Ultra-wideband pulse waveform space-time synchronization compression method and system, medium and product

Abstract

The invention relates to the technical field of signal compression, in particular to an ultra-wideband pulse waveform space-time synchronous compression method and system, a medium and a product. The method comprises the following steps: executing multi-dimensional orthogonal projection on an original ultra-wideband pulse signal by using a spatio-temporal joint perception dictionary to extract a spatio-temporal characteristic coefficient; adjusting an observation matrix according to the channel state information, and performing compressed sampling on the spatial-temporal characteristic coefficient by using the adjusted observation matrix to generate a compressed observation vector; performing manifold learning dimensionality reduction mapping on the compressed observation vector to construct a space-time manifold topological structure, and obtaining manifold feature coordinates based on the space-time manifold topological structure; introducing weighted least square constraint to carry out nonlinear reconstruction on the flow shape characteristic coordinates to generate a sparse pulse sequence; and performing variational Bayesian iterative reasoning on the sparse pulse sequence to obtain an optimal quantization parameter, and performing entropy coding on the sparse pulse sequence by using the optimal quantization parameter to generate a transmission code stream. The method can guarantee that the details of the pulse signals are complete and distortionless.

Description

Technical Field

[0001] This invention relates to the field of signal compression technology, specifically to an ultra-wideband pulse waveform spatiotemporal synchronous compression method, system, medium, and product. Background Technology

[0002] With the rapid evolution of fifth-generation mobile communication technology and next-generation wireless communication systems, ultra-wideband pulse radio technology, with its extremely high transmission rate, centimeter-level positioning accuracy, and excellent anti-interference performance, has shown broad prospects in fields such as high-speed sensor networks, precise ground-penetrating imaging, and panoramic industrial monitoring. However, ultra-wideband pulse signals have extremely wide spectral bandwidth and nanosecond or even picosecond-level narrow pulse characteristics in the time domain. According to the Nyquist sampling theorem, direct digital acquisition of them requires a sampling rate on the order of gigahertz, which places extremely stringent requirements on the hardware bandwidth, sampling accuracy, and throughput capacity of the analog-to-digital converter and subsequent data processing units.

[0003] To overcome this sampling bottleneck, compressed sensing theory is widely used in existing technologies. This theory leverages the sparsity of signals to achieve data acquisition and compression at rates far below the Nyquist frequency. Conventional compressed sensing methods typically use a pre-set fixed observation matrix (such as a random Gaussian matrix or Bernoulli matrix) to process the projected low-dimensional observation vector, focusing on recovering the original ultra-wideband pulse signal through algorithms during the reconstruction stage. However, in real-world complex electromagnetic transmission environments, ultra-wideband pulse signals often exhibit complex statistical characteristics that are non-stationary, non-Gaussian, and time-varying in terms of space-frequency correlation. Their inherent structural information is not only contained within the sparse domain but also hidden within the geometric topology of the high-dimensional data manifold.

[0004] Existing technologies typically treat observation data as a discrete set of points in Euclidean space, employing linear dimensionality reduction and processing methods. This ignores the nonlinear manifold structure distribution characteristics of signals exhibited under channel fading conditions such as multipath effects and Doppler shift. This processing approach often struggles to achieve optimal matching between the observation matrix and the sparse characteristics of the channel when the channel state changes dynamically. Consequently, the observation data contains a large amount of redundant environmental noise and lacks sufficient essential signal characteristics. Furthermore, during linear reconstruction, noise amplification easily leads to distortion of details in weak pulse waveforms.

[0005] Therefore, it is necessary to improve one or more of the problems existing in the above-mentioned related technical solutions.

[0006] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] This disclosure proposes an ultra-wideband pulse waveform spatiotemporal synchronization compression method and related apparatus, aiming to overcome at least one of the defects existing in the prior art.

[0008] To achieve the above objectives, the technical solution disclosed in this invention is as follows: According to a first aspect of the present disclosure, a method for spatiotemporal synchronization compression of ultra-wideband pulse waveforms is provided, comprising the following steps: A spatiotemporal joint sensing dictionary is constructed, and the original ultrawideband pulse signal is subjected to multidimensional orthogonal projection using the spatiotemporal joint sensing dictionary to extract spatiotemporal feature coefficients. The row vector distribution of the observation matrix is ​​adjusted according to the channel state information, and the spatiotemporal feature coefficients are compressed and sampled using the adjusted observation matrix to generate a compressed observation vector. Manifold learning dimensionality reduction mapping is performed on the compressed observation vector to construct a spatiotemporal manifold topology, and manifold feature coordinates are obtained based on the spatiotemporal manifold topology; A weighted least squares constraint is introduced to nonlinearly reconstruct the feature coordinates of the manifold, generating a sparse pulse sequence. Variational Bayesian iterative inference is performed on the sparse pulse sequence to obtain the optimal quantization parameters, and the optimal quantization parameters are used to entropy encode the sparse pulse sequence to generate a transmission code stream.

[0009] In one exemplary embodiment of this disclosure, the step of constructing a spatiotemporal joint sensing dictionary and using the spatiotemporal joint sensing dictionary to perform multidimensional orthogonal projection on the original ultrawideband pulse signal to extract spatiotemporal feature coefficients includes: Acquire raw ultrawideband pulse signals and extract their covariance matrices in the time and spatial dimensions; Construct a KL transform basis based on the diagonalized eigenvectors of the covariance matrix; The Gabor atom library is subjected to tensor product operation with the KL transform basis to generate the spatiotemporal joint sensing dictionary; The original ultrawideband pulse signal is sparsely decomposed using the spatiotemporal joint sensing dictionary to calculate the matching coefficients of each atom in the Gabor atom library; Atom combinations with matching coefficients greater than a preset threshold are selected to generate the spatiotemporal feature coefficients.

[0010] In an exemplary embodiment of this disclosure, the step of adjusting the row vector distribution of the observation matrix according to channel state information, and using the adjusted observation matrix to compress and sample the spatiotemporal feature coefficients to generate a compressed observation vector includes: The feedback information received from the channel includes multipath delay spread and Doppler frequency shift parameters; The time-domain sparsity of the channel is calculated based on the multipath delay spread, and the frequency-domain sparsity of the channel is calculated based on the Doppler frequency shift parameter. A cost function with the channel state information as a parameter is constructed, and the row vector weights of the observation matrix are iteratively updated using the gradient descent method. The spatiotemporal feature coefficients are weighted and projected using the updated observation matrix to generate the compressed observation vector.

