 H U2}
[0049] The eigenvalue can be obtained, denoted as Where eig means taking the eigenvalue.
[0050] S4.3, then according to the formula:
[0051]
[0052] Get λ j The phase angle is denoted as Where angle represents the angle at which the eigenvalue is taken, θ j Indicates the true angle, is θ j The estimated value of .
[0053] S4.4. Order Get f j The estimate of f j is the actual occupied frequency point,
[0054] The present invention also discloses a Prony-Kung broadband spectrum sensing system based on Hankel matrix, comprising the following modules:
[0055] Sampling module: uses a filter to sample the broadband spectrum to obtain the received signal after acquisition;
[0056] Matrix construction module: constructs the Hankel matrix according to the received signal and the set parameter information;
[0057] Singular value decomposition and primary user number determination module: Perform singular value decomposition on the Hankel matrix, set the threshold, separate the signal space and the noise space, and determine the number of primary user signals occupying the frequency band.
[0058] Specific frequency determination module: Use the Prony-Kung method to estimate the frequency position occupied by the primary user.
[0059] The present invention supports adaptive adjustment of frequency, power, bandwidth, waveform and other frequency parameters, laying a foundation for a dynamic access success rate greater than 95%. The present invention can not only achieve the above objectives, but also provide an effective technology for fast and dynamic full-frequency monitoring. BRIEF DESCRIPTION OF THE DRAWINGS
[0060] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0061] Figure 1 is a flow chart of a wideband spectrum sensing method of Prony-Kung based on Hankel matrix in Example 1;
[0062] Figure 2 This is an overall structural framework diagram of a wideband spectrum sensing system of Prony-Kung based on Hankel matrix in Example 2;
[0063] Figure 3 It is a simulation diagram of the accuracy of estimating the number of primary users by the method of the present invention;
[0064] Figure 4It is a simulation diagram of the average error probability of the actual occupied frequency points and the estimated frequency points under different signal-to-noise ratios under the condition that the number of primary users is known according to the method of the present invention.
[0065] Figure 5 It is a simulation diagram of the average error probability of the actual occupied frequency points and the estimated frequency points under different signal-to-noise ratios under the condition that the number of primary users is unknown and the detection is correct according to the method of the present invention. DETAILED DESCRIPTION
[0066] The following embodiments illustrate the method of the present invention through specific examples. The present invention further improves the broadband spectrum sensing technology of multiple main users and one sensing user (a drone), and the spectrum sensing capability is improved, thereby further improving the ability to find spectrum holes.
[0067] Example 1
[0068] See also Figure 1 , assuming that there are a total of P (randomly selected) primary users transmitting signals and a sensing user N r (A drone) senses the primary user signal. Table 1 below gives general parameter settings, and the broadband spectrum can be sensed based on Table 1.
[0069] Table 1 Parameter settings
[0070]
[0071]
[0072] The wideband spectrum sensing method of Prony-Kung of the Hankel matrix in this embodiment specifically includes the following steps:
[0073] S1. Use a filter to sample the wideband spectrum to obtain the collected received signal.
[0074] S2. Construct a Hankel matrix according to the received signal and the parameter information set above.
[0075] S3. Perform singular value decomposition (SVD) on the Hankel matrix and separate the signal space from the noise space using thresholds set at high and low signal-to-noise ratios to determine the number of primary user signals.
[0076] S4. Use the Prony-Kung method to estimate the specific frequency position occupied by the primary user.
[0077] This embodiment uses a method of converting a single-dimensional time series into a matrix with multi-dimensional spatial characteristics, namely, a Hankel matrix. When there is only noise, a threshold value is determined under a low signal-to-noise ratio, and when there is a mixed signal, a threshold value is determined under a high signal-to-noise ratio. The number of primary user signals is determined in this way, and then the specific frequency point position occupied by the primary user is estimated. The accuracy of the estimation is greatly improved under a low signal-to-noise ratio, and the broadband spectrum perception capability is improved.
