A method and device for integrated signal generation and multi-target clustering identification

Abstract

The application discloses a kind of integrated signal generation and multi-target clustering identification method and device of feeling.This application uses unified AFDM waveform as bottom transmission mechanism, realizes full diversity gain in time-frequency plane by adjusting affine parameter, effectively suppresses the inter-subcarrier interference under high dynamic scene.In the receiving end, the distance-Doppler heat map is reconstructed by fast cyclic correlation radar algorithm, and the original detection point cloud is extracted combined with 2D-CFAR detection.Introduce the DBSCAN clustering algorithm based on dimensionless weight, and the scattered point is condensed into an independent entity target in RDM physical space, and the radial velocity characteristics of clustering centroid are used to construct automatic decision engine, to realize the real-time classification and identification of static obstacles and dynamic moving targets.The application solves the technical pain points of traditional ISAC system in high frequency, high mobility scene, such as sensing point fragmentation and confusion of static and dynamic targets.

Description

Technical Field

[0001] This invention relates to the field of integrated sensing and communications (ISAC) technology, and in particular to a method and apparatus for integrated sensing signal generation and multi-target clustering identification. Background Technology

[0002] With the evolution of B5G (Beyond 5G) and 6G mobile communication technologies, spectrum resources are expanding into millimeter-wave (mmWave) and even terahertz (THz) bands. In these bands, communication and radar sensing exhibit natural convergence: the overlap between communication base stations and radar sensors in RF front-ends, antenna arrays, and signal processing units is increasing; the traditional "communication first, sensing later" model leads to spectrum fragmentation. ISAC aims to achieve simultaneous information transmission and target detection using a single waveform. However, existing ISAC research largely focuses on waveform design, lacking robust engineering solutions for how to efficiently and automatically extract entity target information from noise at the back end.

[0003] Orthogonal Frequency Division Multiplexing (OFDM) is the core of 4G / 5G communication and the preferred waveform in early ISAC research. However, in highly mobile scenarios (such as high-speed rail, drones, and vehicle-mounted radar), OFDM reveals fatal weaknesses. High-speed movement leads to inter-carrier interference (ICI), disrupting subcarrier orthogonality. In the delay-Doppler plane, the discreteness of OFDM symbols causes severe velocity ambiguity when detecting high-speed targets. Simultaneously, sensing waveforms require high-power transmission to ensure detection range; OFDM's high peak-to-average power ratio (PAPR) limits the efficiency of power amplifiers, reducing the maximum sensing coverage. Current sensing implementations mainly rely on "embedding" sensing waveforms within a single communication time slot or frequency band. This discrete design mechanism inevitably induces vicious competition for spectrum resources, redundant hardware overhead, and severe inter-system intermodulation interference, making it difficult to maintain stable performance in high-frequency bands such as millimeter waves / terahertz and in highly mobile environments.

[0004] Therefore, developing a unified waveform with "intrinsic sensing" characteristics, enabling it to deeply integrate high-speed information transmission, sub-meter-level high-precision detection, and multi-dimensional anti-interference characteristics at the physical layer, and constructing a resource scheduling framework that supports dynamic decoupling and on-demand allocation of performance parameters, has become a key path and core technology requirement for breaking through the bottleneck of 6G integrated sensing systems.

[0005] To overcome the shortcomings of OFDM, Affine Fourier Transform Division Multiplexing (AFDM) has been proposed as a novel multicarrier modulation technique. AFDM maps symbols onto the slanted axes of the time-frequency plane using the affine Fourier transform. This waveform is essentially a set of orthogonal chirp signals, exhibiting a high degree of similarity to the linear frequency modulation (LFM) used in traditional radar. By adjusting the parameters... and AFDM can convert Doppler frequency shift into cyclic shift between subcarriers, thus achieving complete diversity in the time-delay-Doppler domain. This allows AFDM to maintain lower complexity while providing more accurate parameter estimation when dealing with highly mobile targets.

[0006] Currently, ISAC research typically stops at generating RDMs. However, in practical applications, there is a significant gap between RDMs and "target decision-making." While CFAR detection algorithms can filter out some noise, they often generate multiple adjacent detection points for a single physical target (such as a car). If output directly, the backend controller will not be able to identify that these belong to the same target. Urban environments contain a large number of static objects (walls, fences, trees). Although traditional Moving Target Indication (MTI) filters can filter out stationary targets, in ISAC scenarios, we often need to perceive static obstacles for path planning. Existing ISAC systems lack an automated mechanism that can distinguish between "stationary background" and "moving threats" in the environment in real time.

