NEMS geomagnetic sensor rear-end data acquisition and processing system

Abstract

The invention relates to the technical field of sensor data acquisition, and discloses an NEMS geomagnetic sensor rear-end data acquisition and processing system comprising a sensor array module used for acquiring geomagnetic field signals; the dynamic sparse decoupling module is used for carrying out sparse modeling and decoupling processing on the signals acquired by the sensor array module and separating geomagnetic signals from interference signals; the physical constraint signal reconstruction module is used for carrying out nonlinear reconstruction on the separated signals in combination with the physical distribution rule of the geomagnetic field; and the data output module is used for normalizing the reconstructed geomagnetic signal and outputting the geomagnetic signal to terminal equipment. By adopting a dynamic sparse decoupling technology and combining a sparse decomposition model, precise separation of complex aliasing signals is realized, and the decoupling precision of the target geomagnetic signal is further improved. According to the technical scheme, the effect of extracting the core signal under the multi-field-source interference condition is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of sensor data acquisition, and in particular to a NEMS geomagnetic sensor back-end data acquisition and processing system. Background Art

[0002] The acquisition and processing of geomagnetic signals is an important research direction in the fields of navigation positioning, geomagnetic monitoring, etc. In the existing technology, geomagnetic signals are mainly obtained by single-point observation or simple array arrangement, but these solutions have great limitations in complex magnetic field environments. The aliasing of multi-source signals, the interference of noise, and the lack of decoupling accuracy have always been technical difficulties in the acquisition and processing of geomagnetic signals.

[0003] Traditional sensor layout schemes mostly use fixed arrays or regular distribution. The design of this method does not fully consider the orthogonality and sparseness of the observation matrix, resulting in a large amount of redundant information in the signals collected by the sensors, and the coupling problem between signals is also difficult to avoid. In a multi-source interference environment, the layout of the sensor array cannot effectively separate the target signal from the interference signal, which seriously affects the accuracy of subsequent processing.

[0004] Existing signal decoupling technologies are mostly based on a single sparse decomposition method, which usually assumes that the sparse characteristics of the signal are fixed in time and space. However, in practical applications, the sparsity of geomagnetic signals usually changes dynamically and is directly related to the degree of aliasing of noise and interference signals. Traditional decoupling methods lack dynamic adjustment capabilities and are difficult to adapt to the needs of multi-field source signal separation in complex environments, resulting in incomplete extraction of target signals or serious errors.

[0005] In the process of signal reconstruction, existing technologies often adopt pure data-driven processing methods or simple empirical models, lacking in-depth integration with the physical laws of the geomagnetic field. This method can easily result in signal reconstruction results that deviate from the actual geomagnetic field distribution, especially in scenarios with strong nonlinear interference or high noise levels. Due to the lack of physical constraints, the reconstructed signal has a poor match with the real geomagnetic signal, affecting the credibility of the data and the effectiveness of subsequent applications.

[0006] The existing technology also has many deficiencies in the stability and applicability of signal output. In some technical solutions, the signal output is often provided in a single format or in the form of unprocessed raw signals, which makes it difficult to meet the compatibility requirements of multiple terminal devices. At the same time, for high-noise environments or dynamic application scenarios, the existing solutions fail to fully reduce noise and optimize the signals, and the quality of the output signals is not high, which cannot directly meet the actual use needs. Summary of the invention

[0007] In view of the shortcomings of the prior art, the present invention provides a NEMS geomagnetic sensor back-end data acquisition and processing system, which solves the problem that most traditional sensor layout schemes adopt a fixed array or regular distribution method. The design of this method does not fully consider the orthogonality and sparse characteristics of the observation matrix, resulting in a large amount of redundant information in the signals collected by the sensor.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions: A NEMS geomagnetic sensor back-end data acquisition and processing system, comprising: A sensor array module for collecting geomagnetic field signals; A dynamic sparse decoupling module, used for performing sparse modeling and decoupling processing on the signals collected by the sensor array module, and separating the geomagnetic signal from the interference signal; Physical constraint signal reconstruction module, used to perform nonlinear reconstruction of the separated signal in combination with the physical distribution law of the geomagnetic field; The data output module is used to normalize the reconstructed geomagnetic signal and output it to the terminal device.

[0009] Preferably, the sensor array adopts an optimized distributed NEMS sensor array, and describes the coupling relationship between the sensor and the geomagnetic field signal through an observation matrix. The sensor array modules are sparsely distributed, and the physical position and sensitivity distribution of the sensors are optimized to ensure that the coupling relationship between the sensor array and the target signal meets specific restricted equidistant conditions, so as to achieve sparse decomposition of the observation signal.

[0010] Preferably, the restricted equidistance condition is achieved by optimizing the orthogonality of the measurement matrix, and the goal of orthogonality optimization is to minimize the deviation between the measurement matrix and the unit matrix.

[0011] Preferably, the dynamic sparse decoupling module realizes signal decoupling based on a joint sparse decomposition model, and the optimization goal of the model is to minimize the error between the sensor observation signal and the target signal while satisfying the sparsity constraint.

