Integrated deep aliasing spectrum signal parameter identification method

Abstract

The invention discloses an integrated depth aliasing spectrum signal parameter identification method, which comprises the following steps of: designing a three-layer depth decomposition structure, and simultaneously realizing category, position and power estimation of aliasing spectrum signals; traditional NMF is single-layer decomposition, and the decomposition result is not unique, in the stage of constructing an overall objective function, a deep NMF (Dee-NMF) mode is adopted, signal parameters are decomposed layer by layer, signal type and power estimation and positioning recognition is carried out, different layer factors are decomposed and designed to enable each layer to have different feature expressions, and the feature expression is more accurate. And each layer can obtain corresponding signal characteristic parameters. By adopting the technical scheme of the invention, the accuracy and robustness of signal identification can be effectively improved, the spectrum resource management is optimized, and the overall performance of a wireless communication system is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of aliased spectrum signal recognition, and in particular to an integrated deep aliased spectrum signal parameter recognition method. Background Art

[0002] With the widespread adoption of the internet and the rapid development of the Internet of Things (IoT), wireless communication systems have become widely used worldwide. Advances in wireless communication and the integration of new wireless technologies have led to the possibility of multiple signals occupying the same frequency band, causing mutual interference. This phenomenon is known as aliased spectrum signals. The presence of aliased spectrum signals not only affects communication quality but can also cause communication interruptions. This is particularly true in the unlicensed Industrial, Scientific, and Medical (ISM) bands, where numerous IoT devices communicate simultaneously on the same frequency band. Therefore, effectively identifying and managing OSS (Surveillance-Oriented Scrambling) (OSS) to improve the performance and reliability of wireless communication systems has become a pressing technical challenge in the wireless communication field.

[0003] The identification and management of aliased spectrum signals involves multiple aspects of signal processing and spectrum management. Traditional signal processing methods, such as independent component analysis (ICA), principal component analysis (PCA), and non-negative matrix factorization (NMF), have achieved some success in signal identification and separation. In research [1], the collaborative NMF decomposition of the channel state perception of multiple unlicensed users was used to achieve joint estimation and prediction of the spectrum occupancy status of licensed users.

[0004] However, they have inherent limitations. For example, PCA only retains the component with the largest variance and ignores other potentially valuable components. ICA and PCA lack the ability to accurately determine the number of mixed signals and maintain signal scaling. Although NMF performs well in signal category recognition and separation, its decomposition results may not be unique, which makes the interpretation and selection of results complicated. In addition, deep learning methods such as convolutional neural networks (CNN) and recurrent neural networks (RNN) have been studied [2] to achieve recognition based on power spectrum data of 6 types of signals by constructing a fully connected neural network classifier model. Although progress has been made in automatically learning signal feature representation, they require a large amount of labeled data and computing resources and remain a challenge when encountering new or random aliased spectrum signal combinations. In terms of multi-parameter OSS identification, signal location and transmission power estimation are crucial for radio monitoring. Existing methods such as time difference of arrival (TDOA), received signal strength (RSS) and angle of arrival (AOA) have been studied [3] to analyze the Cramer-Rao lower bound for location estimation using multiple RF sensors to monitor and separate multiple signal sources, and pointed out that its estimation accuracy improves with the increase in the number of monitoring nodes. Research [4] analyzed the sparsity problem in the joint estimation of the position and power of multiple target nodes. Research [5] proposed a network search method to estimate the position and power of the signal source. Research [6] summarized the received signal strength (RSS) of wireless transmission into a linear propagation model and analyzed the positioning model in four cases: the signal transmission energy is known and unknown, and the background noise is known and unknown.

[0005] Although there are certain applications in signal location estimation, they are usually independent of signal separation and category recognition tasks, which limits the comprehensive understanding of signal characteristics and behavior. Therefore, integrating these tasks into a unified analysis framework is of great significance to improve the accuracy and efficiency of signal recognition.

[0006] [1] Kim, SJ, and GGBGiannakis. "Cognitive radio spectrum prediction using dictionary learning." in Proceedings of the IEEE Global Communications Conference 2013, pp. 3206-3211.

[0007] [2] Tian Haomin, Yin Liang, Ma Yue, Li Shufang. Signal recognition method based on signal power spectrum fitting feature extraction. Journal of Beijing University of Posts and Telecommunications, 2018, v.41(02):118-122.

[0008] [3] E. Testi and A. Giorgetti, "RSS-based Localization of Multiple RadioTransmitters via Blind Source Separation," in IEEE Communications Letters.

