High-precision direction finding method and system based on amplitude-phase joint feature fusion

Abstract

The application provides a high-precision direction finding method and system based on amplitude-phase joint feature fusion, comprising: collecting radio signals and preprocessing; calculating the amplitude difference DA and the phase difference of adjacent elements, and constructing a joint feature vector; dynamically weighting and fusing the amplitude and phase features based on the real-time signal-to-noise ratio to obtain a fused direction feature; calculating the spatial spectrum function by using the spatial spectrum estimation technology, and determining the incoming direction angle by searching for the peak value; and finally, performing dynamic smoothing filtering on the continuous measurement results to output a stable direction angle estimate. Through the deep fusion and intelligent weighting of the amplitude and phase features, the direction finding precision and robustness in complex environments such as low signal-to-noise ratio and multipath interference are significantly improved, and the application is suitable for fields such as spectrum monitoring, radio positioning and electronic reconnaissance.

Description

Technical Field

[0001] This invention relates to the field of radio direction finding and signal processing technology, specifically a high-precision direction finding method and system based on amplitude-phase joint feature fusion. Background Technology

[0002] The Watson-Watt direction finding method, a classic direction finding technique, estimates the direction of arrival by using the amplitude ratio or phase difference of signals received by orthogonally arranged antenna elements. However, traditional implementations primarily rely on a single amplitude or phase difference characteristic for calculation, revealing many inherent flaws in complex engineering applications.

[0003] 1. Highly sensitive to multipath propagation interference: In urban, mountainous, or indoor environments, radio wave propagation is easily blocked and reflected by buildings, terrain, etc., causing the receiver to simultaneously receive a direct wave and one or more reflected waves. Signal superposition causes severe distortion of signal amplitude and phase. Traditional single-feature direction finding methods struggle to distinguish between direct and reflected waves, easily misjudging the stronger reflected wave as the main signal direction, leading to significant deviations or even complete errors in the direction finding results.

[0004] 2. Performance Degrades Sharply in Low Signal-to-Noise Ratio (SNR) Environments: For weak signals transmitted over long distances, such as emergency beacons or deep-space communication signals, the signal strength is close to or lower than the ambient noise level. In this situation, both the amplitude and phase information of the signal are severely overwhelmed by noise. Traditional algorithms lack effective noise suppression mechanisms, and their reliable operation typically requires a high SNR threshold (e.g., greater than 20dB), making it impossible to effectively extract accurate directional features under low SNR conditions. The amplitude and phase of the weak signal are overwhelmed by noise, making it impossible to effectively extract directional features.

[0005] 3. Limited azimuth resolution and weak anti-interference capability: When there are two or more radiation sources with similar azimuth angles in space, the amplitude or phase difference characteristics they produce are very small, making it difficult for traditional methods to effectively distinguish them, resulting in insufficient direction-finding resolution. In addition, the presence of narrowband interference at the same frequency will severely contaminate the single characteristic measurement value, causing random jumps in the direction-finding results and making the output extremely unstable.

[0006] To address the above problems, existing technologies have proposed some improvements, but each has its own limitations:

[0007] Hardware enhancement solutions: Increasing the number of antenna elements to improve the spatial sampling rate and thus enhance resolution. However, this approach significantly increases the system's hardware complexity, physical size, and manufacturing cost, and places higher demands on backend signal processing capabilities, making it unsuitable for many size- and cost-sensitive applications.

[0008] Single-feature optimization schemes: Some schemes focus on improving the measurement accuracy of a single feature, such as using high-precision phase synchronization circuits to reduce phase measurement errors, or designing complex filters to suppress noise in amplitude measurements. However, these schemes fail to fundamentally address the complementarity and contradictions between amplitude and phase information in multipath scenarios. When the optimized feature itself fails due to interference, system performance still degrades significantly. Schemes optimizing amplitude measurement are ineffective in scenarios with amplitude fading caused by strong multipath; while schemes optimizing phase measurement are vulnerable to phase ambiguity and rapid phase changes caused by multipath.

