A signal intelligent recognition and filtering method based on deep learning

Abstract

The present invention belongs to the field of signal processing and relates to a signal intelligent identification and filtering method based on deep learning. The method analyzes the abnormal deviation of a target waveform relative to its standard form by comprehensively considering the influence of the industrial Internet of Things environment in which the target device is currently located, and preliminarily detects whether the target waveform has signs of potential communication interference. If the preliminary detection shows that there is, the corresponding characteristics of the target waveform in time domain, frequency domain and time-frequency domain are deeply analyzed, and the probability of carrying an interference signal in the communication signal of the target device currently received is calculated. The authenticity of the potential communication interference signs of the target waveform is further verified, and an adaptation filter is selected for filtering the interference signal attribute in the communication signal of the current target device to effectively evaluate the interference filtering effect of the communication signal of the target device, greatly improving the flexibility, accuracy and reliability of interference identification of communication signals of industrial Internet of Things devices, and promoting the continuous advancement of interference filtering technology of industrial Internet of Things communication signals.

Description

Technical Field

[0001] The present invention belongs to the field of signal processing and relates to a signal intelligent recognition and filtering method based on deep learning. Background Art

[0002] With the continuous leap forward of information technology, signal recognition technology has been widely infiltrated into multiple fields such as communications, radar, electronic countermeasures and industrial IoT equipment, showing strong application potential. In the operation system of industrial IoT equipment, communication signal interference recognition has become one of the core elements to ensure its stable operation. Its importance is not only reflected in the solid support for the stability and reliability of the communication system, but also has a profound impact on the security line and efficiency optimization of information transmission. However, in the face of the challenges of the increasingly complex industrial environment and the diversity of interference types, there is an urgent need to develop a more efficient and intelligent signal recognition solution.

[0003] There are also some related solutions in the prior art that can be applied to the identification of communication signal interference of industrial Internet of Things devices. For example, a communication interference intelligent identification method, system and terminal with Chinese patent publication number CN113435247B first intelligently characterizes the received communication interference signal, extracts the time-frequency distribution, fractional Fourier transform and constellation diagram of the communication interference signal as the deep network input, and then builds a distributed network and constructs a sub-network model based on small sample learning. Finally, the distributed network is trained through federated learning to obtain the global optimal output model and complete the communication interference type identification. It can complete the intelligent identification of the communication interference type with the support of a small sample set, and has good recognition accuracy, generalization performance and sample privacy protection.

[0004] However, the above scheme still has the limitations common in the existing technology, which are specifically manifested as follows: 1. The existing technology ignores the complex influence of industrial environmental factors in the identification of interference with the communication signals of industrial Internet of Things equipment. As a result, even if the communication signals of the industrial Internet of Things equipment are not actually interfered with, they may be mistakenly identified as being interfered with due to slight deviations from the standard state. This not only reduces the accuracy of signal recognition, but may also cause unnecessary fault alarms or affect the normal operation of the communication system, thereby adversely affecting the continuity and efficiency of industrial production.

[0005] 2. The existing technology for interference identification and analysis of industrial Internet of Things equipment communication signals fails to comprehensively integrate the three levels of time domain, frequency domain and time-frequency domain, and often only focuses on one or two levels for analysis, resulting in inaccurate and unreliable identification of interference signals.

[0006] 3. The existing technology fails to conduct a detailed analysis of the specific properties and specific processing of interference signals carried in the communication signals of industrial Internet of Things devices, and fails to provide decision-making ideas for the effective suppression of interference signals, thereby failing to further improve the stability and reliability of industrial Internet of Things device communications. Summary of the invention

[0007] In view of this, in order to solve the problems raised in the above background technology, a signal intelligent recognition and filtering method based on deep learning is proposed.

[0008] The purpose of the present invention can be achieved through the following technical solutions: The present invention provides a signal intelligent recognition and filtering method based on deep learning, including: S1. Receive the time domain waveform of the current target device communication signal and mark it as the target waveform, based on the current industrial Internet of Things environment in which the target device is located and the preset standard communication signal time domain waveform of the target device stored in the WEB cloud, evaluate the abnormal deviation of the target waveform relative to its standard form, so as to preliminarily detect whether the target waveform has signs of potential communication interference.