[0011] In an exemplary embodiment of this disclosure, the step of performing manifold learning dimensionality reduction mapping on the compressed observation vector to construct a spatiotemporal manifold topology, and obtaining manifold feature coordinates based on the spatiotemporal manifold topology, includes: Construct a neighborhood graph of the compressed observation vector and calculate the geodesic distance between each data point in the neighborhood graph; The neighborhood graph is decomposed using the Laplacian feature map algorithm to obtain a low-dimensional embedding vector. The low-dimensional embedding vector is mapped to the Riemannian manifold space, and the principal curvature directions of the Riemannian manifold space are calculated. The data points in the neighborhood graph are rearranged along the principal curvature direction to generate the spatiotemporal manifold topology. The coordinates of stable nodes in the spatiotemporal manifold topology are extracted and used as the manifold feature coordinates.

[0012] In one exemplary embodiment of this disclosure, the step of introducing weighted least squares constraints to nonlinearly reconstruct the manifold feature coordinates and generate a sparse pulse sequence includes: Establish an objective optimization function that includes signal sparsity constraints and data fidelity constraints; A weighted least squares term is introduced into the objective optimization function, and the weight coefficients of the weighted least squares term are adjusted according to the energy density of the local region of the original ultra-wideband pulse signal. The objective optimization function is solved iteratively using the conjugate gradient method to obtain the optimal sparse solution; Based on the optimal sparse solution, the pulse amplitude parameters and time delay parameters of the corresponding original ultra-wideband pulse signal are restored, and the sparse pulse sequence is generated.

[0013] In one exemplary embodiment of this disclosure, the step of performing variational Bayesian iterative inference on the sparse pulse sequence to obtain optimal quantization parameters, and using the optimal quantization parameters to perform entropy coding on the sparse pulse sequence to generate a transmission bitstream, includes: A prior distribution model for initializing quantization parameters, wherein the prior distribution model includes the quantization step size and the probability density function; Using the sparse pulse sequence as observation data, the posterior probability distribution of the quantization parameters is calculated using the variational Bayesian method; Based on the principle of maximizing the expected value of the posterior probability distribution, the quantization parameters are iteratively updated until convergence to obtain the optimal quantization parameters; The sparse pulse sequence is scalar quantized and arithmetic encoded using the optimal quantization parameters to generate the transmission code stream.

[0014] In one exemplary embodiment of this disclosure, performing manifold learning dimensionality reduction mapping on the compressed observation vector further includes the following steps: Calculate the k nearest neighbors for each sample point and construct a local reconstruction weight matrix; The local reconstruction weight matrix is ​​solved by minimizing the reconstruction error; The high-dimensional data is embedded into the low-dimensional space using the local reconstruction weight matrix, and a regularization term is introduced during the embedding process to control the sparse distribution density of the sample points in order to generate the spatiotemporal manifold topology.

[0015] According to a second aspect of the present disclosure, an ultra-wideband pulse waveform spatiotemporal synchronization compression system is provided. The ultra-wideband pulse waveform spatiotemporal synchronization compression system is used to implement the ultra-wideband pulse waveform spatiotemporal synchronization compression method as described in any of the preceding claims. The ultra-wideband pulse waveform spatiotemporal synchronization compression system includes: The spatiotemporal feature extraction module is used to construct a spatiotemporal joint sensing dictionary and use the spatiotemporal joint sensing dictionary to perform multidimensional orthogonal projection on the original ultrawideband pulse signal to extract spatiotemporal feature coefficients. An adaptive compressed sampling module is used to adjust the row vector distribution of the observation matrix according to the channel state information, and to compress and sample the spatiotemporal feature coefficients using the adjusted observation matrix to generate a compressed observation vector. The manifold topology processing module is used to perform manifold learning dimensionality reduction mapping on the compressed observation vector to construct a spatiotemporal manifold topology structure, and obtain manifold feature coordinates based on the spatiotemporal manifold topology structure; The nonlinear reconstruction module is used to introduce weighted least squares constraints to nonlinearly reconstruct the manifold feature coordinates and generate a sparse pulse sequence. The Bayesian inference coding module is used to perform variational Bayesian iterative inference on the sparse pulse sequence to obtain the optimal quantization parameters, and to use the optimal quantization parameters to perform entropy coding on the sparse pulse sequence to generate a transmission code stream.

[0016] According to a third aspect of the present disclosure, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the ultrawideband pulse waveform spatiotemporal synchronization compression method as described in any of the preceding claims.

[0017] According to a fourth aspect of the present disclosure, an electronic product is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the ultra-wideband pulse waveform spatiotemporal synchronization compression method as described in any of the preceding claims.

[0018] The beneficial effects of this invention are as follows: This invention utilizes a spatiotemporal joint sensing dictionary to perform multidimensional orthogonal projection on the original ultra-wideband pulse signal, capturing the transient abrupt changes in the time domain and the statistical correlation of the spatial dimension. Based on channel state information, it adjusts the row vector distribution of the observation matrix in real time, enabling the observation matrix to automatically match the sparsity characteristics of the current channel. This maximizes the acquisition information entropy while reducing the sampling rate, avoiding performance redundancy or information loss of the observation matrix under dynamic channels. Furthermore, by performing manifold learning dimensionality reduction mapping on the compressed observation vector and obtaining manifold feature coordinates based on the constructed spatiotemporal manifold topology, it can deeply mine the inherent low-dimensional geometric manifold distribution of high-dimensional data, effectively removing transmission environment noise. By introducing weighted least squares constraints to nonlinearly reconstruct the manifold feature coordinates, a sparse pulse sequence is generated, which can correct waveform distortion caused by compressed sampling. By performing variational Bayesian iterative inference on the sparse pulse sequence to obtain the optimal quantization parameters, and using the optimal quantization parameters to entropy encode the sparse pulse sequence, it can further approach the source entropy limit, achieving minimization of transmission delay under band-limited conditions. The spatiotemporal synchronization compression method for ultra-wideband pulse waveforms proposed in this invention overcomes the limitations of existing compressed sensing technology in processing non-stationary ultra-wideband pulse signals. While ensuring the integrity and distortion-free nature of ultra-wideband pulse signal details, it minimizes transmission delay under band-limited conditions and can adapt to the transmission requirements of ultra-wideband pulse signals in band-limited transmission scenarios.