[0009] In summary, existing technologies lack a high-precision, robust direction finding solution that can intelligently integrate and utilize the multi-dimensional information of signal amplitude and phase, and adapt to changes in complex electromagnetic environments. Summary of the Invention

[0010] The primary objective of this invention is to overcome the aforementioned deficiencies of existing Wattson-Watt direction finding technology and to provide a direction finding method and system with high accuracy, strong environmental adaptability, and good anti-interference performance.

[0011] Another objective of this invention is to provide a direction finding method that can operate stably under low signal-to-noise ratio conditions.

[0012] Another objective of this invention is to provide a direction finding scheme that can effectively suppress multipath effects.

[0013] A direction-finding method based on amplitude-phase joint feature fusion includes the following steps:

[0014] The system collects radio signals in space, amplifies, down-converts, and performs analog-to-digital conversion on the radio signals to obtain multi-channel digital signals.

[0015] Bandpass filtering, amplitude-phase consistency calibration, and time delay compensation are performed on the multi-channel digital signal to obtain the preprocessed digital signal.

[0016] Based on the preprocessed digital signal, the amplitude difference feature ΔA and phase difference feature of adjacent array elements are calculated. The amplitude difference feature ΔA and the phase difference feature are compared. A joint feature vector is constructed by combining features, and dynamic weighted fusion based on channel state information is used to obtain the fused directional features. Based on the fused directional features, a spatial spectrum function is calculated using spatial spectrum estimation techniques, and the direction angle of arrival is determined by searching for the peak value of the spatial spectrum function.

[0017] The incoming wave direction angles obtained from multiple consecutive measurements are dynamically smoothed and filtered to output a final stable direction angle estimate.

[0018] Furthermore, the calculation method for the amplitude difference feature ΔA is as follows: ,in and The signal amplitudes of the i-th and j-th array elements are respectively; the phase difference feature The calculation method is as follows ,in and The phases are the signal phases of the i-th and j-th array elements, respectively, and the phase difference is subjected to phase unwinding processing to eliminate 2π periodic ambiguity.

[0019] Furthermore, the dynamic weighted fusion specifically involves: dynamically adjusting the amplitude difference feature ΔA and the phase difference feature based on the real-time estimated signal-to-noise ratio. Weights during the fusion process: when the signal-to-noise ratio is below the first threshold, the amplitude difference feature is assigned a higher weight; when the signal-to-noise ratio is above the second threshold, the phase difference feature is assigned a higher weight.

[0020] Furthermore, the spatial spectrum estimation technique employs a multiple signal classification algorithm.

[0021] Furthermore, the dynamic smoothing filtering process employs a Kalman filter.

[0022] Furthermore, it also includes a multipath interference mitigation step: by analyzing the amplitude difference characteristic ΔA and the phase difference characteristic Inconsistencies in time series data are identified and anomalous observation data caused by multipath reflections are removed.

[0023] A direction-finding system based on amplitude-phase joint feature fusion, used to implement the above method, the system comprising:

[0024] Antenna and RF front-end module are used to collect radio signals in space, process the radio signals, and obtain multi-channel digital signals.

[0025] The signal preprocessing module is used to perform bandpass filtering, amplitude and phase consistency calibration, and time delay compensation on multi-channel digital signals to obtain preprocessed digital signals.

[0026] The core direction-finding processing module is used to calculate the amplitude difference feature ΔA and phase difference feature between adjacent array elements based on the preprocessed digital signal. The amplitude difference feature ΔA and the phase difference feature are compared. A joint feature vector is constructed by combining features, and dynamic weighted fusion based on channel state information is used to obtain the fused directional features. Based on the fused directional features, a spatial spectrum function is calculated using spatial spectrum estimation techniques, and the direction angle of arrival is determined by searching for the peak value of the spatial spectrum function.

[0027] The dynamic smoothing module is used to perform dynamic smoothing filtering on the incoming wave direction angle obtained from multiple consecutive measurements, and output the final stable direction angle estimate.

[0028] Furthermore, the antenna and radio frequency front-end module includes:

[0029] A multi-channel antenna array used to collect radio signals in space;

[0030] The radio frequency front-end module is connected to each antenna element and is used to amplify, down-convert, and convert radio signals to obtain multi-channel digital signals.