[0007] Meanwhile, while DBSCAN is widely used in machine learning, its integration into ISAC physical layer waveform processing remains immature. The most direct reason is that performing 3D clustering directly in the Cartesian coordinate system results in enormous computational costs. Second, the numerical ranges of distance (meters) and velocity (meters / second) differ significantly in the RDM coordinate system; directly calculating Euclidean distance would cause the clustering results to be severely skewed in one dimension. Finally, the signal-to-noise ratio of radar echoes fluctuates wildly, and clustering with a fixed radius often leads to the loss of small targets or the formation of noise clusters.

[0008] In summary, existing ISAC schemes suffer from bottlenecks such as weak waveform anti-Doppler capability, fragmented sensing output, and difficulty in automatically sorting moving and static targets. This invention aims to propose a complete ISAC system based on AFDM. Leveraging the inherent adaptability of AFDM waveforms to highly dynamic environments, and combining it with a DBSCAN clustering algorithm optimized in the RDM domain, a closed-loop processing from the original waveform to the target list is achieved by establishing dynamic threshold logic. This not only improves the robustness of sensing but also solves the engineering challenge of parallel static environment sensing and dynamic target tracking, demonstrating significant practical value. Summary of the Invention

[0009] This invention addresses the technical challenges of existing ISAC (Integrated Communication and Sensing) systems in high-dynamic, high-frequency scenarios, including limited communication rates, scattered sensing points, confusion between dynamic and static targets, and low waveform resource utilization. It proposes a complete integrated processing architecture based on AFDM waveforms. This invention aims to achieve synergy between high-gain communication and refined clustering sensing through deep endogenous fusion of a single waveform, and to solve the problem of automated classification from bottom-level detection points to upper-level target entity recognition.

[0010] The objective of this invention is achieved through the following technical solution: a method for integrated sensory signal generation and multi-target clustering recognition, comprising the following steps: A unified integrated waveform signal is constructed by using an affine Fourier transform orthogonal frequency division multiplexing modulation mechanism. The waveform signal is then modulated into a transmit signal by using post-chirp parameters and pre-chirp parameters for frequency domain double phase modulation and time domain double phase rotation. The communication symbols are recovered using the linear minimum mean square error method. The echo signal reflected by the target is received. The communication signal at the transmitting end is used as a sensing reference template. The received signal is processed in the transform domain by the fast cyclic correlation radar algorithm to generate a range-Doppler heat map. Two-dimensional constant false alarm rate detection is performed on the distance-Doppler heatmap, and the detection threshold is adaptively calculated based on the background noise power. Pixels exceeding the detection threshold are extracted as the original detection point cloud. A density-based spatial clustering algorithm is applied to the extracted raw detection point cloud. A weighted distance metric formula is set according to the resolution of the distance dimension and the velocity dimension. Scattered points belonging to the same physical entity are aggregated into independent target clusters in the physical quantity space. The radial velocity of the centroid of each target cluster is calculated and compared with preset dynamic and static critical velocity thresholds to achieve automatic classification of the detected targets and distinguish between static obstacles and dynamic moving targets.

[0011] Furthermore, the specific method for constructing a unified integrated waveform signal using an affine Fourier transform orthogonal frequency division multiplexing modulation mechanism is as follows: in It is a frequency domain sign vector. It is a DFT matrix. and These are two modulated chirp diagonal matrices, represented as ; and These are the pre-chirp parameters and the post-chirp parameters.

[0012] Furthermore, the specific steps of using post-chirp parameters and pre-chirp parameters for frequency-domain quadratic phase modulation and time-domain quadratic phase rotation are as follows: After channel coding and constellation mapping, the communication data is placed on a grid in the transform domain, and a post-chirp parameter is applied to the symbols in the frequency domain. The precoded symbol stream is transformed to the time domain by a second phase modulation and then converted by an N-point IDFT operator. At the discrete time-domain point n of the IDFT output, a parameter based on the pre-chirp is applied. The time-domain phase is rotated twice; finally, a cyclic prefix is ​​added to the leading edge of the generated discrete time-domain symbol block and modulated onto the transmitting antenna for transmission.

[0013] Furthermore, the pre-chirp parameters The configuration follows these principles: Where N is the total number of subcarriers, The number of taps corresponding to the maximum channel delay spread. This represents the maximum integer Doppler offset supported by the system.

[0014] Furthermore, the generation of the distance-Doppler heatmap specifically includes: The received signal is coherently accumulated by frequency domain multiplication with the locally backed-up transmitted waveform block, and pulse compression is performed to obtain the distance dimension information; In adjacent Perform Discrete Fourier Transform (DFT) between symbols to extract the accumulated Doppler frequency shift of the target and obtain velocity dimension information; By combining windowing to suppress sidelobes, a two-dimensional matrix is ​​finally constructed with radial distance and radial velocity as the coordinate axes.