[0019] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Attached Figure Description

[0020] Figure 1 This is a flowchart of the ultra-wideband pulse waveform spatiotemporal synchronization compression method of the present invention; Figure 2 This is a schematic diagram of the original ultra-wideband pulse signal of the present invention; Figure 3 This is a schematic diagram of the spatiotemporal joint sensing dictionary of the present invention; Figure 4 This is a schematic diagram of the spatiotemporal characteristic coefficients of the present invention; Figure 5 This is a schematic diagram of the covariance matrix of the present invention; Figure 6 This is a schematic diagram of the signal reconstruction and residual analysis of the present invention; Figure 7 This is a schematic diagram of the time-frequency sparsity distribution of the present invention; Figure 8 This is a schematic diagram illustrating the gradient descent optimization of the observation matrix according to the present invention; Figure 9 This is a schematic diagram illustrating the information entropy collected in this invention; Figure 10 This is a schematic diagram of the high-dimensional compressed observation space of the present invention; Figure 11 This is a schematic diagram of the neighborhood map and the distance between the geodesic lines in this invention; Figure 12 This is a schematic diagram of the Laplace feature mapping of the present invention; Figure 13 This is a schematic diagram of the energy density-based adaptive weighted reconfiguration mechanism of the present invention; Figure 14 This is a schematic diagram illustrating the convergence of the variational Bayesian iterative parameters of the present invention. Figure 15 This is a schematic diagram comparing the entropy coding efficiency of the present invention; Figure 16 This is a schematic diagram comparing the waveform spectrum recovery fidelity of the original ultra-wideband pulse signal of the present invention. Detailed Implementation

[0021] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0022] In embodiments of the present invention, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" or "for example" in embodiments of the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Rather, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0023] This example implementation first provides a method for spatiotemporal synchronization compression of ultra-wideband pulse waveforms, such as... Figure 1 As shown, the ultra-wideband pulse waveform spatiotemporal synchronization compression method provided in this embodiment includes the following steps S1 to S5.

[0024] Step S1: Construct a spatiotemporal joint sensing dictionary, and use the spatiotemporal joint sensing dictionary to perform multidimensional orthogonal projection on the original ultrawideband pulse signal to extract spatiotemporal feature coefficients, which are used to achieve signal dimensionality reduction while preserving the fine structure of the pulse waveform of the original ultrawideband pulse signal.

[0025] Step S2: Adjust the row vector distribution of the observation matrix according to the channel state information, and use the adjusted observation matrix to compress and sample the spatiotemporal feature coefficients to generate a compressed observation vector, which is used to match the channel sparsity to maximize the entropy of the collected information.

[0026] Step S3: Perform manifold learning dimensionality reduction mapping on the compressed observation vector to construct a spatiotemporal manifold topology, and obtain manifold feature coordinates based on the spatiotemporal manifold topology to remove transmission environmental noise.

[0027] Step S4: Introduce weighted least squares constraints to nonlinearly reconstruct the manifold feature coordinates and generate a sparse pulse sequence to correct waveform distortion caused by compressed sampling.

[0028] Step S5: Perform variational Bayesian iterative inference on the sparse pulse sequence to obtain the optimal quantization parameters, and use the optimal quantization parameters to entropy encode the sparse pulse sequence to generate a transmission code stream, which is used to minimize the transmission delay under band-limited conditions.

[0029] The ultra-wideband pulse waveform spatiotemporal synchronous compression method provided in this embodiment utilizes a spatiotemporal joint sensing dictionary to perform multidimensional orthogonal projection on the original ultra-wideband pulse signal. This captures the transient change details in the time domain and the statistical correlation of the spatial dimension of the original ultra-wideband pulse signal. Based on channel state information, the row vector distribution of the observation matrix is ​​adjusted in real time, enabling the observation matrix to automatically match the sparsity characteristics of the current channel. This maximizes the acquisition information entropy while reducing the sampling rate, avoiding performance redundancy or information loss of the observation matrix under dynamic channels. Furthermore, by performing manifold learning dimensionality reduction mapping on the compressed observation vector and obtaining manifold feature coordinates based on the constructed spatiotemporal manifold topology, the low-dimensional geometric manifold distribution inherent in high-dimensional data can be deeply mined, effectively removing transmission environment noise. By introducing weighted least squares constraints to nonlinearly reconstruct the manifold feature coordinates, a sparse pulse sequence is generated, which can correct the waveform distortion caused by compressed sampling. By performing variational Bayes iterative inference on the sparse pulse sequence to obtain the optimal quantization parameters and using the optimal quantization parameters to entropy encode the sparse pulse sequence, the source entropy limit can be further approximated, achieving minimization of transmission delay under band-limited conditions.

[0030] The spatiotemporal synchronization compression method for ultra-wideband pulse waveforms proposed in this invention overcomes the limitations of existing compressed sensing technology in processing non-stationary ultra-wideband pulse signals. While ensuring the integrity and distortion-free nature of ultra-wideband pulse signal details, it minimizes transmission delay under band-limited conditions and can adapt to the transmission requirements of ultra-wideband pulse signals in band-limited transmission scenarios.

[0031] In one embodiment, step S1 includes the following steps: The original ultra-wideband pulse signal is acquired and its covariance matrix in time and space is extracted. A KL transform basis is constructed based on the diagonalized eigenvectors of the covariance matrix. The Gabor atom library and the KL transform basis are subjected to tensor product operation to generate a spatiotemporal joint sensing dictionary. The original ultra-wideband pulse signal is sparsely decomposed using the spatiotemporal joint sensing dictionary to calculate the matching coefficient of each atom in the Gabor atom library. Atom combinations with matching coefficients greater than a preset threshold are selected to generate spatiotemporal feature coefficients.

[0032] The above embodiments remove redundant information from the original ultra-wideband pulse signal, achieving signal dimensionality reduction while preserving the fine structure of the pulse waveform. This provides a more accurate sparse pulse sequence basis for subsequent compressed sampling, avoiding redundant information from occupying computational and transmission resources in subsequent processing, and ensuring the integrity of signal features. It should be noted that the sparse pulse sequence retains the fine structure of the pulse waveform and complete signal features of the original ultra-wideband pulse signal, eliminating redundant components, and exhibits higher sparsity and more concentrated effective information than the original ultra-wideband pulse signal.

[0033] Example, Figure 2 The original ultra-wideband pulse signal acquired is shown, targeting Figure 2 The original ultra-wideband pulse signal is shown. The covariance matrices of the time and spatial dimensions of the original ultra-wideband pulse signal are extracted to obtain... Figure 5 The covariance matrix is ​​shown. A KL transform basis is constructed based on the diagonalized eigenvectors of the covariance matrix. Then, a tensor product operation is performed between the Gabor atom library and the KL transform basis to generate... Figure 3 The spatiotemporal joint sensing dictionary shown can simultaneously capture the time-shift and frequency-shift characteristics of the original ultra-wideband pulse signal. It should be explained that the Gabor atom library in this embodiment is a set of basis functions adapted to the time-domain characteristics of the original ultra-wideband pulse signal. It consists of multiple Gabor atoms with different time shifts, scales, and frequencies, capable of matching the transient change characteristics of the original ultra-wideband pulse signal in the time domain, achieving multi-dimensional sparse matching in the signal's time domain. It should also be explained that the KL transform basis is an orthogonal basis constructed based on the diagonalized eigenvectors of the covariance matrix, capable of removing redundant information in the spatial dimension of the signal, adapting to the statistical correlation characteristics of the original ultra-wideband pulse signal in the spatial domain.

