6G communication and inductance integration method for multi-dimensional radio frequency characteristic intelligent analysis

Abstract

The invention relates to the technical field of mobile telecommunication services such as new-generation mobile communication base station equipment, new-generation mobile communication core network and access network construction, networking and the like, in particular to a 6G communication and sensing integration method for multi-dimensional radio frequency characteristic intelligent analysis, which comprises the following steps of: S1, establishing a ground multi-antenna base station, and collecting radio frequency characteristics of a target; s2, establishing a self-adaptive path loss model, and realizing accurate estimation of a target distance by analyzing received signal strength data; s3, estimating an angle of arrival of the target signal in combination with an improved MUSIC-ESPRIT fusion algorithm based on the radio frequency characteristics; s4, based on the frequency change information of the target signal, estimating the distance position and the motion state of the target relative to the ground multi-antenna base station according to a Doppler frequency shift theory in combination with a hybrid filtering architecture driven by environmental perception; and S5, performing information fusion of different levels on the distance position and the motion state obtained by signal feature analysis to obtain a final target tracking result.

Description

Technical Field

[0001] The present invention relates to the fields of manufacturing of communication system equipment such as a new generation of mobile communication base station equipment, digital program-controlled switches, construction of a new generation of mobile communication core networks and access networks, networking, and other mobile telecommunication services, as well as other telecommunication service technologies such as mobile data communication services. Specifically, it relates to a 6G synesthesia integration method for intelligent analysis of multi-dimensional radio frequency characteristics. Background Art

[0002] In recent years, radio frequency sensing technology has rapidly developed in the fields of target tracking and interawareness integration, demonstrating unique advantages in non-line-of-sight target detection in complex electromagnetic environments. However, existing target tracking technologies based on radio frequency signals still face significant technical bottlenecks in the coordinated application of drone swarms.

[0003] First, traditional RF positioning methods rely heavily on single characteristic parameters, such as signal strength RSSI (Reference Signal Strength Index) or time difference of arrival (TDOA). These methods are susceptible to signal attenuation, multipath interference, and noise superposition in dynamic multipath propagation environments, leading to cumulative errors in target position calculations. Especially in densely built-up or vegetated areas, the time-varying nature of signal reflection paths can severely destabilize RF characteristics, causing tracking to break or shift.

[0004] Secondly, existing RF tracking systems lack adaptability to target motion characteristics. When a target changes speed, direction, or briefly stops, conventional Doppler shift analysis methods struggle to accurately interpret this change in motion, leading to false alarms and tracking loss. Furthermore, the intermingling of different RF signatures in multi-target scenarios can lead to target identity confusion. The lack of effective feature separation and association mechanisms severely impacts the reliability of multi-target parallel tracking. Summary of the Invention

[0005] The present invention provides a 6G synaesthesia integrated method for intelligent analysis of multi-dimensional radio frequency features that can efficiently and accurately complete target tracking tasks, which can solve at least one of the above technical problems.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0007] The 6G synaesthesia integrated method for multi-dimensional radio frequency feature intelligent analysis includes the following steps:

[0008] S1. Establish a ground multi-antenna base station to collect the target's radio frequency characteristics;

[0009] S2. Establish an adaptive path loss model that can adapt to different communication environments and scenarios and accurately estimate the target distance by analyzing the received signal strength data;

[0010] S3, based on the radio frequency characteristics of S1, combined with the improved MUSIC-ESPRIT fusion algorithm, estimate the arrival angle of the target signal;

[0011] S4. Based on the frequency change information of the target signal, according to the Doppler frequency shift theory and combined with the hybrid filtering architecture driven by environmental perception, the target position and motion state relative to the ground multi-antenna base station are estimated;

[0012] S5. Perform information fusion at different levels on the target position and motion state obtained by analyzing the radio frequency features in S4 to obtain the final target tracking result.

[0013] Furthermore, in S2, the received signal expression is:

[0014] r(t)=∑ k A k s(t-τ k )+n(t)

[0015] Among them, r(t) is the received signal, A k is the amplitude attenuation factor of the kth signal, indicating the attenuation of the signal during propagation, s(t) is the transmitted signal, τ k is the time delay of the kth signal, which indicates the time delay from signal transmission to reception, and n(t) is the noise signal;

[0016] Taking into account multi-source environmental factors, the explicit environmental factor superposition method is adopted to convert different environmental factors into independent influence coefficients. An adaptive enhanced path loss formula is constructed through linear combination, which decomposes the environmental factors into quantifiable correction terms. The distance term and frequency term in the free space loss formula are directly corrected. The corrected formula is expressed as follows:

[0017] L FS =32.15+20log 10 (f eff )+20log 10 (d eff )

[0018] f eff =f×K weather

[0019] d eff =d LJ ×Γ terrain ×Γ humidity

[0020] Among them, L FS is the path loss, f eff is the equivalent frequency, d eff is the equivalent distance, K weather is the comprehensive meteorological correction factor, Γterrain is the geographical environment correction factor, Γ humidity is the humidity-temperature coupling correction factor.

[0021] Furthermore, in S2, the calculation methods of the multiple correction factors are respectively:

[0022] Comprehensive meteorological correction factor K weather :

[0023]

[0024] A rain =0.18R d 1.2

[0025] A fog =0.43C n

[0026] A dust =0.015D d 0.7

[0027] Among them, A rain is the rainfall correction, R d is the rainfall intensity in millimeters per hour, A fog is the fog attenuation correction, C n is visibility, the unit is kilometers, A dust Corrected for dust, D d is the dust concentration in milligrams per cubic meter;

[0028] Geographical environment correction factor Γ terrain :

[0029]

[0030] Among them, α and β are altitude compensation parameters, h msl is the mean altitude, γ is the latitude curvature compensation, and lat is the geographic latitude;

[0031] Humidity-temperature coupling correction factor Γ humidity :

[0032]

[0033] Where μ is the relative humidity and T is the temperature.