[0031] Furthermore, the calculation method for the amplitude difference feature ΔA is as follows: ,in and The signal amplitudes of the i-th and j-th array elements are respectively; the phase difference feature The calculation method is as follows ,in and The phases are the signal phases of the i-th and j-th array elements, respectively, and the phase difference is subjected to phase unwinding processing to eliminate 2π periodic ambiguity.

[0032] Furthermore, the dynamic weighted fusion specifically involves: dynamically adjusting the amplitude difference feature ΔA and the phase difference feature based on the real-time estimated signal-to-noise ratio. Weights during the fusion process: when the signal-to-noise ratio is below the first threshold, the amplitude difference feature is assigned a higher weight; when the signal-to-noise ratio is above the second threshold, the phase difference feature is assigned a higher weight.

[0033] This invention proposes a core technical approach of "amplitude-phase joint feature fusion and auxiliary solution." This scheme constructs multi-dimensional signal features, utilizes their feature complementarity, and introduces dynamic weighted fusion and advanced signal processing algorithms to achieve high-precision and robust direction-of-arrival estimation. Compared with existing technologies, this invention has the following significant advantages and beneficial effects:

[0034] 1. Significantly improved direction finding accuracy: By combining and complementing amplitude and phase characteristics and employing high-resolution algorithms, the direction finding error can be controlled within ±3° in typical application scenarios, which is far superior to traditional methods.

[0035] 2. Excellent low signal-to-noise ratio performance: The innovative dynamic weighted fusion mechanism enables the system to extract effective information and maintain reliable direction finding capability even under harsh conditions with a signal-to-noise ratio of less than 10dB through feature complementarity and Kalman filtering.

[0036] 3. Strong anti-multipath interference capability: For multipath interference: Through the feature contradiction identification mechanism, it can effectively identify and eliminate data points that are severely affected by multipath interference, and the direction finding stability of the system in multipath environment is significantly improved compared with traditional solutions.

[0037] 4. High flexibility and adaptability: The system hardware platform supports wideband operation (verified 100MHz~512MHz). By adjusting the antenna element type, array layout, and software algorithm parameters, it can be easily adapted to different frequency bands such as shortwave, ultra-shortwave, and microwave, as well as various application scenarios such as ground fixed stations, vehicle-mounted, ship-mounted, and airborne.

[0038] 5. Stable and smooth output: The introduced dynamic smoothing post-processing module effectively suppresses instantaneous jumps in the direction finding results, providing a smoother, more continuous direction trajectory that conforms to the actual movement law of the target, greatly improving the user experience and the convenience of subsequent data processing. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the high-precision direction finding system based on amplitude-phase joint feature fusion according to the present invention.

[0040] Figure 2 This is a flowchart of the high-precision direction finding method based on amplitude-phase joint feature fusion of the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] Please see Figure 1 This invention provides a direction finding system based on amplitude-phase joint feature fusion, comprising the following modules:

[0043] Multi-channel antenna array: The receiving array is composed of a 4-element omnidirectional composite microstrip patch antenna, which is used to collect radio signals in space.

[0044] RF front-end module: Each antenna element is independently connected to an RF receiving channel, and each RF receiving channel includes, in sequence:

[0045] Low-noise amplifiers are used to amplify weak signals with a noise figure better than 1.5dB to maintain the signal-to-noise ratio.

[0046] A downconverter is used to convert the received radio frequency signal to a fixed intermediate frequency for easy bandpass sampling;

[0047] High-speed analog-to-digital converter with an ADC sampling rate greater than twice the intermediate frequency bandwidth to ensure distortion-free signal sampling;

[0048] Signal preprocessing module: This module preprocesses the multi-channel digital signals output by the ADC, mainly including:

[0049] Bandpass filtering: Design a digital filter based on the target signal frequency band to retain the useful signal and suppress out-of-band noise and interference.

[0050] Amplitude and phase consistency calibration: During system initialization or periodic operation, a known calibration signal is injected to measure and compensate for amplitude and phase inconsistencies introduced by hardware differences between channels.

[0051] Delay compensation: Accurately calculate and compensate for the transmission path delay differences introduced by the signal arriving at array elements at different spatial locations, ensuring the accuracy of subsequent feature extraction.