[0015] Furthermore, the step of setting the weighted distance metric formula based on the resolution of the distance and velocity dimensions specifically includes: in, For distance coordinates, For velocity coordinates, and These are the proportional weighting factors set according to the system's distance resolution and velocity resolution, used to normalize the influence of different dimensions on the clustering results.

[0016] Furthermore, the agglomeration of scattered points belonging to the same physical entity into independent target clusters within the physical quantity space includes: Set search radius and minimum neighborhood number ; Traverse the probe point cloud; if the number of points in the neighborhood of a certain point is not less than... If so, it is determined to be a core point and a new cluster is started; Recursively group all density-accessible core points and their boundary points into the same target cluster, and remove noise interference points.

[0017] Furthermore, the comparison of the radial velocity of the center of mass with preset dynamic and static critical velocity thresholds specifically includes: Extract the first Centroid coordinates of the target cluster ;like If so, the target cluster is classified as a static obstacle; if If it is, then it is determined to be a dynamically moving target; among which, To account for the redundant speed threshold after considering system phase noise and clock jitter.

[0018] Furthermore, the method also includes: while performing sensing processing, the receiver obtains demodulated data by performing an inverse affine transformation on the received signal, and uses the range-Doppler spectrum obtained by sensing to assist in constructing a channel estimation matrix, thereby achieving communication equalization and data recovery.

[0019] According to another aspect of the specification, a sensor-integrated signal generation and multi-target clustering recognition device is also provided, including a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it implements the sensor-integrated signal generation and multi-target clustering recognition method.

[0020] The beneficial effects of this invention are: (1) High spectral efficiency: Since the sensing reference signal and the communication data symbol share the same waveform frame, the present invention does not need to open up an additional sensing silence period or a dedicated sensing frequency band like the traditional system, which significantly improves the spectral efficiency per unit area of ​​the system.

[0021] (2) Extremely high data cleanliness: Compared with directly outputting CFAR detection points, the clustering mechanism of this invention improves the perception accuracy from "pixel level" to "target level", which greatly reduces the computational load of the back-end tracking algorithm.

[0022] (3) Scene Adaptability: The dynamic classification logic allows the system to accurately eliminate stationary interference in strong clutter environments (such as tunnels and dense forests) while retaining sensitivity to low-speed dynamic targets.

[0023] (4) Doppler robustness: The affine transformation characteristics of AFDM ensure that the target detection results will not be affected by energy dissipation due to carrier frequency shift at 24 GHz or even higher frequencies, thus improving the detection probability.

[0024] (5) Breaking through the Doppler extension bottleneck: In mobile communication with relative speeds of up to 300 km / h and above, this invention utilizes the frequency modulation characteristics of AFDM to maintain an extremely low bit error rate, avoiding the problem of frequent communication link interruptions in traditional schemes. Attached Figure Description

[0025] Figure 1 This is a block diagram of the transceiver system of the integrated communication and sensing system in this invention; Figure 2 These are waveforms of AFDM in the time and frequency domains in this invention; Figure 3 This is a time-frequency-space three-dimensional waveform diagram of different subcarriers of the AFDM waveform in this invention; Figure 4 This is a two-dimensional time-frequency-space waveform diagram of different subcarriers of the AFDM waveform in this invention. Figure 5 This is a simulation diagram of the AFDM waveform in this invention regarding communication performance; Figure 6 These are the Doppler frequency shift profile and range profile of the AFDM waveform with respect to sensing in this invention; Figure 7 This is a schematic diagram of the sensing RDM of the AFDM waveform in this invention; Figure 8 This is the system comprehensive simulation and performance analysis panel for the integrated AFDM waveform sensing scenario in this invention; Figure 9 This is a schematic diagram of a sensor-integrated signal generation and multi-target clustering recognition device according to the present invention. Detailed Implementation

[0026] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0027] This invention provides a method and apparatus for integrated sensing signal generation and multi-target clustering recognition, characterized by integrating physical layer AFDM modulation, adaptive FCCR receiver, 2D-CFAR detection, and DBSCAN automatic classification algorithm based on RDM domain optimization. This not only achieves high-precision sensing but also establishes a high-speed communication link at the underlying level.