[0009] [4] P. Qian, Y. Guo, N. Li, and Z. Xu, "Block variational Bayesian algorithm for multiple target localization with unknown and time-varying transmitpowers in WSNs," IEEE Access, vol.7, pp.54796-54808, 2019.

[0010] [5] P.Zuo, T.Peng, K.You, W.Guo and W.Wang, "RSS-Based Localization ofMultiple Directional Sources With Unknown Transmit Powers and Orientations," in IEEE Access, vol.7, pp.88756-88767, 2019.

[0011] [6] Y.Hu, J.Liu and B.Zhang, "Localization Using Blind RSSMeasurements," in IEEE Wireless Communications Letters, vol.8, no.2, pp.464-467, April 2019. Summary of the Invention

[0012] The purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art and to provide an integrated method for identifying parameters of deep aliasing spectrum signals.

[0013] The purpose of the present invention is achieved through the following technical solutions:

[0014] An integrated method for identifying parameters of deep aliasing spectrum signals comprises the following steps:

[0015] Step 1: Construct the overall objective function

[0016] (1) Assume that the aliased spectrum signal received by the mth monitoring node is R after short-time Fourier transform. m (f), for R m (f) Perform deep NMF decomposition and define R m (f)≈B(f)G m ≈B(f)D m P m K m ; Where B(f) is the spectrum feature matrix, G m is the category of the aliased spectrum signal, D m is the distance of the aliased spectrum signal, P m is the power of the aliased spectrum signal, K m is the radian of the aliased spectrum signal;

[0017] (2) Take B(f) as the basis matrix Z of the first layer signal decomposition 1,m , G m The coefficient matrix A of the first layer signal decomposition 1,m ;definition The consensus result matrix of the first-layer signal decomposition;

[0018] (3) Using particle swarm optimization algorithm to optimize D m Initialize; take D m As the basis matrix Z of the second-level decomposition 2,m , P m K m The coefficient matrix A of the second layer decomposition 2,m ;definition The consensus result matrix of the second-layer signal decomposition;

[0019] (4) To A 2,m Decompose it and get P m As the basis matrix Z of the third layer 3,m , the obtained K m As the coefficient matrix A of the third layer 3,m ;definition The consensus result matrix of the third-layer signal decomposition;

[0020] (5) Introduce the consensus result matrix of each layer to construct the overall objective function

[0021]

[0022] Z 1,m ≥0,Z 2,m ≥0,Z 3,m ≥0,

[0023]

[0024] A 2,m =Z 3,m A 3,m ,

[0025] A 1,m =Z 2,m A 2,m ;

[0026] Among them, the matrix It means that the nth type of signal is not detected at the mth monitoring point, and On the contrary;

[0027] (6) To Z l,m and A l,m Iterate

[0028] Step 2: Take the aliased spectrum signal as Input, Output A 1,m 、Z 2,m 、Z 3,m 、A 3,m G corresponding to the aliased spectrum signal m 、D m 、P m , K m ;D m , K m The location of the signal that constitutes the aliased spectrum.

[0029] In step (6), the alternating direction multiplication method is used to multiply Z l,m and A l,m The purpose of the iteration is to optimize the overall decomposition problem. The use of the alternating direction multiplier method can ensure that the solution process is effective and convergent.

[0030] The update rule of the alternating direction multiplier method is as follows:

[0031]

[0032] in Λ l-1,m =Z 1,m …Z l-1,m , when l=1, Λ 0,m is the identity matrix.

[0033] In step (6), Z is calculated based on KL divergence, β divergence or projected gradient. l,m and A l,m Iterate.

[0034] The channel gain η from the nth unknown signal source to the mth monitoring node n,m(t) is calculated as follows:

[0035]

[0036] 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 E t,n are the antenna gains of the mth monitoring node and the nth unknown signal respectively; |h n,m (t)| and d n,m are the channel response and distance between the mth monitoring node and the nth unknown signal respectively.

[0037] At the same time, the present invention provides:

[0038] 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 integrated deep aliasing spectrum signal parameter identification method.

[0039] A computer-readable storage medium stores at least one program, which is loaded and executed by a processor to implement the above-mentioned integrated deep aliasing spectrum signal parameter identification method.

