Adaptive filtering method and system for pulse signal based on entropy value of multi-band signal

Abstract

The application provides a pulse signal adaptive filtering method and system based on multi-band signal entropy value. Wherein, a plurality of different frequency band satellite communication signals are received, and phase offset, amplitude value and frequency value are extracted; a multi-dimensional feature group of each frequency band is generated, and the signal-to-noise ratio and frequency change of the corresponding frequency band are extracted therefrom. Based on the signal-to-noise ratio and frequency change of all frequency bands, the filtering parameters are dynamically adjusted through a pre-set three-level tuning structure, and a filtered signal is obtained. The filtered signal is transmitted to a federal learning framework, and the multi-dimensional feature group is combined to calculate the multi-band signal entropy value; the adaptive filtering parameters are optimized based on the entropy value, and are globally aggregated after differential privacy processing to generate an aggregated feature map group. Through the channel and spatial attention mechanism, the effective information of the high-entropy value frequency band is enhanced, and the noise of the low-entropy value frequency band is suppressed, and the pulse signal adaptive filtering based on the multi-band signal entropy value is completed. The technical scheme provided by the application improves the accuracy of pulse signal adaptive filtering.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of adaptive filtering, in particular to a pulse signal adaptive filtering method and system based on multi-band signal entropy values. BACKGROUND

[0002] In scenarios such as satellite communication and radar detection that rely on pulse signal transmission, multi-band pulse signals often face interference from complex electromagnetic environments, and effective signals in high-entropy bands and noise interference in low-entropy bands are often mixed and superimposed. To ensure the stability and accuracy of signal transmission, adaptive filtering technology is needed to dynamically distinguish between effective information and noise, accurately preserve high-entropy band signals, and efficiently suppress low-entropy band noise, while adapting to dynamic changes in signal characteristics in different bands, which puts strict requirements on the real-time performance and relevance of the filtering method.

[0003] Currently, for adaptive filtering of multi-band pulse signals, a common solution is to construct a filtering model based on a single feature (such as signal-to-noise ratio) and adjust the filtering parameters uniformly for signals in each band using a pre-set fixed threshold. This solution first extracts a single feature value of the signal, then adjusts the filtering parameters linearly based on the comparison result of the feature value and the threshold, to achieve filtering of signals in different bands.

[0004] However, this solution has obvious defects: relying on a single feature cannot fully reflect the complex characteristics of multi-band pulse signals, and the differences in characteristics between high-entropy bands and low-entropy bands cannot be accurately captured; the fixed threshold parameter adjustment method lacks dynamic adaptability and cannot match real-time changes in signal characteristics, which can easily lead to excessive filtering of effective information in high-entropy bands or excessive noise remaining in low-entropy bands, ultimately affecting the transmission quality of pulse signals and the accuracy of subsequent processing. SUMMARY

[0005] The present application provides a pulse signal adaptive filtering method and system based on multi-band signal entropy values to solve the problems of incomplete feature reflection, lack of dynamic adaptation, and poor filtering effect in the prior art.

[0006] In a first aspect, the embodiments of the present application provide a pulse signal adaptive filtering method based on multi-band signal entropy values, comprising:

[0007] receiving satellite communication signals in multiple different bands, and extracting phase offsets, amplitude values, and frequency values in the satellite communication signals;

[0008] based on the phase offsets, amplitude values, and frequency values, generating a multi-dimensional feature group corresponding to each band, and extracting the signal-to-noise ratio and frequency change of the corresponding band from each multi-dimensional feature group;

[0009] Based on the signal-to-noise ratio and the frequency variation of all frequency bands, the filter characteristic parameters are dynamically adjusted through a preset three-level tuning structure, and the satellite communication signals of different frequency bands are filtered based on the dynamically adjusted filter characteristic parameters to obtain filtered signals.

[0010] The filtered signals are transmitted to a federated learning framework, and based on the federated learning framework and the multi-dimensional feature group, a multi-band signal entropy value is calculated.

[0011] Based on the multi-band signal entropy value, the adaptive filtering parameters are optimized, and the optimized adaptive filtering parameters are globally aggregated through a differential privacy protection rule to generate an aggregated feature map group.