[0028] This invention uses AFDM as the integrated underlying waveform. Unlike traditional multicarrier technologies, AFDM introduces chirp parameters. and Orthogonality was established in the time-delay-Doppler domain. Communication data is first channel-coded (e.g., LDPC or Polar codes), and then mapped to M-QAM modulation symbols. The input QAM symbol stream... The discrete-time signal is generated by inverse affine Fourier transform (IDAFT). Compared to the IFFT transform of OFDM, this invention introduces a quadratic phase factor. and In the mathematics of communication, this process essentially involves distributing the energy of data symbols along a specific time-frequency slope. Its core operator involves compensation of a quadratic phase term, causing the signal to appear as a chirp cluster with a specific slope in the frequency domain. Simultaneously, through dynamic adjustment... This allows the Doppler frequency shift to produce a controlled linear displacement in the transform domain, thereby eliminating ICI and ensuring communication reliability in high-speed mobile scenarios.

[0029] In terms of sensing, after receiving the echo signal at the sensing end, the system uses the Fast Cyclic Correlation (FCCR) algorithm for processing. The receiver uses the communication symbols from the transmitting end as a "sensing template" and performs point-to-point complex conjugate product in the transform domain. Combined with windowing processing (such as Hamming window) to suppress sidelobes, the time-domain echo is mapped to a range-velocity heatmap (RDM) through a two-dimensional FFT transform. This process utilizes the pulse compression characteristics of the AFDM waveform, which can provide sensing resolution and dynamic range superior to traditional OFDM.

[0030] To extract potential targets from heatmaps, this invention implements a cell-averaged constant false alarm rate (2D-CA-CFAR) detection method with a neighborhood guard band. A reference window and a guard window are set, and the background noise power around the target cell is estimated in real time. Based on preset parameters... The detection threshold is dynamically adjusted. This step outputs a set of "detection point clouds" distributed in the RDM coordinate system, which includes distance coordinates, velocity coordinates, and amplitude information.

[0031] To address the issue that the point clouds output by CFAR are too fragmented and cannot correspond to physical entities, this invention proposes a DBSCAN algorithm that operates in the RDM physical quantity space. Since the distance dimension and velocity dimension have different dimensions, this invention introduces a weight matrix. When calculating the distance between points, a weighted Euclidean distance is used: The system iterates through all detection points by defining a search radius. and minimum neighborhood number If the number of points in the neighborhood of a certain point exceeds... This is defined as a "core point of an entity," and core points with mutually accessible densities are aggregated into a single target cluster. This process effectively filters out isolated noise points and aggregates discrete echo points belonging to the same car or obstacle into a single target.

[0032] After acquiring the target clusters, the system executes feature extraction logic to distinguish between "environmental background" and "moving target." Specifically, it extracts the center coordinates of each cluster. The system records classification results for multiple consecutive frames. If the average radial velocity of a target remains below a threshold, it is marked as a "Static Obstacle"; otherwise, it is marked as a "Moving Target." The real-time situation is displayed in the UI using different colors and icons.

[0033] In communications, traditional OFDM communication subcarriers suffer from severe inter-carrier interference (ICI) in high-Doppler environments due to the challenges posed by high-frequency bands (such as millimeter waves) and high dynamics (such as high-speed vehicle travel). This invention addresses this by adjusting the affine parameters of AFDM, ensuring that the Doppler shift at the receiver is represented as an integer multiple of the subcarrier index cyclic shift. The original information can be recovered using a simple frequency domain equalizer, effectively improving bit error rate (BER) performance in high-speed scenarios. Furthermore, to combat the delay spread of the wireless channel, this invention adds a length of [missing information - likely a unit of length] before each AFDM symbol block. The cyclic prefix. Through proper configuration... This allows the system to fully cover the maximum delay of multipath channels while maintaining high spectral efficiency.

[0034] refer to Figure 1 As shown, the specific implementation steps of the present invention are as follows: (1) First, the foundation of AFDM consists of AFT and DAFT. AFT, also known as linear canonical transform or generalized Fresnel transform, is a continuous transform that maps a continuous-time signal s(t) to S(f), as shown below: Here, it is assumed that the parameters (a, b, c, d) form an invertible matrix. Its determinant should be equal to 1. The AFT is an integral transform that encompasses many standard transforms, such as the Fourier transform (0, 1, -1, 0), the Laplace transform (0,j, j, 0), and... Fractional Fourier Transform The Gauss-Weierstrass, Fresnel, and Bargmann transforms are also special cases.

[0035] First, the system sets the basic physical parameters. Let the carrier frequency be... The number of subcarriers is N, and the number of symbols (time slots) is The uniqueness of AFDM waveforms lies in the introduction of two affine parameters. and These factors determine the tilt of the signal in the time-frequency plane: in It is a frequency domain sign vector. It is a DFT matrix of length L. and It consists of two modulated chirp diagonal matrices, represented as .