[0012] Preferably, the dynamic sparse decoupling module adopts an alternating optimization method to alternately optimize the sparse matrix and the target signal matrix respectively, so as to improve the signal decoupling accuracy and computational efficiency.

[0013] Preferably, the physical constraint signal reconstruction module constructs a constraint model in combination with the dipole distribution law of the geomagnetic field, wherein the geomagnetic field strength varies inversely with the distance of the position, and is used to correct the nonlinear characteristics of the signal.

[0014] Preferably, the constraint model achieves high-precision calibration of the signal by fitting the deviation between the reconstructed geomagnetic signal and the theoretical geomagnetic field signal, and a deviation fitting error minimization model is established based on an analytical equation of the geomagnetic field distribution.

[0015] Preferably, the physical constraint signal reconstruction module completes the nonlinear reconstruction of the signal through a deep learning model, and the deep learning model corrects the signal in combination with the physical laws of the geomagnetic field. The optimization goal of the signal correction is to simultaneously minimize the weighted sum of the data fitting error and the physical law constraint error.

[0016] Preferably, the deep learning model includes an input layer, a hidden layer and an output layer, and the hidden layer embeds physical laws as regularization constraints to correct deviations of nonlinear characteristics in the signal.

[0017] Preferably, the data output module processes the signal through normalization and filtering noise reduction to improve the quality of the output signal, and outputs the signal to a terminal device used for geomagnetic monitoring, navigation positioning or security monitoring.

[0018] The present invention provides a NEMS geomagnetic sensor back-end data acquisition and processing system, which has the following beneficial effects: 1. The present invention adopts dynamic sparse decoupling technology and a joint sparse decomposition model to achieve accurate separation of complex aliasing signals, further improve the decoupling accuracy of the target geomagnetic signal, and achieve the effect of extracting the core signal under multi-field source interference conditions. Compared with the single sparse decomposition or linear separation method in the prior art, it solves the problems of insufficient sparsity modeling of dynamic signals and incomplete signal separation in complex environments.

[0019] 2. The present invention adopts a technical solution of optimizing the sensor array and the observation matrix. Through sparse distribution design and orthogonality optimization, the independence and separation capability of geomagnetic signal acquisition are improved, and the technical effects of high-sensitivity acquisition and low coupling interference are achieved. Compared with the technical solution of fixed or unoptimized sensor arrangement in the prior art, the problem of high redundancy of acquisition signals and severe coupling interference is solved.

[0020] 3. The present invention introduces the dipole distribution law of the geomagnetic field and combines it with a deep physical constraint model to perform nonlinear reconstruction and correction of the signal, thereby achieving the technical effect of high-precision reconstructed signals consistent with the actual geomagnetic field distribution. Compared with the prior art signal processing solutions that rely on empirical modeling or pure data-driven, this solves the defects of insufficient accuracy of the reconstruction results and difficulty in matching the geomagnetic physical characteristics.

[0021] 4. The data output module designed in the present invention integrates multiple functions such as signal normalization, noise reduction, format conversion and real-time output, which significantly improves the applicability and output quality of the signal, and achieves the technical effect of being compatible with multi-terminal requirements and optimizing signal stability. Compared with the method of only providing a single format output or simple noise reduction in the prior art, it solves the problems of poor output signal stability and limited application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a system framework diagram of the present invention. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] Please refer to the attached Figure 1 The embodiment of the present invention provides a NEMS geomagnetic sensor back-end data acquisition and processing system, including: A sensor array module for collecting geomagnetic field signals; A dynamic sparse decoupling module, used for performing sparse modeling and decoupling processing on the signals collected by the sensor array module, and separating the geomagnetic signal from the interference signal; Physical constraint signal reconstruction module, used to perform nonlinear reconstruction of the separated signal in combination with the physical distribution law of the geomagnetic field; The data output module is used to normalize the reconstructed geomagnetic signal and output it to the terminal device.

[0025] System overall architecture and module design The present invention discloses a NEMS geomagnetic sensor back-end data acquisition and processing system. The system mainly includes four modules: sensor array module, dynamic sparse decoupling module, physical constraint signal reconstruction module and data output module. Through optimized design, each module works in coordination to achieve accurate acquisition, processing and output of geomagnetic signals. The beneficial effect of the present invention is that it can accurately decouple geomagnetic signals in a complex magnetic field environment, effectively suppress noise and interference, and provide reliable data support for high-precision geomagnetic measurement and navigation positioning.

[0026] Sensor array module This module uses multiple NEMS sensors to construct a two-dimensional sparsely distributed array. The sensors are laid out through an optimized observation matrix to ensure that the array has high sensitivity and low coupling to geomagnetic signals. The observation matrix is ​​optimized to meet the restricted isometry (RIP) condition for sparse signal decomposition. This design greatly improves the signal separation capability.

