A remote signaling detection method and device

Abstract

The embodiment of the application provides a kind of telesignalling detection method and device, belong to telesignalling detection field;The method and device include by chaos mapping generation first non-uniform sampling point;Obtain the first telesignalling signal of the first non-uniform sampling point, input the joint feature vector into reinforcement learning model, output sampling density adjustment coefficient;Adjust the sampling interval of first non-uniform sampling point, obtain second non-uniform sampling point, utilize improved orthogonal matching pursuit algorithm to carry out second telesignalling signal reconstruction, obtain second telesignalling signal reconstruction signal;Based on the second telesignalling signal reconstruction signal carries out abnormal discrimination, completes telesignalling detection.The application generates non-uniform sampling point by chaos mapping, extracts signal time-frequency feature in combination with deep learning model, and optimizes sampling density dynamically using reinforcement learning, finally realizes the adaptive reconstruction and abnormal detection of telesignalling signal through fractal dimension and energy analysis, can improve the efficiency and accuracy of power equipment telesignalling signal detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote signaling detection, and particularly relates to a remote signaling detection method and device. BACKGROUND

[0002] Remote signaling detection is mainly used for collecting and transmitting various protection and switching information, such as the switching state of equipment, alarm information, etc., so as to monitor the running state of the equipment and determine whether the equipment is normally running or has failed. Specifically, remote signaling detection is mainly used for monitoring the running state of the equipment, fault signals, alarm signals, etc. binary information (such as the on / off state of the switch, the action of the protection device, etc.), and transmitting these information to the monitoring center or dispatching center, so that the operating personnel can master the running state of the system in real time. With the progress of technology, remote signaling detection is continuously developing in the aspects of intelligence and digitization, providing important support for improving the safety and operation efficiency of the system.

[0003] In the field of power system monitoring, remote signaling detection technology realizes device state perception by collecting electrical quantity signals such as voltage and current. Traditional methods mostly use uniform sampling strategies, which are difficult to balance the details capture in the signal mutation stage and the resource consumption in the steady state stage. Although existing non-uniform sampling technologies (such as adaptive sampling) can dynamically adjust the sampling rate, their adjustment rules are mostly based on fixed thresholds or simple statistics, which cannot effectively cope with the complex and variable time-frequency characteristics of power signals, and are prone to cause loss of key information or storage redundancy.

[0004] The purpose of the embodiments of the present application is to provide a remote signaling detection method and device for improving the efficiency and accuracy of remote signaling signal detection. SUMMARY

[0005] In order to achieve the above-mentioned purpose, the embodiments of the present application provide a remote signaling detection method, comprising: setting a first chaotic mapping parameter, and generating a first non-uniform sampling point through chaotic mapping; acquiring a first remote signaling signal of the first non-uniform sampling point, and extracting a first remote signaling signal feature to obtain a joint feature vector; inputting the joint feature vector into a reinforcement learning model to output a sampling density adjustment coefficient; adjusting the sampling interval of the first non-uniform sampling point according to the sampling density adjustment coefficient to obtain a second non-uniform sampling point, performing signal collection based on the second non-uniform sampling point to obtain a second remote signaling signal; constructing an observation matrix, introducing the joint feature vector into an orthogonal matching pursuit algorithm to obtain an improved orthogonal matching pursuit algorithm, and using the improved orthogonal matching pursuit algorithm to reconstruct the second remote signaling signal to obtain a second remote signaling signal reconstruction signal; performing abnormality discrimination based on the second remote signaling signal reconstruction signal to complete remote signaling detection.

[0006] Optionally, the setting the first chaotic mapping parameter and generating the first non-uniform sampling point through chaotic mapping comprises: setting an initial parameter of the chaotic mapping; generating a chaotic sequence through the chaotic mapping iteratively, and discarding the chaotic sequence in a preheating stage; performing normalization processing on the retained chaotic sequence, and mapping the normalized chaotic sequence to an initial telesignaling signal time range; calculating an initial telesignaling signal complexity, and adjusting a sampling start time according to the initial telesignaling signal complexity; filling a blank area in the initial telesignaling signal time range, and removing a sampling point with a repeated timestamp to obtain the first non-uniform sampling point.

[0007] Optionally, the extracting the first telesignaling signal feature to obtain a joint feature vector comprises: extracting a local time domain feature of the first telesignaling signal using a convolutional neural network; extracting a long-term time sequence feature of the first telesignaling signal using a bidirectional long short-term memory network; and performing feature splicing on the local time domain feature and the long-term time sequence feature of the first telesignaling signal to form the joint feature vector.