[0036] Based on the Doppler coverage requirements in high dynamic scenarios, parameters The value of is a sufficiently small irrational number, and the parameter in this invention... The value is ,parameter The design formula is: in, This represents the normalized number of taps corresponding to the maximum channel delay spread. This is the integer offset corresponding to the maximum Doppler frequency shift. This parameter ensures the orthogonality of different time delays and Doppler components in the transform domain, avoiding energy aliasing.

[0037] (2) Waveform Transmission and Wireless Channel Propagation: At the transmitting end, the AFDM integrated data stream with added guard intervals enters the wireless space. During channel transmission, which features delay spread and Doppler spread, the signal undergoes waveform distortion. The echo signal reflected from the target, after being superimposed with channel noise and environmental interference, is discretized by the receiver front end. Communication Data Symbols (The complex symbol after QAM modulation) is mapped onto At the time-frequency grid points. The modulation process of AFDM is characterized by the following equation: in, For time-domain discrete indexes, This is the transform domain index.

[0038] The formula first performs a frequency domain symbol based on... The second phase encoding is then transformed to the time domain via IDFT, and finally a time-domain-based... The time-domain phase is rotated twice. This makes the AFDM signal essentially a set of orthogonal linear frequency modulated (Chirp) signals. To combat inter-symbol interference (ISI) caused by multipath channels, a length of [missing information] is added before each symbol block. The cyclic prefix is ​​used to obtain the final transmitted signal. .

[0039] (3) Receiver signal processing and functional decoupling: After acquiring the signal, the receiver's primary task is to strip away protection components such as the cyclic prefix in order to extract clean, effective data payload. Subsequently, based on a predefined demodulation and sensing link, the system performs functional decoupling processing for communication information restoration and radar feature extraction: (3.1) The communication section uses DAFT to transform the received signal and the Least Mean Square Error (LMMSE) algorithm for demodulation. Specifically, the user-received signal is represented as: in It is the normalized channel gain coefficient for the p-th path, and we have . It is additive white Gaussian noise. and It is the normalized time delay and Doppler frequency offset. This is the number of subcarriers. After removing the CPP of the received signal and performing a DAFT transform, the result is obtained. in , Represents the DAFT matrix. It is the conjugate transpose of the transformation matrix. This represents the impact matrix / truncation matrix related to the removal of CPP operations. This represents the diagonal phase rotation matrix caused by the Doppler frequency offset. Represents normalized delay The resulting time delay shift matrix. The matrix after DAFT transformation can be represented as: Then for the linear relationship ,in The data estimated using LMMSE can be expressed as follows: Obtain a soft estimate of the transmitted data symbols. The receiver will then use this data for subsequent demodulation and decoding operations. Specifically, First, the data is fed into the symbol demapper, and then this soft information is sent to the channel decoder for error correction, ultimately recovering the original transmitted bit stream. Simultaneously, during the simulation phase, this data is also used to calculate the system's bit error rate (BER) to evaluate the receive performance.

[0040] (3.2) Sensing part of the echo signal After being transmitted through the channel, the signal is captured by the receiver. The core task of the sensing module is to utilize the known transmitted signal. With received signal Perform cross-correlation calculations.

[0041] The system uses the FCCR algorithm for distance estimation. First, it processes the data in the transform domain: in, Represents complex conjugation.

[0042] Pulse compression was achieved by performing complex conjugate multiplication in the frequency domain and returning to the time domain. The row indices of the matrix directly correspond to the target's time delay information (Range Bins).

[0043] Perform an FFT transformation on the Z matrix along the sign dimension (column dimension): in A windowing function is applied. This step generates a distance-Doppler heatmap (RDM). Peak energy coordinates on the heatmap are shown. Physical properties of the corresponding target: Distance calculation: Speed ​​calculation: (4) Two-dimensional constant false alarm rate (CFAR) detection: To automatically extract target points, this implementation uses the cell average constant false alarm rate (CA-CFAR) algorithm. For each cell in the RDM... The logic behind the judgment is as follows: in For the number of reference units, It is based on the preset false alarm rate Calculated threshold factor.

[0044] when At this point, the point is identified as a potential target detection point. This step transforms the continuous heatmap into a discrete "detection point cloud".

[0045] (5) DBSCAN clustering in physical quantity space: Since a physical target (such as a large truck) may generate multiple CFAR detection points, this invention uses the DBSCAN algorithm to cluster the points.

[0046] To eliminate the influence of the different dimensions of distance units (m) and speed units (m / s), a weighted Euclidean distance is defined: in and Here, represents the system's distance and velocity resolution, respectively, and in the formula, i and j represent the i-th and j-th CFAR detectors in the dataset, respectively.