[0034] Furthermore, in S3, in a scenario with multiple signal sources and a large signal-to-noise ratio, the MUSIC multi-target positioning and signal parameter estimation algorithm is used to estimate the target signal arrival angle. The process is as follows:

[0035] S3.a.1. Assume that the received data matrix is ​​Xr , the linear array consists of M array elements and receives K far-field narrowband signals. The expression of the data matrix is:

[0036] X r (t)=A(θ)S(t)+N(t)

[0037] A(θ)=[a(θ1),a(θ2),…,a(θ k )]

[0038]

[0039] Where A(θ)∈C M×K is the array flow matrix, a(θ k ) is the steering vector of the kth signal, d is the array element spacing, λ is the signal wavelength, S(t)∈C K×1 is the complex envelope of K incident signals, N(t)∈C M×1 is additive noise;

[0040] S3.a.2. The covariance matrix R of the received data is expressed as:

[0041]

[0042] Perform eigenvalue decomposition on the covariance matrix R:

[0043]

[0044] Among them, E is the eigenvector matrix, D is the diagonal eigenvalue matrix, and the eigenvector is divided into signal subspace and noise subspace according to the size of the eigenvalue. The signal subspace is represented by U s Represented by U, the noise subspace is represented by n express;

[0045] For each possible signal arrival angle θ, calculate the steering vector a(θ) and calculate the projection of the steering vector on the noise subspace:

[0046]

[0047] Where P(θ) is the MUSIC spatial spectrum of the signal, and the peak of P(θ) is searched within the range of θ, where θ corresponding to the peak is the arrival angle of the signal.

[0048] Furthermore, in S3, in scenarios with low signal-to-noise ratio, real-time processing, or limited computing resources, the ESPRIT algorithm is used to estimate signal parameters using rotational invariance technology. The process is as follows:

[0049] S3.b.1. Assume a uniform linear array with M elements, spaced d apart, receiving K far-field narrowband signals with wavelength λ. The array receives the data matrix X.r ∈C M×1 The expression is:

[0050] X r (t)=A(θ)S(t)+N(t)

[0051] Where A(θ)∈C M×K is the array flow matrix, a(θ k ) is the steering vector of the kth signal, d is the array element spacing, λ is the signal wavelength, S(t)∈C K×1 is the complex envelope of K incident signals, N(t)∈C M×1 is additive noise;

[0052] S3.b.2. The covariance matrix R of the received data is expressed as:

[0053]

[0054] Perform eigenvalue decomposition on the covariance matrix:

[0055] R=UΣU H

[0056] Where Σ=diag(σ1,…,σ M ) is a diagonal matrix with eigenvalues ​​arranged in descending order, and U is the corresponding eigenvector matrix;

[0057] S3.b.3. Select the eigenvectors corresponding to the first K largest eigenvalues ​​to form the signal subspace U s :

[0058] U S =[u1,…u k ]∈C M×K

[0059] The signal subspace U s Split into two sub-matrices, corresponding to the two translation sub-arrays:

[0060] U1=U s (1:M-1,:)∈C (M-1)×K

[0061] U2=U s (2:M,:)∈C (M-1)×K

[0062] Construct the augmented matrix and perform singular value decomposition:

[0063] C 增广 =[U1,U2]∈C (M-1)×2K

[0064]

[0065] V c Divide into 4 K×K sub-blocks and calculate the rotation matrix:

[0066]

[0067] Among them, V c is a right singular matrix, Ψ is a rotation matrix;

[0068] Perform eigendecomposition on Ψ:

[0069] Ψ=TΦT -1

[0070] Where Φ=diag(φ1,…,φ K ) is a diagonal matrix with eigenvalues ​​arranged in descending order, T is the eigenvector matrix, from which the arrival angle is calculated by the phase of the eigenvalue:

[0071]

[0072] Furthermore, in S3, the advantages of the MUSIC and ESPRIT algorithms are combined to construct a three-level progressive architecture of "coarse estimation + fine search + joint optimization". This reduces computational complexity while improving estimation accuracy to adapt to more complex and changing external environments. The process is as follows:

[0073] S3.c.1, coarse estimation, divide the linear array into two translation sub-arrays, extract the signal subspace U s And solve for rotation invariance:

[0074] X1=A1S,X2=A2S=A1ΦS

[0075]

[0076] Among them, X1 and X2 are two sub-arrays, A1 and A2 are the flow matrix of the array, S is the complex amplitude of the signal, R is the covariance matrix, U s and U n The signal subspace and noise subspace are respectively extracted, and the signal subspace U s and divided into and Corresponding to subarray 1 and subarray 2 respectively, solve the rotation matrix, calculate the eigenvalue, and get a rough estimate of the angle

[0077]

[0078] Where Ψ is the rotation matrix, ψ k is the eigenvalue of Ψ;

[0079] S3.c.2. Based on the rough estimation result, a local fine search is performed near the rough estimation, introducing the focusing factor β and the rough estimation result. Construct a weighted spatial spectrum:

[0080]

[0081] Among them, the numerator represents the signal subspace projection, the denominator is the traditional MUSIC noise subspace orthogonality measure, and a(θ) is the signal steering vector. Therefore, only in Searching for spectral peaks within a certain range can reduce the amount of calculation. Δ is the constraint range, which is determined by the array beam width.

[0082] Dynamically adjust the focusing factors β and σ according to the real-time estimated confidence and environmental parameters rough,k And the constraint range Δ, to achieve intelligent local search, the process is:

[0083] Calculate the standard deviation σ of the rough estimate of the ESPRIT algorithm rough,k , define the confidence index C of the rough estimate k , the calculation expression is:

[0084]

[0085] Dynamically adjust the focus factor based on the confidence index:

[0086]

[0087] Among them, β k is the dynamic focusing factor, β0 is the basic focusing strength;

[0088] Adaptively adjust the constraint range based on the confidence index:

[0089]

[0090] Among them, Δ b ase is the baseline constraint range;

[0091] When the confidence level C k When it is low, the estimation is considered unreliable, and the search range is automatically expanded. Based on the dynamic focusing factor and adaptive constraint range obtained from the above formula, a more accurate local refined search result is calculated.

[0092] S3.c.3. Joint optimization based on coarse estimation and local fine search is performed. The closed-form solution of the ESPRIT algorithm and the spectral information of the MUSIC algorithm are integrated into the optimization framework to further improve the accuracy. The objective function is designed as:

[0093]

[0094] The Gauss-Newton iterative method is used to update the angle estimation until convergence, and the final joint optimization angle result is obtained.

[0095] Furthermore, the S4 further includes:

[0096] S4.1. Use a single antenna or array to synthesize the beam to receive the target signal and record the baseband signal s b (t):

[0097]

[0098] Where A(t) is the complex envelope of the signal, n(t) is the noise, φ dop (t) includes phase changes caused by target motion;

[0099] The target signal is extracted by matched filtering or low-pass filtering, noise is suppressed, and the results are obtained under high signal-to-noise ratio and low signal-to-noise ratio conditions respectively.