[0008] Optionally, the extracting the first telesignaling signal feature to obtain a joint feature vector further comprises: segmenting the first telesignaling signal into time sequence segments according to a fixed time window; performing normalization processing on the time sequence segments, and performing two-dimensional time-frequency feature construction on the normalized time sequence segments to obtain a two-dimensional time-frequency matrix, and inputting the two-dimensional time-frequency matrix into a convolutional neural network.

[0009] Optionally, the inputting the joint feature vector into a reinforcement learning model to output a sampling density adjustment coefficient comprises: designing a reward function of the reinforcement learning model, and the reward function comprises a signal fidelity term, a sampling uniformity term and a time deviation penalty term.

[0010] Optionally, the adjusting a sampling interval of the first non-uniform sampling point according to the sampling density adjustment coefficient comprises: shortening the sampling interval of the first non-uniform sampling point when the sampling density adjustment coefficient is greater than a set coefficient; keeping the sampling interval of the first non-uniform sampling point unchanged when the sampling density adjustment coefficient is equal to the set coefficient; and expanding the sampling interval of the first non-uniform sampling point when the sampling density adjustment coefficient is less than the set coefficient.

[0011] Optionally, the reconstructing the second telesignaling signal using the improved orthogonal matching pursuit algorithm comprises: inputting the observation matrix and the second telesignaling signal into the improved orthogonal matching pursuit algorithm, and iteratively solving sparse coefficients.

[0012] Optionally, the performing abnormality discrimination based on the second telesignaling signal reconstruction signal comprises: calculating a fractal dimension of the second telesignaling signal reconstruction signal when the sparse coefficients are iteratively solved; and stopping iteration when the fractal dimension of the second telesignaling signal reconstruction signal is within a preset interval.

[0013] Optionally, the abnormality discrimination based on the second telesignaling signal reconstruction signal further includes: if the fractal dimension of the second telesignaling signal reconstruction signal is not within a set fractal dimension range, or the current second telesignaling signal reconstruction signal energy is greater than a set energy threshold, it is determined as abnormal, otherwise it is determined as normal.

[0014] In another aspect, the present application provides a telesignaling detection device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the telesignaling detection method.

[0015] Beneficial effects: the present application generates non-uniform sampling points through chaotic mapping, extracts signal time-frequency features combined with deep learning model, and dynamically optimizes sampling density using reinforcement learning, finally realizes adaptive reconstruction and abnormality detection of telesignaling signal through fractal dimension and energy analysis, which can improve the efficiency and accuracy of power equipment telesignaling signal detection.

[0016] Other features and advantages of the embodiments of the present application will be described in detail in the following specific implementation part. BRIEF DESCRIPTION OF DRAWINGS

[0017] The accompanying drawings are included to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the embodiments of the present application together with the following specific implementation, but do not constitute a limitation on the embodiments of the present application. In the drawings:

[0018] Figure 1 is the telesignaling detection flowchart of the present application.

[0019] Figure 2 is the reinforcement learning flowchart of the present application. DETAILED DESCRIPTION

[0020] The following will be described in detail in combination with the accompanying drawings. Figure 1 - the accompanying drawings Figure 2 The specific implementation of the embodiments of the present application will be described in detail. It should be understood that the specific implementation described herein is only used to illustrate and explain the embodiments of the present application, and is not used to limit the embodiments of the present application.

[0021] It should be noted that the acquisition, transmission, storage, use, processing and the like of data in the technical solutions of the present application comply with the relevant provisions of national laws and regulations. In the embodiments of the present application, some existing industry solutions may be mentioned, such as software, components, models and the like, which should be considered as exemplary, and the purpose is only to illustrate the feasibility of the implementation of the technical solutions of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0022] The present inventors found in the process of implementing the present application that the existing remote signaling detection method has the disadvantages of resource waste and feature loss caused by fixed sampling strategy, difficulty in adapting to non-stationary signal characteristics, dependence on single dimension analysis for feature extraction, lack of time-frequency joint representation ability, poor system flexibility caused by parameter configuration solidification, and the need for repeated debugging.

[0023] Embodiment 1

[0024] Reference Figures 1-2 For the first embodiment of the present application, the embodiment provides a remote signaling detection method, comprising:

[0025] S100: setting a first chaotic mapping parameter, and generating a first non-uniform sampling point through chaotic mapping.