[0052] The core direction finding processing module includes:

[0053] Joint feature extraction unit: synchronously captures the preprocessed digital signals of each array element and calculates the amplitude difference features of adjacent array element pairs. ) and phase difference characteristics ( ).

[0054] Dynamic weighted fusion unit: combines the extracted amplitude difference and phase difference features into a joint feature vector F = [ , The core innovation of this unit lies in introducing a fusion framework based on a Kalman filter. This filter can dynamically adjust the weights of amplitude and phase differences in the fusion process based on the real-time estimated channel state (especially the signal-to-noise ratio, SNR). Specifically, under low SNR conditions, the amplitude difference is less affected by phase noise and is given a higher weight; under high SNR conditions, the phase difference provides higher direction-finding accuracy and is also given a higher weight. Through the prediction and update mechanism of the Kalman filter, optimal estimation of the fused directional characteristics is achieved.

[0055] Spatial spectrum estimation unit: Based on the fused high-reliability directional characteristics, an improved MUSIC algorithm spectral analysis algorithm is used to construct the spatial spectrum function P(θ). This function reflects the possibility that the signal comes from different directional angles θ.

[0056] Dynamic Smoothing Module: This module introduces another Kalman filter to smooth the preliminary direction finding results at multiple consecutive time points. This filter utilizes the continuity of target motion (such as the rate of change of direction angle) to make an optimal estimate of the current direction finding result and predict the possible position at the next time point, thereby effectively suppressing jumps in direction finding results caused by instantaneous interference or noise, and outputting a smooth and stable direction angle sequence.

[0057] Multipath interference identification and rejection module: Utilizing the inconsistencies in amplitude and phase changes between direct waves and reflected waves (direct waves are the actual signals needed for direction finding, with stable amplitude and phase; reflected waves are multipath signals reflecting the wrong direction, with rapidly fading amplitude and relatively stable phase changes), a contradiction detection logic is designed. When a significant contradiction is detected between the amplitude difference and phase difference characteristics, the data at that moment (reflected wave data) is marked as abnormal and rejected, prioritizing the use of data that conforms to the characteristics of direct waves for direction finding.

[0058] Please see Figure 2 This invention also provides a direction finding method based on amplitude-phase joint feature fusion, comprising the following steps:

[0059] Step S1, Signal Acquisition and Preprocessing: The multi-channel antenna array receives space radio signals, which are then down-converted and digitized by the RF front-end. The signal preprocessing module performs bandpass filtering, amplitude and phase consistency calibration, and time delay compensation to obtain synchronized and calibrated baseband or intermediate frequency digital signals for each array element.

[0060] Step S2, Multi-dimensional Feature Extraction: For each pair of preset adjacent antenna elements (i, j):

[0061] Calculate the amplitude difference characteristics: ,in and These are the amplitude values ​​of the two array element signals, respectively.

[0062] Calculate phase difference characteristics: ,in and The phase value is extracted using digital signal processing techniques such as quadrature demodulation or FFT, and 2π periodic ambiguity is eliminated.

[0063] Step S3, Joint Feature Dynamic Weighted Fusion: Combine ΔA obtained in step S2 with... These are combined to form an eigenvector F. Based on the real-time estimated system signal-to-noise ratio (SNR), the amplitude difference weighting factor w_A and the phase difference weighting factor are dynamically determined. (satisfy The weighted features are input into a Kalman filter, which, based on the system state model and the observation model, outputs an optimally estimated fusion directional feature quantity that is more robust to noise.

[0064] The implementation method of joint feature fusion is as follows:

[0065] 1) Establishment of the state-space model:

[0066] The core of Kalman filtering is the state-space model. The desired "true" directional feature is defined as a state vector, which contains intermediate observations determined by both amplitude and phase.

[0067] The state vector is defined as: ;

[0068] in: It is an intermediate variable related to the direction of arrival, a normalized direction phase that combines amplitude and phase information; It is the incoming wave direction angle at time k, which is the target to be solved; the subscript k indicates the discrete-time index.

[0069] State equation process model:

[0070] Assuming the target direction changes steadily over a short period of time, a simple random walk model is used:

[0071] ;

[0072] Where: F is the state transition matrix. For the random walk model, F is the identity matrix I; w_k is the process noise, which follows a Gaussian distribution with zero mean and covariance matrix Q, i.e., w_k ~ N(0, Q).