[0100] Under high signal-to-noise ratio, the instantaneous frequency f is directly extracted through the signal phase d (t):

[0101]

[0102] Under low signal-to-noise ratio, short-time Fourier transform is used to calculate the time-frequency matrix:

[0103]

[0104] Where S(t,f) is the time-frequency matrix, w(t) is the Hamming window function;

[0105] S4.2. Based on Doppler shift theory, the radial velocity and range of the target are modeled as follows:

[0106]

[0107]

[0108] Among them, v r (t) is the radial velocity, d dop (t) is the distance between the target and the receiving antenna, f d (t) is the Doppler shift, λ is the signal wavelength;

[0109] If the initial distance d0 is unknown, select the frequency observation values ​​f at two moments d1 and f d2 Simultaneous equations:

[0110]

[0111] Among them, d1 and d2 are the distances between the target and the receiving antenna during the two observations, Δt is the observation interval, and after eliminating d0, we can get v r1 、v r2 , d1 and d2;

[0112] If the initial distance d0 is known, the real-time distance can be directly obtained by integrating the Doppler shift:

[0113]

[0114] S4.3. To achieve real-time dynamic tracking of the target's motion state, an environment-aware driven hybrid filtering architecture is introduced. By monitoring channel characteristics in real time, the optimal filtering strategy is dynamically selected to achieve matching between the model and the environment. The specific process is as follows:

[0115] S4.3.1. Extract environmental features and calculate the instantaneous signal-to-noise ratio SNR(t):

[0116]

[0117] Among them, s b (t) is the baseband signal, n(t) is the noise;

[0118] S4.3.2. Determine the multipath strength by analyzing the signal cyclostationary characteristics or delay correlation, and design the following hybrid filtering strategy:

[0119] In low-noise scenarios, the standard Kalman filter (KF) is enabled;

[0120] In high-noise / multipath scenarios, the robust particle filter (PF) is used to combat non-Gaussian noise through Monte Carlo sampling.

[0121] In sudden interference scenarios, the sliding window least squares (LS) method is triggered to avoid filter divergence.

[0122] The following adaptive switching logic is obtained:

[0123]

[0124] Furthermore, the S5 further includes:

[0125] S5.1. In order to perform multi-level information fusion of the target position and motion state obtained by RF feature analysis, a fusion framework based on extended Kalman filtering is adopted, and the target state vector is defined as position and velocity:

[0126] X k =[x k ,y k ,v x,k ,v y,k ] T

[0127] Among them, X k is the state vector, x k and y k are the positions in the x and y directions respectively, v x,k and v y,k are the velocities in the x and y directions respectively;

[0128] The process model is established as follows:

[0129] X k|k-1 =FX k-1 +w k

[0130]

[0131] Among them, F is the state transfer matrix, w k is the process noise, w k ~N(0,Q),σ x , σ y 、 and is the noise of each state component, Q is the process noise covariance matrix;

[0132] S5.2, based on the distance d1,k obtained in S2 and the arrival angle θ obtained in S3 k , the observation model is established as follows:

[0133]

[0134] Among them, Z 1，k is the observation source, H1 is the Jacobian matrix, and R1 is the observation noise covariance;

[0135] S5.3, based on the radial velocity v obtained in S4 r,k and distance d 2,k , the observation model is established as follows:

[0136]

[0137] Among them, Z 2，k is the observation source, H2 is the Jacobian matrix, and R2 is the observation noise covariance;

[0138] S5.3. Perform extended Kalman filtering based on the observation model in S5.2 and S5.3. The prediction steps are as follows:

[0139] X k|k-1 =FX k-1

[0140] P k|k-1 =FP k-1 F T +Q

[0141] Among them, X k|k-1 is the state prediction value, P k|k-1 is the covariance prediction value, F is the state transfer matrix, and Q is the process noise covariance matrix;

[0142] S5.4. For each observation source, calculate the Kalman gain and update the state and covariance:

[0143]

[0144] X k =X k|k-1 +K i (z i -h i (X k|k-1 ))

[0145] P k =(IK i H i )P k|k-1

[0146] Among them, K i is the Kalman gain, X k is the updated state, P k is the updated covariance, I is the identity matrix;

[0147] The final output state estimation vector is:

[0148] X k =[x k ,y k ,v x,k ,v y,k ] T ;

[0149] S5.5. After fusing multi-source RF features through extended Kalman filtering, the target state estimation contains the following core information:

[0150] Position: The Cartesian coordinates (x, y) of the target in the two-dimensional plane;

[0151] Velocity: The velocity components of the target in the x and y directions (v x , v y );

[0152] Direction: the target's direction of movement φ;

[0153] According to the state estimation vector finally output in S5.4, the target position is the first two components of the state estimation vector, the target velocity vector is the last two components of the state estimation vector, and the target motion direction is the direction of the velocity vector, that is, the azimuth of the velocity vector. The expressions are:

[0154]

[0155] The beneficial effects of the present invention are embodied in:

[0156] 1. This invention targets complex environments, comprehensively considers influencing factors such as weather, temperature and humidity, and geographical location, and constructs an adaptive path loss model. It can accurately estimate the target distance through received power in different environments, ultimately achieving more accurate tracking of targets in complex scenarios.

[0157] 2. The present invention adopts an improved MUSIC-ESPRIT fusion algorithm, which can accurately calculate the arrival angle of the target signal under different signal-to-noise ratio environments and limited computing resources, and ultimately achieve more efficient target tracking in complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0158] The drawings described herein are used to provide further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute improper limitations on the present application.

[0159] Figure 1 It is a schematic diagram of the overall flow of the method according to an embodiment of the present invention.

[0160] Figure 2 It is a schematic diagram of a specific flow chart of the method according to an embodiment of the present invention.

[0161] Figure 3 It is a schematic diagram of fitting results for different incoming wave directions according to an embodiment of the present invention.

[0162] Figure 4 It is a schematic diagram of the simulation of the motion trajectory tracking effect of the method in an embodiment of the present invention.

[0163] Figure 5 It is a structural block diagram of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0164] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. In the absence of conflict, the embodiments in this application and the features in the embodiments can be combined with each other. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0165] It should be noted that the meaning of "and / or" appearing throughout the text includes three parallel solutions. Taking "A and / or B" as an example, it includes solution A, solution B, or solutions in which both A and B are satisfied. In addition, "multiple" refers to more than two. In addition, the technical solutions between the various embodiments can be combined with each other, but this must be based on the fact that ordinary technicians in this field can implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0166] See also Figure 1 , an embodiment of the present invention provides a 6G synaesthesia integrated method for multi-dimensional radio frequency feature intelligent analysis, comprising the following steps:

[0167] S1. Establish a ground base station with multiple antennas. Since the target being tracked radiates radio frequency signals when communicating externally, the base station receives the target signal and analyzes its radio frequency characteristics.