[0027] The sensor position is calculated using an optimization algorithm based on matrix orthogonality optimization of the Frobenius norm, which ultimately determines the optimal layout of the sensors. During the calibration process, a standard geomagnetic field signal source is introduced to ensure the response consistency and calibration accuracy of each sensor. By optimizing the sensor spacing and sensitivity, the system has good adaptability to geomagnetic signals of different intensities.

[0028] The optimized array design reduces signal redundancy while improving acquisition efficiency. The calibration of sensor sensitivity further reduces measurement errors and ensures the accuracy of observation data.

[0029] Dynamic sparse decoupling module The dynamic sparse decoupling module is used to process the aliased signals collected by the sensor array. This module is based on the joint sparse decomposition technology, combined with the sparse characteristics in time and space, to decouple and separate the signals. The observed signal is modeled as:

[0030] in, To collect the signal matrix, is the observation matrix, is a dynamic sparse matrix, is the target signal, For the noise signal decoupling process: a joint sparse optimization model is used to minimize the error between the observed signal and the target signal under the condition of sparsity constraints. The optimization goal is:

[0031] in, is the constraint term of the sparse matrix, is the regularization parameter. During the solution process, the alternating optimization method is used to alternately optimize the sparse matrix and the target signal matrix. The sparse matrix is ​​fixed and the target signal is optimized. Then the target signal is fixed and the sparse matrix is ​​optimized. The iteration is continued until convergence.

[0032] After decoupling, the signal separation is significantly improved, the background noise and interference signals are effectively eliminated, and the target geomagnetic signal is accurately restored. The dynamic sparse decoupling module achieves efficient separation of aliased signals through joint sparse optimization technology. The module can adapt to the sparse changes of geomagnetic signals in time and space, and improves the adaptability of the system in dynamic environments.

[0033] Physical constraint signal reconstruction module The physical constraint signal reconstruction module performs nonlinear reconstruction on the separated target signal based on the dipole distribution characteristics of the geomagnetic field. The distribution law of the geomagnetic field can be expressed as:

[0034] in, is the geomagnetic field strength at the target location, is the geomagnetic dipole moment, is the position vector, Based on the above physical laws, this module introduces a constraint model to correct the signal by fitting the deviation between the target signal and the theoretical geomagnetic field signal. The model objectives are:

[0035] in, is the reconstructed signal, In addition, this module combines the deep learning model with the physics-guided deep learning framework (PINNs) to further improve the signal reconstruction accuracy. The training loss function of the deep learning model includes the weighted sum of the data fitting error and the physical law constraint error:

[0036] in, Constrained by physical laws, The physical constraint signal reconstruction module introduces the constraints of geomagnetic physics laws as weight parameters, which greatly improves the accuracy of signal reconstruction and the interpretability of the model. Combined with deep learning technology, the module can effectively process nonlinear signals and improve the adaptability and robustness of the system.

[0037] Data output module The data output module normalizes and reduces noise on the reconstructed signal to ensure that the output signal has good usability. The signal is smoothed by a digital filter to remove any residual high-frequency noise components. The final signal is transmitted to the terminal device for application scenarios such as geomagnetic monitoring, navigation positioning or security monitoring.

[0038] Normalization and noise reduction processing improve the stability and quality of the signal. The output signal has high accuracy and reliability and can be directly used in practical applications.

[0039] Module 1: Sensor Array Module The sensor array module of the present invention is the core foundation of the entire NEMS geomagnetic sensor back-end data acquisition and processing system. Through this module, accurate acquisition of geomagnetic signals can be achieved, providing a reliable input data source for subsequent dynamic sparse decoupling and physical constraint signal reconstruction. In the connection with subsequent modules, the sensor array module ensures the integrity and independence of the signal transmission process through an optimized observation matrix, thereby laying the foundation for signal processing in a complex geomagnetic field environment.

[0040] In this embodiment, the sensor array module adopts a two-dimensional sparsely distributed array. Generally, the array is composed of multiple highly sensitive NEMS sensors, each of which can sense and capture weak geomagnetic field change signals. The arrangement of the sensors is determined by an optimization algorithm to meet the restricted isometry (RIP) condition of sparse signal decomposition, so that the observed signal can be better decomposed and reconstructed in the mathematical model.

[0041] Specifically, the location of the sensor is determined by optimizing the observation matrix Sure, Represents the coupling relationship between the sensor array and the target signal. The optimization goal is to minimize the matrix orthogonality deviation. The optimization function form is:

[0042] in, is the identity matrix, Represents the Frobenius norm of the matrix. Through this design, it is possible to ensure that the signals observed by each sensor are highly independent, thereby reducing the interference problem caused by signal coupling. As an option, when calibrating the sensor array, a standard geomagnetic field signal source can be introduced for calibration. Specifically, by exposing each sensor to a uniform magnetic field environment of known intensity, the sensitivity and response deviation of the sensor are measured to correct the individual characteristic differences of each sensor. In a possible implementation method, the nonlinear response of the sensor can also be fitted during the calibration process, thereby improving the measurement accuracy of the entire array.