[0012] Based on the channel and spatial attention mechanism, the aggregated feature map group is weighted to enhance the effective information of high-entropy frequency bands and suppress the noise interference of low-entropy frequency bands, and the multi-band signal entropy value-based pulse signal adaptive filtering is completed.

[0013] Optionally, based on the signal-to-noise ratio and the frequency variation of all frequency bands, the filter characteristic parameters are dynamically adjusted through a preset three-level tuning structure, including:

[0014] Based on the signal-to-noise ratio of all frequency bands, the filter bandwidth adjustment parameters of the corresponding frequency bands are calculated according to a preset inverse proportional correspondence relationship to construct a first-level tuning structure.

[0015] Based on the frequency variation of all frequency bands, the center frequency adjustment parameters of the corresponding frequency bands are calculated according to a preset proportional correspondence relationship to construct a second-level tuning structure.

[0016] The signal-to-noise ratio and the frequency variation of all frequency bands are integrated, and based on the integration result, a signal comprehensive evaluation parameter is obtained through a preset weighting calculation method.

[0017] Based on the signal comprehensive evaluation parameter, the signal gain adjustment parameters of the corresponding frequency bands are determined according to a preset correspondence relationship to construct a third-level tuning structure.

[0018] The first-level tuning structure, the second-level tuning structure, and the third-level tuning structure are integrated to obtain a preset three-level tuning structure.

[0019] Based on the three-level tuning structure, the filter bandwidth adjustment parameters, the center frequency adjustment parameters, and the signal gain adjustment parameters are mapped to the filter characteristics of the corresponding frequency bands to dynamically adjust the filter characteristic parameters and obtain the dynamically adjusted filter characteristic parameters.

[0020] Optionally, the filter bandwidth adjustment parameter, the center frequency adjustment parameter and the signal gain adjustment parameter are mapped into filter characteristics of a corresponding frequency band based on the three-level tuning structure to dynamically adjust filter characteristic parameters, and dynamically adjusted filter characteristic parameters are obtained, including:

[0021] A basic parameter set of each frequency band is determined, and the basic parameter set includes a basic bandwidth parameter, a basic center frequency parameter and a basic gain parameter.

[0022] The filter bandwidth adjustment parameter in the three-level tuning structure is combined with the basic bandwidth parameter of a corresponding frequency band based on a preset bandwidth mapping rule to obtain a dynamic bandwidth parameter.

[0023] The center frequency adjustment parameter in the three-level tuning structure is combined with the basic center frequency parameter of a corresponding frequency band based on a preset frequency mapping rule to obtain a dynamic center frequency parameter.

[0024] The signal gain adjustment parameter in the three-level tuning structure is combined with the basic gain parameter of a corresponding frequency band based on a preset gain mapping rule to obtain a dynamic gain parameter.

[0025] The dynamic bandwidth parameter, the dynamic center frequency parameter and the dynamic gain parameter are mapped into filter characteristics of a corresponding frequency band to generate dynamically adjusted filter characteristic parameters of the corresponding frequency band.

[0026] Optionally, the filter signal is transmitted to a federated learning framework, and a multi-frequency band signal entropy value is calculated based on the federated learning framework and the multi-dimensional feature group, including:

[0027] The filter signal is divided into a plurality of sub-signals based on a preset frequency band division standard, and the plurality of sub-signals are respectively transmitted to nodes corresponding to each frequency band in the federated learning framework, so that each node receives a sub-signal of a corresponding frequency band, establishes an association between the sub-signal and the multi-dimensional feature group, and based on the association result, performs feature parameter calculation on the associated data in the sub-signal and the multi-dimensional feature group according to a preset local calculation rule to obtain a signal feature distribution parameter of a corresponding frequency band.

[0028] Based on the federated learning framework and a preset cross-node cooperation rule, the signal feature distribution parameters of each node corresponding to a frequency band are integrated to form a global feature distribution parameter.

[0029] The global feature distribution parameter is calculated according to a preset entropy value calculation rule to obtain a multi-frequency band signal entropy value.