[0168] S2. Establish an adaptive path loss model that can adapt to different communication environments and scenarios and accurately estimate the target distance by analyzing the received signal strength data;

[0169] S3, based on the radio frequency characteristics of S1, combined with the improved MUSIC-ESPRIT fusion algorithm, estimate the arrival angle of the target signal;

[0170] S4. Based on the frequency change information of the target signal, according to the Doppler frequency shift theory and combined with the hybrid filtering architecture driven by environmental perception, the target position and motion state relative to the ground multi-antenna base station are estimated;

[0171] S5. Perform information fusion at different levels on the target position and motion state obtained by analyzing the radio frequency features in S4 to obtain the final target tracking result.

[0172] See also Figure 2 In this embodiment, in S2, the received signal expression is:

[0173] r(t)=∑ k A k s(t-τ k )+n(t)

[0174] Among them, r(t) is the received signal, A k is the amplitude attenuation factor of the kth signal, which indicates the attenuation of the signal during propagation. s(t) is the transmitted signal, for example, it can be a sinusoidal signal s(t) = sin(2πf c t), where f c is the carrier frequency, τ kis the time delay of the kth signal, which indicates the time delay from the transmission to the reception of the signal, and n(t) is the noise signal, which is usually assumed to be Gaussian noise;

[0175] Taking into account multiple environmental factors, such as weather, geographic location, temperature and humidity, an explicit environmental factor superposition method is used to convert different environmental factors into independent influencing coefficients. An adaptive enhanced path loss formula is constructed through linear combination, which decomposes environmental factors into quantifiable correction terms. The distance and frequency terms in the free space loss formula are directly corrected. The corrected formula is expressed as follows:

[0176] L FS =32.15+20log 10 (f eff )+20log 10 (d eff )

[0177] f eff =f×K weather

[0178] d eff =d LJ ×Γ terrain ×Γ humidity

[0179] Among them, L FS is the path loss, f eff is the equivalent frequency, d eff is the equivalent distance, K weather is the comprehensive meteorological correction factor, Γ terrain is the geographical environment correction factor, Γ humidity is the humidity-temperature coupling correction factor.

[0180] See also Figure 2 In this embodiment, in S2, the calculation methods of the multiple correction factors are:

[0181] Comprehensive meteorological correction factor K weather :

[0182]

[0183] A rain =0.18R d 1.2

[0184] A fog =0.43C n

[0185] A dust =0.015D d 0.7

[0186] Among them, A rain is the rainfall correction, R d is the rainfall intensity in millimeters per hour, A fog is the fog attenuation correction, C n is visibility, the unit is kilometers, A dust Corrected for dust, D d is the dust concentration in milligrams per cubic meter;

[0187] Geographical environment correction factor Γ terrain :

[0188]

[0189] Among them, α and β are altitude compensation parameters, h msl is the mean altitude, γ is the latitude curvature compensation, and lat is the geographic latitude;

[0190] Humidity-temperature coupling correction factor Γ humidity :

[0191]

[0192] Where μ is the relative humidity and T is the temperature.

[0193] By adding the above correction factors, the improved adaptive path loss model can have good environmental adaptability under different weather conditions and geographical locations, thereby more accurately calculating the target distance.

[0194] See also Figure 2 In this embodiment, in S3, in a scenario with multiple signal sources and a large signal-to-noise ratio, the MUSIC multi-target positioning and signal parameter estimation algorithm is used to estimate the target signal arrival angle. The process is as follows:

[0195] S3.a.1. Assume that the received data matrix is ​​X r , the linear array consists of M array elements and receives K far-field narrowband signals. The expression of the data matrix is:

[0196] X r (t)=A(θ)S(t)+N(t)

[0197] A(θ)=[a(θ1),a(θ2),…,a(θ k )]

[0198]

[0199] Where A(θ)∈C M×K is the array flow matrix, a(θ k) is the steering vector of the kth signal, d is the array element spacing, λ is the signal wavelength, S(t)∈C K×1 is the complex envelope of K incident signals, N(t)∈C M×1 is additive noise;

[0200] S3.a.2. The covariance matrix R of the received data is expressed as:

[0201]

[0202] Perform eigenvalue decomposition on the covariance matrix R:

[0203]

[0204] Among them, E is the eigenvector matrix, D is the diagonal eigenvalue matrix, and the eigenvector is divided into signal subspace and noise subspace according to the size of the eigenvalue. The signal subspace is represented by U s Represented by U, the noise subspace is represented by n express;

[0205] For each possible signal arrival angle θ, calculate the steering vector a(θ) and calculate the projection of the steering vector on the noise subspace:

[0206]

[0207] Where P(θ) is the MUSIC spatial spectrum of the signal, and the peak of P(θ) is searched within the range of θ, where θ corresponding to the peak is the arrival angle of the signal.

[0208] See also Figure 2 In this embodiment, in S3, in scenarios with low signal-to-noise ratio, real-time processing, or limited computing resources, the ESPRIT algorithm is used to estimate signal parameters using rotational invariance technology. The process is as follows:

[0209] S3.b.1. Assume a uniform linear array with M elements, spaced d apart, receiving K far-field narrowband signals with wavelength λ. The array receives the data matrix X. r ∈C M×1 The expression is:

[0210] X r (t)=A(θ)S(t)+N(t)

[0211] Where A(θ)∈C M×K is the array flow matrix, a(θ k ) is the steering vector of the kth signal, d is the array element spacing, λ is the signal wavelength, S(t)∈C K×1 is the complex envelope of K incident signals, N(t)∈C M×1 is additive noise;

[0212] S3.b.2. The covariance matrix R of the received data is expressed as:

[0213]

[0214] Perform eigenvalue decomposition on the covariance matrix:

[0215] R=UΣU H

[0216] Where Σ=diag(σ1,…,σ M ) is a diagonal matrix with eigenvalues ​​arranged in descending order, and U is the corresponding eigenvector matrix;

[0217] S3.b.3. Select the eigenvectors corresponding to the first K largest eigenvalues ​​to form the signal subspace U s :

[0218] U S =[u1,…u k ]∈C M×K

[0219] The signal subspace U s Split into two sub-matrices, corresponding to the two translation sub-arrays:

[0220] U1=U s (1:M-1,:)∈C (M-1)×K

[0221] U2=U s (2:M,:)∈C (M-1)×K

[0222] Construct the augmented matrix and perform singular value decomposition:

[0223] C 增广 =[U1,U2]∈C (M-1)×2K

[0224]

[0225] V c Divide into 4 K×K sub-blocks and calculate the rotation matrix:

[0226]

[0227] Among them, V c is a right singular matrix, Ψ is a rotation matrix;

[0228] Perform eigendecomposition on Ψ:

[0229] Ψ=TΦT -1

[0230] Where Φ=diag(φ1,…,φK ) is a diagonal matrix with eigenvalues ​​arranged in descending order, T is the eigenvector matrix, from which the arrival angle is calculated by the phase of the eigenvalue:

[0231]

[0232] See also Figure 2 In this embodiment, due to the complexity of actual application scenarios, in S3, the advantages of the MUSIC and ESPRIT algorithms are combined to construct a three-level progressive architecture of "coarse estimation + fine search + joint optimization". This architecture reduces computational complexity while improving estimation accuracy to adapt to more complex and changing external environments. The process is as follows:

[0233] S3.c.1, coarse estimation, divide the linear array into two translation sub-arrays, extract the signal subspace U s And solve for rotation invariance:

[0234] X1=A1S,X2=A2S=A1ΦS

[0235]

[0236] Among them, X1 and X2 are two sub-arrays, A1 and A2 are the flow matrix of the array, S is the complex amplitude of the signal, R is the covariance matrix, U s and U n The signal subspace and noise subspace are respectively extracted, and the signal subspace U s and divided into and Corresponding to subarray 1 and subarray 2 respectively, solve the rotation matrix, calculate the eigenvalue, and get a rough estimate of the angle

[0237]

[0238] Where Ψ is the rotation matrix, ψ k is the eigenvalue of Ψ;

[0239] S3.c.2. Based on the rough estimation result, a local fine search is performed near the rough estimation, introducing the focusing factor β and the rough estimation result. Construct a weighted spatial spectrum:

[0240]

[0241] Among them, the numerator represents the signal subspace projection, the denominator is the traditional MUSIC noise subspace orthogonality measure, and a(θ) is the signal steering vector. Therefore, only in Searching for spectral peaks within a certain range can greatly reduce the amount of calculation. Δ is the constraint range, which is determined by the array beam width.

[0242] Dynamically adjust the focusing factors β and σ based on the real-time estimated confidence and environmental parameters (such as SNR and signal coherence) rough,k And the constraint range Δ, to achieve intelligent local search, the process is:

[0243] Calculate the standard deviation σ of the rough estimate of the ESPRIT algorithm rough,k , define the confidence index C of the rough estimate k , the calculation expression is:

[0244]

[0245] Dynamically adjust the focus factor based on the confidence index:

[0246]

[0247] Among them, β k is the dynamic focusing factor, β0 is the basic focusing strength (e.g. β0 = 10);

[0248] β0 is adjusted according to SNR:

[0249]

[0250] Adaptively adjust the constraint range based on the confidence index:

[0251]

[0252] Among them, Δ b ase is the baseline constraint range (e.g., 15°);

[0253] When the confidence level C k When it is low, the estimation is considered unreliable, and the search range is automatically expanded. Based on the dynamic focusing factor and adaptive constraint range obtained from the above formula, a more accurate local refined search result is calculated.

[0254] S3.c.3. Joint optimization based on coarse estimation and local fine search is performed. The closed-form solution of the ESPRIT algorithm and the spectral information of the MUSIC algorithm are integrated into the optimization framework to further improve the accuracy. The objective function is designed as:

[0255]

[0256] The Gauss-Newton iterative method is used to update the angle estimation until convergence, and the final joint optimization angle result is obtained.

[0257] In this embodiment, the S4 further includes:

[0258] S4.1. Use a single antenna or array to synthesize the beam to receive the target signal and record the baseband signal sb (t):

[0259]

[0260] Where A(t) is the complex envelope of the signal, n(t) is the noise, φ dop (t) includes phase changes caused by target motion;

[0261] The target signal is extracted by matched filtering or low-pass filtering, noise is suppressed, and the results are obtained under high signal-to-noise ratio and low signal-to-noise ratio conditions respectively.

[0262] Under high signal-to-noise ratio, the instantaneous frequency f is directly extracted through the signal phase d (t):

[0263]

[0264] Under low signal-to-noise ratio, short-time Fourier transform is used to calculate the time-frequency matrix:

[0265]

[0266] Where S(t,f) is the time-frequency matrix, w(t) is the Hamming window function;

[0267] S4.2. Based on Doppler shift theory, the radial velocity and range of the target are modeled as follows:

[0268]

[0269] Among them, v r (t) is the radial velocity, d dop (t) is the distance between the target and the receiving antenna, f d (t) is the Doppler shift, λ is the signal wavelength;

[0270] If the initial distance d0 is unknown, select the frequency observation values ​​f at two moments d1 and f d2 Simultaneous equations:

[0271]

[0272] Among them, d1 and d2 are the distances between the target and the receiving antenna during the two observations, Δt is the observation interval, and after eliminating d0, we can get v r1 、v r2 , d1 and d2;

[0273] If the initial distance d0 is known, the real-time distance can be directly obtained by integrating the Doppler shift:

[0274]

[0275] When the system is started and the target is lost and then recaptured, since the initial distance d0 is unknown, the above-mentioned simultaneous equations are used to solve the initial state of the target or fit the motion parameters through multi-time observation data. After calibrating the initial value, the target distance can be calculated by integrating the Doppler shift.

[0276] S4.3. To achieve real-time dynamic tracking of the target's motion state, an environment-aware driven hybrid filtering architecture is introduced. By monitoring channel characteristics in real time, the optimal filtering strategy is dynamically selected to achieve matching between the model and the environment. The specific process is as follows:

[0277] S4.3.1. Extract environmental features and calculate the instantaneous signal-to-noise ratio SNR(t):

[0278]

[0279] Among them, s b (t) is the baseband signal, n(t) is the noise;

[0280] S4.3.2. Determine the multipath strength by analyzing the signal cyclostationary characteristics or delay correlation, and design the following hybrid filtering strategy:

[0281] In low-noise scenarios, the standard Kalman filter (KF) is enabled;

[0282] In high-noise / multipath scenarios, the robust particle filter (PF) is used to combat non-Gaussian noise through Monte Carlo sampling.