[0040] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0041] 1. This invention effectively addresses the technical challenges of identifying and managing aliased spectrum signals. Its core lies in proposing a unified, multi-task collaborative analysis framework. By integrating tasks such as signal separation, category recognition, location estimation, and power analysis, it overcomes the problems of information fragmentation, computational redundancy, and poor generalization in existing technologies. This unified analysis framework utilizes Deep-NMF decomposition to design different layer factors, resulting in distinct feature representations for each layer, enabling the generation of corresponding signal characteristic parameters for each layer.

[0042] 2. Traditional NMF is a single-layer decomposition, and the decomposition results are not unique. This paper adopts a deep NMF (Deep-NMF) method to decompose signal parameters layer by layer to identify signal types, power estimation, and positioning. The participation of the three consensus result matrices in this paper is an important part of the integrated / unified analysis framework, among which the overall objective function is the most important. Using an overall objective function for solution does not require segmentation into different sub-problems.

[0043] 3. The multi-parameter monitoring node cooperation mechanism of the present invention can be dynamically adjusted according to the characteristic parameters of different aliased spectrum signals, thereby improving the adaptability of the system in various complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 Flowchart for constructing the overall objective function.

[0045] Figure 2 Schematic diagram of the spectrum aliasing signal monitoring model.

[0046] Figure 3 Flowchart of the integrated deep aliasing spectrum signal parameter identification method.

[0047] Figure 4 The following is a comparison chart of signal type recognition results under different algorithms.

[0048] Figure 5 The figure below is a comparison chart of signal power estimation results under different algorithms.

[0049] Figure 6 The following is a comparison chart of signal positioning results under different algorithms. DETAILED DESCRIPTION

[0050] The present invention will be described in further detail below with reference to the embodiments and drawings, but the embodiments of the present invention are not limited thereto.

[0051] like Figure 1 , an integrated deep aliasing spectrum signal parameter identification method, which constructs the overall objective function includes the following steps:

[0052] S1, using the deep NMF method to transform R m (f) decomposed into:

[0053] R m (f)≈B(f)G m

[0054] G m ≈D m C m

[0055] C m ≈P m K m

[0056] R m (f)≈B(f)D m P m K m

[0057] S2, take the spectrum feature matrix B(f) as the basis matrix Z of the first layer signal decomposition 1,m , G m is the decomposed coefficient matrix A 1,m . Define the matrix Indicates signal reception and introduces the consensus result matrix To achieve multi-parameter and multi-angle coupled consensus.

[0058] S3. Define the distance representation matrix D of the monitoring node m , using the particle swarm optimization (PSO) algorithm as the initialization algorithm for D m Initialize. Take D m As the basis matrix Z of the second-level decomposition 2,m , the coupling coefficient matrix C of signal power and position information m is the coefficient matrix A 2,m , introduce the consensus result matrix

[0059] S4, against A 2,m Decompose and get the power coefficient matrix P m As the basis matrix Z of the third layer 3,m , radian information matrix K m As the coefficient matrix A of the third layer 3,m , introduce the consensus result matrix The overall problem description is as follows:

[0060]

[0061] Z 1,m ≥0,Z 2,m ≥0,Z 3,m ≥0,

[0062]

[0063] A 2,m =Z 3,m A 3,m ,

[0064] A 1,m =Z 2,m A 2,m ;

[0065] S5, use the alternating direction multiplication method to Z l,m and A l,m Iterate and update the rules as follows:

[0066]

[0067] Specifically, an integrated deep aliasing spectrum signal parameter identification method is implemented as follows:

[0068] The time it takes for each monitoring node to complete the detection of a frequency band is T s , the frequency band spectrum data return time is T d , then the total time to complete a frequency band monitoring task is T0 = T s +T d. Assume that there are N unknown wireless signal sources in the same frequency band within this monitoring area, and the monitoring node has no prior information about their signal type, location, transmission power, etc. M monitoring nodes are evenly distributed within the communication range, and each monitoring node can transmit the spectrum data generated by spectrum scanning to the edge computing node by wired or wireless means, and the separation, positioning and power estimation of the aliased signals are completed in the edge computing node. The edge computing node receives the relevant data of each monitoring node. The monitoring node does not judge and process the aliased signal, but the edge computing node performs statistics and analysis on the data to make the final decision. The monitoring network model is as follows: Figure 2 shown.

[0069] The spectrum mixed signal received by the mth monitoring node in the tth spectrum sampling time slot is:

[0070]

[0071] where r m (t) is the aliased signal received by the mth monitoring node, s n (t) is the original signal of the nth transmitting source, p n (t) is the transmission power of the nth unknown signal source in time slot t, η n,m (t) is the channel gain from the nth unknown signal source to the mth monitoring node, z m (t) is the noise at the mth monitoring node and obeys Gaussian distribution.