[0030] Optionally, based on the association result, a feature parameter of the sub-signal and the association data in the multi-dimensional feature group is calculated according to a preset local calculation rule to obtain a signal feature distribution parameter of the corresponding frequency band, including:

[0031] The phase offset of the association data in the multi-dimensional feature group is compared with a preset reference phase value to generate a phase difference value sequence;

[0032] The maximum difference value and the minimum difference value are extracted from the phase difference value sequence, and a phase fluctuation range parameter is calculated based on the maximum difference value and the minimum difference value;

[0033] The amplitude value of the association data is calculated based on a time domain distribution statistical rule in the preset local calculation rule to obtain an amplitude change rate parameter;

[0034] The frequency value in the association data is calculated by difference with a preset reference frequency value to obtain a frequency offset parameter;

[0035] According to the phase fluctuation range parameter, the amplitude change rate parameter and the frequency offset parameter, a feature parameter set is constructed;

[0036] The feature parameter set is analyzed for distribution characteristics, and based on the result of the distribution characteristic analysis, a signal feature distribution parameter of the corresponding frequency band is generated.

[0037] Optionally, based on the multi-frequency band signal entropy value, an adaptive filtering parameter is optimized, and the optimized adaptive filtering parameter is globally aggregated by a differential privacy protection rule to generate an aggregated feature map group, including:

[0038] According to the size of the multi-frequency band signal entropy value, an adjustment weight of the adaptive filtering parameter is determined;

[0039] Based on the adjustment weight, the initial adaptive filtering parameter is step-by-step optimized to obtain the optimized adaptive filtering parameter;

[0040] Based on the differential privacy protection rule, a privacy perturbation value meeting a preset threshold condition is superimposed on the optimized adaptive filtering parameter to obtain a target adaptive filtering parameter;

[0041] Based on a preset parameter fusion ratio, the target adaptive filtering parameter is input to an aggregation node in the federated learning framework for global integration to generate a global parameter;

[0042] The global parameter is reorganized according to a preset feature dimension to generate an aggregated feature map group.

[0043] Optionally, the channel and space-based attention mechanism is used to weight the aggregated feature map group to enhance the effective information of high-entropy frequency bands and suppress the noise interference of low-entropy frequency bands, to complete adaptive filtering of the pulse signal based on the multi-band signal entropy value, comprising:

[0044] extracting initial channel features and initial space features corresponding to each frequency band from the aggregated feature map group;

[0045] determining first weighting coefficients and second weighting coefficients for the initial channel features and the initial space features based on the multi-band signal entropy value of each frequency band and the attention mechanism;

[0046] performing enhancement or attenuation processing on the initial channel features based on the first weighting coefficients to obtain target channel features;

[0047] performing regional weight allocation on the initial space features based on the second weighting coefficients to obtain target space features;

[0048] superimposing and fusing the target channel features and the target space features to generate a weighted aggregated feature map group;

[0049] using the weighted aggregated feature map group to enhance the effective information of high-entropy frequency bands and suppress the noise interference of low-entropy frequency bands to complete adaptive filtering of the pulse signal based on the multi-band signal entropy value.

[0050] In a second aspect, the embodiments of the present application provide a pulse signal adaptive filtering system based on a multi-band signal entropy value, comprising:

[0051] a receiving module configured to receive satellite communication signals of multiple different frequency bands, and extract phase offsets, amplitude values and frequency values in the satellite communication signals;

[0052] a first generating module configured to generate a multi-dimensional feature group corresponding to each frequency band based on the phase offsets, the amplitude values and the frequency values, and extract a signal-to-noise ratio and a frequency change of the corresponding frequency band from each multi-dimensional feature group;

[0053] an adjusting module configured to dynamically adjust filter characteristic parameters through a preset three-level tuning structure based on the signal-to-noise ratios and the frequency changes of all frequency bands, and perform filter processing on the satellite communication signals of different frequency bands based on the dynamically adjusted filter characteristic parameters to obtain a filter signal;

[0054] a transmission module configured to transmit the filter signal to a federated learning framework, and calculate a multi-band signal entropy value based on the federated learning framework and the multi-dimensional feature group;

[0055] a second generation module configured to optimize adaptive filtering parameters based on the multi-band signal entropy value, and to generate an aggregated feature map group by globally aggregating the optimized adaptive filtering parameters based on a differential privacy protection rule;

[0056] a weighting module configured to weight the aggregated feature map group based on a channel and spatial attention mechanism to enhance effective information of a high-entropy value band and suppress noise interference of a low-entropy value band, and to complete the multi-band signal entropy value-based adaptive filtering of the pulse signal.