[0047] Definition for a point If it is within the radius The number of points inside is greater than ,but Using the core point as the core, all density-reachable core points and their neighboring points are grouped into the same cluster. This process utilizes the principle that "entity targets have local density characteristics in the RDM domain" to aggregate scattered points belonging to the same target, while identifying and eliminating low-density random noise as interference.

[0048] In obtaining clusters Then, the system calculates the average radial velocity of each cluster: The classification logic is set as follows: Static target recognition: If If the speed is 1.0 m / s, then the target corresponding to the cluster is marked as "Static Obstacle".

[0049] Dynamic target recognition: If If so, it is marked as "Moving Target".

[0050] This step utilizes the high precision of AFDM in the Doppler dimension, which can effectively separate the target from the stationary background even if the target moves at an extremely slow speed.

[0051] (6) System simulation and performance comparison verification: With a fixed channel, the communication bit error rate (BER) and sensing accuracy of the AFDM waveform are tested to verify the effectiveness and superiority of the invention in a complex environment with high dynamics and high interference.

[0052] The functions and effects of this invention are further illustrated and demonstrated through the following simulation experiments: Figure 2 This intuitively illustrates the physical characteristic that the subcarriers are mutually orthogonal under AFDM modulation. From the perspective of time-domain discrete sampling, the expression for the nth subcarrier at the kth discrete sampling point is: when Then, continue to examine whether the inner product between any two subcarriers in the AFDM waveform is 0: After sorting, we can obtain Therefore, it can be proven that the orthogonality between any two subcarriers in AFDM is satisfied: Figure 3 and Figure 4 Three-dimensional time-frequency energy distribution plots of four representative AFDM subcarrier basis functions with indices 2, 7, 12, and 17 are presented. As can be seen from the plots, each subcarrier basis function does not present a flat rectangular grid on the time-frequency plane like traditional OFDM, but rather exhibits a chirp shape with a specific slope. The four subplots show the translation of the basis functions with different indices along the frequency axis on the time-frequency plane. Although they overlap in both the time and frequency domains, the subcarriers remain strictly orthogonal in the transform domain due to the phase orthogonality resulting from their affine transformation.

[0053] The two-dimensional top-down time-frequency evolution plot of the aforementioned basis functions more clearly depicts the trajectory of energy distribution, with bright areas representing concentrated energy regions. It can be observed that the energy trajectories of all subcarriers exhibit parallel, sloping straight lines. This slope is determined by the core parameters of AFDM. The mathematical essence of this decision is the rotation angle of the waveform in the time-frequency plane. This tilted energy distribution means that when there is a Doppler shift in the wireless channel, the frequency shift is transformed into a cyclic displacement of the subcarrier in the transform domain, rather than directly causing ICI as in OFDM. As the subcarrier index increases (from 2 to 17), the energy center shifts sequentially along the frequency axis, which verifies that AFDM can fully utilize the entire system bandwidth to achieve full diversity gain.

[0054] Figure 5The reliability of the communication link was demonstrated, and the communication performance of AFDM as the unified waveform of ISAC was verified. The bit error rate performance was presented as a curve showing the system's bit error rate (BER) as a function of signal-to-noise ratio (SNR). It can be seen that as SNR increases, the BER exhibits a smooth and rapid decreasing trend. When the SNR reaches approximately 30 dB, the bit error rate has decreased to 10. -4 The following demonstrates that the AFDM receiver proposed in this invention can effectively suppress channel fading and noise.

[0055] The received constellation diagram shows the received signal constellation at SNR = 40dB. It can be seen that the 16 signal points of 16-QAM are clearly clustered with distinct boundaries, without obvious phase rotation or severe amplitude distortion. This indicates that even in complex ISAC scenarios, the original communication data can still be accurately recovered.

[0056] Figure 6 The system's range sensing performance and velocity (Doppler) sensing performance are demonstrated. The range sensing performance graph shows the power spectrum distribution (Distance Profile) of the system along the range dimension, used to verify ranging accuracy. The blue solid line in the graph represents the amplitude response curve after processing by the sensing algorithm. It can be seen that a sharp pulse peak appears on the curve near the left side (starting point) of the range axis. The red dashed line in the graph represents the true actual distance to the target, and the green dotted line represents the estimated value calculated by the system algorithm. The green estimated line almost completely overlaps with the red true value line, and the sharp peak indicates that this invention utilizes the pulse compression characteristics of AFDM waveforms to accurately locate the physical distance of the target from background noise, with extremely low time delay estimation error.