[0026] Specifically, a chaotic mapping is selected, and initial parameters of the chaotic mapping are set, such as initial values and control parameters r of Logistic mapping. The chaotic mapping is iterated multiple times to obtain a chaotic sequence.

[0027] Preferably, step S100 can include: setting initial parameters of the chaotic mapping; generating a chaotic sequence through iteration of the chaotic mapping, discarding chaotic sequences in a preheating stage; performing normalization processing on the remaining chaotic sequences, mapping the normalized chaotic sequences to a time range of an initial remote signaling signal; calculating a complexity of the initial remote signaling signal, and adjusting a sampling start time according to the complexity of the initial remote signaling signal; filling blank areas in the time range of the initial remote signaling signal, and eliminating sampling points with repeated timestamps to obtain the first non-uniform sampling point.

[0028] Further, before formal sampling, the first 1000 values of the chaotic sequence are discarded, which is a “preheating stage”, to make the system enter a stable chaotic state.

[0029] Preferably, the remaining effective chaotic sequences are normalized, the normalized chaotic sequences are mapped to the time range of the initial remote signaling signal, the complexity of the initial remote signaling signal is calculated, and the position of the first effective sampling point is dynamically adjusted according to the complexity of the initial remote signaling signal.

[0030] Preferably, if the complexity of the initial remote signaling signal is higher than a set complexity threshold, such as the presence of high-frequency noise, the sampling start point is advanced (for example, sampling starts at 0.08 seconds). If the complexity of the initial remote signaling signal is lower than the set complexity threshold, sampling starts at the default 0.1 (1001th chaotic value) second.

[0031] Further, the distribution of the dynamically adjusted sampling points in the actual remote signaling signal time range is checked through statistical analysis, such as a histogram, to ensure that there is no obvious blank area, and the first non-uniform sampling point is obtained.

[0032] Preferably, the pseudo-randomness and ergodicity of the chaotic sequence are used to generate more uniformly distributed non-uniform sampling points, avoiding the problem that local features may be missed by traditional uniform sampling. The preheating stage (discarding the first 1000 values) is set to eliminate the influence of initial conditions on the chaotic sequence, ensure the stability of the sampling points, and adaptively adjust the starting sampling position according to the signal complexity, better adaptability; using histogram to check the distribution of sampling points, avoiding time blank area, ensuring signal full period coverage, reducing the risk of missing detection.

[0033] S200: Obtain the first telesign signal of the first non-uniform sampling point, and extract the first telesign signal feature to obtain a joint feature vector.

[0034] Specifically, the first non-uniform sampling point is used for sampling to obtain the first telesign signal.

[0035] Preferably, step S200 can include: segmenting the first telesign signal according to a fixed time window to form a time sequence segment; normalizing the time sequence segment, and constructing a two-dimensional time-frequency feature of the normalized time sequence segment to obtain a two-dimensional time-frequency matrix, and inputting the two-dimensional time-frequency matrix into a convolutional neural network.

[0036] Preferably, the normalized time sequence segment is subjected to short-time Fourier transform or wavelet transform to generate a two-dimensional time-frequency matrix (time-frequency-amplitude) as an input of a CNN (Convolutional Neural Network).

[0037] Further, a lightweight CNN is used to extract local time-frequency features of the two-dimensional time-frequency matrix; the lightweight CNN (such as 3 layers of convolutional layers + pooling layers) has 64 3x3 convolutional kernels in the first layer, 28 5x5 convolutional kernels in the second layer, and a maximum pooling layer 2x2 in the third layer; and the output is a time-frequency feature vector.

[0038] Further, the first telesign signal is input into a Bi-LSTM to capture the long-term dependence relationship (such as trend change and periodic fluctuation) of the first telesign signal, and the Bi-LSTM outputs a time sequence state vector, which represents the dynamic trend (rise / fall rate and mutation probability) of the first telesign signal.

[0039] Further, the time-frequency feature vector of the CNN and the time sequence state vector of the Bi-LSTM are spliced to form a joint feature vector.

[0040] Preferably, the joint feature vector is compressed to 128 dimensions through a fully connected layer to generate a compact state representation as the state input of the reinforcement learning model.