[0043] As an option, when calibrating the sensor array, a standard geomagnetic field signal source can be introduced for calibration. Specifically, by exposing each sensor to a uniform magnetic field environment of known strength, the sensitivity and response deviation of the sensor are measured to correct the individual characteristic differences of each sensor. In a possible implementation, the nonlinear response of the sensor can also be fitted during the calibration process to improve the measurement accuracy of the entire array. In some embodiments, in order to further improve the rationality of the sensor layout, the present invention optimizes the spatial layout of the sensor array based on the distribution law of geomagnetic dipoles. In general, the geomagnetic field strength changes inversely proportional to the cubic power of the distance, and its analytical formula is:

[0044] in, is the geomagnetic field strength vector, is the distance vector from the sensor to the geomagnetic dipole, For its model length, is the vacuum permeability, is the magnetic moment of the geomagnetic dipole.

[0045] In the specific implementation, the above formula is used to optimize the position of each sensor to ensure that the array can respond uniformly under different magnetic field strength conditions. For example, when a sensor is in a high magnetic field area, its distance parameter is adjusted. To reduce the dynamic range of its output signal, thereby avoiding signal overload; when in a low magnetic field area, the sensitivity of signal acquisition is improved by adjusting the position density of the sensor.

[0046] As another possible implementation, the sensor array of the present invention can also operate in extreme environments. Generally, in order to ensure the reliability of the array in high temperature, high humidity or strong electromagnetic interference environments, the sensor packaging material can be selected. For example, the use of corrosion-resistant materials and electromagnetic shielding design can effectively improve the environmental adaptability of the sensor.

[0047] The sensor array also pre-processes the raw signal through an internal signal conditioning circuit. In some embodiments, the conditioning circuit includes a filter, a signal amplifier, and an analog-to-digital conversion module, which is used to convert the analog signal into a digital signal and transmit it to a subsequent module in a standardized format. As an option, the filter can use a high-pass or band-pass filter to suppress low-frequency noise and high-frequency interference, thereby improving the effective signal-to-noise ratio of the signal.

[0048] In a possible implementation, the present invention also considers the power consumption of sensors. Since NEMS sensors have low power consumption characteristics, the overall energy consumption of the system can be kept at a low level even in a large-scale sensor array. At the same time, through a dynamic start-stop strategy, sensors in specific areas are activated only when needed, thereby further reducing the operating power consumption of the system.

[0049] In certain specific scenarios, the sensor array module of the present invention can also be linked with external auxiliary equipment. For example, in the geomagnetic anomaly detection scenario, the geomagnetic signal error collected by the sensor array can be corrected by linking with an inertial navigation device or a GNSS module. This multi-sensor fusion method can significantly improve the overall accuracy and robustness of the system.

[0050] Through the above design, this module realizes high-precision acquisition of geomagnetic signals and provides high-quality data input for subsequent dynamic sparse decoupling and physical constraint signal reconstruction. Generally, the optimized sensor array module can be applied to a variety of complex environments, including earthquake monitoring, space magnetic field detection and other fields, thereby further expanding the application scope of the present invention.

[0051] Module 2: Dynamic Sparse Decoupling Module The dynamic sparse decoupling module in the present invention is used to separate and decouple the aliased signals collected by the sensor array module. Through sparse modeling and optimization methods, the coupling relationship between the geomagnetic signal and the interference signal is processed, the core characteristics of the target geomagnetic signal are retained, and a reliable input is provided for the subsequent physical constraint signal reconstruction. When connected with the sensor array module, the module directly receives the signal data defined by the observation matrix, uses the dynamic sparse method to extract the sparse components of the target signal, and separates the background noise and other interference signals.

[0052] In this embodiment, the dynamic sparse decoupling module adopts a joint sparse decomposition technology, combining the dynamic sparse characteristics of time and space to establish a signal decoupling model. In general, the aliased signal collected by the sensor array can be expressed as:

[0053] in, is the observation signal matrix collected by the sensor, is the observation matrix, which describes the mapping relationship between the sensor and the field source; is a dynamic sparse matrix, which represents the sparse characteristics of the target signal over time; is the target geomagnetic signal; is Gaussian noise or other random interference signal, specifically, the dynamic sparse matrix The sparsity depends on the representation characteristics of the target signal in the transform domain (such as the wavelet domain or the discrete cosine transform domain). In general, the geomagnetic signal has limited non-zero components in these transform domains, so the target signal can be extracted by the sparse optimization method. In some embodiments, the wavelet transform is used to enhance the sparsity of the time-frequency domain, so as to more efficiently separate the target signal. As an option, the module introduces a joint sparse optimization model for decomposing the signals collected by multiple sensors. The model optimization goal is to minimize the fitting error under the sparsity constraint. The specific formula is as follows:

[0054] in, For sparse matrices - norm, used for sparsity constraints; Represents the fitting error between the observed signal and the target signal; is a regularization parameter used to balance sparsity and fitting error.