[0009] S2. If the target waveform is initially detected to have signs of potential communication interference, the target waveform is gradually analyzed in time domain, frequency domain, and time-frequency domain to determine the probability of an interference signal being carried in the communication signal of the target device currently received, so as to verify the authenticity of the signs of potential communication interference in the target waveform.

[0010] S3. If the target waveform is verified to have potential communication interference signs, then analyze the interference signal properties in the communication signal of the current target device and select an adaptation filter to filter it.

[0011] S4. Perform secondary analysis on the filtered target waveform to evaluate the interference filtering effect of the target device communication signal and provide feedback.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) The present invention uses simulated signals and historical signals of the same environment as references, comprehensively considers the influence of the current industrial Internet of Things environment in which the target device is located, conducts an abnormal deviation analysis of the target waveform relative to its standard form, and preliminarily detects whether the target waveform has signs of potential communication interference, thereby greatly improving the flexibility, accuracy and reliability of communication signal interference detection of industrial Internet of Things devices.

[0013] (2) The present invention conducts an in-depth analysis of the characteristics of the target waveform in the time domain, frequency domain, and time-frequency domain, accurately calculates the probability of the interference signal being carried in the currently received communication signal of the target device, and further verifies the authenticity of the potential communication interference signs of the target waveform, thereby providing a strong basis for subsequent interference filtering processing and effectively improving the overall performance and stability of industrial Internet of Things device communications.

[0014] (3) The present invention analyzes the properties of interference signals carried in the communication signal of the current target device and selects an adaptation filter to filter it, thereby achieving accurate suppression of interference in the communication signal of the industrial Internet of Things device, thereby ensuring the communication quality of the industrial Internet of Things device.

[0015] (4) The present invention evaluates the interference filtering effect of the communication signal of the target device after filtering processing, and effectively feeds back the communication signal filtering processing capability, thereby helping to promote the continuous advancement and upgrading of the industrial Internet of Things communication signal interference filtering technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0017] Figure 1 The present invention is a flowchart of the steps for implementing the method. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments 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.

[0019] See also Figure 1 As shown, the present invention provides a signal intelligent identification and filtering method based on deep learning, including: S1. receiving the time domain waveform of the current target device communication signal and marking it as the target waveform, based on the industrial Internet of Things environment in which the target device is currently located and the preset standard communication signal time domain waveform of the target device stored in the WEB cloud, evaluating the abnormal deviation of the target waveform relative to its standard form, so as to preliminarily detect whether the target waveform has signs of potential communication interference.

[0020] Specifically, the evaluation of the abnormal deviation of the target waveform relative to its standard form includes: S11. Importing the target waveform and the preset standard communication signal time domain waveform of the target device into Matlab software for comparison, and analyzing the deviation of the target waveform relative to its standard form. .

[0021] S12. Collect the layout information and meteorological information of the industrial Internet of Things environment where the target device is currently located, build a communication propagation simulation environment in which the target device is currently located in the industrial Internet of Things environment without interference sources, simulate and capture the time domain waveforms of the communication signals of a preset number of target devices, record them as each simulation waveform, analyze the deviation of each simulation waveform relative to its standard form and calculate the mean, and use the mean calculation result as the first-order reasonable deviation of the target waveform relative to its standard form .

[0022] S13. Retrieve the time domain waveforms of each historical communication signal recorded in the historical communication log of the target device under the condition that the meteorological information of the current industrial Internet of Things environment is consistent, record them as each historical waveform, analyze the deviation of each historical waveform relative to its standard form, take the historical waveform with a deviation less than the preset benchmark deviation as the reference waveform, and calculate the mean value of the deviation of each reference waveform relative to its standard form as the second-order reasonable deviation of the target waveform relative to its standard form .

[0023] It should be noted that the analysis and calculation process of the deviation of each simulation waveform relative to its standard form and each historical waveform relative to its standard form is consistent with the analysis and calculation process of the deviation of the target waveform relative to its standard form.

[0024] S14. By computational model Evaluate the abnormal deviation of the target waveform from its standard shape ,in express , They represent the preset weights corresponding to the first and second order reasonable deviations, .

[0025] For example, the above The specific values ​​can be 0.5, 0.5.