[0078] The specific perception process is as follows:
[0079] In the broadband spectrum sensing, there are P primary users sending signals and one secondary user, i.e., the sensing user (drone). The specific steps in S1 are as follows:
[0080] A filter is used to sample the frequency band B = 20 MHz at equal intervals. The sampling time interval Δt is (5×10 -8 )s; this wideband spectrum is divided into (0-20MHz / 2):(0+20MHz / 2), that is, dd=(-10MHz:10MHz). There are P primary users sending signals in this bandwidth. The center frequency of each primary user signal is in the spectrum range of dd. The bandwidth of the transmitted signal is unknown. Within the time range of 300×Δt, the transmitted sampling data {y1,y2,…,y N}, these data come from a symbol of the user and can be expressed as follows:
[0081]
[0082] Where N represents the number of sample data, h1,h2,…,h P represents the channel from which P users arrive at the detection party, where {h1,h2,…,h P} obeys a complex Gaussian distribution with a mean of 0 and a variance of 1. The values are randomly generated. s1, s2, …, sP represent random emission signals. Randomly select f1,f2,…f P represents the frequency position of P primary users in the broadband, Δt is the sampling interval of collecting data signals, N is the number of transmitting antennas, and w N is complex Gaussian noise, obeying CN(0,δ 2 ), where δ 2 =1, j represents an imaginary unit. In this embodiment j 2 =-1.
[0083] The specific process of step S2 is as follows:
[0084] In S1, the mathematical expression of the collected signal in the model is written in the following matrix form:
[0085] Y300×1 =[y1,y2,…,y 300 ] T
[0086] Among them, T represents transpose, which is represented by the following matrix:
[0087]
[0088] make
[0089]
[0090] Get the final model expression:
[0091]
[0092] The single-dimensional time series, namely Y 300×1 The vector is converted into a matrix with multi-dimensional space and written in Hnakel matrix form as follows:
[0093]
[0094] According to the above formula, Q+L-1=300, Q≥P and L≥P, Q is greater than or equal to An integer, where Q is 151 and L is 150.
[0095] The specific process of step S3 is as follows:
[0096] S3.1. For the Hankel matrix H 151×150 Perform singular value decomposition (SVD), which has the following form:
[0097] H 151×150 =UDV H
[0098] Among them, U and V are unitary matrices of size 151×150 and 150×150 respectively, and D is a diagonal matrix of size 150×150, D=diag(d1,d2,…,d 150 ), where d1≥d2≥…≥d P >>d6≥…≥d 150 ≥0, the first P singular values are much larger than the following singular values, and the following 150-P singular values are very small.
[0099] S3.2. Setting threshold λ at low signal-to-noise ratio w1 To separate the signal singular values from the noise singular values
[0100] S3.2.1. According to the binary hypothesis model in probability statistics, when only noise exists, the matrix expression obtained by collecting the received signal is as follows:
[0101]
[0102] Among them, W N×1 is a 300×1 vector, N is the number of samples, W N×1 Obey a complex Gaussian distribution with mean 0 and variance 1, w1,w2,…,w 300 is a randomly generated complex value from a complex Gaussian distribution.
[0103] S3.2.2, construct the Hankel matrix:
[0104] The W in step S3.2.1 above N×1 Converted into a matrix with multi-dimensional space characteristics, as follows:
[0105]
[0106] That is, the Hankel matrix form in which only noise exists is obtained.
[0107] S3.2.3. Perform SVD transformation on the above Hankel matrix:
[0108] H 151×150 =U w D w V w H
[0109] in, D w is a 150×150 diagonal matrix, D w =diag(d w·1 ,d w·2 ,…,d w·150 ), and d w·1 ,d w·2 ,…,d w·150 Yes D w The diagonal elements of w·1 ≥d w·2 ≥…≥d w·150 ≥ 0. Then the Vandermonde experiment was performed 10,000 times, and the largest singular value d was taken each time. w·1 After the experiment, we get 10,000 largest singular values, sort them from large to small, and take the 500th one as the final threshold λ w1 (P f =0.05).