[0034] Understandably, tensor product operations can fuse the basic features of the original ultra-wideband pulse signal in both the time and spatial domains, preserving the feature representation capabilities of the original ultra-wideband pulse signal in different dimensions. This construction method can simultaneously capture the transient change characteristics of the original ultra-wideband pulse signal in the time domain and the statistical correlation characteristics in the spatial domain, ensuring that there is no defect that a single basis function can only adapt to a single dimension of features. This solves the problem that a single basis function cannot simultaneously match the time and spatial features of the original ultra-wideband pulse signal.

[0035] For example, a spatiotemporal joint sensing dictionary is used to perform sparse decomposition on the original ultrawideband pulse signal to calculate the matching coefficients of each atom in the Gabor atom library; combinations of atoms with matching coefficients greater than a preset threshold are selected to generate spatiotemporal feature coefficients. Figure 4 The spatiotemporal characteristic coefficients are shown. It is important to understand that the sparse decomposition process only retains the matching coefficients corresponding to atoms with high matching degrees. This can eliminate redundant information, compress the amount of data from the signal expression stage, and at the same time, it does not lose the core features of the signal.

[0036] It should be explained that after completing the sparse decomposition of the original ultrawideband pulse signal, signal reconstruction and residual analysis can be performed based on the obtained spatiotemporal characteristic coefficients, such as... Figure 6 As shown, the dimensionality reduction fidelity is confirmed through signal reconstruction and residual analysis.

[0037] It is understandable that the KL transform itself is an orthogonal basis built on the statistical properties of signals, which can adapt to the correlation characteristics of signals in the spatial dimension and remove the redundancy in the spatial dimension. Meanwhile, the Gabor atom library can achieve multi-scale and multi-frequency signal matching in the time domain, adapting to the transient change characteristics of the original ultra-wideband pulse signal. The combination of the two makes the precision of sparse representation higher than that of using any one of the bases alone.

[0038] In one embodiment, step S2 includes the following steps: The system receives feedback information from the receiving channel, which includes multipath delay spread and Doppler frequency shift parameters. It calculates the temporal sparsity of the channel based on the multipath delay spread and the frequency sparsity based on the Doppler frequency shift parameters. A cost function with channel state information as parameters is constructed, and the row vector weights of the observation matrix are iteratively updated using gradient descent. The updated observation matrix is ​​then used to perform a weighted projection of the spatiotemporal feature coefficients to generate a compressed observation vector.

[0039] After obtaining the spatiotemporal characteristic coefficients, the above embodiments do not transmit them directly. Instead, they dynamically adjust the observation matrix according to the channel state information and then use the adjusted observation matrix to compress and sample the spatiotemporal characteristic coefficients. This allows the sampling focus direction to be automatically adjusted according to the changes in the channel state during the compression sampling process, and only samples the sparse regions in the channel where energy is concentrated, avoiding invalid sampling of the noise-dominated subspace. This solves the problem that traditional fixed observation matrices cannot adapt to dynamic channel states and are prone to invalid sampling.

[0040] For example, the feedback information returned by the receiving channel includes the multipath delay spread and Doppler frequency shift parameters of the current channel; the time-domain sparsity of the channel is calculated based on the multipath delay spread, and the frequency-domain sparsity of the channel is calculated based on the Doppler frequency shift, thus obtaining the time-frequency sparsity distribution of the current channel, such as... Figure 7 As shown in the figure, time-domain sparsity and frequency-domain sparsity are used to characterize the channel state information representing the sparse distribution characteristics of the current channel's time-frequency dimension. It is understandable that the time-frequency sparsity distribution of the channel differs under different channel environments. The stronger the multipath effect, the more extended the time domain, and the lower the sparsity; the larger the Doppler shift, the more extended the frequency domain, and the lower the sparsity. The time-frequency sparsity calculated based on channel parameters can accurately reflect the distribution state of useful information in the current channel.

[0041] For example, a cost function with channel state information as parameters is constructed, and the row vector weights of the observation matrix are iteratively updated using gradient descent to obtain an observation matrix adapted to the current channel state, such as... Figure 8 The gradient descent optimization of the observation matrix is ​​illustrated. After the row vector weights of the observation matrix are updated, the obtained spatiotemporal feature coefficients are weighted and projected using the updated observation matrix to generate a compressed observation vector. This step matches the current channel sparsity and maximizes the entropy of the collected information, such as... Figure 9 The diagram shows the acquisition information entropy corresponding to the observation matrix (i.e., adaptive observation matrix) of this embodiment and the traditional fixed observation matrix. It reflects that the observation matrix updated iteratively by the gradient descent method in this embodiment can ensure that the limited observation data contains sufficient original ultra-wideband pulse signal feature information, thereby improving transmission efficiency.

[0042] Optionally, the observation matrix can be pre-optimized using a hybrid gradient optimization approach. The observation matrix is ​​initialized as a linear combination of a partial Hadamard matrix and a random Gaussian matrix. The cross-correlation function between the observation matrix and the spatiotemporal joint sensing dictionary is calculated. Based on the cross-correlation function, a hybrid objective function containing incoherence constraints and equidistant constraints is constructed. The hybrid objective function is iteratively optimized using the steepest descent method and the Newton method. The observation matrix is ​​then updated. The spatiotemporal feature coefficients are compressed and sampled using the updated observation matrix to generate a compressed observation vector.

[0043] It is important to understand that the lower the cross-correlation between the observation matrix and the spatiotemporal joint sensing dictionary, the better it meets the theoretical requirements of compressed sensing, and the easier it is to ensure the uniqueness and stability of the reconstruction results. Hybrid gradient optimization combines the advantages of the steepest descent method's strong global search capability and Newton's method's fast local convergence speed, enabling the observation matrix that satisfies the constraints to be obtained more quickly. This avoids the problem of unstable reconstruction performance of traditional random observation matrices and reduces computational redundancy.

[0044] In one embodiment, step S3 includes the following steps: A neighborhood graph of compressed observation vectors is constructed, and the geodesic distance between data points in the neighborhood graph is calculated. The Laplace eigenmap algorithm is used to perform eigenvalue decomposition on the neighborhood graph to obtain low-dimensional embedding vectors. The low-dimensional embedding vectors are mapped to the Riemannian manifold space, and the principal curvature direction of the Riemannian manifold space is calculated. The data points in the neighborhood graph are rearranged along the principal curvature direction to generate a spatiotemporal manifold topology that describes the intrinsic geometric structure of the original ultra-wideband pulse signal. The coordinate values ​​of stable nodes in the spatiotemporal manifold topology are extracted as manifold feature coordinates.

[0045] This embodiment mines the inherent low-dimensional structure of the compressed observation vector, effectively removes environmental noise introduced during transmission, preserves the essential characteristics of the original ultra-wideband pulse signal, provides a foundation for the accurate reconstruction of the original ultra-wideband pulse signal, reduces the computational complexity of the subsequent reconstruction process, and adapts to the ultra-wideband pulse signal transmission requirements in low-bandwidth transmission scenarios.