[0057] The velocity-axis (Doppler) sensing performance plot shows the power spectrum distribution of the system along the velocity dimension at a determined target distance. The horizontal axis represents the radial velocity of the target, and the vertical axis represents the normalized amplitude. Near 30 m / s, the blue profile curve exhibits extremely strong energy gain, forming a single, distinct spectral peak. Similarly, the true value of the red dashed line and the estimated value of the green dotted line perfectly match at the 30 m / s position. This demonstrates that the AFDM waveform passes through the affine parameters. The introduced linear frequency modulation characteristic can effectively resist Doppler spread even in complex wireless channels, enabling fine-grained detection of the velocity of highly dynamic moving targets.

[0058] Figure 7The 3D RDM map generated by the system after processing with the FCCR algorithm is shown, used to verify its multi-target detection and point extraction capabilities. The vertical axis represents the normalized amplitude, and the horizontal planes represent distance and velocity, respectively. Three significant energy peaks are clearly visible, corresponding to three different physical targets in the scene. The red dots represent the target estimation centers determined by the system based on the energy peaks. The green plus signs represent the original detected point cloud extracted after 2D-CFAR detection. The results demonstrate that this invention can effectively distinguish multiple targets from complex noisy backgrounds and provide accurate candidate points for the subsequent DBSCAN clustering algorithm.

[0059] Figure 8 Composed of four subgraphs, the system's effectiveness was verified from three dimensions: spatial location, parameter clustering, and motion characteristics. The physical scene modeling diagram (top left) shows the relative positions of the transmitting base station (green star) and multiple targets. The diagram includes three dynamic targets (T1, T2, and T3, represented by colored squares and dashed tracks, respectively) and two static obstacles (S1 and S2, represented by gray rhombuses). The tracks reflect the real-time motion vectors of the dynamic targets in two-dimensional space (XY), providing a physical truth reference for subsequent perception algorithms.

[0060] The clustering and static / dynamic classification results (top right) show the processing results of the range-radial velocity (Range-Radial Vel) plane, with the background being a heatmap processed by FCCR. It can be seen that after the DBSCAN clustering algorithm, the original detection points are clustered into independent target entities. The target attributes are automatically labeled using borders of different colors and shapes: magenta boxes represent identified dynamic targets, distributed in the non-zero velocity region; cyan rhombuses represent identified static targets, precisely aligned on the axis with a velocity of 0 m / s; the red crosses highly coincide with the cluster centers, verifying the system's high accuracy in classifying dynamic and static targets.

[0061] The dynamic target rate tracking plot (bottom left) shows the system's real-time estimation results of the rates of three dynamic targets over time. The solid line represents the true values, and the scatter points represent the estimated values. It can be seen that even in the presence of noise interference, the system can still stably track the rate changes of the targets, with the error remaining within an extremely low range, demonstrating the high-resolution characteristics of the AFDM waveform at the Doppler frequency.

[0062] The dynamic target heading angle tracking graph (bottom right) shows the curve of the target heading angle changing over time. This graph demonstrates that the system can accurately capture sudden changes in the target's heading during travel (such as a 90-degree turn).

[0063] Corresponding to the aforementioned embodiment of a synesthetic signal generation and multi-target clustering recognition method, the present invention also provides an embodiment of a synesthetic signal generation and multi-target clustering recognition device.

[0064] See Figure 9 The present invention provides a sensory integrated signal generation and multi-target clustering recognition device, including a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it is used to implement a sensory integrated signal generation and multi-target clustering recognition method in the above embodiment.

[0065] The embodiment of the integrated sensing signal generation and multi-target clustering recognition device provided by this invention can be applied to any device with data processing capabilities, such as a computer. The device embodiment can be implemented through software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of any data processing device loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 9 The diagram shown is a hardware structure diagram of any device with data processing capabilities, which is an integrated sensor signal generation and multi-target clustering recognition device provided by the present invention. Except for... Figure 9 In addition to the processor, memory, network interface, and non-volatile memory shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0066] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0067] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the present invention according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0068] This invention also provides a computer-readable storage medium storing a program thereon, which, when executed by a processor, implements a method for generating integrated sensor signals and multi-target clustering recognition as described in the above embodiments.

[0069] The computer-readable storage medium can be an internal storage unit of any data processing device described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device of any data processing device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of any data processing device. The computer-readable storage medium is used to store the computer program and other programs and data required by the data processing device, and can also be used to temporarily store data that has been output or will be output.

[0070] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements the aforementioned method for integrated sensor signal generation and multi-target clustering recognition.

[0071] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this application are indicated by the claims.