[0041] Preferably, the short-time Fourier transform or wavelet transform generates a time-frequency matrix, captures the joint characteristics of the time domain and frequency domain of the signal, and enhances the representation ability of complex patterns such as mutations and periodicity. The lightweight CNN is used to efficiently extract local time-frequency features, reduce computational overhead, and be suitable for real-time processing; the Bi-LSTM captures the long-term trend of the signal (such as the rising / falling rate and periodic fluctuation), and can improve the detection ability of slow degradation or gradual failure.

[0042] Preferably, the local features of the CNN and the global dynamic features of the Bi-LSTM are combined to form a comprehensive state representation, which can improve the accuracy of subsequent reinforcement learning decisions.

[0043] S300: inputting the joint feature vector into a reinforcement learning model and outputting a sampling density adjustment coefficient.

[0044] Figure 2 The reinforcement learning process to which the embodiments of the present application are applied is shown. In combination with the above Figure 2 For example, the reward function of the reinforcement learning model is designed, and the reward function includes a signal fidelity term, a sampling uniformity term, and a time deviation penalty term; the calculation formula of the reward function is as follows:

[0045]

[0046] wherein R represents the reward function, , , are weight coefficients, MSE is the mean square error between the signal reconstructed by the first non-uniform sampling point and the first remote signal, is the maximum allowed error, which is determined by experiment, is the standard deviation of the sampling time interval, is the maximum allowed sampling interval fluctuation scale, which is used to normalize the standard deviation; P is the time deviation penalty term.

[0047]

[0048] wherein P is the time deviation penalty term; represents the actual starting sampling time of the first non-uniform sampling point; is the ideal starting sampling time, which is dynamically set according to the complexity of the first remote signal; is a normalization factor, which is determined according to the total time length of the first remote signal.

[0049] It should be noted that is the signal fidelity term, is the sampling uniformity term, and P is the time deviation penalty term.

[0050] Further, the parameters of the reinforcement learning model (DQN, Deep Q-Network) are initialized, and an experience replay buffer is set. In each sampling period, the current joint feature vector is input, and the DQN outputs the Q value corresponding to the discretized value of the sampling density adjustment coefficient. The policy selection action (sampling density adjustment coefficient) is adopted. In the early stage, random exploration (high value, = 1.0) is adopted, and in the later stage, gradual transition to utilization (low value, = 0.01) is adopted, and the sampling density adjustment coefficient is output.

[0051] Preferably, the reinforcement learning realizes the balance among signal fidelity, resource efficiency, and abnormal detection robustness through dynamic sampling optimization and multi-objective reward mechanism, and significantly improves the adaptive ability and applicability of the remote signaling detection system.

[0052] Preferably, the reconstruction error is constrained by MSE to ensure that the sampling adjustment does not lose key information, to control the sampling interval fluctuation to avoid analysis deviation caused by excessive encryption or sparseness, and to dynamically calibrate the starting sampling point to reduce the error introduced by complexity misjudgment.

[0053] S400: Adjust the sampling interval of the first non-uniform sampling point according to the sampling density adjustment coefficient to obtain a second non-uniform sampling point, and perform signal collection based on the second non-uniform sampling point to obtain a second remote signaling signal.

[0054] When the sampling density adjustment coefficient > set coefficient, the sampling interval of the first non-uniform sampling point is shortened;

[0055] When the sampling density adjustment coefficient = set coefficient, the sampling interval of the first non-uniform sampling point remains unchanged;

[0056] When the sampling density adjustment coefficient < set coefficient, the sampling interval of the first non-uniform sampling point is expanded.

[0057] Preferably, the set coefficient is 1.

[0058] Further, after adjusting the sampling interval of the first non-uniform sampling point, a second non-uniform sampling point is obtained; and sampling is performed based on the second non-uniform sampling point to obtain a second remote signaling signal.

[0059] Preferably, the sampling density is adaptively adjusted, the sampling rate can be increased in signal mutation or high frequency area to enhance the abnormal capture ability, and the sampling rate can be reduced in a smooth area to reduce data redundancy and transmission load.

[0060] S500: Construct an observation matrix, introduce the joint feature vector into the orthogonal matching pursuit algorithm, obtain an improved orthogonal matching pursuit algorithm, use the improved orthogonal matching pursuit algorithm to reconstruct the second telesignaling signal, and obtain a second telesignaling signal reconstruction signal.