[0073] Observation equations and measurement models:

[0074] The observed values ​​are the amplitude and phase differences obtained by direct observation and calculation:

[0075] Z_k = H(X_k) + v_k;

[0076] in: Let X be the observation vector at time k, representing the measured amplitude and phase differences; H(X_k) is the observation model function, which maps the state vector to the observation space. For a specific antenna pair (i, j), the relationship between its theoretical observations and the orientation angle θ is:

[0077] ;

[0078] Where d is the element spacing and λ is the signal wavelength.

[0079] ΔA_theoretical is related to the antenna pattern G(θ), ΔA_theoretical = G_i(θ) - G_j(θ) (in dB).

[0080] Therefore, H(X_k) can be expressed as:

[0081] ;

[0082] Here, v_k is the observation noise, which follows a Gaussian distribution with zero mean and covariance matrix R_k, i.e., v_k ~ N(0, R_k). R_k is the key to realizing dynamic weighting.

[0083] 2) Implementation of the dynamic weighting strategy:

[0084] Dynamic weighting is achieved by adjusting the observation noise covariance matrix R_k in real time. When the signal-to-noise ratio (SNR) is high, the phase difference observation is considered more reliable and should be assigned lower uncertainty (i.e., smaller noise variance); when the SNR is low, the opposite is true.

[0085] The observation noise covariance matrix R_k is defined as follows:

[0086] ;

[0087] This is a diagonal matrix, and its diagonal elements and These are the observation noise variances of the amplitude difference and phase difference, respectively, which are functions of the current signal-to-noise ratio (SNR).

[0088] Variance-Signal-Noise Ratio Function Design:

[0089] Amplitude difference noise variance: ;

[0090] This is a decreasing function; the higher the SNR, the smaller the variance, and the higher the reliability. γ_A is the lower limit, indicating that even with a high SNR, there is still a fundamental measurement error.

[0091] Phase difference noise variance:

[0092] ;

[0093] Similarly decreasing, but usually and The large value indicates that at low SNR, the phase difference variance increases sharply, much larger than the amplitude difference variance. This feature is specifically designed to simulate the rapid degradation of phase measurements at low SNR.

[0094] Through the above function design, the Kalman gain of the Kalman filter will automatically assign higher weights to observation components with smaller noise variances during the calculation. For example, when the SNR is low, Much larger The filter will place more trust in the amplitude difference observation ΔA_m.

[0095] 3) The Kalman filter iterative process is as follows:

[0096] State prediction: ;

[0097] Error covariance prediction: ;

[0098] Calculate the Kalman gain:

[0099] ;

[0100] in: It is an observation model exist The Jacobian matrix at that point is used for linearization;

[0101] Status Update: ;

[0102] Error covariance update: ;

[0103] After iterative Kalman filtering, the resulting state estimate In This is already a more accurate orientation angle estimate that has been fused and smoothed. It is then used as the initial estimate in the next step of spectral estimation.

[0104] Step S4: High-precision azimuth angle calculation: Using the fused features obtained in step S3, a spatial spectrum function P(θ) is constructed based on the MUSIC algorithm. Within a preset azimuth angle search range, θ is scanned at certain steps, and the P(θ) value corresponding to each θ is calculated. The search is conducted to find the angle that maximizes P(θ) globally. This angle is the initial estimate of the incoming wave direction angle obtained from the final calculation.

[0105] The specific implementation method is as follows:

[0106] 1) Construct the fused data vector matrix:

[0107] For M array elements, L non-repeating baseline antenna pairs can be formed. For the l-th baseline, the fused directional features obtained by Kalman filtering are used to construct an estimate of the fused covariance matrix R_fused in the updated state X_k.

[0108] Construct the instantaneous fused data vector at time k:

[0109] ;

[0110] Here Instead of a direct measurement, the phase is a calibrated and smoothed "theoretical" phase obtained by inversely calculating the updated state θ_{k|k} using a Kalman filter. The amplitude A_m(k) uses the filtered amplitude value. Each element of this vector y(k) already contains high-precision directional information from the Kalman filter fusion result in its phase.

