Frequency spectrum aliasing wireless signal depth identification method

Abstract

The invention discloses a spectrum aliasing wireless signal depth identification method, which comprises the following steps of: performing non-negative matrix factorization on M monitored spectrum aliasing signals, constructing an objective function Q1 and solving the objective function Q1 to obtain a coefficient matrix Am and a signal type consensus matrix A * containing type information of the mth spectrum aliasing signal; and a target function Q2 is constructed, the Am is further decomposed, the position and the transmitting power of each signal in the spectrum aliasing signals are obtained, and the positions of the signals are grid positions obtained through grid division of the monitoring area. According to the invention, signal identification, signal position estimation and transmitting power estimation are integrated in a depth identification framework, and decoupling estimation is carried out on multiple parameters of a spectrum aliasing signal through a multi-layer NMF; according to the invention, deep identification can be carried out on the spectrum aliasing signal, and the type, the position and the transmitting power of the unknown signal in the spectrum aliasing signal can be identified.

Description

Technical Field

[0001] The present invention relates to the field of spectrum aliasing signal recognition, and in particular to a spectrum aliasing wireless signal deep recognition method. Background Art

[0002] With the widespread use of wireless communication devices and the application of flexible radio spectrum strategies, such as 2.4 GHz unlicensed spectrum, spectrum sharing, dynamic frequency selection, etc., spectrum aliasing signals appear more and more frequently, and some illegal spectrum aliasing signals will bring more serious interference.

[0003] At present, the research on the identification of spectral aliasing signals usually uses blind source separation technology, such as independent component analysis (ICA), principal component analysis (PCA) and singular value decomposition (SVD), etc., but these technologies cannot accurately determine the number of mixed signals and it is difficult to keep the original scale of the signal unchanged, and it is no longer possible to locate the signal and estimate the transmission power. The non-negative matrix factorization (NMF) algorithm can effectively decouple the aliasing signal through matrix decomposition, and can simultaneously identify the signal type and determine the number of mixed signals. Due to its good separation structure, the multi-layer NMF algorithm composed of multiple NMFs can perform multi-parameter decoupling estimation, and can locate the signal and estimate the transmission power on the basis of signal type identification. In addition, most studies on blind source separation technology only focus on static scenes. Fixed radio monitoring systems are limited by their static positions and signal reception coverage, so radio monitoring systems need to have flexible deployment capabilities.

[0004] For example, the structured NMF model [1] can separate the spectra of overlapping signals transmitted on different carriers without knowing the number of unknown signal sources. This is a blind source separation technology, but it cannot identify multiple types of signals with aliased spectra.

[0005] Using an unmanned aerial vehicle (UAV) equipped with a spectrum analyzer to dynamically analyze radio signals in a specified area is a more flexible way. Currently, the radio signal identification and positioning technology based on UAVs mainly relies on the characteristics of received and transmitted signals. UAVs move widely with their maneuverability, analyze signal types based on spectrum analyzers, and perform direction finding and positioning based on parameters such as time, angle, and signal strength. However, most designs and research on UAV-based monitoring systems focus on locating a single type of signal, which is challenging in environments containing multiple categories of unknown signals.

[0006] For example, research [2] proposed a direct positioning method for simultaneous positioning of multiple transmitters based on mobile drone monitoring, which locates the radiation source through the collaborative direction finding of multiple drones. However, this solution is also unable to identify multiple types of signals with spectral aliasing.

[0007] [1]

[0008] [2]Li, Y.He, Summary of the invention

[0009] The purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art and to provide a method for deep identification of spectral aliased wireless signals.