[0046] It should be explained that the compressed observation vector retains most of the features of the original ultra-wideband pulse signal, but its dimensionality remains high, and it also includes noise and outliers introduced during transmission. For example... Figure 10 As shown, a high-dimensional compressed observation space is constructed based on the compressed observation vectors obtained by compressed sampling. It can be seen that the compressed observation vectors are usually distributed on a specific low-dimensional manifold in the high-dimensional compressed observation space. Therefore, it is necessary to extract effective low-dimensional features by dimensionality reduction.

[0047] For example, construct a neighborhood graph of the compressed observation vectors and calculate the geodesic distance between each data point in the neighborhood graph, such as... Figure 11 The diagram illustrates the neighborhood graph and the geodesic distance. It is understandable that Euclidean distance cannot reflect the spatial distances inherent in the manifold of high-dimensional data, while geodesic distance can calculate the actual distance between two data points along the surface of the manifold, better reflecting the inherent geometric structure of high-dimensional data.

[0048] For example, the Laplacian eigenmap algorithm is used to perform feature decomposition on the constructed neighborhood graph to obtain low-dimensional embedding vectors, such as... Figure 12As shown, Laplacian eigenmaps can preserve the local neighborhood structure of high-dimensional data. After mapping data originally distributed in a high-dimensional compressed observation space to a low-dimensional space, the original topological relationships are still maintained. It is important to understand that the effective features of the original ultra-wideband pulse signal are themselves distributed on a low-dimensional manifold in the high-dimensional compressed observation space. Ordinary linear dimensionality reduction methods cannot explore this nonlinear manifold structure; they can only achieve simple data dimensionality compression and cannot remove noise interference. However, the nonlinear dimensionality reduction based on Laplacian eigenmaps used in this embodiment can explore the intrinsic geometric structure of high-dimensional data.

[0049] For example, the low-dimensional embedding vector is mapped to the Riemannian manifold space, and the principal curvature directions of the Riemannian manifold space are calculated. Data points in the neighborhood graph are rearranged along the principal curvature directions to generate a spatiotemporal manifold topology describing the intrinsic geometry of the original ultra-wideband pulse signal. The coordinates of stable nodes in the spatiotemporal manifold topology are extracted as manifold feature coordinates. It is important to understand that the principal curvature directions reflect the directions of most significant data changes in the spatiotemporal manifold topology. Rearranging along the principal curvature directions more clearly reveals the manifold topology. Stable nodes correspond to the core feature points of the original ultra-wideband pulse signal, while unstable noise interference points are eliminated. Therefore, the obtained manifold feature coordinates can retain the essential topological distribution characteristics of the original ultra-wideband pulse signal while removing additional noise and outliers.

[0050] In one embodiment, step S4 includes the following steps: The entire reconstruction process is based on manifold characteristic coordinates, establishing an objective optimization function that includes signal sparsity constraints and data fidelity constraints. A weighted least squares term is introduced into the objective optimization function, and the weight coefficients of the weighted least squares term are adjusted according to the energy density of the local region of the original ultra-wideband pulse signal to complete the constraint model for nonlinear reconstruction. The conjugate gradient method is used to iteratively solve the objective optimization function to obtain the optimal sparse solution. Based on the optimal sparse solution, the pulse amplitude parameters and time delay parameters of the corresponding original ultra-wideband pulse signal are restored, generating a sparse pulse sequence and completing the entire nonlinear reconstruction process.

[0051] The above embodiments avoid both filtering out weak signal components as noise under traditional unified constraints and avoiding the loss of signal details caused by over-constraints, thus adapting to the signal reconstruction requirements under different channel environments.

[0052] For example, an objective optimization function containing signal sparsity constraints and data fidelity constraints is established to constrain the search space of the objective optimization function solution process, ensuring that the solution result of the objective optimization function conforms to the inherent sparsity characteristics of the original ultra-wideband pulse signal itself.

[0053] For example, a weighted least squares term is introduced into the objective optimization function, and the weight coefficients of the weighted least squares term are dynamically adjusted according to the energy density of local regions of the original ultra-wideband pulse signal. It is important to understand that by introducing a weighted least squares term and dynamically adjusting its weight coefficients, a differentiated error penalty strategy can be implemented for signal components of different energy levels during the reconstruction of the original ultra-wideband pulse signal. This strategy focuses on suppressing the noise amplification effect in low-energy regions and correcting the distortion of weak signal components caused by nonlinear compression sampling, thereby improving the accuracy of pulse detail reconstruction while maintaining overall waveform similarity.

[0054] For example, the weighting coefficients of the weighted least squares term can be dynamically adjusted based on the energy density of a local region of the original ultra-wideband pulse signal, which can be combined with... Figure 13 The energy density-based adaptive weighted reconstruction mechanism is as follows: The energy density gradient of a local region in the original ultra-wideband pulse signal is calculated to identify strong and weak support regions. Smaller weight coefficients are set in the strong support regions to relax the constraints on the fitting error and preserve high-frequency details of the signal. Larger weight coefficients are set in the weak support regions to tighten the constraints on the fitting error and suppress the spread of background noise. It is understandable that the signal component characteristics differ between the strong and weak support regions. The strong support regions correspond to the pulse body with concentrated energy in the original ultra-wideband pulse signal, while the weak support regions correspond to the pulse edges and background areas. The degree of noise influence differs between the strong and weak support regions. It should be noted that after adjusting the weight coefficients of the weighted least squares term, the objective optimization function after adjusting the weight coefficients is smoothed to ensure the convergence stability of the optimization process.

[0055] It is important to understand that by adjusting the weight coefficients of the weighted least squares term according to the energy density gradient difference, an adaptive balance between noise and detail can be achieved during signal reconstruction. This effectively filters out random noise in flat areas while protecting the key high-frequency edge information of the original ultra-wideband pulse signal.

[0056] For example, the conjugate gradient method is used to iteratively solve the objective function to obtain the optimal sparse solution. The input data in the high-dimensional compressed observation space is dimensionality-reduced and then substituted into the adjusted objective function for iterative processing. Each iteration updates the error direction based on the weight coefficient adjustment results, avoiding the search process from getting trapped in local optima. Finally, based on the optimal sparse solution, the pulse amplitude and delay parameters of the corresponding original ultra-wideband pulse signal are reconstructed, generating a sparse pulse sequence.

[0057] It should be explained that the original ultra-wideband pulse signal itself has a wide bandwidth, and the pulse details contain a large amount of positioning and communication information. Weak signal components often carry the key features required for multipath resolution. Traditional unified constraint reconstruction methods tend to filter out weak signal components as noise or introduce a large amount of noise in low-energy regions. However, this embodiment identifies strong and weak support regions through energy density gradients and adjusts the constraint strength for different support regions. This preserves the effective details of the pulse while controlling the spread of noise. The smoothing process avoids oscillations in the optimization process caused by abrupt changes in weight coefficients, thus improving the stability of the iterative solution of the objective optimization function. In addition, this embodiment uses a differentiated weight coefficient adjustment strategy to adapt to the energy distribution characteristics of different regions of the original ultra-wideband pulse signal. This avoids the problem of weak signal components being submerged by noise under traditional unified constraints and also avoids the loss of signal details due to over-constraint. It can adapt to the signal reconstruction requirements under different channel environments and maintain the convergence stability of the reconstruction process.