[0026] Specifically, the The specific analysis process includes: aligning the start and end points of the target waveform and the preset standard communication signal time domain waveform of the target device, respectively constructing the best fitting function corresponding to the target waveform and the preset standard communication signal time domain waveform of the target device, recorded as , according to the formula Analyze the deviation of the target waveform from its standard form in the time domain ,in Respectively indicate that the start and end points correspond to the preset time points.

[0027] The target waveform and the preset standard communication signal time domain waveform of the target device are converted into frequency domain modes, the target waveform spectrum and its comparison standard spectrum are obtained, the main lobe frequency range is defined, and the main lobe area and side lobe area of ​​the target waveform spectrum and its comparison standard spectrum are screened, and the main lobe difference index and side lobe difference index of the target waveform spectrum with respect to its comparison spectrum are calculated, and the accumulated value of the two is used as the deviation of the target waveform relative to its standard form in the frequency domain.

[0028] It should be noted that the mainlobe difference index of the above-mentioned target waveform spectrum diagram with respect to its comparison spectrum diagram is calculated by substituting the signal amplitudes corresponding to the same frequency points in the mainlobe area of ​​the target waveform spectrum diagram and its comparison standard spectrum diagram into the root mean square error formula. Similarly, the sidelobe difference index of the target waveform spectrum diagram with respect to its comparison spectrum diagram is obtained.

[0029] The deviation of the target waveform from its standard form at the time domain level and the frequency domain level is accumulated with the product of the corresponding preset weights to obtain the deviation of the target waveform from its standard form.

[0030] Specifically, the condition for the preliminary detection of the target waveform for the presence of potential communication interference signs is that the abnormal deviation of the target waveform relative to its standard form is greater than or equal to the preset abnormal deviation warning threshold. .

[0031] The embodiment of the present invention uses simulated signals and historical signals of the same environment as references, comprehensively considers the impact of the current industrial Internet of Things environment in which the target device is located, conducts abnormal deviation analysis of the target waveform relative to its standard form, and preliminarily detects whether the target waveform has signs of potential communication interference, thereby greatly improving the flexibility, accuracy and reliability of communication signal interference detection of industrial Internet of Things devices.

[0032] S2. If the target waveform is initially detected to have signs of potential communication interference, the target waveform is gradually analyzed in time domain, frequency domain, and time-frequency domain to determine the probability of an interference signal being carried in the communication signal of the target device currently received, so as to verify the authenticity of the signs of potential communication interference in the target waveform.

[0033] Specifically, the method of determining the probability of an interference signal being carried in the communication signal of the currently received target device includes: S21. acquiring the signal amplitude of each unit sampling point on the target waveform, calculating the absolute difference in signal amplitude between each unit sampling point on the target waveform and its adjacent unit sampling points, retrieving each burr sampling point on the target waveform to evaluate the degree of mutation in the time domain communication of the target waveform, calculating the overall standard deviation corresponding to the signal amplitude of each unit sampling point on the target waveform, and using it as the degree of fluctuation in the time domain communication of the target waveform.

[0034] The degree of time domain communication mutation of the comprehensive target waveform and time domain communication fluctuation , analyze the time domain quality coefficient of the target device communication signal currently received.

[0035] It should be noted that the search criteria for each glitch sampling point on the target waveform is as follows: if the absolute difference in signal amplitude between a unit sampling point on the target waveform and its adjacent unit sampling point is greater than or equal to a preset signal amplitude deviation allowable threshold between adjacent sampling points, then the unit sampling point is used as a glitch sampling point.

[0036] It should be noted that the specific calculation formula for the time domain quality coefficient of the target device communication signal currently received is: .

[0037] S22. Obtain the harmonic characteristic information, frequency characteristic interval and bandwidth value in the time domain waveform of the preset standard communication signal of the target device and the spectrum diagram of the target waveform converted to the frequency domain mode respectively, wherein the harmonic characteristic information includes the frequency multiples and amplitude difference of each harmonic relative to the fundamental wave, compare and analyze the relative growth degree of the harmonics, the intensity of the additional frequency proportion and the relative expansion degree of the bandwidth of the target waveform frequency domain communication, so as to evaluate the frequency domain quality coefficient of the currently received target device communication signal.