[0010] The purpose of the present invention is achieved through the following technical solutions:

[0011] A method for deep identification of spectrum aliased wireless signals, comprising the following steps:

[0012] (1) The M monitored spectral aliasing signals are decomposed by non-negative matrix, the objective function Q1 is constructed and solved, and the coefficient matrix A containing the type information of the mth spectral aliasing signal is obtained. m , signal type consensus matrix A * ; Among them A * To combine all A m ;

[0013] The objective function Q1 is as follows:

[0014]

[0015] In the formula, R m is the mth spectrum aliasing signal monitored, 1≤m≤M; B s is a dictionary of radio signal spectrum, is the signal type determination matrix of the mth spectrum aliasing signal, indicating whether it is this type of signal through 0 and 1;

[0016] (2) Construct the objective function Q2, and mFurther decomposition is performed to obtain the position and transmission power of each signal in the spectrum aliasing signal, wherein the position of the signal is the grid position obtained by gridding the monitoring area;

[0017] The objective function Q2 is as follows:

[0018]

[0019] Where P m is the diagonal matrix of the mth spectrum aliasing signal, and the values ​​on the diagonal correspond to the transmission power of different signals; L m is the position indication matrix of the mth spectrum aliasing signal, which indicates whether there is an nth signal at the gth grid by 0 and 1. m is the signal path loss fading matrix of the mth spectrum aliasing signal, L * , P * They are respectively m , P m The consensus matrix.

[0020] In step (1), the objective function Q1 is solved alternately by m With A * is convex, respectively for A m With A * The derivative obtains the update rule:

[0021]

[0022] Where T is the matrix transpose operation; is the transposed spectrum dictionary of the radio signal;

[0023] Get an estimate of the number of unknown signal sources for:

[0024]

[0025] in is the matrix A * The nth element in .

[0026] In step (2), objective functions Q3 and Q4 are constructed, and the objective function Q2 is transformed and then solved:

[0027] Process the diagonal matrix restrictions and 0-1 integer constraints in Q2, and let H m =P m L m , the constraint in Q2 is transformed into: ||(H m ) n ||1=||(H m ) n ||2,(Hm ) n Yes H m The nth row vector of , the objective function Q3 is:

[0028]

[0029] In the formula, H m is the matrix of the position and transmission power of the unknown signal estimated from the mth perspective; H * It is the consensus matrix of the position and transmission power of the unknown signal obtained from all viewing angles;

[0030] The non-convex constraints are transformed through the penalty term to obtain the objective function Q4:

[0031]

[0032] In the formula, λ is H * The penalty factor; ρ is the penalty coefficient; Indicates the matrix transposition of a matrix with all elements being 1;

[0033] Since Q4 satisfies the KL condition, it is solved by the proximal alternating linear minimization algorithm. m With H * The update rules are as follows:

[0034]

[0035] Solve to get H * , H * The nth row of represents the position and transmission power of the nth spectrum aliasing signal, where the subscript corresponding to the maximum element of the row is the grid number where the nth spectrum aliasing signal is located, and the maximum element is the transmission power of the nth spectrum aliasing signal; T is a matrix transpose operation; 1 G×G is a G×G matrix of all 1s.

[0036] Define the spectrum mixed signal received by the mth monitoring node in the tth spectrum sampling time slot as s m (t):

[0037]

[0038] in J m is the number of frequency sampling points of the mth monitoring node in the frequency band; is the n0th unknown signal in J m The standard power spectral density function at the frequency point, and A standard spectrum dictionary for edge nodes to identify signals; is the transmission power of the n0th unknown signal source in time slot t; is the channel gain coefficient from the n0th unknown signal source to the mth monitoring node, z m (t) is the noise at the mth monitoring node and obeys Gaussian distribution.

[0039] Said Defined as:

[0040]

[0041] Among them, the channel constant f is the operating center frequency of the unknown signal, c is the speed of light, E r,m and are the antenna gains of the mth monitoring node and the n0th unknown signal respectively; and are the channel response and distance between the mth monitoring node and the n0th unknown signal respectively.

[0042] Said Defined as:

[0043]

[0044] in and (x r,m ,y r,m ,h m ) are the coordinates of the n0th unknown signal and the mth monitoring node in the coordinate system with the edge node as the origin; the maximum horizontal coordinate and maximum vertical coordinate of the signal monitoring range of the monitoring network are ±x max ±y max ; In T s The link loss from the gth grid to the mth monitoring node remains unchanged;

[0045] At the same time, the present invention provides:

[0046] A server includes a processor and a memory, wherein the memory stores at least one program, and the program is loaded and executed by the processor to implement the above-mentioned spectrum aliasing wireless signal deep identification method.