[0110] S3.2.4. Determine the number of primary user signals as follows:
[0111] Threshold λ w1After the determination is made, all the singular values d1, d2, ..., d 150 With threshold λ w1 Compare and set the value less than the threshold λ w1 The singular values of are set to zero, and those greater than the threshold λ are retained. w1 The singular values of w1 The number of singular values of is regarded as the number of primary user signals The number of primary users is determined. The actual number of primary users P in Table 1 is compared with the estimated number of primary users Compare, if That is, the estimation is correct. The Monte Carlo simulation experiment is performed 10,000 times, and the average accuracy of the estimated number of primary users is calculated.
[0112] S3.3. Setting threshold λ at high signal-to-noise ratio w2 To separate the signal singular values from the noise singular values
[0113] Construct a Hankel matrix for the received signal of the one-dimensional time series, perform singular value decomposition on the Hankel matrix, and obtain the maximum singular value, denoted as d1, satisfying Obeying the Tracy-Widom distribution, here Using the P set in Table 1 above f , we can get the threshold λ under high signal-to-noise ratio w2 , Estimated value of the number of users P The first p-value that satisfies the inequality: d 2·p ≤d1 / λ w2 . Compare the actual number of primary users P in Table 1 with the estimated number of primary users Compare, if That is, the estimation is correct. The Monte Carlo simulation experiment is performed 10,000 times, and the average accuracy of the estimated number of primary users is calculated.
[0114] Step S4 is specifically as follows:
[0115] When the estimated number of primary users is the same as the number of primary users actually occupying the broadband spectrum, that is, When the primary user occupies a certain frequency in the broadband spectrum, the next step can be carried out to estimate the specific frequency point.
[0116] S4.1. For U, take rows 1 to 150 and columns 1 to P, record it as U1; for U, take rows 2 to 151 and columns 1 to P, record it as U2, that is:
[0117]
[0118] S4.2、Calculate (U1 H U1) -1 U1 H The eigenvalue of U2 is:
[0119] eig{(U1 H U1) -1 U1 H U2}
[0120] The characteristic value can be obtained Where eig means taking the eigenvalue, yes The estimated value of .
[0121] S4.3, then according to the following formula:
[0122]
[0123] get The phase angle is denoted as Among them, angle represents the angle at which the eigenvalue is taken. yes The estimated value of .
[0124] S4.4. Next, estimate the specific occupied frequency points according to the formula:
[0125]
[0126] The specific frequency points occupied in the broadband spectrum can be obtained, where Yes j Then use the average estimation accuracy to estimate the error probability of the estimated frequency points:
[0127]
[0128] Reference Figure 3 As shown, the accuracy of the number of primary users is detected. As the signal-to-noise ratio continues to increase, the accuracy of the method of the present invention estimates the number of primary users. At 6 dB, the detection probability can reach 100%, and the ability to find spectrum holes is improved. Figure 4 As shown in FIG. 1 , the average estimated error probability of the frequency point is detected under the assumption that the number of primary users is unknown. It can be seen that the method of the present invention decreases with the gradual increase of the signal-to-noise ratio. At -14 dB, the average error probability is 6%, and it can be reduced to 0.7% at 4 dB. Moreover, as the signal-to-noise ratio increases, the average error probability will continue to decrease. Figure 5As shown, the average estimated error probability of the frequency point is detected under the condition that the number of primary users is unknown. At a low signal-to-noise ratio of -14dB, the average error probability of the method of the present invention is 4%, and as the signal-to-noise ratio continues to increase, its average error probability continues to decrease. When the signal-to-noise ratio is 4dB, its error probability can be reduced to 0.2%, and there is still room for decline. The conclusions drawn from the above three situations verify the effectiveness of the method of the present invention.