[0038] It should be noted that the specific comparative analysis process of the relative growth degree of harmonics, the intensity of the additional frequency ratio and the relative expansion degree of bandwidth of the above-mentioned target waveform frequency domain communication is as follows: after the target waveform is converted into a frequency domain mode, the same-frequency harmonics and different-frequency harmonics relative to its standard form are retrieved, and the frequency multiples of the different-frequency harmonics relative to the fundamental wave are calculated. and amplitude difference , the difference in the growth rate of each harmonic of the same frequency relative to the fundamental wave in its standard form ,in Indicates the number of each harmonic frequency. , Indicates the number of each harmonic of the same frequency, , according to the formula Analyze the relative growth of harmonics in frequency domain communications of target waveforms ,in It is the preset reference amplitude difference of the harmonic relative to the fundamental wave.

[0039] It should be noted in particular that there are two situations for the values ​​of the growth amplitude difference of the above-mentioned same-frequency harmonics relative to the fundamental wave in their standard forms. First, if the amplitude difference of a certain same-frequency harmonic relative to the fundamental wave after the target waveform is converted to the frequency domain mode is less than or equal to the amplitude difference of the same-frequency harmonic relative to the fundamental wave after the target device preset standard communication signal time domain waveform is converted to the frequency domain mode, then the growth amplitude difference of the same-frequency harmonic relative to the fundamental wave in its standard form is set to 0; if the amplitude difference of a certain same-frequency harmonic relative to the fundamental wave after the target waveform is converted to the frequency domain mode is greater than the amplitude difference of the same-frequency harmonic relative to the fundamental wave after the target device preset standard communication signal time domain waveform is converted to the frequency domain mode, then the absolute difference between the two is taken as the growth amplitude difference of the same-frequency harmonic relative to the fundamental wave in its standard form.

[0040] Compare the frequency characteristic intervals in the time domain waveform of the preset standard communication signal of the target device and the frequency spectrum diagram of the target waveform converted to the frequency domain mode, and retrieve the spurious frequencies and their amplitudes in the frequency spectrum diagram of the target waveform converted to the frequency domain mode relative to the frequency spectrum diagram of the preset standard communication signal of the target device ,in Indicates the number of each spurious frequency, , obtain the deviation frequency span of each spurious frequency relative to the frequency characteristic interval in the spectrum diagram of the frequency domain mode converted from the time domain waveform of the preset standard communication signal of the target device , according to the formula Analyze the additional frequency ratio intensity of the target waveform frequency domain communication ,in Preset the frequency span of the frequency characteristic interval in the frequency spectrum diagram of the frequency domain mode of the standard communication signal time domain waveform conversion for the target device, is the preset reference signal amplitude.

[0041] Subtract the bandwidth value in the frequency spectrum of the target waveform conversion frequency domain mode from the bandwidth value in the frequency spectrum of the target device preset standard communication signal time domain waveform conversion frequency domain mode. If the calculated difference is greater than 0, perform a ratio analysis on the calculated difference and the bandwidth value in the frequency spectrum of the target device preset standard communication signal time domain waveform conversion frequency domain mode to obtain the relative expansion degree of the bandwidth of the target waveform frequency domain communication. , if the calculated difference is less than or equal to 0, the relative expansion degree of the bandwidth of the target waveform frequency domain communication is set to 0.

[0042] It should also be noted that the frequency domain quality coefficient of the target device communication signal currently received is The specific calculation formula is: .

[0043] S23. Apply short-time Fourier transform to the target waveform, and construct the time-frequency domain function corresponding to the target waveform according to the preset window function and window length, denoted as ,in Represent the time variable and frequency variable respectively, and calculate the Shannon entropy of the communication signal of the target device currently received ,in Indicates the target device communication signal currently being received Time variables, The probability distribution of energy at the frequency variable, , compare the Shannon entropy of the target device communication signal currently received with the reasonable Shannon entropy interval of the target device communication standard signal stored in the WEB cloud. If the Shannon entropy of the target device communication signal currently received is within the reasonable Shannon entropy interval of the target device communication standard signal, then set the time-frequency domain quality coefficient of the target device communication signal currently received to 1. If the Shannon entropy of the target device communication signal currently received is outside the reasonable Shannon entropy interval of the target device communication standard signal, then calculate the deviation ratio of the Shannon entropy of the target device communication signal currently received relative to the reasonable Shannon entropy interval of the target device communication standard signal. , according to the formula Evaluate the time-frequency domain quality coefficient of the target device communication signal currently received .