[0111] Calculate the fusion covariance matrix:

[0112] Where H represents the conjugate transpose, and N is the number of snapshots used for averaging.

[0113] 2) Construct the spatial spectral function P(θ) and perform a search, specifically including:

[0114] Eigenvalue decomposition: Perform eigenvalue decomposition on R_fused: ;

[0115] U is a unitary matrix composed of eigenvectors; Σ is a diagonal matrix composed of eigenvalues. .

[0116] Signal subspace and noise subspace partitioning: For a space containing D signal sources (D < M), the eigenvectors corresponding to the first D large eigenvalues ​​span the signal subspace U_S, and the eigenvectors corresponding to the remaining MD small eigenvalues ​​span the noise subspace U_N.

[0117] Constructing the MUSIC spatial spectrum: Where a(θ) is the array manifold vector, .

[0118] Spectral peak search: A one-dimensional search is performed within the range of θ, calculating the P_MUSIC(θ) value corresponding to each θ. The θ values ​​corresponding to the D peaks of the spectral function P_MUSIC(θ) are the estimated D incoming wave direction angles. .

[0119] Step S5: Dynamic Smoothing and Output of Results: The initial direction angle estimates obtained in step S4 are input into the Kalman filter of the dynamic smoothing module. The current direction finding results are smoothed, and the direction at the next moment is predicted. Finally, a stable, smooth, and highly accurate incoming wave direction angle sequence with significantly reduced jumps is output.

[0120] Implementation Case: Shipborne High-Precision Direction Finding System in the Ultra-Shortwave Band (100~512MHz)

[0121] Application scenario: High-precision orientation of communication signals or interference sources in the 100MHz ~ 512MHz VHF band on ships in complex electromagnetic environments.

[0122] The hardware configuration is as follows: Antenna array; 4-element wideband microstrip patch antenna array. RF front end: LNA gain of 20dB per channel, noise figure of 1.2dB; downconverted to 175MHz intermediate frequency, 40MHz bandwidth; ADC sampling rate of 100MSPS, resolution of 16 bits.

[0123] The direction finding process is as follows: After the system is powered on, amplitude and phase consistency calibration is first performed; the target VHF signal is received, the preprocessing module performs bandpass filtering (passband 100-512MHz), and applies calibration parameters; the core processing module calculates ΔA and ΔA of adjacent array elements. The current environmental signal-to-noise ratio is estimated to be 8dB. The fusion algorithm automatically adjusts the weights, favoring the amplitude difference feature. After Kalman fusion, robust direction features are obtained. Using the fused features, the spatial spectrum P(θ) is calculated based on the MUSIC algorithm, and the initial azimuth angle is obtained by searching for the peak. The single direction finding error can be controlled within ±1.2° after testing. The dynamic smoothing module performs Kalman filtering on 5 consecutive direction finding results, and the final output azimuth angle fluctuation range can be stabilized within ±0.5°.

[0124] Actual performance: In a simulated environment with a signal-to-noise ratio of 10dB and significant multipath reflections, the azimuth direction finding error of the method of this invention is ≤ ±3°. In contrast, under the same conditions, the error of the traditional single amplitude difference direction finding method is ≥ ±5°, and the results fluctuate drastically.

[0125] This invention significantly improves the performance of traditional Watson-Watt direction finding systems in harsh environments such as multipath interference, low signal-to-noise ratio, and coexistence of signals at the same frequency by deeply fusing the amplitude and phase difference characteristics of signals and combining spatial spectrum estimation theory with dynamic filtering techniques. This invention can be widely applied in various technical fields such as spectrum monitoring, radio positioning, emergency communication, electronic reconnaissance, and countermeasures.