[0058] In one embodiment, step S5 includes the following steps: The prior distribution model of the quantization parameters is initialized, which includes the quantization step size and probability density function. The sparse pulse sequence is used as the observation data, and the posterior probability distribution of the quantization parameters is calculated using the variational Bayesian method. The quantization parameters are iteratively updated until convergence according to the principle of maximizing the expected value of the posterior probability distribution to obtain the optimal quantization parameters. The scalar quantization and arithmetic coding of the sparse pulse sequence are performed using the optimal quantization parameters to generate the transmission code stream.

[0059] The above embodiments adjust the quantization parameters through probabilistic reasoning, which can adapt to the changes in the statistical characteristics of the original ultra-wideband pulse signal under different scenarios. While ensuring the fidelity of the recovered signal, the redundancy of the transmitted code stream is reduced, the utilization rate of bandwidth resources is improved, and the transmission requirements of the original ultra-wideband pulse signal under different bandwidth limitations are adapted.

[0060] For example, the prior distribution model for initializing the quantization parameters includes the quantization step size and probability density function. The initialization of the prior distribution model is based on setting initial values ​​according to the general statistical properties of the original ultra-wideband pulse signal, providing a basic search starting point for subsequent iterative updates. It should be explained that the general statistical properties include the statistical properties of the pulse amplitude, multipath delay, spatiotemporal sparsity, and energy distribution of the original ultra-wideband pulse signal.

[0061] For example, the sparse pulse sequence obtained after dimensionality reduction and reconstruction is used as observation data, and the posterior probability distribution of the quantization parameters is calculated using the variational Bayesian method. It's important to understand that the variational Bayesian method, by approximating the posterior probability distribution, can perform probability estimation of the quantization parameters with lower computational complexity. Compared to the traditional maximum likelihood estimation method, it is better suited to the changing statistical characteristics of sparse pulse sequences. It's understandable that after sparse projection and dimensionality reduction, the statistical characteristics of the sparse pulse sequence will fluctuate with changes in the channel environment and the original ultra-wideband pulse signal. Fixed quantization parameters cannot accommodate these fluctuations, easily leading to accumulated quantization errors or increased coding redundancy.

[0062] For example, based on the principle of maximizing the expected value of the posterior probability distribution, the quantization parameters are iteratively updated until convergence to obtain the optimal quantization parameters. Figure 14 As shown, during the variational Bayesian iteration process, the quantization step size parameter gradually converges, and the uncertainty of the posterior probability distribution gradually decreases, ultimately yielding a stable optimal quantization parameter. It is important to understand that using variational Bayesian iterative inference to obtain the optimal quantization parameter can adaptively mine changes in the statistical characteristics of the signal from a probabilistic statistical perspective. This overcomes the performance bottleneck of traditional fixed quantization strategies in non-stationary ultra-wideband pulse signal transmission. By dynamically adjusting the quantization precision and coding strategy, it achieves efficient lossless or low-loss compression approaching the source entropy limit under strictly limited bandwidth resources.

[0063] For example, after obtaining the optimal quantization parameters, the sparse pulse sequence is scalar quantized and arithmetic encoded using these parameters to generate a transmission bitstream. Figure 15 As shown, the encoding efficiency of this embodiment has a lower output code rate compared to the traditional fixed quantization encoding method; as Figure 16 As shown, the ultra-wideband waveform spectrum recovered after encoding and decoding in this embodiment has a higher similarity to the original signal spectrum (i.e., the original ultra-wideband pulse signal spectrum).

[0064] Understandably, during the initialization of the prior distribution model for quantization parameters, this prior distribution model can be adjusted based on prior experience of the actual application scenario. For example, for indoor dense multipath ultra-wideband transmission scenarios, the prior distribution model can be biased towards a smaller quantization step size, while for outdoor open scenarios, the prior distribution model can be biased towards a larger quantization step size, further improving the speed of iterative convergence, reducing the number of iterations, and lowering the computational complexity of the encoding process.

[0065] In one embodiment, performing manifold learning dimensionality reduction mapping on the compressed observation vector in step S3 may further include the following steps: Calculate the k nearest neighbors for each sample point and construct a local reconstruction weight matrix; solve the local reconstruction weight matrix by minimizing the reconstruction error; use the local reconstruction weight matrix to embed high-dimensional data into low-dimensional space, and introduce a regularization term during the embedding process to control the sparse distribution density of sample points in order to generate a spatiotemporal manifold topology.

[0066] The above embodiments introduce an optimization step based on local linear embedding during the manifold learning dimensionality reduction mapping process of compressed observation vectors. This can better adapt to the characteristics of the original ultra-wideband pulse signal high-dimensional compressed observation data, improve the stability of the dimensionality reduction results while preserving the inherent manifold structure of the signal, and thus solve the problem of poor generalization ability of the dimensionality reduction results caused by insufficient preservation of local geometric structure and overfitting during the manifold learning dimensionality reduction mapping process.

[0067] For example, the k nearest neighbors of each sample point are calculated, and a local reconstruction weight matrix is ​​constructed. The sample point is the vertex data point of the neighborhood graph. It needs to be explained that the dimensionality reduction process first constructs the neighborhood relationships of the sample points in the high-dimensional compressed observation space. The neighborhood relationship is the proximity association between each sample point and its k nearest neighbors in the high-dimensional compressed observation space. The neighborhood of each sample point reflects its local geometric relationship in the manifold space. Figure 11 Discrete nodes are used to represent sample points, and all nodes are distributed on the surface of the manifold space. Furthermore, the local reconstruction weight matrix is ​​solved by minimizing the reconstruction error. The local reconstruction weight matrix represents the weight relationship that each sample point can be linearly represented by other sample points in its neighborhood. This representation can capture the local geometric structure characteristics of the manifold in the high-dimensional compressed observation space.

[0068] It is important to understand that the core assumption of local linear embedding is that the manifold is locally approximated as linear. Therefore, each sample point can be linearly reconstructed from the sample points in its neighborhood. By keeping the local reconstruction weights unchanged in the low-dimensional space, high-dimensional data can be embedded into the low-dimensional space while preserving the original local geometric relationships.