[0055] In one possible implementation, sparse optimization is achieved through an alternating optimization method. In general, the optimization process is divided into two steps: fixing the sparse matrix , optimize the target signal , the optimization problem is simplified to:

[0056] Fixed target signal , optimize the sparse matrix , then the optimization problem is:

[0057] Through the above-mentioned alternating iterations, the dynamic sparse decoupling module can accurately separate the target signal from the background noise. In some embodiments, in order to further improve the computing efficiency, the optimization process can be accelerated by combining the gradient descent method or the sparse reconstruction algorithm (such as the iterative threshold algorithm).

[0058] In another implementation, the module also supports adaptive adjustment of sparse parameters. Specifically, when the sparsity of the target signal changes over time or space, the module can dynamically update parameters to adapt to different signal characteristics, for example, when the background noise level is high, increase To strengthen the sparsity constraint, thereby reducing the impact of noise on the decoupling results.

[0059] As an extension, the dynamic sparse decoupling module also supports the joint processing of multi-band signals. By decomposing the sparse representation of the observed signal in multiple frequency bands, multiple source signals can be extracted simultaneously. For example, in some environments, the geomagnetic signal may be affected by both low-frequency background interference and high-frequency electromagnetic noise. At this time, the module can construct sparse optimization models for low and high frequencies respectively, and independently decouple and reconstruct signals in different frequency bands.

[0060] Generally, the module also includes signal normalization and noise filtering submodules for further processing the decoupled signal. In some embodiments, a bandpass filter can be used to remove high-frequency noise, and the gain difference between the sensor arrays can be eliminated through the normalization process, thereby improving the stability of the decoupled signal.

[0061] Under certain circumstances, the dynamic sparse decoupling module can also be linked with external computing devices. For example, in scenarios with high computational complexity, the sparse optimization process can be accelerated by transmitting the observation signal to an external GPU device. This method is particularly suitable for real-time signal processing requirements in large-scale sensor array environments.

[0062] Through the above design, the dynamic sparse decoupling module realizes the efficient separation and decoupling of complex aliasing signals. The module has significant advantages in processing dynamic sparse characteristics and multi-band signals, and provides high-quality input signals for subsequent physical constraint signal reconstruction. Combined with the module's various implementation methods, the module can adapt to the geomagnetic signal processing needs in different scenarios and has broad application potential.

[0063] Module 3: Physical Constraint Signal Reconstruction Module The physical constraint signal reconstruction module is one of the core functional modules in the system of the present invention. This module receives the target signal output by the dynamic sparse decoupling module and performs further nonlinear reconstruction and optimization processing on it. By introducing the physical distribution law of the geomagnetic field, this module corrects the nonlinear error in the decoupling signal to ensure that the output signal is more in line with the distribution characteristics of the actual geomagnetic field. In general, the main task of this module is to combine the physical laws of the geomagnetic field, establish a reconstruction model, and correct and restore the signal through an optimization algorithm, ultimately providing high-precision geomagnetic signals for subsequent terminal applications.

[0064] In this embodiment, the physical constraint signal reconstruction module constructs a constraint model based on the dipole distribution characteristics of the geomagnetic field. The distribution of the geomagnetic field in space can be expressed by the following formula:

[0065] in, For location The geomagnetic field strength vector at is the position vector, is the magnitude of the position vector, is the magnetic moment of the geomagnetic dipole, is the vacuum magnetic permeability. Specifically, the module first generates the theoretical geomagnetic field distribution through the above formula as a benchmark for signal correction. In a possible implementation, the module will perform a preliminary analysis of the input signal to determine the degree of fit between the signal and the theoretical geomagnetic field, and establish an optimization objective function based on the fitting error. The optimization goal is:

[0066] in, The geomagnetic signal reconstructed for the module, is the theoretical geomagnetic signal generated based on the above physical laws. It is the square of the second norm, which is used to measure the error of the signal.

[0067] In general, in order to improve the applicability of signal reconstruction, the module is combined with a deep learning model to further enhance the nonlinear processing capability of the signal. The structure of the deep learning model includes an input layer, a hidden layer, and an output layer. The physical law is embedded in the hidden layer as a regularization constraint. In this way, the module can integrate physical constraints on the basis of data-driven and significantly improve the accuracy of signal correction. In some embodiments, the loss function of the deep learning model is designed as follows:

[0068] in, is the fitting error based on the input signal, is the weight parameter, For physical law constraints, by adjusting the weight parameters , the module can strike a balance between the fitting accuracy of signal data and the consistency of physical laws.

[0069] As an option, the module uses the gradient descent method to solve the optimization objective function. During the optimization process, the module dynamically adjusts the learning rate to avoid overfitting or slow convergence. For example, when the noise level of the signal is high, the module will reduce the learning rate to improve the stability of the optimization process.

[0070] In another possible implementation, the module also supports signal reconstruction based on multi-source data fusion. Specifically, external auxiliary data (such as spatial physical parameters or environmental noise data) can be used as additional input to participate in the training and optimization of the deep learning model. For example, in certain complex magnetic field environments, the module can introduce time series data of geomagnetic storm events to further enhance the robustness of signal reconstruction.