[0072] This embodiment mainly considers the large-scale fading of wireless propagation in an outdoor rich scattering environment. Therefore, the channel gain coefficient is defined as:

[0073]

[0074] 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 E t,n are the antenna gains of the mth monitoring node and the nth unknown signal respectively. n,m (t)| and d n,m are the channel response and distance between the mth monitoring node and the nth unknown signal, respectively, and:

[0075]

[0076] Among them, (x n ,y n ) and (x m ,y m ) are the coordinates of the mth monitoring node and the nth unknown signal source in the D×D monitoring range in the coordinate system, xn ∈(0,D),y n ∈(0,D).

[0077] Therefore, let the polar coordinates of the nth unknown signal source be (θ n ,r n ),in In T s Through the short-time Fourier transform (STFT), a typical time-frequency transform analysis method, the time series is multiplied by a window function to convert the signal into the time-frequency domain. m (t) is R after short-time Fourier transform m (f), It is n The result of the short-time Fourier transform of (t), that is, the power spectrum density of the unknown signal source n, defines the standard spectrum dictionary for identifying signals and in is the spectrum characteristic matrix of the nth signal, where J is the number of sampling frequency points of the monitoring point in the frequency band, d n is the spectral feature representation dimension and d n =1, N0 is the number of signal templates, and S n (f)∈B n (f).

[0078] like Figure 3 ,Based on the deep NMF theory, this paper proposes an integrated deep ,signal recognition (CMSR) architecture.

[0079] Aiming at the problem of identifying the category, location and power estimation of multi-parameter aliased spectrum signals, a deep NMF method is used to extract multiple features from the aliased spectrum signals received by all monitoring nodes. m (f) decomposed into:

[0080] R m (f)≈B(f)G m ;

[0081] G m ≈D m C m ;

[0082] C m ≈P m K m ;

[0083] R m (f)≈B(f)D m P m K m ;

[0084] Use matrix B(f) as the basis matrix Z for signal decomposition 1,m Considering the aliased signal received by the mth monitoring point, let the coefficient matrix of the aliased signal be A 1,m , the condition is in Represents a non-negative value, A 1,m As the characterization matrix of the aliased spectrum signal type. 1,m It is multi-parameter and multi-angle coupled, and it is difficult to reach a consensus directly. Therefore, a new matrix was redesigned in, It means that the nth type of signal is not detected at the mth monitoring point, and It means that the n-th type of signal is detected at the m-th monitoring point. Defined as:

[0085]

[0086] Where ε is the characteristic index threshold. In the process of signal classification and recognition, signals with too low characteristic index do not exist, and vice versa.

[0087] In addition, let the matrix To represent the consistent results of signal category recognition. Define the diagonal matrix As the distance representation matrix of the mth monitoring node, Represents the arc angle of the nth URS estimated by the mth monitoring node. It is a A function used to extract the arc features from the aggregated position information.

[0088] Order Z 2,m =D m , becomes the basis matrix of the positioning problem, The coupling coefficient matrix representing signal power and position information.

[0089] In order to speed up the optimization of the model, it is very necessary and important to pre-train and initialize the parameters of each layer. Since the basis matrix represents the attributes of each layer, its physical meaning and the choice of dimension have a great influence on the optimization results. At the same time, a reasonable choice of the basis matrix can also reduce the convergence time of the model. The particle swarm optimization (PSO) algorithm is used as the initialization algorithm to initialize the basis matrix D. m For a multi-transmitter system, the objective function is:

[0090]

[0091] The x obtained by optimization n ,y nUsed to get the arc of the nth URS estimated by the mth monitoring node Thus completing the Initialization.

[0092] Continue to A 2,m Decompose and get the power coefficient matrix As the basis matrix of the third layer, the radian information matrix The coefficient matrix of the third layer. The overall problem can be described as:

[0093]

[0094] Z 1,m ≥0,Z 2,m ≥0,Z 3,m ≥0,

[0095]

[0096] A 2,m =Z 3,m A 3,m ,

[0097] A 1,m =Z 2,m A 2,m .

[0098] Use the alternating direction multiplication method to calculate Z l,m and A l,m Perform iterations and update rules as follows:

[0099]

[0100] in Λ l-1,m =Z 1,m …Z l-1,m , when l=1, Λ 0,m is the identity matrix.