[0126] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A direction-finding method based on amplitude-phase joint feature fusion, characterized in that, Includes the following steps: The system collects radio signals in space, amplifies, down-converts, and converts the radio signals from analog to digital to obtain multi-channel digital signals. Bandpass filtering, amplitude-phase consistency calibration, and time delay compensation are performed on the multi-channel digital signal to obtain the preprocessed digital signal. Based on the preprocessed digital signal, the amplitude difference feature ΔA and phase difference feature of adjacent array elements are calculated. The amplitude difference feature ΔA and the phase difference feature are compared. A joint feature vector is constructed by combining the features, and dynamic weighted fusion based on channel state information is used to obtain the fused directional features. Based on the fused directional features, a spatial spectrum function is calculated using spatial spectrum estimation techniques, and the incoming wave direction angle is determined by searching for the peak value of the spatial spectrum function. The incoming wave direction angle obtained from multiple consecutive measurements is dynamically smoothed and filtered to output a final stable direction angle estimate. The dynamic weighted fusion specifically involves: dynamically adjusting the amplitude difference feature ΔA and the phase difference feature based on the real-time estimated signal-to-noise ratio. Weights during the fusion process: When the signal-to-noise ratio is below the first threshold, the amplitude difference feature is assigned a higher weight; when the signal-to-noise ratio is above the second threshold, the phase difference feature is assigned a higher weight.

2. The direction finding method based on amplitude-phase joint feature fusion according to claim 1, characterized in that, The amplitude difference feature ΔA is calculated as follows: ,in and The signal amplitudes of the i-th and j-th array elements are respectively; the phase difference feature The calculation method is as follows ,in and The phases are the signal phases of the i-th and j-th array elements, respectively, and the phase difference is subjected to phase unwinding processing to eliminate 2π periodic ambiguity.

3. The direction finding method based on amplitude-phase joint feature fusion according to claim 1, characterized in that, The spatial spectrum estimation technique employs a multiple signal classification algorithm.

4. The direction finding method based on amplitude-phase joint feature fusion according to claim 1, characterized in that, The dynamic smoothing filtering process employs a Kalman filter.

5. The direction finding method based on amplitude-phase joint feature fusion according to claim 1, characterized in that, It also includes a multipath interference mitigation step: by analyzing the amplitude difference characteristic ΔA and the phase difference characteristic Inconsistencies in time series data are identified and anomalous observation data caused by multipath reflections are removed.

6. A direction-finding system based on amplitude-phase joint feature fusion, used to implement the method described in any one of claims 1 to 5, characterized in that, The system includes: Antenna and RF front-end module are used to collect radio signals in space, process the radio signals, and obtain multi-channel digital signals. The signal preprocessing module is used to perform bandpass filtering, amplitude and phase consistency calibration, and time delay compensation on multi-channel digital signals to obtain preprocessed digital signals. The core direction-finding processing module is used to calculate the amplitude difference feature ΔA and phase difference feature between adjacent array elements based on the preprocessed digital signal. The amplitude difference feature ΔA and the phase difference feature are compared. A joint feature vector is constructed by combining features, and dynamic weighted fusion based on channel state information is used to obtain the fused directional features. Based on the fused directional features, a spatial spectrum function is calculated using spatial spectrum estimation techniques, and the direction angle of arrival is determined by searching for the peak value of the spatial spectrum function. The dynamic smoothing module is used to perform dynamic smoothing filtering on the incoming wave direction angle obtained from multiple consecutive measurements, and output the final stable direction angle estimate. The dynamic weighted fusion specifically involves: dynamically adjusting the amplitude difference feature ΔA and the phase difference feature based on the real-time estimated signal-to-noise ratio. Weights during the fusion process: when the signal-to-noise ratio is below the first threshold, the amplitude difference feature is assigned a higher weight; when the signal-to-noise ratio is above the second threshold, the phase difference feature is assigned a higher weight.

7. The direction finding system based on amplitude-phase joint feature fusion according to claim 6, characterized in that, The antenna and radio frequency front-end module includes: A multi-channel antenna array used to collect radio signals in space; The radio frequency front-end module is connected to each antenna element and is used to amplify, down-convert, and convert radio signals to obtain multi-channel digital signals.

8. The direction finding system based on amplitude-phase joint feature fusion according to claim 6, characterized in that, The amplitude difference feature ΔA is calculated as follows: ,in and The signal amplitudes of the i-th and j-th array elements are respectively; the phase difference feature The calculation method is as follows ,in and The phases are the signal phases of the i-th and j-th array elements, respectively, and the phase difference is subjected to phase unwinding processing to eliminate 2π periodic ambiguity.

Images