[0071] In some embodiments, the module can also smooth the output signal to remove residual high-frequency noise. Specifically, by introducing a bandpass filter, the interference components in the signal can be effectively suppressed while retaining the main features of the target signal. This method is suitable for the final optimization of the reconstructed signal to make it smoother and more stable.

[0072] In specific scenarios, the module can also adapt to dynamic magnetic field environments. Generally, dynamic magnetic fields will cause the frequency of changes of input signals to increase. The module models and corrects the time-varying characteristics of signals by introducing time-related constraint models. For example, timing analysis technology can be used to model the time correlation of signals, thereby better describing the changing laws of dynamic magnetic fields.

[0073] Through the above design, the physical constraint signal reconstruction module can comprehensively correct the errors and nonlinear characteristics in the input signal and output a highly consistent target signal. Combined with the dynamic sparse decoupling module, this module exhibits extremely high signal processing capabilities in complex magnetic field environments. Its design not only improves the reconstruction accuracy of the signal, but also ensures the consistency of the signal with the laws of geomagnetic physics, providing strong support for subsequent data analysis and terminal applications.

[0074] Module 4: Data Output Module The data output module is the key link to achieve the final signal quality improvement and output in the system of the present invention. This module receives the target signal generated by the physical constraint signal reconstruction module, and further normalizes, optimizes noise reduction and converts the signal to meet different terminal requirements. In general, the data output module mainly ensures the stability and applicability of the signal, so that it can be directly applied to navigation positioning, geomagnetic monitoring or other fields. The module is closely connected with the aforementioned physical constraint signal reconstruction module, and the actual application value of the signal is improved by post-processing the reconstructed signal.

[0075] In this embodiment, the data output module first normalizes the received signal. Specifically, the normalization process normalizes the amplitude of the signal to make it suitable for the subsequent digital signal processing system. In general, the normalization formula can be expressed as:

[0076] in, is the normalized signal, is the input signal, and are the minimum and maximum values ​​of the signal, respectively. In some embodiments, the normalization process can also be combined with the adjustment of the dynamic range of the signal amplitude. For example, for weak geomagnetic signals in specific scenarios, an amplitude amplification factor can be set to enhance the expressiveness of the signal, thereby avoiding processing errors caused by too small a signal. As an option, the data output module also performs noise reduction on the signal. Generally, noise reduction uses digital filtering technology to remove high-frequency interference components in the signal. Specifically, a bandpass filter is used in this embodiment to achieve noise suppression of the target signal by setting the passband range of a specific frequency band. The transfer function of the filter can be expressed as:

[0077] in, is the frequency response of the filter, and are the lower and upper frequency limits of the bandpass filter respectively.

[0078] Specifically, in some embodiments, the parameters of the bandpass filter can be dynamically adjusted according to the characteristics of the target signal. For example, for low-frequency geomagnetic field signals, a lower passband range can be selected to suppress high-frequency noise in the environment; for wideband signals, the passband range can be expanded to retain more signal details.

[0079] In one possible implementation, the data output module also performs format conversion on the signal to adapt to the requirements of different terminal devices. Generally, the signal can be converted from the original floating-point format to the fixed-point format, thereby reducing storage space and transmission bandwidth. In some embodiments, the module supports switching of multiple output formats, such as standardized binary format or text format. This design enhances the flexibility of signal output and makes the system compatible with more application scenarios.

[0080] In addition, to improve the real-time performance of the signal, the module also supports block processing and streaming output functions. Specifically, the input signal is divided into signal blocks of fixed length, and each signal block is output immediately after processing without waiting for the processing of the entire signal to be completed. This method is suitable for application scenarios that require high real-time performance, such as dynamic navigation and rapid geomagnetic monitoring.

[0081] In some special cases, the data output module also has a signal calibration function. Generally, this function is used to fine-tune the offset or gain of the signal before output. For example, when there is still a small error in the target signal, the module can calibrate the signal to the expected range through linear adjustment. This function is of great significance in high-precision demand scenarios.

[0082] As an extension, this module also supports multi-channel combined output of signals. Specifically, the module can receive reconstructed signals from multiple sensors and generate a comprehensive signal through weighted averaging or other fusion methods. For example, in some complex geomagnetic field environments, the fusion of multi-channel signals can effectively improve the overall output accuracy of the system.

[0083] Generally, the data output module will transmit the optimized signal to the terminal device for further processing or analysis by the user. As a possible implementation method, the output signal can be directly used in the geomagnetic navigation system to complete high-precision position information solution by matching the geomagnetic signal with the magnetic field map.

[0084] Through the above design, the data output module realizes multiple optimization processing of the signal, and combines normalization, noise reduction, format conversion and real-time output functions, so that the signal has higher stability and adaptability, and can meet the needs of various complex application scenarios. Combined with other modules of the present invention, the data output module provides the system with powerful signal post-processing capabilities, ensuring the reliability and practicality of the overall system.