[0047] A computer-readable storage medium stores at least one program, which is loaded and executed by a processor to implement the above-mentioned spectrum aliasing wireless signal deep identification method.

[0048] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0049] The present invention integrates signal recognition, signal position estimation and transmission power estimation in a deep recognition framework, and decouples and estimates multiple parameters of spectrum aliasing signals through multi-layer NMF; the present invention can perform deep recognition of spectrum aliasing signals and identify the type, position and transmission power of unknown signals in the spectrum aliasing signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 The present invention is a flow chart of a method for deep identification of spectrum aliased wireless signals.

[0051] Figure 2 This is a comparison chart of signal type recognition results under different algorithms.

[0052] Figure 3 This is a comparison chart of signal position estimation results under different algorithms. DETAILED DESCRIPTION

[0053] The present invention is further described in detail below in conjunction with embodiments and drawings, but the embodiments of the present invention are not limited thereto.

[0054] Assuming that M monitoring nodes are evenly distributed within the communication range, the spectrum data generated by the spectrum scanning of each monitoring node is transmitted back to the edge computing node via wireless. The time it takes for each monitoring node to complete the detection of a frequency band is T s , the frequency spectrum data transmission time of this frequency band is T d , then the total time to complete a frequency band monitoring task is T0 = T s +T d Assume that there are N0 unknown wireless signal sources in the same frequency band within the monitoring area, and the monitoring node has no prior information about their signal type, location, transmission power, etc.

[0055] The spectrum mixed signal received by the mth monitoring node in the tth spectrum sampling time slot is:

[0056]

[0057] in J m is the number of frequency sampling points of the mth monitoring node in the frequency band. is the n0th unknown signal in J m The standard power spectral density function at the frequency point, and It is a standard spectrum dictionary for edge nodes to identify signals. is the transmission power of the n0th unknown signal source in time slot t, is the channel gain from the n0th unknown signal source to the mth monitoring node, z m (t) is the noise at the mth monitoring node and obeys Gaussian distribution. We mainly consider the large-scale fading of wireless propagation in outdoor rich scattering environment, so the channel gain coefficient is defined as:

[0058]

[0059] Among them, the channel constant f is the operating center frequency of the unknown signal, c is the speed of light, E r,m and are the antenna gains of the mth monitoring node and the n0th unknown signal respectively. and are the channel response and distance between the mth monitoring node and the n0th unknown signal respectively, and

[0060]

[0061] in and (x r,m ,y r,m ,h m ) are the coordinates of the n0th unknown signal and the mth monitoring node in the coordinate system with the edge node as the origin. The maximum horizontal coordinate and maximum vertical coordinate of the signal monitoring range of the monitoring network are (±x max ,±y max ). In T s The link loss from the gth grid to the mth monitoring node is A dictionary of signal fading within a defined area And W m =[q 1,m ,q 2,m ……q G,m ],in is the signal fading template matrix of the mth monitoring node.

[0062] like Figure 1 , a spectrum aliasing wireless signal deep identification method, comprising the following steps:

[0063] (1) The M monitored spectral aliasing signals are decomposed by non-negative matrix, the objective function Q1 is constructed and solved, and the coefficient matrix A containing the type information of the mth spectral aliasing signal is obtained. m , signal type consensus matrix A * ; Among them A * To combine all A m ;

[0064] The objective function Q1 is as follows:

[0065]

[0066] In the formula, Rm is the mth spectrum aliasing signal monitored, 1≤m≤M; B s is a dictionary of radio signal spectrum, is the signal type determination matrix of the mth spectrum aliasing signal, indicating whether it is this type of signal through 0 and 1;

[0067] (2) Construct the objective function Q2, and m Further decomposition is performed to obtain the position and transmission power of each signal in the spectrum aliasing signal, wherein the position of the signal is the grid position obtained by gridding the monitoring area;

[0068] The objective function Q2 is as follows:

[0069]

[0070] Where P m is the diagonal matrix of the mth spectrum aliasing signal, and the values ​​on the diagonal correspond to the transmission power of different signals; L m is the position indication matrix of the mth spectrum aliasing signal, which indicates whether there is an nth signal at the gth grid by 0 and 1. m is the signal path loss fading matrix of the mth spectrum aliasing signal, L * , P * They are respectively m , P m The consensus matrix.