[0044] It should be noted that the above The specific calculation process is as follows: if the Shannon entropy of the target device communication signal currently received is less than the lower limit of the reasonable Shannon entropy interval of the target device communication specification signal, then the lower limit of the interval is subtracted from the Shannon entropy of the target device communication signal currently received, and the ratio of the calculated difference to the lower limit of the interval is further used as On the contrary, if the Shannon entropy of the target device communication signal currently received is greater than the upper limit of the reasonable Shannon entropy interval of the target device communication standard signal, the upper limit of the interval is subtracted from the Shannon entropy of the target device communication signal currently received, and the absolute value of the calculated difference is further divided by the ratio of the upper limit of the interval as .

[0045] S24. Based on the preset reference evaluation of the industrial IoT device communication signal stored in the WEB cloud, the time domain quality coefficient, frequency domain quality coefficient and time-frequency domain quality coefficient corresponding to the communication signal are compared and analyzed, and the interference impact of the currently received target device communication signal relative to the time domain level, frequency domain level and time-frequency domain level are accumulated. The products of each with the corresponding preset weight are added up, and the accumulated value is substituted into the hyperbolic tangent function as the function variable to output the probability of the currently received target device communication signal carrying an interference signal.

[0046] It should be noted that the specific comparison and analysis process of the interference influence degree of the currently received target device communication signal relative to the time domain level, frequency domain level and time-frequency domain level is as follows: the time domain quality coefficient of the currently received target device communication signal is Time domain quality coefficient corresponding to preset reference evaluation of IIoT device communication signals For comparison, if , then the interference impact of the currently received target device communication signal relative to the time domain level is set to 0. , then by the formula Calculate the interference impact of the currently received target device communication signal relative to the time domain level Similarly, the interference impact of the currently received target device communication signal at the frequency domain level and the time-frequency domain level is compared and analyzed.

[0047] Specifically, the evaluation of the time domain communication mutation degree of the target waveform includes: sorting out the overall over-radiation value of the target waveform burr sampling point according to the signal amplitude and time point of each burr sampling point of the target waveform , duration of continuous superradiation span and the proportion of super radiation frequency , according to the formula Analyze the time domain communication mutation degree of the target waveform ,in Indicates the overall preset time span of the target waveform. Indicates the first The signal amplitude of each unit sampling point is Indicates the number of each unit sampling point, .

[0048] It should be noted that the process of obtaining the overall over-radius value of the burr sampling point of the above-mentioned target waveform is: subtract the absolute difference in signal amplitude between each burr sampling point of the target waveform and its adjacent unit sampling point from the preset signal amplitude deviation allowable threshold between adjacent sampling points, and add up the calculated differences, and the accumulated value is used as the overall over-radius value of the burr sampling point of the target waveform.

[0049] The process of obtaining the continuous super-radial span duration of the above-mentioned target waveform burr sampling points is: collecting the time points corresponding to each burr sampling point of the target waveform, sorting out the continuous time periods corresponding to the target waveform burr sampling points, and taking the maximum span duration in each continuous time period as the continuous super-radial span duration of the target waveform burr sampling point.

[0050] The process of obtaining the over-radius frequency ratio of the above-mentioned target waveform burr sampling points is as follows: counting the number of target waveform burr sampling points, and taking the ratio of the number of the target waveform burr sampling points to the total number of unit sampling points on the target waveform as the over-radius frequency ratio of the target waveform burr sampling points.

[0051] Specifically, the condition for verifying the authenticity of the potential communication interference signs of the target waveform is that the probability of the interference signal being carried in the currently received communication signal of the target device is greater than or equal to a preset warning probability threshold.

[0052] The embodiment of the present invention conducts an in-depth analysis of the characteristics of the target waveform in the time domain, frequency domain, and time-frequency domain, accurately calculates the probability of carrying an interference signal in the currently received communication signal of the target device, further verifies the authenticity of the potential communication interference signs of the target waveform, provides a strong basis for subsequent interference filtering processing, and effectively improves the overall performance and stability of industrial Internet of Things device communications.

[0053] S3. If the target waveform is verified to have potential communication interference signs, then analyze the interference signal properties in the communication signal of the current target device and select an adaptation filter to filter it.