[0085] Example 2: Geomagnetic signal acquisition based on optimized sensor array This embodiment uses an optimized NEMS sensor array to collect geomagnetic signals to solve the problems of severe signal coupling and excessive redundant information in the prior art. The specific steps are as follows: Sensor layout design According to the principle of sparse distribution, the sensor array is optimally arranged. The Frobenius norm is used as the optimization target to calculate the position of each sensor to ensure that the orthogonality of the measurement matrix reaches the optimal state. The optimized measurement matrix satisfies the restricted equidistance condition and improves the independence of the sensor signal.

[0086] Sensor array calibration The sensors are calibrated one by one using a standard geomagnetic signal source. The sensitivity and response deviation of each sensor are calibrated in a uniform magnetic field to ensure the measurement consistency of all sensors in the array. At the same time, the response curve of each sensor under different magnetic field strengths is recorded for subsequent data correction.

[0087] Signal acquisition The sensor array with optimized layout is placed in the target magnetic field area. The weak geomagnetic signals are captured by NEMS sensors and transmitted to subsequent modules. The collected signals contain target geomagnetic signals and environmental interference signals, providing high-quality input data for subsequent dynamic sparse decoupling.

[0088] Through the above steps, this embodiment solves the problem of poor signal acquisition quality caused by unreasonable sensor array design in the prior art. At the same time, the optimized observation matrix effectively reduces signal coupling and improves the accuracy of subsequent signal processing.

[0089] Example 3: Aliasing signal processing based on dynamic sparse decoupling This embodiment combines dynamic sparse modeling and joint sparse decomposition technology to accurately decouple the collected aliased signals, solving the problems of incomplete signal separation and insufficient dynamic sparsity modeling in the prior art. The specific steps are as follows: Dynamic Sparse Modeling Perform sparsity analysis on the signals output by the sensor array and construct a dynamic sparse matrix ,The signal is decomposed in time and frequency domain by wavelet transform,sparse features are extracted, and the dynamic sparse matrix represents the sparse characteristics of the target signal in different time and space,providing a basis for subsequent optimization.

[0090] Signal decoupling optimization The aliased signal is separated by using the joint sparse decomposition model, and the optimization objective function is established:

[0091] in, The signal collected by the sensor, is the target geomagnetic signal, is the regularization parameter, the sparse matrix and the target signal matrix are fixed, the solution is iteratively solved, and finally the decoupled target signal is obtained.

[0092] Result Verification The separated signal is compared with the theoretical model to verify the decoupling accuracy. For the signal part with insufficient sparsity, the value is dynamically adjusted to improve the sparse constraint effect. Through the above steps, this embodiment realizes the precise separation of complex aliasing signals. The dynamic sparse modeling technology improves the flexibility of signal decoupling and solves the problem that the fixed sparse decomposition method in the prior art is difficult to adapt to the dynamic environment.

[0093] Example 4: Signal nonlinear reconstruction based on physical constraints This embodiment introduces the physical laws of the geomagnetic field to perform nonlinear reconstruction on the separated target signal to solve the problem of insufficient signal reconstruction accuracy and difficulty in matching the real physical distribution in the prior art. The specific implementation is as follows: Establishing a model of the geomagnetic field distribution According to the dipole distribution law of the geomagnetic field, a theoretical model of the geomagnetic field in space is constructed. This model describes the relationship between the strength of the geomagnetic field and the spatial position. The characteristic is that the strength of the geomagnetic field changes inversely with the distance, and the direction is determined by the magnetic moment. Through this theoretical model, physical constraints can be provided for signal reconstruction.

[0094] Signal correction and fitting The decoupled signal is compared and analyzed with the theoretical geomagnetic field distribution to identify and correct the deviation. Specifically, by comparing the difference between the reconstructed signal and the theoretical distribution, the characteristic parameters of the signal are adjusted to gradually approach the actual distribution characteristics of the geomagnetic field.

[0095] Optimization objectives and iterative solutions The error between the signal and the theoretical model is set as the optimization target, and the various components of the signal are adjusted through repeated iterative calculations. For example, the strength and direction of the signal will be gradually corrected according to the theoretical geomagnetic field until the deviation is reduced to an allowable range. The optimization process is combined with dynamic adjustment of the step size to avoid overfitting or non-convergence of calculations.

[0096] Combining Deep Learning Models In the process of signal reconstruction, a physics-guided deep learning framework was introduced to further improve the ability to process nonlinear signals. This deep learning model adds the constraints of physical laws to traditional data fitting to ensure that the signal achieves a balance between data consistency and physical matching.

[0097] This embodiment greatly improves the accuracy and reliability of signal reconstruction by combining physical constraint models and deep learning technology. Compared with the existing technology, this embodiment no longer relies on empirical models, but is based on physical laws, solving the problems of large deviation and poor applicability of signal reconstruction results.