[0283] In sudden interference scenarios, the sliding window least squares (LS) method is triggered to avoid filter divergence.

[0284] The following adaptive switching logic is obtained:

[0285]

[0286] The above hybrid filtering strategy can realize real-time dynamic tracking and filtering of the target state obtained by Doppler calculation.

[0287] See also Figure 2 In this embodiment, S5 further includes:

[0288] S5.1. In order to perform multi-level information fusion of the target position and motion state obtained by RF feature analysis, a fusion framework based on extended Kalman filtering is adopted, and the target state vector is defined as position and velocity:

[0289] X k =[x k ,y k ,v x,k ,v y,k ] T

[0290] Among them, X k is the state vector, x k and y k are the positions in the x and y directions respectively, v x,k and v y,k are the velocities in the x and y directions respectively;

[0291] The process model is established as follows:

[0292] X k|k-1 =FX k-1 +w k

[0293]

[0294] Among them, F is the state transfer matrix, w k is the process noise, w k ~N(0,Q),σ x , σ y 、 and is the noise of each state component, Q is the process noise covariance matrix;

[0295] S5.2, based on the distance d1,k obtained in S2 and the arrival angle θ obtained in S3 k , the observation model is established as follows:

[0296]

[0297]

[0298] Among them, Z 1，k is the observation source, H1 is the Jacobian matrix, and R1 is the observation noise covariance;

[0299] S5.3, based on the radial velocity v obtained in S4 r,k and distance d 2,k , the observation model is established as follows:

[0300]

[0301] Among them, Z 2，k is the observation source, H2 is the Jacobian matrix, and R2 is the observation noise covariance;

[0302] S5.3. Perform extended Kalman filtering based on the observation model in S5.2 and S5.3. The prediction steps are as follows:

[0303] X k|k-1 =FX k-1

[0304] P k|k-1 =FP k-1 F T +Q

[0305] Among them, X k|k-1 is the state prediction value, P k|k-1 is the covariance prediction value, F is the state transfer matrix, and Q is the process noise covariance matrix;

[0306] S5.4. For each observation source, calculate the Kalman gain and update the state and covariance:

[0307]

[0308] X k =X k|k-1 +K i (z i -h i (X k|k-1 ))

[0309] P k =(IK i H i )P k|k-1

[0310] Among them, K i is the Kalman gain, X k is the updated state, P k is the updated covariance, I is the identity matrix;

[0311] The final output state estimation vector is:

[0312] X k =[x k ,y k ,v x,k ,v y,k ] T ;

[0313] S5.5. After fusing multi-source RF features through extended Kalman filtering, the target state estimation contains the following core information:

[0314] Position: Cartesian coordinates (x, y) of the target in the two-dimensional plane, unit: meter;

[0315] Velocity: The velocity components of the target in the x and y directions (v x , v y ), unit: m / s;

[0316] Direction: The target's direction of motion φ, in radians or degrees, relative to the coordinate system;

[0317] According to the state estimation vector finally output in S5.4, the target position is the first two components of the state estimation vector, the target velocity vector is the last two components of the state estimation vector, and the target motion direction is the direction of the velocity vector, that is, the azimuth of the velocity vector. The expressions are:

[0318]

[0319] Through the above steps, the hierarchical fusion of multi-source RF features (angle, distance, speed) is achieved, significantly improving the target tracking accuracy.

[0320] like Figure 3 As shown in the figure, the simulation results of the angle estimation algorithm under different incoming wave directions in space are shown. For the true direction of the signal source, the spatial spectrum will reach its maximum value. The peak value in the spatial spectrum is the estimated incident angle of the signal source. The simulation of this method shows the fitting results of incident angles of 40°, 60°, 240° and 320° in turn.

[0321] like Figure 4 As shown in the figure, the target motion trajectory perception result obtained by information fusion is shown. The green line is the target source motion trajectory, and the red line is the estimation of the target motion trajectory obtained by the information fusion algorithm. It can be seen that the two curves have a high degree of fit and the simulation effect is good.

[0322] An embodiment of the present invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor performs the steps of the 6G synaesthesia integration method for intelligent analysis of multi-dimensional radio frequency features as described above.

[0323] See also Figure 5 An embodiment of the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the 6G synaesthesia integration method for multi-dimensional radio frequency feature intelligent analysis as described above.

[0324] An embodiment of the present invention also provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the steps of the 6G synaesthesia integration method for intelligent analysis of multi-dimensional radio frequency features as described above.

[0325] It is understandable that the system, device and storage medium provided by the embodiments of the present invention correspond to the method provided by the embodiments of the present invention. The explanation, examples and beneficial effects of the relevant content can refer to the corresponding parts of the 6G synaesthesia integration method of multi-dimensional radio frequency feature intelligent analysis mentioned above.

[0326] It should be noted that those skilled in the art will understand that all or part of the steps implemented in the embodiments of the present invention can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using hardware, it can be implemented in whole or in part in the form of purchased standard parts or modified parts. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiments of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated therein. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a DVD), or a semiconductor medium (eg, a solid state disk (SSD)).

[0327] In summary, the present invention estimates the target position and motion state by analyzing multiple features of the radio frequency signal, and finally realizes the perception of the communication target position and motion state by fusing and analyzing the estimation results of different dimensions, thereby greatly improving the accuracy and efficiency of target tracking.

[0328] It should be understood that the examples and implementation methods described herein are for illustrative purposes only and are not intended to limit the present invention. Those skilled in the art may make various modifications or changes based on them. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A 6G synaesthesia integrated method for multi-dimensional radio frequency feature intelligent analysis, characterized in that: The following steps are involved: S1. Establish a ground multi-antenna base station to collect the target's radio frequency characteristics; S2. Establish an adaptive path loss model that can adapt to different communication environments and scenarios and accurately estimate the target distance by analyzing the received signal strength data; S3, based on the radio frequency characteristics of S1, combined with the improved MUSIC-ESPRIT fusion algorithm, estimate the arrival angle of the target signal; S4. Based on the frequency change information of the target signal, according to the Doppler frequency shift theory and combined with the hybrid filtering architecture driven by environmental perception, the target position and motion state relative to the ground multi-antenna base station are estimated; S5. Perform information fusion at different levels on the target position and motion state obtained by analyzing the radio frequency features in S4 to obtain the final target tracking result.