[0061] Specifically, an observation matrix is constructed, and the dimension of the observation matrix is MxN, where M is the number of second non-uniform sampling points (much smaller than N), and N is the theoretical total sampling point number of the signal under traditional uniform sampling, which is determined by the signal length and the sampling frequency.

[0062] Preferably, the observation matrix is subjected to wavelet multi-resolution decomposition, and the global trend and local mutation of the signal are matched in different levels to adapt to the multi-scale characteristics of non-uniform sampling.

[0063] Further, on the basis of the traditional OMP (Orthogonal Matching Pursuit, Orthogonal Matching Pursuit algorithm), the joint feature vector obtained before is introduced, the atoms of the observation matrix are weighted in combination with the joint feature vector, the observation matrix and the second telesignaling signal are input into the improved orthogonal matching pursuit algorithm, and the sparse coefficient is iteratively solved.

[0064] Further, the fractal dimension and signal energy of the second telesignaling signal reconstruction signal are calculated, and the abnormality is judged according to the fractal dimension and signal energy of the second telesignaling signal reconstruction signal.

[0065] Preferably, when the sparse coefficient is iteratively solved, the fractal dimension of the second telesignaling signal reconstruction signal is calculated; when the fractal dimension of the second telesignaling signal reconstruction signal enters a preset interval, the iteration is stopped to avoid overfitting. If the fractal dimension is abnormal or not within the set range, the iteration is extended to capture more details.

[0066] Further, Dmin represents the lower limit of the fractal dimension, Dmax represents the upper limit of the fractal dimension, Dcurrent represents the fractal dimension of the current signal to be detected after reconstruction; Ecurrent represents the current signal energy, Ethreshold represents the energy threshold set according to the upper limit of the normal telesignaling signal energy.

[0067] If [ ] or , it is determined that the detection result is abnormal, otherwise it is determined that the detection result is normal;

[0068] For example, It may be that the device motion is blocked, causing the signal fluctuation complexity to increase, such as mechanical jamming;

[0069] Possibly signal degeneration into a fixed value or a simple periodic signal, such as sensor failure;

[0070] Possibly a short circuit / overload caused by the sudden surge of current or voltage signal energy, and possibly high-frequency noise caused by communication interference.

[0071] Preferably, the wavelet multi-resolution decomposition matches the global trend and local mutation of the signal, which can improve the accuracy of signal reconstruction under non-uniform sampling; the introduction of the joint feature vector can enhance the reconstruction weight of key features (such as high-frequency noise and mutation points) and reduce missed detection. The fractal dimension can quantify the complexity of the signal and distinguish between normal fluctuations and abnormalities (such as simple periodic signals caused by sensor failure); through the fractal dimension, iteration can be terminated to avoid overfitting noise and improve detection efficiency; through the calculation of the energy threshold, the sudden surge of energy caused by short circuit / overload can be detected, covering multiple fault types, making the remote signaling detection result more accurate, and the number of detectable abnormalities more.

[0072] Embodiment 2

[0073] Reference Figures 1-2 As a second embodiment of the present application, based on the core architecture of embodiment 1, a differentiated remote signaling detection method is provided, comprising:

[0074] Specifically, the Tent chaotic mapping is selected, the initial parameters of the Tent chaotic mapping are set, the chaotic sequence is generated, and the generated chaotic sequence is normalized to the standard time interval through piecewise linear transformation.

[0075] Further, the complexity of the initial remote signaling signal is evaluated using multi-scale entropy, and the position of the first effective sampling point is dynamically adjusted according to the complexity of the initial remote signaling signal.

[0076] Preferably, the distribution of the sampling points is analyzed by kernel density estimation, a heat map model is established for the historical sampling point distribution, and when a continuous blank area exceeds a set threshold, a gradient descent algorithm is started to insert a supplementary sampling point, ensuring that key events are not missed, filling the time blank area, and obtaining the first non-uniform sampling point.

[0077] Further, the first non-uniform sampling point is used for sampling to obtain the first remote signaling signal, and the first remote signaling signal is subjected to continuous wavelet transform to generate a time-frequency matrix. The Morlet wavelet basis function is selected, the balance between the number of oscillations and the decay rate of its waveform is adjusted, the main lobe width is adapted to the duration of the common transient process, and the scale parameter is set to correspond to the actual frequency range of 10-100Hz, which can cover the typical fault feature frequency band of power equipment (such as mechanical vibration, electromagnetic interference, etc.).

[0078] Preferably, a more intensive scale grading is set in the mechanical vibration sensitive area (30-60Hz).