[0098] Example 5: Data output optimization based on multifunctional processing This embodiment performs multifunctional processing on the reconstructed geomagnetic signal, including normalization, noise reduction, format conversion and real-time output, to solve the problems of poor signal output stability and insufficient applicability in the prior art. The specific implementation is as follows: Normalization The amplitude of the reconstructed signal may have a large dynamic range difference. In order to ensure the consistency of the signal in different devices and applications, the normalization processing technology is used to standardize the signal amplitude. The basic principle of normalization is to adjust the amplitude of the signal to a fixed range so that it is not limited by the original signal value range.

[0099] Noise reduction Digital filtering technology is used to reduce the noise of the reconstructed signal. The low-frequency interference and high-frequency noise in the signal are filtered out, and the main frequency band of the target signal is retained. By setting the passband range of the filter, the filtering parameters can be dynamically adjusted according to application requirements, thereby retaining the key features of the signal to the greatest extent.

[0100] Format Conversion In order to meet the needs of different terminal devices, signals are converted into multiple output formats. For example, on devices with limited storage, signals can be converted from floating-point format to fixed-point format, thereby saving storage space and transmission bandwidth. At the same time, for different application scenarios, signals can also be output in binary or standard text format for subsequent data processing.

[0101] Real-time streaming output For dynamic monitoring scenarios, the module supports streaming output function. The signal is output in the form of data blocks immediately after processing, without waiting for the processing of all signals to be completed. This function can meet the application requirements of real-time navigation or rapid monitoring, ensuring that the delay between signal processing and output is minimized.

[0102] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A NEMS geomagnetic sensor back-end data acquisition and processing system, characterized in that: include: A sensor array module for collecting geomagnetic field signals; A dynamic sparse decoupling module, used for performing sparse modeling and decoupling processing on the signals collected by the sensor array module, and separating the geomagnetic signal from the interference signal; Physical constraint signal reconstruction module, used to perform nonlinear reconstruction of the separated signal in combination with the physical distribution law of the geomagnetic field; The data output module is used to normalize the reconstructed geomagnetic signal and output it to the terminal device.

2. A NEMS geomagnetic sensor back-end data acquisition and processing system according to claim 1, characterized in that: The sensor array adopts an optimized distributed NEMS sensor array, and describes the coupling relationship between the sensor and the geomagnetic field signal through an observation matrix. The sensor array module adopts a sparse distribution arrangement, and by optimizing the physical position and sensitivity distribution of the sensor, it is ensured that the coupling relationship between the sensor array and the target signal meets specific restricted equidistant conditions to achieve sparse decomposition of the observation signal.

3. A NEMS geomagnetic sensor back-end data acquisition and processing system according to claim 2, characterized in that: The restricted equidistance condition is achieved by optimizing the orthogonality of the measurement matrix, and the goal of orthogonality optimization is to minimize the deviation between the measurement matrix and the unit matrix.

4. A NEMS geomagnetic sensor back-end data acquisition and processing system according to claim 1, characterized in that: The dynamic sparse decoupling module realizes signal decoupling based on a joint sparse decomposition model, and the optimization goal of the model is to minimize the error between the sensor observation signal and the target signal while satisfying the sparsity constraint.

5. The NEMS geomagnetic sensor back-end data acquisition and processing system according to claim 1, characterized in that: The dynamic sparse decoupling module adopts an alternating optimization method to perform alternating optimization on the sparse matrix and the target signal matrix respectively, so as to improve the signal decoupling accuracy and calculation efficiency.

6. A NEMS geomagnetic sensor back-end data acquisition and processing system according to claim 1, characterized in that: The physical constraint signal reconstruction module constructs a constraint model in combination with the dipole distribution law of the geomagnetic field. The dipole distribution law is that the geomagnetic field strength varies inversely with the distance of the position, and is used to correct the nonlinear characteristics of the signal.

7. A NEMS geomagnetic sensor back-end data acquisition and processing system according to claim 1, characterized in that: The constraint model achieves high-precision signal calibration by fitting the deviation between the reconstructed geomagnetic signal and the theoretical geomagnetic field signal, and a deviation fitting error minimization model is established based on an analytical equation of geomagnetic field distribution.

8. The NEMS geomagnetic sensor back-end data acquisition and processing system according to claim 1, characterized in that: The physical constraint signal reconstruction module completes the nonlinear reconstruction of the signal through a deep learning model. The deep learning model corrects the signal in combination with the physical laws of the geomagnetic field. The optimization goal of the signal correction is to simultaneously minimize the weighted sum of the data fitting error and the physical law constraint error.

9. A NEMS geomagnetic sensor back-end data acquisition and processing system according to claim 8, characterized in that: The deep learning model includes an input layer, a hidden layer and an output layer. The hidden layer embeds physical laws as regularization constraints to correct the deviation of nonlinear characteristics in the signal.

10. A NEMS geomagnetic sensor back-end data acquisition and processing system according to claim 1, characterized in that: The data output module processes the signal through normalization and filtering noise reduction to improve the quality of the output signal, and outputs the signal to a terminal device for geomagnetic monitoring, navigation positioning or security monitoring.

Images