[0069] For example, a local reconstruction weight matrix can be used to embed high-dimensional data into a low-dimensional space while preserving the local neighborhood geometry. Figure 10 As shown, sample points in the high-dimensional compressed observation space are mapped to a low-dimensional space after embedding, such as... Figure 12As shown, the dimensionality reduction yields a low-dimensional feature mapping result that preserves the original topological structure. It is understandable that compressing observation vectors often results in uneven sample distribution and noise perturbation. Directly performing local linear embedding can easily lead to overfitting, causing large fluctuations in the manifold structure obtained from dimensionality reduction and poor generalization ability for new samples. Therefore, this embodiment introduces a regularization term during the embedding process to control the sparse distribution density of sample points, generating a smooth spatiotemporal manifold topology to prevent overfitting.

[0070] It's important to understand that the optimization process based on local linear embedding and the introduction of a regularization term not only preserves the local manifold structure of the original ultra-wideband pulse signal during dimensionality reduction mapping, but also prevents overfitting through regularization constraints. This enhances the robustness of the manifold structure to fluctuations in sample data, ensuring that the extracted manifold feature coordinates have generalization ability and stability. The regularization term constrains the magnitude of the local reconstruction weight matrix, preventing excessive fluctuations. For sparsely distributed sample points, the regularization term smooths the neighborhood reconstruction results, preventing disturbances from a single sample point from having an excessive impact on the overall manifold structure. The number of nearest neighbors, k, can be adjusted based on the dimension of the input compressed observation vector and the number of samples. Higher dimensions and larger sample numbers allow for a relatively larger value for k, ensuring the accuracy of local geometric structure estimation.

[0071] Furthermore, the low-dimensional features obtained through regularization optimization retain the spatiotemporal manifold topology of the high-dimensional compressed observation space. These features can be used to solve for the spatiotemporal feature coefficients and reconstruct the original ultrawideband pulse signal, thereby reducing the computational complexity of subsequent processing and improving the stability of the reconstruction results.

[0072] This embodiment also provides an ultra-wideband pulse waveform spatiotemporal synchronous compression system for implementing the ultra-wideband pulse waveform spatiotemporal synchronous compression method of any of the above embodiments. The ultra-wideband pulse waveform spatiotemporal synchronous compression system provided in this embodiment includes a spatiotemporal feature extraction module, an adaptive compression sampling module, a manifold topology processing module, a nonlinear reconstruction module, and a Bayesian inference coding module.

[0073] The spatiotemporal feature extraction module is used to construct a spatiotemporal joint sensing dictionary. The spatiotemporal joint sensing dictionary is used to perform multidimensional orthogonal projection on the original ultrawideband pulse signal to extract spatiotemporal feature coefficients.

[0074] The adaptive compressed sampling module is used to adjust the row vector distribution of the observation matrix according to the channel state information, and to compress and sample the spatiotemporal feature coefficients using the adjusted observation matrix to generate compressed observation vectors.

[0075] The manifold topology processing module is used to perform manifold learning dimensionality reduction mapping on compressed observation vectors to construct a spatiotemporal manifold topology and obtain manifold feature coordinates based on the spatiotemporal manifold topology.

[0076] The nonlinear reconstruction module is used to introduce weighted least squares constraints to nonlinearly reconstruct the feature coordinates of the manifold, generating a sparse pulse sequence.

[0077] The Bayesian inference coding module is used to perform variational Bayesian iterative inference on sparse pulse sequences to obtain optimal quantization parameters, and then use the optimal quantization parameters to entropy encode the sparse pulse sequences to generate a transmission code stream.

[0078] The ultra-wideband pulse waveform spatiotemporal synchronous compression system provided in this embodiment utilizes a spatiotemporal joint sensing dictionary to perform multidimensional orthogonal projection on the original ultra-wideband pulse signal. This captures the transient change details in the time domain and the statistical correlation of the spatial dimension of the original ultra-wideband pulse signal. Based on channel state information, it adjusts the row vector distribution of the observation matrix in real time, enabling the observation matrix to automatically match the sparsity characteristics of the current channel. This maximizes the acquisition information entropy while reducing the sampling rate, avoiding performance redundancy or information loss of the observation matrix under dynamic channels. Furthermore, by performing manifold learning dimensionality reduction mapping on the compressed observation vector and obtaining manifold feature coordinates based on the constructed spatiotemporal manifold topology, it can deeply mine the inherent low-dimensional geometric manifold distribution of high-dimensional data and effectively remove transmission environment noise. By introducing weighted least squares constraints to nonlinearly reconstruct the manifold feature coordinates, a sparse pulse sequence is generated, which can correct the waveform distortion caused by compressed sampling. By performing variational Bayes iterative inference on the sparse pulse sequence to obtain the optimal quantization parameters and using the optimal quantization parameters to entropy encode the sparse pulse sequence, it can further approach the source entropy limit and minimize transmission delay under band-limited conditions.

[0079] Regarding the ultra-wideband pulse waveform spatiotemporal synchronization compression system in the above embodiments, the specific methods by which each module performs its operations have been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0080] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units. Components shown as modules or 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 this disclosure according to actual needs. Those skilled in the art can understand and implement this without any inventive effort.

[0081] This embodiment also provides a computer-readable storage medium. The computer-readable storage medium provided in this embodiment stores a computer program, which, when executed by a processor, can implement all the processing steps of the ultra-wideband pulse waveform spatiotemporal synchronization compression method as described in any of the above embodiments.

[0082] Furthermore, the computer-readable storage medium is implemented using non-volatile storage media, which can permanently retain program code and related configuration parameters even when power is off, ensuring no loss of program information due to power failure, and facilitating portability and deployment across different hardware platforms. It's important to understand that non-volatile storage media can also be selected in different specifications to meet the storage capacity requirements of the hardware platform, adapting to various processing platforms from low-power embedded devices to high-performance servers, satisfying deployment needs in different scenarios.

[0083] Furthermore, computer programs stored on computer-readable storage media are compiled and packaged in a modular manner. Each processing stage corresponds to an independent program module, and the modules interact with each other through standardized interfaces. This facilitates the modification and upgrading of individual modules without affecting the overall program operation. Understandably, modular program design reduces the cost of subsequent maintenance and upgrades. Modules can be replaced for different application scenarios to meet customized needs without requiring the redevelopment of a complete program.

[0084] This embodiment also provides an electronic product. The electronic product provided in this embodiment includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it can implement all the processing steps of the ultra-wideband pulse waveform spatiotemporal synchronization compression method of any of the above embodiments.

[0085] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein.

Claims

1. A method for spatiotemporal synchronization compression of ultrawideband pulse waveforms, characterized in that, Includes the following steps: A spatiotemporal joint sensing dictionary is constructed, and the original ultrawideband pulse signal is subjected to multidimensional orthogonal projection using the spatiotemporal joint sensing dictionary to extract spatiotemporal feature coefficients. The row vector distribution of the observation matrix is ​​adjusted according to the channel state information, and the spatiotemporal feature coefficients are compressed and sampled using the adjusted observation matrix to generate a compressed observation vector. Manifold learning dimensionality reduction mapping is performed on the compressed observation vector to construct a spatiotemporal manifold topology, and manifold feature coordinates are obtained based on the spatiotemporal manifold topology; A weighted least squares constraint is introduced to nonlinearly reconstruct the feature coordinates of the manifold, generating a sparse pulse sequence. Variational Bayesian iterative inference is performed on the sparse pulse sequence to obtain the optimal quantization parameters, and the optimal quantization parameters are used to entropy encode the sparse pulse sequence to generate a transmission code stream.