[0129] Example 2
[0130] like Figure 2 As shown, this embodiment, based on the method of Embodiment 1, discloses a spectrum sensing system based on the Prony-Kung of the Hankel matrix, including the following modules:
[0131] Sampling module: uses a filter to sample the broadband spectrum to obtain the received signal after acquisition;
[0132] Matrix construction module: constructs the Hankel matrix according to the received signal and the set parameter information;
[0133] Singular value decomposition and primary user number determination module: Perform singular value decomposition on the Hankel matrix, set the threshold, separate the signal space and the noise space, and determine the number of primary user signals occupying the frequency band.
[0134] Specific frequency determination module: Use the Prony-Kung method to estimate the frequency position occupied by the primary user.
[0135] For other contents of this embodiment, please refer to Embodiment 1.
[0136] The specific embodiments described herein are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications to the specific embodiments described or replace them with similar methods without departing from the spirit of the present invention or exceeding the scope defined by the appended claims.
Claims
1. Prony-Kung's broadband spectrum sensing method based on Hankel matrix, which is characterized by The steps include: S1, use the filter to sample the wideband spectrum to obtain the received signal after acquisition; S2. Constructing a Hankel matrix according to the received signal and the set parameter information; S3, perform singular value decomposition on the Hankel matrix, separate the signal space from the noise space using the thresholds set under high and low signal-to-noise ratios, and determine the number of primary user signals occupying the frequency band; S4, using the Prony-Kung method to estimate the frequency position occupied by the primary user; In step S4, the frequency point f p The estimation of is carried out using the Prony-Kung method, where p = 1, 2, ..., P, where P is the number of users sending signals in the bandwidth, as follows: S4.
1. Take the first row to the Q-1 row and the first column to the Q-1 column of the U matrix. Column, denoted as U1, take the 2nd to Qth row of the U matrix, the 1st to the The column is denoted as U2: S4.2, calculation matrix (U1 H U1) -1 U1 H The eigenvalue of U2 is: own{(U1 H U1) -1 U1 H U2} Get the eigenvalue, denoted as Among them, eig means taking the eigenvalue; S4.3, according to the following formula: Get λ j The phase angle is denoted as Among them, angle represents the angle of taking eigenvalue, θ j Indicates the true angle, is θ j An estimated value of S4.
4. Order Get f j The estimate of f j is the actual occupied frequency point, Δt is the sampling interval for collecting signal data.
2. The wideband spectrum sensing method based on the Prony-Kung of Hankel matrix as claimed in claim 1, characterized in that: In step S1, let the detected bandwidth be B, in which P users are sending signals, and the rate point of the Pth user is f P , the signal bandwidth is B P , use a filter to sample the signal; The exponential and mathematical expression of the acquired signal are: Among them, ρ represents the received signal-to-noise ratio, y N is the received signal after acquisition, h p represents the channel from p users to the detection party, s p represents the pth random transmission signal, f p represents the frequency point position occupied by the pth primary user in the broadband, N represents the number of sampling samples, and w N is a complex Gaussian white noise, obeying CN(0,δ 2 ).
3. The wideband spectrum sensing method based on the Prony-Kung of Hankel matrix as claimed in claim 2, characterized in that: Step S2 is specifically as follows: The data collected within N·Δt time is {y1,y2,…,y N }, written in the form of the following vector: The exponents and mathematical expressions in step S1 are written as the following matrix expressions: Among them, c P =h P ·s P , represents the set of all complex numbers with size N rows and 1 column, j represents the imaginary unit; from the above formula, if the noise is not considered, the sample vector Y is in the support vector The generated space is completely composed of occupied frequency points f1, f2, ..., f P Therefore, the spectrum sensing problem is the problem of determining the subspace; make Get the final model expression: The one-dimensional time series, namely Y N×1 Converted to a matrix with multi-dimensional space, written in Hankel matrix form as follows: Where Q and L are two positive integers, satisfying the conditions Q+L-1=N, Q≥K, L≥K, and Q≥L+1; when N is an even number, then take An integer of, when N is an odd number, then take An integer that satisfies L=N-Q+1.