2. The 6G synaesthesia integration method for multi-dimensional radio frequency feature intelligent analysis according to claim 1, characterized in that: In S2, the received signal expression is: r(t)=∑ k A k s(t-τ k )+n(t) Among them, r(t) is the received signal, A k is the amplitude attenuation factor of the kth signal, indicating the attenuation of the signal during propagation, s(t) is the transmitted signal, τ k is the time delay of the kth signal, which indicates the time delay from signal transmission to reception, and n(t) is the noise signal; Taking into account multi-source environmental factors, the explicit environmental factor superposition method is adopted to convert different environmental factors into independent influence coefficients. An adaptive enhanced path loss formula is constructed through linear combination, which decomposes the environmental factors into quantifiable correction terms. The distance term and frequency term in the free space loss formula are directly corrected. The corrected formula is expressed as follows: L FS =32.15+20log 10 (f eff )+20log 10 (d eff ) f eff =f×K weather d eff =d LJ ×Γ terrain ×Γ humidity Among them, L FS is the path loss, f eff is the equivalent frequency, d eff is the equivalent distance, K weather is the comprehensive meteorological correction factor, Γ terrain is the geographical environment correction factor, Γ humidity is the humidity-temperature coupling correction factor.

3. The 6G synaesthesia integrated method for multi-dimensional radio frequency feature intelligent analysis according to claim 2, characterized in that: In S2, the calculation methods of multiple correction factors are: Comprehensive meteorological correction factor K weather : A rain =0.18R d 1.2 A fog =0.43C n A dust =0.015D d 0.7 Among them, A rain is the rainfall correction, R d is the rainfall intensity in millimeters per hour, A fog is the fog attenuation correction, C n is visibility, the unit is kilometers, A dust Corrected for dust, D d is the dust concentration in milligrams per cubic meter; Geographical environment correction factor Γ terrain : Among them, α and β are altitude compensation parameters, h msl is the mean altitude, γ is the latitude curvature compensation, and lat is the geographic latitude; Humidity-temperature coupling correction factor Γ humidity : Where μ is the relative humidity and T is the temperature.

4. The 6G synaesthesia integration method for multi-dimensional radio frequency feature intelligent analysis according to claim 1, characterized in that: In S3, in a scenario with multiple signal sources and a large signal-to-noise ratio, the MUSIC multi-target positioning and signal parameter estimation algorithm is used to estimate the target signal arrival angle. The process is as follows: S3.a.

1. Assume that the received data matrix is ​​X r , the linear array consists of M array elements and receives K far-field narrowband signals. The expression of the data matrix is: X r (t)=A(θ)S(t)+N(t) A(θ)=[a(θ1),a(θ2),…,a(θ k )] Where A(θ)∈C M×K is the array flow matrix, a(θ k ) is the steering vector of the kth signal, d is the array element spacing, λ is the signal wavelength, S(t)∈C K×1 is the complex envelope of K incident signals, N(t)∈C M×1 is additive noise; S3.a.

2. The covariance matrix R of the received data is expressed as: Perform eigenvalue decomposition on the covariance matrix R: Among them, E is the eigenvector matrix, D is the diagonal eigenvalue matrix, and the eigenvector is divided into signal subspace and noise subspace according to the size of the eigenvalue. The signal subspace is represented by U s Represented by U, the noise subspace is represented by n express; For each possible signal arrival angle θ, calculate the steering vector a(θ) and calculate the projection of the steering vector on the noise subspace: Where P(θ) is the MUSIC spatial spectrum of the signal, and the peak of P(θ) is searched within the range of θ, where θ corresponding to the peak is the arrival angle of the signal.

5. The 6G synaesthesia integration method for multi-dimensional radio frequency feature intelligent analysis according to claim 4, characterized in that: In S3, in scenarios with low signal-to-noise ratio, real-time processing, or limited computing resources, the ESPRIT algorithm is used to estimate signal parameters using rotational invariance technology. The process is as follows: S3.b.

1. Assume a uniform linear array with M elements, spaced d apart, receiving K far-field narrowband signals with wavelength λ. The array receives the data matrix X. r ∈C M×1 The expression is: X r (t)=A(θ)S(t)+N(t) Where A(θ)∈C M×K is the array flow matrix, a(θ k ) is the steering vector of the kth signal, d is the array element spacing, λ is the signal wavelength, S(t)∈C K×1 is the complex envelope of K incident signals, N(t)∈C M×1 is additive noise; S3.b.

2. The covariance matrix R of the received data is expressed as: Perform eigenvalue decomposition on the covariance matrix: R=UΣU H Where Σ=diag(σ1,…,σ M ) is a diagonal matrix with eigenvalues ​​arranged in descending order, and U is the corresponding eigenvector matrix; S3.b.

3. Select the eigenvectors corresponding to the first K largest eigenvalues ​​to form the signal subspace U s : U S =[u1,…u k ]∈C M×K The signal subspace U s Split into two sub-matrices, corresponding to the two translation sub-arrays: U1=U s (1:M-1,:)∈C (M-1)×K U2=U s (2:M,:)∈C (M-1)×K Construct the augmented matrix and perform singular value decomposition: C 增广 =[U1,U2]∈C (M-1)×2K V c Divide into 4 K×K sub-blocks and calculate the rotation matrix: Among them, V c is a right singular matrix, Ψ is a rotation matrix; Perform eigendecomposition on Ψ: Ψ=TΦT -1 Where Φ=diag(φ1,…,φ K ) is a diagonal matrix with eigenvalues ​​arranged in descending order, T is the eigenvector matrix, from which the arrival angle is calculated by the phase of the eigenvalue:

6. The 6G synaesthesia integrated method for multi-dimensional radio frequency feature intelligent analysis according to claim 5, characterized in that: In S3, the advantages of the MUSIC and ESPRIT algorithms are combined to build a three-level progressive architecture of "coarse estimation + fine search + joint optimization". This reduces computational complexity while improving estimation accuracy to adapt to more complex and changing external environments. The process is as follows: S3.c.1, coarse estimation, divide the linear array into two translation sub-arrays, extract the signal subspace U s And solve for rotation invariance: X1=A1S,X2=A2S=A1ΦS Among them, X1 and X2 are two sub-arrays, A1 and A2 are the flow matrix of the array, S is the complex amplitude of the signal, R is the covariance matrix, U s and U n The signal subspace and noise subspace are respectively extracted, and the signal subspace U s and divided into and Corresponding to subarray 1 and subarray 2 respectively, solve the rotation matrix, calculate the eigenvalue, and get a rough estimate of the angle Where Ψ is the rotation matrix, ψ k is the eigenvalue of Ψ; S3.c.