[0054] Specifically, the analysis of the interference signal attribute carried in the current target device communication signal includes: extracting the energy probability distribution of the currently received target device communication signal at each time variable and each frequency variable, accumulating the corresponding energy probability distributions of all time variables with the same frequency, and obtaining the overall energy probability distribution of the currently received target device communication signal at each frequency variable; if the overall energy probability distribution at a certain frequency variable is greater than or equal to a preset reasonable energy probability distribution threshold, then the frequency variable is identified as an interference frequency variable, thereby determining the interference frequency points corresponding to the interference signal carried in the current target device communication signal, and integrating the starting frequency, ending frequency and bandwidth of each continuous interference frequency band as the interference signal attribute carried in the current target device communication signal.

[0055] Specifically, the specific analysis process of selecting the adaptive filter includes: extracting the starting frequency, ending frequency and bandwidth of a continuous interference frequency band in the interference signal carried in the communication signal of the current target device, and comparing it with the standard frequency band presented after the time domain waveform of the preset standard communication signal of the target device is converted into the frequency domain mode; if the continuous interference frequency band overlaps with the standard frequency band, then selecting the band-stop filter as the adaptive filter for the continuous interference frequency band.

[0056] If there is no overlap between the continuous interference frequency band and the standard frequency band, the spectral positions of the continuous interference frequency band and the standard frequency band are further determined. If the continuous interference frequency band is located in the low frequency band and the standard frequency band is located in the high frequency band, a high-pass filter is selected as the adaptation filter of the continuous interference frequency band. If the continuous interference frequency band is located in the high frequency band and the standard frequency band is located in the low frequency band, a low-pass filter is selected as the adaptation filter of the continuous interference frequency band. On the contrary, if the spectral positions of the continuous interference frequency band and the standard frequency band are both low frequency bands or high frequency bands, and the bandwidth of the continuous interference frequency band is less than the preset bandwidth, a bandpass filter is selected as the adaptation filter of the continuous interference frequency band.

[0057] Thus, an adaptive filter is selected for each continuous frequency segment of the interference signal carried in the communication signal of the current target device.

[0058] The embodiment of the present invention analyzes the interference signal properties carried in the communication signal of the current target device, selects an adaptation filter to filter it, and achieves accurate suppression of interference in the communication signal of the industrial Internet of Things device, thereby ensuring the communication quality of the industrial Internet of Things device.

[0059] S4. Perform secondary analysis on the filtered target waveform to evaluate the interference filtering effect of the target device communication signal and provide feedback.

[0060] Specifically, the evaluation of the interference filtering effect of the target device communication signal includes: obtaining the abnormal deviation of the target waveform after filtering relative to its standard form; , combined with the abnormal deviation of the target waveform relative to its standard form before filtering , calculate the interference filtering effect evaluation index of the target device communication signal , .

[0061] It should be noted that the above analysis process of the abnormal deviation of the target waveform after filtering relative to its standard form is consistent with the analysis process of the abnormal deviation of the target waveform before filtering relative to its standard form, and will not be repeated here.

[0062] The embodiment of the present invention evaluates the interference filtering effect of the communication signal of the target device after filtering processing, and effectively feeds back the communication signal filtering processing capability, thereby helping to promote the continuous advancement and upgrading of the industrial Internet of Things communication signal interference filtering technology.

[0063] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they shall all fall within the protection scope of the present invention.

Claims

1. A signal intelligent recognition and filtering method based on deep learning, characterized in that: The method includes: S1. Receive the current target device communication signal time domain waveform and mark it as the target waveform. Based on the current industrial IoT environment of the target device and the preset standard communication signal time domain waveform of the target device stored in the WEB cloud, evaluate the abnormal deviation of the target waveform relative to its standard form, so as to preliminarily detect whether the target waveform has signs of potential communication interference; S2. If the target waveform is initially detected to have potential communication interference signs, then the time domain, frequency domain, and time-frequency domain feature analysis is gradually performed on the target waveform to determine the probability of the interference signal being carried in the communication signal of the target device currently received, so as to verify whether the target waveform has potential communication interference signs or not; S3. If the target waveform is verified to have potential communication interference signs, then analyze the interference signal properties in the communication signal of the current target device and select an adaptation filter to filter it; S4. Perform secondary analysis on the filtered target waveform to evaluate the interference filtering effect of the target device communication signal and provide feedback.