4. The wideband spectrum sensing method based on the Prony-Kung of Hankel matrix as claimed in claim 3, characterized in that: Step S3 is as follows: S3.
1. Perform singular value decomposition on the Hankel matrix in step S2, which has the following form: H Q×L =UDV H Among them, U and V are unitary matrices of size Q×L and L×L respectively, D is a L×L diagonal matrix, D=diag(d1,d2,…,d L ), diag means taking diagonal elements, d1, d2, ..., d L All are diagonal elements, and d1≥d2≥…≥d P ≥d P+1 ≥d P+2 ≥…≥d L ≥0; S3.
2. Determine the threshold to separate signal singular values from noise singular values: S3.2.
1. Determining the threshold λ in the case of low signal-to-noise ratio w1 Using the binary hypothesis model, when only noise exists, the false alarm probability P f To set the threshold λ w1 ; The binary hypothesis model is as follows: Among them, ρ is the signal-to-noise ratio generated in the channel from the transmitter to the receiver, N0 indicates when only noise exists, N1 indicates that both signal and noise exist, S is the primary user's random transmission signal, W1 and W2 are both complex Gaussian noises with a mean of 0 and a variance of 1. The false alarm probability is used. When only noise exists, the Hankel matrix is constructed for the noise signal, and the singular value decomposition is performed. The Monte Carlo simulation experiment is performed to determine the threshold. The false alarm probability is controlled between (0-0.1). When the threshold is determined as λ w1 After that, the singular values obtained after singular value decomposition are arranged from large to small and compared with the threshold λ w1 Compare them respectively, set the singular values smaller than the threshold to zero, and retain those larger than the threshold λ w1 The singular value of is greater than the threshold λ w1 The number of singular values of is the number of primary user signals, denoted as S3.2.
2. Determining the threshold λ under high signal-to-noise ratio w2 Construct a Hankel matrix for the received signal of the one-dimensional time series, perform singular value decomposition on the Hankel matrix, and obtain the maximum singular value, denoted as d1, satisfying δ W is the noise standard deviation, and Obeying the Tracy-Widom distribution, here Setting P f Then, we get the threshold λ under high signal-to-noise ratio w2 , Estimated value of the number of users P The first p-value that satisfies the inequality: d 2·p ≤d1 / λ w2 .
5. Prony-Kung's broadband spectrum sensing system based on Hankel matrix, characterized by Includes the following modules: Sampling module: uses a filter to sample the broadband spectrum to obtain the received signal after acquisition; Matrix construction module: constructs the Hankel matrix according to the received signal and the set parameter information; Singular value decomposition and primary user number determination module: Perform singular value decomposition on the Hankel matrix, set the threshold, separate the signal space and the noise space, and determine the number of primary user signals occupying the frequency band; Specific frequency determination module: Use the Prony-Kung method to estimate the frequency position occupied by the primary user; frequency f p The estimation of is carried out using the Prony-Kung method, (p = 1, 2, ..., P, P is the number of users sending signals in the bandwidth, as follows: S4.
1. Take the first row to the Q-1 row and the first column to the Q-1 column of the U matrix. Column, denoted as U1, take the 2nd to Qth row of the U matrix, the 1st to the The column is denoted as U2: S4.2, calculation matrix (U1 H U1) -1 U1 H The eigenvalue of U2 is: own{(U1 H U1) -1 U1 H U2} Get the eigenvalue, denoted as Among them, eig means taking the eigenvalue; S4.3, according to the following formula: Get λ j The phase angle is denoted as Among them, angle represents the angle of taking eigenvalue, θ j Indicates the true angle, is θ j An estimated value of S4.
4. Order Get f j The estimate of f j is the actual occupied frequency point, Δt is the sampling interval for collecting signal data.