2. The ultra-wideband pulse waveform spatiotemporal synchronization compression method as described in claim 1, characterized in that, The construction of the spatiotemporal joint sensing dictionary, and the use of the spatiotemporal joint sensing dictionary to perform multidimensional orthogonal projection on the original ultrawideband pulse signal to extract spatiotemporal feature coefficients, includes: Acquire raw ultrawideband pulse signals and extract their covariance matrices in the time and spatial dimensions; Construct a KL transform basis based on the diagonalized eigenvectors of the covariance matrix; The Gabor atom library is subjected to tensor product operation with the KL transform basis to generate the spatiotemporal joint sensing dictionary; The original ultrawideband pulse signal is sparsely decomposed using the spatiotemporal joint sensing dictionary to calculate the matching coefficients of each atom in the Gabor atom library; Atom combinations with matching coefficients greater than a preset threshold are selected to generate the spatiotemporal feature coefficients.

3. The ultra-wideband pulse waveform spatiotemporal synchronization compression method as described in claim 1, characterized in that, The step of adjusting the row vector distribution of the observation matrix according to the channel state information, and using the adjusted observation matrix to compress and sample the spatiotemporal feature coefficients to generate a compressed observation vector includes: The feedback information received from the channel includes multipath delay spread and Doppler frequency shift parameters; The time-domain sparsity of the channel is calculated based on the multipath delay spread, and the frequency-domain sparsity of the channel is calculated based on the Doppler frequency shift parameter. A cost function with the channel state information as a parameter is constructed, and the row vector weights of the observation matrix are iteratively updated using the gradient descent method. The spatiotemporal feature coefficients are weighted and projected using the updated observation matrix to generate the compressed observation vector.

4. The ultra-wideband pulse waveform spatiotemporal synchronization compression method as described in claim 1, characterized in that, The step of performing manifold learning dimensionality reduction mapping on the compressed observation vector to construct a spatiotemporal manifold topology, and obtaining manifold feature coordinates based on the spatiotemporal manifold topology, includes: Construct a neighborhood graph of the compressed observation vector and calculate the geodesic distance between each data point in the neighborhood graph; The neighborhood graph is decomposed using the Laplacian feature map algorithm to obtain a low-dimensional embedding vector. The low-dimensional embedding vector is mapped to the Riemannian manifold space, and the principal curvature directions of the Riemannian manifold space are calculated. The data points in the neighborhood graph are rearranged along the principal curvature direction to generate the spatiotemporal manifold topology. The coordinates of stable nodes in the spatiotemporal manifold topology are extracted and used as the manifold feature coordinates.

5. The ultra-wideband pulse waveform spatiotemporal synchronization compression method as described in claim 1, characterized in that, The introduction of weighted least squares constraints to nonlinearly reconstruct the manifold feature coordinates and generate a sparse pulse sequence includes: Establish an objective optimization function that includes signal sparsity constraints and data fidelity constraints; A weighted least squares term is introduced into the objective optimization function, and the weight coefficients of the weighted least squares term are adjusted according to the energy density of the local region of the original ultra-wideband pulse signal. The objective optimization function is solved iteratively using the conjugate gradient method to obtain the optimal sparse solution; Based on the optimal sparse solution, the pulse amplitude parameters and time delay parameters of the corresponding original ultra-wideband pulse signal are restored, and the sparse pulse sequence is generated.

6. The ultra-wideband pulse waveform spatiotemporal synchronization compression method as described in claim 1, characterized in that, The step of performing variational Bayesian iterative inference on the sparse pulse sequence to obtain optimal quantization parameters, and then using the optimal quantization parameters to perform entropy encoding on the sparse pulse sequence to generate a transmission bitstream, includes: A prior distribution model for initializing quantization parameters, wherein the prior distribution model includes the quantization step size and the probability density function; Using the sparse pulse sequence as observation data, the posterior probability distribution of the quantization parameters is calculated using the variational Bayesian method; Based on the principle of maximizing the expected value of the posterior probability distribution, the quantization parameters are iteratively updated until convergence to obtain the optimal quantization parameters; The sparse pulse sequence is scalar quantized and arithmetic encoded using the optimal quantization parameters to generate the transmission code stream.

7. The ultra-wideband pulse waveform spatiotemporal synchronization compression method as described in claim 1, characterized in that, The process of performing manifold learning dimensionality reduction mapping on the compressed observation vector further includes the following steps: Calculate the k nearest neighbors for each sample point and construct a local reconstruction weight matrix; The local reconstruction weight matrix is ​​solved by minimizing the reconstruction error; The high-dimensional data is embedded into the low-dimensional space using the local reconstruction weight matrix, and a regularization term is introduced during the embedding process to control the sparse distribution density of the sample points in order to generate the spatiotemporal manifold topology.

8. A spatiotemporal synchronization compression system for ultrawideband pulse waveforms, characterized in that, The ultra-wideband pulse waveform spatiotemporal synchronization compression system is used to implement the ultra-wideband pulse waveform spatiotemporal synchronization compression method as described in any one of claims 1-7, and the ultra-wideband pulse waveform spatiotemporal synchronization compression system comprises: The spatiotemporal feature extraction module is used to construct a spatiotemporal joint sensing dictionary and use the spatiotemporal joint sensing dictionary to perform multidimensional orthogonal projection on the original ultrawideband pulse signal to extract spatiotemporal feature coefficients. An adaptive compressed sampling module is used to adjust the row vector distribution of the observation matrix according to the channel state information, and to compress and sample the spatiotemporal feature coefficients using the adjusted observation matrix to generate a compressed observation vector. The manifold topology processing module is used to perform manifold learning dimensionality reduction mapping on the compressed observation vector to construct a spatiotemporal manifold topology structure, and obtain manifold feature coordinates based on the spatiotemporal manifold topology structure; The nonlinear reconstruction module is used to introduce weighted least squares constraints to nonlinearly reconstruct the manifold feature coordinates and generate a sparse pulse sequence. The Bayesian inference coding module is used to perform variational Bayesian iterative inference on the sparse pulse sequence to obtain the optimal quantization parameters, and to use the optimal quantization parameters to perform entropy coding on the sparse pulse sequence to generate a transmission code stream.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the ultrawideband pulse waveform spatiotemporal synchronization compression method as described in any one of claims 1 to 7.

10. An electronic product comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the ultrawideband pulse waveform spatiotemporal synchronization compression method as described in any one of claims 1 to 7.

Images