2. The signal intelligent identification and filtering method based on deep learning according to claim 1 is characterized in that: The method of evaluating the abnormal deviation of the target waveform relative to its standard form includes: S11. importing the target waveform and the preset standard communication signal time domain waveform of the target device into Matlab software for comparison, and analyzing the deviation of the target waveform relative to its standard form. ; S12. Collect the layout information and meteorological information of the industrial Internet of Things environment where the target device is currently located, build a communication propagation simulation environment in which the target device is currently located in the industrial Internet of Things environment without interference sources, simulate and capture the time domain waveforms of the communication signals of a preset number of target devices, record them as each simulation waveform, analyze the deviation of each simulation waveform relative to its standard form and calculate the mean, and use the mean calculation result as the first-order reasonable deviation of the target waveform relative to its standard form ; S13. Retrieve the time domain waveforms of each historical communication signal recorded in the historical communication log of the target device under the condition that the meteorological information of the current industrial Internet of Things environment is consistent, record them as each historical waveform, analyze the deviation of each historical waveform relative to its standard form, take the historical waveform with a deviation less than the preset benchmark deviation as the reference waveform, and calculate the mean value of the deviation of each reference waveform relative to its standard form as the second-order reasonable deviation of the target waveform relative to its standard form ; S14. By computational model Evaluate the abnormal deviation of the target waveform from its standard shape ,in express , They represent the preset weights corresponding to the first and second order reasonable deviations, .

3. The signal intelligent identification and filtering method based on deep learning according to claim 2 is characterized in that: Said The specific analysis process includes: aligning the start and end points of the target waveform and the preset standard communication signal time domain waveform of the target device, respectively constructing the best fitting function corresponding to the target waveform and the preset standard communication signal time domain waveform of the target device, recorded as , according to the formula Analyze the deviation of the target waveform from its standard form in the time domain ,in Respectively indicate that the starting and ending points correspond to the preset time points; Convert the target waveform and the preset standard communication signal time domain waveform of the target device into frequency domain mode, obtain the target waveform spectrum and its comparison standard spectrum, define the main lobe frequency range, and use it to screen the main lobe area and side lobe area of ​​the target waveform spectrum and its comparison standard spectrum, calculate the main lobe difference index and side lobe difference index of the target waveform spectrum with respect to its comparison spectrum, and use the accumulated value of the two as the deviation degree of the target waveform relative to its standard form in the frequency domain; The deviation of the target waveform from its standard form at the time domain level and the frequency domain level is accumulated with the product of the corresponding preset weights to obtain the deviation of the target waveform from its standard form.

4. The signal intelligent identification and filtering method based on deep learning according to claim 2 is characterized in that: The condition for the preliminary detection of the target waveform for potential communication interference is that the abnormal deviation of the target waveform relative to its standard form is greater than or equal to the preset abnormal deviation warning threshold. .

5. The signal intelligent identification and filtering method based on deep learning according to claim 1 is characterized in that: The method of determining the probability of the interference signal being carried in the communication signal of the target device currently received comprises: S21. acquiring the signal amplitude of each unit sampling point on the target waveform, calculating the absolute difference between the signal amplitude of each unit sampling point on the target waveform and its adjacent unit sampling points, retrieving each burr sampling point on the target waveform to evaluate the degree of sudden change in the time domain communication of the target waveform, calculating the overall standard deviation corresponding to the signal amplitude of each unit sampling point on the target waveform, and using it as the degree of fluctuation in the time domain communication of the target waveform; The time domain quality coefficient of the communication signal of the target device currently received is analyzed by comprehensively analyzing the time domain communication mutation degree and the time domain communication fluctuation degree of the target waveform; S22. Obtain the harmonic characteristic information, frequency characteristic interval and bandwidth value in the time domain waveform of the preset standard communication signal of the target device and the spectrum diagram of the target waveform converted to the frequency domain mode, wherein the harmonic characteristic information includes the frequency multiples and amplitude difference of each harmonic relative to the fundamental wave, and compare and analyze the relative growth degree of the harmonics, the additional frequency proportion intensity and the relative expansion degree of the bandwidth of the target waveform frequency domain communication, so as to evaluate the frequency domain quality coefficient of the target device communication signal currently received; S23. Apply short-time Fourier transform to the target waveform, and construct the time-frequency domain function corresponding to the target waveform according to the preset window function and window length, denoted as ,in Represent the time variable and frequency variable respectively, and calculate the Shannon entropy of the communication signal of the target device currently received ,in Indicates the target device communication signal currently being received Time variables, The probability distribution of energy at the frequency variable, , compare the Shannon entropy of the target device communication signal currently received with the reasonable Shannon entropy interval of the target device communication specification signal stored in the WEB cloud, and evaluate the time-frequency domain quality coefficient of the target device communication signal currently received; S24. Based on the preset reference evaluation of the industrial IoT device communication signal stored in the WEB cloud, the time domain quality coefficient, frequency domain quality coefficient and time-frequency domain quality coefficient corresponding to the communication signal are compared and analyzed, and the interference impact of the currently received target device communication signal relative to the time domain level, frequency domain level and time-frequency domain level are accumulated. The products of each with the corresponding preset weight are added up, and the accumulated value is substituted into the hyperbolic tangent function as the function variable to output the probability of the currently received target device communication signal carrying an interference signal.