[0071] In step (1), the objective function Q1 is solved alternately by m With A * is convex, respectively for A m With A * The derivative obtains the update rule:

[0072]

[0073] Where T is the matrix transpose operation; is the transposed spectrum dictionary of the radio signal;

[0074] Get an estimate of the number of unknown signal sources for:

[0075]

[0076] in is the matrix A * The nth element in .

[0077] In step (2), objective functions Q3 and Q4 are constructed, and the objective function Q2 is transformed and then solved:

[0078] Process the diagonal matrix restrictions and 0-1 integer constraints in Q2, and let H m =P m L m , the constraint in Q2 is transformed into: ||(H m ) n ||1=||(H m ) n ||2,(H m ) n Yes H m The nth row vector of , the objective function Q3 is:

[0079]

[0080] In the formula, H m is the matrix of the position and transmission power of the unknown signal estimated from the mth perspective; H * It is the consensus matrix of the position and transmission power of the unknown signal obtained from all viewing angles;

[0081] The non-convex constraints are transformed through the penalty term to obtain the objective function Q4:

[0082]

[0083] In the formula, λ is H * The penalty factor; ρ is the penalty coefficient; Indicates the matrix transposition of a matrix with all elements being 1;

[0084] Since Q4 satisfies the KL (Kurdyka-Lojasiewicz) condition, it is solved by the proximal alternating linear minimization (PLAM) algorithm. m With H * The update rules are as follows:

[0085]

[0086] Solve to get H * , H * The nth row of represents the position and transmission power of the nth spectrum aliasing signal, where the subscript corresponding to the maximum element of the row is the grid number where the nth spectrum aliasing signal is located, and the maximum element is the transmission power of the nth spectrum aliasing signal; T is a matrix transpose operation; 1 G×G is a G×G matrix of all 1s.

[0087] The objective function Q2 is difficult to solve, but after transformation and deformation through Q3 and Q4, the difficulty of solving is reduced.

[0088] At the same time, the present invention provides:

[0089] A server includes a processor and a memory, wherein the memory stores at least one program, and the program is loaded and executed by the processor to implement the above-mentioned spectrum aliasing wireless signal deep identification method.

[0090] A computer-readable storage medium stores at least one program, which is loaded and executed by a processor to implement the above-mentioned spectrum aliasing wireless signal deep identification method.

[0091] like Figure 2 ,In the signal type recognition task, the CMNR method proposed in this ,invention is superior to the traditional BSS method.

[0092] like Figure 3 ,In the signal position estimation task, the CMNR method proposed in this ,invention is superior to the traditional three-sided positioning ,algorithm.

[0093] Figure 2 and Figure 3 In, CMNR means collaborative multi-layer non-negative matrix factorization recognition in Chinese, JMNR means joint multi-layer non-negative matrix factorization recognition in Chinese, CMNR(no consensus) means collaborative multi-layer non-negative matrix factorization recognition without consensus matrix, JMNR(no consensus) means joint multi-layer non-negative matrix factorization recognition without consensus matrix, FastICA means fast independent component analysis in Chinese, PCA means principal component analysis in Chinese, TSVD means truncated singular value decomposition in Chinese, trilateration means trilateration in Chinese, and PSO means particle swarm optimization in Chinese.

[0094] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A method for deep identification of spectrum aliased wireless signals, characterized in that: The following steps are involved: (1) The M monitored spectral aliasing signals are decomposed by non-negative matrix, the objective function Q1 is constructed and solved, and the coefficient matrix A containing the type information of the mth spectral aliasing signal is obtained. m , signal type consensus matrix A * ; Among them A * To combine all A m ; The objective function Q1 is as follows: s.t.A m ≥0,A * ≥0 In the formula, R m is the mth spectrum aliasing signal monitored, 1≤m≤M; B s is a dictionary of radio signal spectrum, is the signal type determination matrix of the mth spectrum aliasing signal, indicating whether it is this type of signal through 0 and 1; (2) Construct the objective function Q2, and m Further decomposition is performed to obtain the position and transmission power of each signal in the spectrum aliasing signal, wherein the position of the signal is the grid position obtained by gridding the monitoring area; The objective function Q2 is as follows: Where P m is the diagonal matrix of the mth spectrum aliasing signal, and the values ​​on the diagonal correspond to the transmission power of different signals; L m is the position indication matrix of the mth spectrum aliasing signal, which indicates whether there is an nth signal at the gth grid by 0 and 1. m is the signal path loss fading matrix of the mth spectrum aliasing signal, L * , P * L m , P m The consensus matrix.