2. Based on the rough estimation result, a local fine search is performed near the rough estimation, introducing the focusing factor β and the rough estimation result. Construct a weighted spatial spectrum: Among them, the numerator represents the signal subspace projection, the denominator is the traditional MUSIC noise subspace orthogonality measure, and a(θ) is the signal steering vector. Therefore, only in Searching for spectral peaks within a certain range can reduce the amount of calculation. Δ is the constraint range, which is determined by the array beam width. Dynamically adjust the focusing factors β and σ according to the real-time estimated confidence and environmental parameters rough,k And the constraint range Δ, to achieve intelligent local search, the process is: Calculate the standard deviation σ of the rough estimate of the ESPRIT algorithm rough,k , define the confidence index C of the rough estimate k , the calculation expression is: Dynamically adjust the focus factor based on the confidence index: Among them, β k is the dynamic focusing factor, β0 is the basic focusing strength; Adaptively adjust the constraint range based on the confidence index: Among them, Δ b ase is the baseline constraint range; When the confidence level C k When it is low, the estimation is considered unreliable, and the search range is automatically expanded. Based on the dynamic focusing factor and adaptive constraint range obtained from the above formula, a more accurate local refined search result is calculated. S3.c.

3. Joint optimization based on coarse estimation and local fine search is performed. The closed-form solution of the ESPRIT algorithm and the spectral information of the MUSIC algorithm are integrated into the optimization framework to further improve the accuracy. The objective function is designed as: The Gauss-Newton iterative method is used to update the angle estimation until convergence, and the final joint optimization angle result is obtained.

7. The 6G synaesthesia integration method for multi-dimensional radio frequency feature intelligent analysis according to claim 1, characterized in that: Said S4 further comprises: S4.

1. Use a single antenna or array to synthesize the beam to receive the target signal and record the baseband signal s b (t): Where A(t) is the complex envelope of the signal, n(t) is the noise, φ dop (t) includes phase changes caused by target motion; The target signal is extracted by matched filtering or low-pass filtering, noise is suppressed, and the results are obtained under high signal-to-noise ratio and low signal-to-noise ratio conditions respectively. Under high signal-to-noise ratio, the instantaneous frequency f is directly extracted through the signal phase d (t): Under low signal-to-noise ratio, short-time Fourier transform is used to calculate the time-frequency matrix: Where S(t,f) is the time-frequency matrix, w(t) is the Hamming window function; S4.

2. Based on Doppler shift theory, the radial velocity and range of the target are modeled as follows: Among them, v r (t) is the radial velocity, d dop (t) is the distance between the target and the receiving antenna, f d (t) is the Doppler shift, λ is the signal wavelength; If the initial distance d0 is unknown, select the frequency observation values ​​f at two moments d1 and f d2 Simultaneous equations: Among them, d1 and d2 are the distances between the target and the receiving antenna during the two observations, Δt is the observation interval, and after eliminating d0, we can get v r1 、v r2 , d1 and d2; If the initial distance d0 is known, the real-time distance can be directly obtained by integrating the Doppler shift: S4.

3. To achieve real-time dynamic tracking of the target's motion state, an environment-aware driven hybrid filtering architecture is introduced. By monitoring channel characteristics in real time, the optimal filtering strategy is dynamically selected to achieve matching between the model and the environment. The specific process is as follows: S4.3.

1. Extract environmental features and calculate the instantaneous signal-to-noise ratio SNR(t): Among them, s b (t) is the baseband signal, n(t) is the noise; S4.3.

2. Determine the multipath strength by analyzing the signal cyclostationary characteristics or delay correlation, and design the following hybrid filtering strategy: In low-noise scenarios, the standard Kalman filter (KF) is enabled; In high-noise / multipath scenarios, the robust particle filter (PF) is used to combat non-Gaussian noise through Monte Carlo sampling. In sudden interference scenarios, the sliding window least squares (LS) method is triggered to avoid filter divergence. The following adaptive switching logic is obtained:

8. The 6G synaesthesia integrated method for multi-dimensional radio frequency feature intelligent analysis according to claim 1, characterized in that: Said S5 further comprises: S5.

1. In order to perform multi-level information fusion of the target position and motion state obtained by RF feature analysis, a fusion framework based on extended Kalman filtering is adopted, and the target state vector is defined as position and velocity: X k =[x k ,y k ,v x,k ,v y,k ] T Among them, X k is the state vector, x k and y k are the positions in the x and y directions respectively, v x,k and v y,k are the velocities in the x and y directions respectively; The process model is established as follows: X k|k-1 =FX k-1 +w k Among them, F is the state transfer matrix, w k is the process noise, w k ~N(0,Q),σ x , σ y 、 and is the noise of each state component, Q is the process noise covariance matrix; S5.2, based on the distance d1,k obtained in S2 and the arrival angle θ obtained in S3 k , the observation model is established as follows: Among them, Z 1，k is the observation source, H1 is the Jacobian matrix, and R1 is the observation noise covariance; S5.3, based on the radial velocity v obtained in S4 r,k and distance d 2,k , the observation model is established as follows: Among them, Z 2，k is the observation source, H2 is the Jacobian matrix, and R2 is the observation noise covariance; S5.

3. Perform extended Kalman filtering based on the observation model in S5.2 and S5.

3. The prediction steps are as follows: X k|k-1 =FX k-1 P k|k-1 =FP k-1 F T +Q Among them, X k|k-1 is the state prediction value, P k|k-1 is the covariance prediction value, F is the state transfer matrix, and Q is the process noise covariance matrix; S5.

4. For each observation source, calculate the Kalman gain and update the state and covariance: X k =X k|k-1 +K i (z i -h i (X k|k-1 )) P k =(I-K i H i )P k|k-1 Among them, K i is the Kalman gain, X k is the updated state, P k is the updated covariance, I is the identity matrix; The final output state estimation vector is: X k =[x k ,y k ,v x,k ,v y,k ] T ; S5.

5. After fusing multi-source RF features through extended Kalman filtering, the target state estimation contains the following core information: Position: The Cartesian coordinates (x, y) of the target in the two-dimensional plane; Velocity: The velocity components of the target in the x and y directions (v x , v y ); Direction: the target's direction of movement φ; According to the state estimation vector finally output in S5.4, the target position is the first two components of the state estimation vector, the target velocity vector is the last two components of the state estimation vector, and the target motion direction is the direction of the velocity vector, that is, the azimuth of the velocity vector. The expressions are:

Images