[0072] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this application. This application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for integrated sensory signal generation and multi-target clustering recognition, characterized in that, Includes the following steps: A unified integrated waveform signal is constructed by using an affine Fourier transform orthogonal frequency division multiplexing modulation mechanism. The waveform signal is then modulated into a transmit signal by using post-chirp parameters and pre-chirp parameters for frequency domain double phase modulation and time domain double phase rotation. The communication symbols are recovered using the linear minimum mean square error method. The echo signal reflected by the target is received. The communication signal of the transmitting end is used as the sensing reference template. The received signal is processed in the transform domain by the fast cyclic correlation radar algorithm to generate a range-Doppler heat map. Two-dimensional constant false alarm rate detection is performed on the distance-Doppler heatmap, and the detection threshold is adaptively calculated based on the background noise power. Pixels exceeding the detection threshold are extracted as the original detection point cloud. A density-based spatial clustering algorithm is applied to the extracted raw detection point cloud. A weighted distance metric formula is set according to the resolution of the distance dimension and the velocity dimension. Scattered points belonging to the same physical entity are aggregated into independent target clusters in the physical quantity space. The radial velocity of the centroid of each target cluster is calculated and compared with preset dynamic and static critical velocity thresholds to achieve automatic classification of the detected targets and distinguish between static obstacles and dynamic moving targets.

2. The method for integrated sensory signal generation and multi-target clustering recognition according to claim 1, characterized in that, The specific method for constructing a unified integrated waveform signal using an affine Fourier transform orthogonal frequency division multiplexing modulation mechanism is as follows: in It is a frequency domain sign vector. It is a DFT matrix. and These are two modulated chirp diagonal matrices, represented as ; and These are the pre-chirp parameters and the post-chirp parameters.

3. The method for integrated sensory signal generation and multi-target clustering recognition according to claim 1, characterized in that, The specific steps for using post-chirp parameters and pre-chirp parameters to perform frequency-domain secondary phase modulation and time-domain secondary phase rotation are as follows: After channel coding and constellation mapping, the communication data is placed on a grid in the transform domain, and a post-chirp parameter is applied to the symbols in the frequency domain. The precoded symbol stream is transformed to the time domain by a second phase modulation and then converted by an N-point IDFT operator. At the discrete time-domain point n of the IDFT output, a parameter based on the pre-chirp is applied. The time-domain phase is rotated twice; finally, a cyclic prefix is ​​added to the leading edge of the generated discrete time-domain symbol block and modulated onto the transmitting antenna for transmission.

4. The method for integrated sensory signal generation and multi-target clustering recognition according to claim 1, characterized in that, The pre-chirp parameter The configuration follows these principles: Where N is the total number of subcarriers, The number of taps corresponding to the maximum channel delay spread. This represents the maximum integer Doppler offset supported by the system.

5. The method for integrated sensory signal generation and multi-target clustering recognition according to claim 1, characterized in that, The generation of the distance-Doppler heatmap specifically includes: The received signal is coherently accumulated by frequency domain multiplication with the locally backed-up transmitted waveform block, and pulse compression is performed to obtain the distance dimension information; In adjacent Perform Discrete Fourier Transform (DFT) between symbols to extract the accumulated Doppler frequency shift of the target and obtain velocity dimension information; By combining windowing to suppress sidelobes, a two-dimensional matrix is ​​finally constructed with radial distance and radial velocity as the coordinate axes.

6. The method for integrated sensory signal generation and multi-target clustering recognition according to claim 1, characterized in that, The specific steps of setting the weighted distance metric formula based on the resolution of the distance and velocity dimensions include: in, For distance coordinates, For velocity coordinates, and These are the proportional weighting factors set according to the system's distance resolution and velocity resolution, used to normalize the influence of different dimensions on the clustering results.

7. The method for integrated sensory signal generation and multi-target clustering recognition according to claim 1, characterized in that, The process of aggregating scattered points belonging to the same physical entity into independent target clusters within the physical quantity space includes: Set search radius and minimum neighborhood number ; Traverse the probe point cloud; if the number of points in the neighborhood of a certain point is not less than... If so, it is determined to be a core point and a new cluster is started; Recursively group all density-accessible core points and their boundary points into the same target cluster, and remove noise interference points.

8. The method for integrated sensory signal generation and multi-target clustering recognition according to claim 1, characterized in that, The comparison of the radial velocity of the center of mass with the preset dynamic and static critical velocity thresholds specifically includes: Extract the first Centroid coordinates of the target cluster ;like If so, the target cluster is classified as a static obstacle; if If it is, then it is determined to be a dynamically moving target; among them, To account for the redundant speed threshold after considering system phase noise and clock jitter.

9. A sensor-integrated signal generation and multi-target clustering recognition device, comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that, When the processor executes the executable code, it implements a method for integrated sensor signal generation and multi-target clustering recognition as described in any one of claims 1-8.

Images