[0079] Preferably, each element of the time-frequency matrix corresponds to the energy intensity of the signal at a specific time (determined by the non-uniform sampling points) and frequency (mapped by the scale parameter), providing a two-dimensional input for subsequent CNN feature extraction.

[0080] Further, a two-channel CNN is designed, the first channel is a 3-layer convolution, the first layer uses a large-size convolution kernel to capture macro features of the time-frequency image, and the subsequent deep layer uses a small convolution kernel to extract local detail patterns; the second channel is a 1D convolution layer; after the two-channel features are spliced, a 64-dimensional time-frequency feature vector is compressed through a fully connected layer, the first telesignaling signal time-frequency feature is extracted, a gated recurrent unit is used to extract the timing feature of the first telesignaling signal; the time-frequency feature vector and the timing feature vector are weighted and fused by attention to generate a joint feature vector.

[0081] Further, the joint feature vector is input into a reinforcement learning model, a frequency domain fidelity constraint term is introduced into the reward function of the reinforcement learning model, and a sampling density adjustment coefficient is output.

[0082] Further, according to the sampling density adjustment coefficient, the sampling interval of the first non-uniform sampling point is adjusted to obtain a second non-uniform sampling point, and signal collection is performed based on the second non-uniform sampling point to obtain a second telesignaling signal.

[0083] Preferably, an observation matrix is constructed, the joint feature vector is introduced into an orthogonal matching pursuit algorithm to obtain an improved orthogonal matching pursuit algorithm, and the second telesignaling signal is reconstructed by using the improved orthogonal matching pursuit algorithm to obtain a second telesignaling signal reconstruction signal.

[0084] A three-level anomaly discrimination is constructed, including: a first level based on energy mutation detection of the second telesignaling signal reconstruction signal, for capturing microsecond-level transients; a second level for analyzing energy distribution anomalies of feature bands in the time-frequency matrix, and a third level for capturing cumulative effects of signal mode deviating from historical normal baseline by using a GRU network.

[0085] Further, a voting decision mechanism is used, and when any two levels alarm at the same time, the final anomaly determination is triggered, and the risk of single detection false alarm is reduced.

[0086] The application also provides a telesignaling detection device for implementing the telesignaling detection method, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the telesignaling detection method.

[0087] The embodiment of the application provides a storage medium having a program stored thereon, and the program is executed by a processor to implement the telesignaling detection method.

[0088] The embodiment of the present application provides a processor, which is used for running a program, wherein the program performs the telesignaling detection method when running.

[0089] The embodiment of the present application provides a device, which comprises a processor, a memory, and a program stored in the memory and capable of running on the processor, and the processor performs the telesignaling detection method when running the program. The device herein can be a server, a PC, a PAD, a mobile phone, or the like.

[0090] The present application also provides a computer program product, which is suitable for performing the program initialized with the method steps of the telesignaling detection method when executed on a data processing device.

[0091] Those skilled in the art should understand that the embodiments of the present application can provide a method, a device, or a computer program product. Therefore, the present application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt a computer program product in the form of being implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

[0092] The present application is described with reference to flowcharts and / or block diagrams of the method, device (system), and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices generate a device implemented in the flowcharts and / or block diagrams. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks.

[0093] These computer program instructions can also be stored in a computer readable storage medium, which can guide the computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer readable storage medium produce a product including instruction devices, which implement the flowcharts and / or block diagrams. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks.

[0094] These computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 Figure 1

[0095] In one typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0096] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) about which the processor can execute instructions. The memory can also include non-volatile memory, such as read only memory (ROM), electrically programmable read only memory (EPROM), electrically erasable programmable read only memory (EEPROM), flash memory, or other memory technologies, about which the processor can execute instructions. The memory is an example of computer readable media.

[0097] Computer readable media includes permanent and non-permanent, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically programmable read only memory (EEPROM), flash memory or other memory technologies, compact disc read only memory (CD-ROM), digital versatile disc (DVD), or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to computing devices. According to the definition herein, computer readable media does not include transitory media, such as modulated data signals and carrier waves.

[0098] It should also be noted that the terms "comprising," "including," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements recited, but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without limitation, an element preceded by "comprises a" does not, without more constraints, foreclose the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0099] ​​The above merely provides an example of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application should be included in the scope of claims of the present application.