[0057] In a third aspect, an embodiment of the present application provides a computing device including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the multi-band signal entropy value-based adaptive filtering of the pulse signal according to any one of the first aspect.

[0058] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program, which, when executed by a computer, implements the multi-band signal entropy value-based adaptive filtering of the pulse signal according to any one of the first aspect.

[0059] In the present application, a multi-band signal entropy value-based adaptive filtering method of a pulse signal is provided, including: receiving satellite communication signals of multiple different frequency bands, and extracting phase offsets, amplitude values and frequency values in the satellite communication signals; generating multi-dimensional feature groups corresponding to respective frequency bands based on the phase offsets, the amplitude values and the frequency values, and extracting signal-to-noise ratios and frequency variation amounts of the corresponding frequency bands from the multi-dimensional feature groups; dynamically adjusting filtering characteristic parameters through a preset three-level tuning structure based on the signal-to-noise ratios and the frequency variation amounts of all the frequency bands, filtering the satellite communication signals of different frequency bands based on the dynamically adjusted filtering characteristic parameters to obtain filtered signals; transmitting the filtered signals to a federated learning framework, calculating multi-band signal entropy values based on the federated learning framework and the multi-dimensional feature groups; optimizing adaptive filtering parameters based on the multi-band signal entropy values, and globally aggregating the optimized adaptive filtering parameters based on a differential privacy protection rule to generate an aggregated feature map group; and weighting the aggregated feature map group based on a channel and spatial attention mechanism to enhance effective information of a high-entropy value band and suppress noise interference of a low-entropy value band, and to complete the multi-band signal entropy value-based adaptive filtering of the pulse signal.

[0060] The application has the following advantages: by receiving satellite communication signals of multiple different frequency bands and extracting phase offset, amplitude value and frequency value, the basic characteristic parameters of the signals can be obtained to provide original data for subsequent feature analysis; by generating a multi-dimensional feature group based on the phase offset, amplitude value and frequency value and extracting the signal-to-noise ratio and frequency change, the basic features can be converted into key indicators reflecting signal quality and dynamic characteristics to provide a basis for adjusting filter parameters; by dynamically adjusting filter characteristic parameters using a preset three-level tuning structure based on the signal-to-noise ratio and frequency change of all frequency bands and obtaining a filtered signal, the filter processing can be initially adapted to the characteristic differences of signals of each frequency band; by transmitting the filtered signal to a federated learning framework and calculating multi-band signal entropy values in combination with the multi-dimensional feature group, the quantization indicators reflecting the complexity of signals of each frequency band can be obtained using a distributed framework; by optimizing adaptive filter parameters based on multi-band signal entropy values and globally aggregating an aggregated feature map group through a differential privacy protection rule, the global features of the multi-node information can be obtained while protecting the privacy of the parameters; by weighting the aggregated feature map group based on the channel and spatial attention mechanism, the effective information of the high-entropy frequency band is enhanced and the noise interference of the low-entropy frequency band is suppressed to complete the filtering, so that the effective signal can be enhanced and the noise can be suppressed accurately, and adaptive filtering of the pulse signal is realized.

[0061] Further, based on the signal-to-noise ratio and frequency change of all frequency bands, a first-level tuning structure containing a filter bandwidth adjustment parameter, a second-level tuning structure containing a center frequency adjustment parameter, and a third-level tuning structure containing a signal gain adjustment parameter are constructed respectively. After integrating the three-level tuning structures, the above adjustment parameters are mapped to the filter characteristics of the corresponding frequency band in combination with the basic parameter set of each frequency band, and the dynamically adjusted filter characteristic parameters are obtained. The application enables the filter characteristic parameters to be dynamically adapted in multiple dimensions according to the signal-to-noise ratio and frequency change of each frequency band, thereby improving the matching degree of the filter characteristics and the signal characteristics of each frequency band.

[0062] These and other aspects of the application will become more apparent from the following description of embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, hereinafter, a brief introduction will be given to the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can also be obtained by those skilled in the art without any creative effort.