[0101] After the overall objective function is constructed, the aliased spectrum signal is used as Input, Output A 1,m 、Z 2,m 、Z 3,m 、A 3,m G corresponding to the aliased spectrum signal m 、D m 、P m , K m ;D m , K m The location of the signal that constitutes the aliased spectrum.

[0102] At the same time, this embodiment provides:

[0103] 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 integrated deep aliasing spectrum signal parameter identification method.

[0104] A computer-readable storage medium stores at least one program, which is loaded and executed by a processor to implement the above-mentioned integrated deep aliasing spectrum signal parameter identification method.

[0105] Experimental results show that the performance of this patent in signal recognition is higher than that of traditional blind source separation algorithms:

[0106] like Figure 4 ,In the task of signal type recognition, the collaborative multi-layer ,signal recognition method proposed in this paper is superior to the ,traditional blind source signal separation methods of PCA, FsatICA, and NMF.

[0107] like Figure 5 ,In the signal power estimation task, the collaborative multi-layer signal recognition ,method proposed in this paper has a significant improvement in performance after ,initialization using the particle swarm algorithm.

[0108] like Figure 6 In the signal positioning task, the collaborative multi-layer signal recognition method proposed in the present invention has a greatly improved performance after being initialized with the particle swarm algorithm. The traditional particle swarm algorithm is superior to the present invention in the positioning task because it requires prior information on power.

[0109] 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 considered as equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. An integrated method for identifying parameters of deep aliasing spectrum signals, characterized in that: The following steps are involved: Step 1: Construct the overall objective function (1) Assume that the aliased spectrum signal received by the mth monitoring node is R after short-time Fourier transform. m (f), for R m (f) Perform deep NMF decomposition and define R m (f)≈B(f)G m ≈B(f)D m P m K m ; Where B(f) is the spectrum feature matrix, G m is the category of the aliased spectrum signal, D m is the distance of the aliased spectrum signal, P m is the power of the aliased spectrum signal, K m is the radian of the aliased spectrum signal; (2) Take B(f) as the basis matrix Z of the first layer signal decomposition 1,m , G m The coefficient matrix A of the first layer signal decomposition 1,m ;definition The consensus result matrix of the first-layer signal decomposition; (3) Using particle swarm optimization algorithm to optimize D m Initialize; take D m As the basis matrix Z of the second-level decomposition 2,m , P m K m The coefficient matrix A of the second layer decomposition 2,m ;definition The consensus result matrix of the second-layer signal decomposition; (4) To A 2,m Decompose it and get P m As the basis matrix Z of the third layer 3,m , the obtained K m As the coefficient matrix A of the third layer 3,m ;definition The consensus result matrix of the third-layer signal decomposition; (5) Introduce the consensus result matrix of each layer to construct the overall objective function Among them, the matrix It means that the nth type of signal is not detected at the mth monitoring point, and On the contrary; (6) To Z l,m and A l,m Iterate Step 2: Use the aliased spectrum signal as Input, Output A 1,m 、Z 2,m 、Z 3,m 、A 3,m G corresponding to the aliased spectrum signal m 、D m 、P m , K m ;D m , K m The location of the signal that constitutes the aliased spectrum.

2. The integrated deep aliasing spectrum signal parameter identification method according to claim 1, characterized in that: In step (6), the alternating direction multiplication method is used to multiply Z l,m and A l,m Iterate.

3. The integrated deep aliasing spectrum signal parameter identification method according to claim 2, characterized in that: The update rule of the alternating direction multiplier method is as follows: in Λ l-1,m =Z 1,m …Z l-1,m , when l=1, Λ 0,m is the identity matrix.

4. The integrated deep aliasing spectrum signal parameter identification method according to claim 1, characterized in that: In step (6), Z is calculated based on KL divergence, β divergence or projected gradient. l,m and A l,m Iterate.

5. The integrated deep aliasing spectrum signal parameter identification method according to claim 1, characterized in that: The channel gain η from the nth unknown signal source to the mth monitoring node n,m (t) is calculated as follows: 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 E t,n are the antenna gains of the mth monitoring node and the nth unknown signal respectively; |h n,m (t)| and d n,m are the channel response and distance between the mth monitoring node and the nth unknown signal respectively.

6. A server comprising a processor and a memory, wherein the memory stores at least one program, characterized in that: The program is loaded and executed by the processor to implement the integrated deep aliasing spectrum signal parameter identification method according to any one of claims 1 to 5.

7. 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 integrated deep aliasing spectrum signal parameter identification method according to any one of claims 1 to 5.

Images