2. According to claim 1, the method for deep identification of spectrum aliasing wireless signals is characterized in that: In step (1), the objective function Q1 is solved alternately m With A * is convex, respectively for A m With A * The derivative obtains the update rule: Where T is the matrix transpose operation; is the transposed spectrum dictionary of the radio signal; Get an estimate of the number of unknown signal sources for: in is the matrix A * The nth element in .

3. The method for deep identification of spectrum aliased wireless signals according to claim 1, characterized in that: In step (2), objective functions Q3 and Q4 are constructed, and the objective function Q2 is transformed and then solved: Process the diagonal matrix restrictions and 0-1 integer constraints in Q2, and let H m =P m L m , the constraint in Q2 is transformed into: ||(H m ) n ||1=||(H m ) n ||2,(H m ) n Yes H m The nth row vector of , the objective function Q3 is: In the formula, H m is the matrix of the position and transmission power of the unknown signal estimated from the mth perspective; H * It is the consensus matrix of the position and transmission power of the unknown signal obtained from all viewing angles; The non-convex constraints are transformed through the penalty term to obtain the objective function Q4: In the formula, λ is H * The penalty factor; ρ is the penalty coefficient; 1 T Indicates the matrix transposition of a matrix with all elements being 1; Since Q4 satisfies the KL condition, it is solved by the proximal alternating linear minimization algorithm. m With H * The update rules are as follows: Solve to get H * , H * The nth row of represents the position and transmission power of the nth spectrum aliasing signal, where the subscript corresponding to the maximum element of the row is the grid number where the nth spectrum aliasing signal is located, and the maximum element is the transmission power of the nth spectrum aliasing signal; T is a matrix transpose operation; 1 G×G is a G×G matrix of all 1s.

4. The method for deep identification of spectrum aliased wireless signals according to claim 1, characterized in that: Define the spectrum mixed signal received by the mth monitoring node in the tth spectrum sampling time slot as s m (t): in J m is the number of frequency sampling points of the mth monitoring node in the frequency band; is the n0th unknown signal in J m The standard power spectral density function at the frequency point, and A standard spectrum dictionary for edge nodes to identify signals; is the transmission power of the n0th unknown signal source in time slot t; is the channel gain coefficient from the n0th unknown signal source to the mth monitoring node, z m (t) is the noise at the mth monitoring node and obeys Gaussian distribution.

5. The method for deep identification of spectrum aliased wireless signals according to claim 4, characterized in that: Said Defined as: Among them, the channel constant f is the operating center frequency of the unknown signal, c is the speed of light, E r,m and are the antenna gains of the mth monitoring node and the n0th unknown signal respectively; and are the channel response and distance between the mth monitoring node and the n0th unknown signal respectively.

6. The method for deep identification of spectrum aliased wireless signals according to claim 5, characterized in that: Said Defined as: in and (x r,m ,y r,m ,h m ) are the coordinates of the n0th unknown signal and the mth monitoring node in the coordinate system with the edge node as the origin; the maximum horizontal coordinate and maximum vertical coordinate of the signal monitoring range of the monitoring network are ±x max ±y max ; In T s The link loss from the gth grid to the mth monitoring node remains unchanged; 7. A server, comprising a processor and a memory, characterized in that: At least one program is stored in the memory, and the program is loaded and executed by the processor to implement the method for deep identification of spectrum-aliased wireless signals as described in any one of claims 1 to 6.

8. A computer-readable storage medium, wherein at least one program is stored in the storage medium, characterized in that: The program is loaded and executed by a processor to implement the method for deep identification of spectrum aliased wireless signals as described in any one of claims 1 to 6.

Images