6. The signal intelligent identification and filtering method based on deep learning according to claim 5 is characterized in that: The evaluation of the time domain communication mutation degree of the target waveform includes: sorting out the overall over-radiation value of the burr sampling point of the target waveform according to the signal amplitude and time point of each burr sampling point of the target waveform , continuous superradiation span duration and the proportion of super radiation frequency , according to the formula Analyze the time domain communication mutation degree of the target waveform ,in Indicates the overall preset time span of the target waveform. Indicates the first The signal amplitude of each unit sampling point is Indicates the number of each unit sampling point, .

7. The signal intelligent identification and filtering method based on deep learning according to claim 5 is characterized in that: The condition for verifying the authenticity of the potential communication interference signs of the target waveform is that the probability of carrying an interference signal in the currently received communication signal of the target device is greater than or equal to a preset warning probability threshold.

8. The signal intelligent identification and filtering method based on deep learning according to claim 5 is characterized in that: The analysis of the interference signal attribute carried in the current target device communication signal includes: extracting the energy probability distribution of the currently received target device communication signal at each time variable and each frequency variable, accumulating the corresponding energy probability distributions of all time variables with the same frequency, obtaining the overall energy probability distribution of the currently received target device communication signal at each frequency variable, if the overall energy probability distribution at a certain frequency variable is greater than or equal to a preset reasonable energy probability distribution threshold, identifying the frequency variable as an interference frequency variable, thereby determining the interference frequency points corresponding to the interference signal carried in the current target device communication signal, and integrating the starting frequency, ending frequency and bandwidth of each continuous interference frequency band as the interference signal attribute carried in the current target device communication signal.

9. The signal intelligent identification and filtering method based on deep learning according to claim 8 is characterized by: The specific analysis process of selecting the adaptive filter includes: extracting the starting frequency, ending frequency and bandwidth of a continuous interference frequency band in the interference signal carried in the communication signal of the current target device, and comparing it with the standard frequency band presented after the time domain waveform of the preset standard communication signal of the target device is converted into the frequency domain mode; if the continuous interference frequency band overlaps with the standard frequency band, then selecting the band-stop filter as the adaptive filter of the continuous interference frequency band; If there is no overlap between the continuous interference frequency band and the standard frequency band, the spectrum positions of the continuous interference frequency band and the standard frequency band are further determined; if the continuous interference frequency band is located in the low frequency band and the standard frequency band is located in the high frequency band, a high-pass filter is selected as the adaptation filter of the continuous interference frequency band; if the continuous interference frequency band is located in the high frequency band and the standard frequency band is located in the low frequency band, a low-pass filter is selected as the adaptation filter of the continuous interference frequency band; conversely, if the spectrum positions of the continuous interference frequency band and the standard frequency band are both low frequency bands or high frequency bands, and the bandwidth of the continuous interference frequency band is less than the preset bandwidth, a bandpass filter is selected as the adaptation filter of the continuous interference frequency band; Thus, an adaptive filter is selected for each continuous frequency segment of the interference signal carried in the communication signal of the current target device.

10. The signal intelligent identification and filtering method based on deep learning according to claim 4 is characterized in that: The evaluation of the interference filtering effect of the target device communication signal includes: obtaining the abnormal deviation degree of the target waveform after filtering relative to its standard form , combined with the abnormal deviation of the target waveform relative to its standard form before filtering , calculate the interference filtering effect evaluation index of the target device communication signal , .

Images