Claims

1. A method of telesignalling detection, characterized in that The method comprises the following steps: setting first chaotic mapping parameters and generating first non-uniform sampling points through chaotic mapping; obtaining first telesignaling signals of the first non-uniform sampling points and extracting features of the first telesignaling signals to obtain a joint feature vector; inputting the joint feature vector into a reinforcement learning model to output a sampling density adjustment coefficient; adjusting a sampling interval of the first non-uniform sampling points according to the sampling density adjustment coefficient to obtain second non-uniform sampling points, collecting signals based on the second non-uniform sampling points to obtain second telesignaling signals; constructing an observation matrix and introducing the joint feature vector into an orthogonal matching pursuit algorithm to obtain an improved orthogonal matching pursuit algorithm, and reconstructing the second telesignaling signals by using the improved orthogonal matching pursuit algorithm to obtain second telesignaling reconstructed signals; wherein the step of reconstructing the second telesignaling signals by using the improved orthogonal matching pursuit algorithm comprises inputting the observation matrix and the second telesignaling signals into the improved orthogonal matching pursuit algorithm and iteratively solving sparse coefficients; performing abnormality discrimination based on the second telesignaling reconstructed signals to complete telesignaling detection.

2. The remote signaling detection method of claim 1, wherein The step of setting first chaotic mapping parameters and generating first non-uniform sampling points through chaotic mapping comprises the following steps: setting initial parameters of chaotic mapping; iteratively generating chaotic sequences through the chaotic mapping and discarding chaotic sequences in a preheating stage; normalizing the retained chaotic sequences and mapping the normalized chaotic sequences to an initial telesignaling signal time range; calculating an initial telesignaling signal complexity and adjusting a sampling start time according to the initial telesignaling signal complexity; filling blank areas in the initial telesignaling signal time range and removing sampling points with repeated time stamps to obtain first non-uniform sampling points.

3. The remote signaling detection method of claim 1, wherein The step of extracting features of the first telesignaling signals to obtain a joint feature vector comprises the following steps: extracting local time domain features of the first telesignaling signals by using a convolutional neural network; extracting long-term time sequence features of the first telesignaling signals by using a bidirectional long short-term memory network; concatenating the local time domain features and the long-term time sequence features of the first telesignaling signals to form a joint feature vector.

4. The remote signaling detection method according to claim 3, wherein The step of extracting features of the first telesignaling signals to obtain a joint feature vector further comprises the following steps: segmenting the first telesignaling signals into time sequence segments according to a fixed time window; normalizing the time sequence segments, constructing two-dimensional time-frequency features of the normalized time sequence segments to obtain a two-dimensional time-frequency matrix, and inputting the two-dimensional time-frequency matrix into a convolutional neural network.

5. The remote signaling detection method of claim 1, wherein The step of inputting the joint feature vector into a reinforcement learning model to output a sampling density adjustment coefficient comprises the following step: designing a reward function of the reinforcement learning model, wherein the reward function comprises a signal fidelity term, a sampling uniformity term and a time deviation penalty term.

6. The remote signaling detection method of claim 1, wherein The step of adjusting a sampling interval of the first non-uniform sampling points according to the sampling density adjustment coefficient comprises the following steps: when the sampling density adjustment coefficient is greater than a set coefficient, shortening the sampling interval of the first non-uniform sampling points; when the sampling density adjustment coefficient is equal to the set coefficient, keeping the sampling interval of the first non-uniform sampling points unchanged; when the sampling density adjustment coefficient is less than the set coefficient, expanding the sampling interval of the first non-uniform sampling points.

7. The remote signaling detection method of claim 1, wherein The abnormality discrimination based on the second telesignaling signal reconstructed signal comprises: In the iteration solving of the sparse coefficient, the fractal dimension of the second telesignaling signal reconstructed signal is calculated; When the fractal dimension of the second telesignaling signal reconstructed signal is in the preset interval, the iteration is stopped.

8. The remote signaling detection method according to claim 7, wherein The abnormality discrimination based on the second telesignaling signal reconstructed signal further comprises: If the fractal dimension of the second telesignaling signal reconstructed signal is not in the set fractal dimension range, or the current second telesignaling signal reconstructed signal energy is greater than the set energy threshold, it is determined as abnormal, otherwise it is determined as normal.

9. A remote signaling detection apparatus, characterized by comprising: The telesignaling detection device comprises a memory, a processor and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the telesignaling detection method according to any one of claims 1-8.

Images