[0064] Figure 1 A flowchart of a pulse signal adaptive filtering method based on multi-band signal entropy values provided by the embodiments of the application;

[0065] Figure 2 A structural schematic diagram of a multi-band signal entropy value-based pulse signal adaptive filtering system provided for an embodiment of the present application is shown in FIG. 1.

[0066] Figure 3 A structural schematic diagram of a computing device provided for an embodiment of the present application is shown in FIG. 4. DETAILED DESCRIPTION

[0067] In order to enable those skilled in the art to better understand the scheme of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0068] In some processes described in the specification and claims of the present application and the above-described drawings, a plurality of operations appearing in a specific order are included, but it should be clearly understood that these operations can be executed or performed in parallel without the order in which they appear in the present text, and the serial numbers of the operations, such as 11, 12, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes can include more or fewer operations, and the operations can be executed in sequence or in parallel. It should be noted that the descriptions such as "first", "second", etc. in the present text are used to distinguish different messages, devices, modules, etc., and do not represent the order of precedence. Also, "first" and "second" are not of different types.

[0069] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0070] In order to solve the problem of poor pulse signal filtering effect caused by the incomplete feature reflection, lack of dynamic adaptation, poor filtering effect, etc. in the prior art, the present embodiment provides a multi-band signal entropy value-based pulse signal adaptive filtering method, which adopts the following concept: through step-by-step processing, the accurate filtering of multi-band satellite pulse signals is realized: from obtaining basic signal information to converting into key indicators, then preliminary filtering through dynamic tuning structure, calculating signal complexity in combination with a federated learning framework, integrating multi-source features while protecting privacy, and finally enhancing effective information and suppressing noise through an attention mechanism. The synergistic effect of each step enables the filtering process to adapt to changes in signal characteristics, balances information retention and noise suppression, and improves the stability and accuracy of pulse signal transmission.

[0071] Figure 1 A flowchart of a multi-band signal entropy value-based pulse signal adaptive filtering method provided for an embodiment of the present application is shown in FIG. 2.Figure 1 The method comprises:

[0072] S11, receiving a plurality of satellite communication signals of different frequency bands, extracting the phase offset, amplitude value and frequency value in the satellite communication signal.

[0073] The plurality of satellite communication signals of different frequency bands refer to signals transmitted by a satellite and in different frequency ranges (different frequency bands have different signal fluctuation speeds) ; the phase offset is the deviation of the signal fluctuation position from the reference position; the amplitude value reflects the strength of the signal; the frequency value is the number of signal fluctuations per second; and the generated result of this step is the phase offset, amplitude value and frequency value of each frequency band.

[0074] In the embodiments of the present application, different frequency band satellite signals are acquired by a receiving device, and a measuring tool is used to sequentially determine the phase offset, amplitude value and frequency value of each signal. For example, after receiving A frequency band (fluctuation range 10-20 million times / s) and B frequency band (20-30 million times / s) signals, it is measured that the A frequency band deviates from the reference position by 25 degrees (phase offset), the strength is 4 units (amplitude value), and the fluctuation is 15 million times per second (frequency value). The corresponding values of the B frequency band are obtained in the same way.

[0075] S12, based on the phase offset, amplitude value and frequency value, generating a plurality of multi-dimensional feature groups corresponding to each frequency band respectively, and extracting the signal-to-noise ratio and frequency change amount of the corresponding frequency band from each multi-dimensional feature group.

[0076] The multi-dimensional feature group is a set of phase offset, amplitude value and frequency value of the same frequency band; the signal-to-noise ratio is the ratio of effective signal to noise, and the larger the ratio, the clearer the effective signal; the frequency change amount is the change amount of the signal frequency in a period of time; and the generated result of this step is the signal-to-noise ratio and frequency change amount of each frequency band.

[0077] In the embodiments of the present application, for each frequency band, the phase offset, amplitude value and frequency value extracted in S11 are integrated into a multi-dimensional feature group, and the signal-to-noise ratio and frequency change amount are calculated from the feature group. For example, the multi-dimensional feature group of the A frequency band contains 25 degrees, 4 units and 15 million times / s. When calculating the signal-to-noise ratio, the effective signal amplitude is 4 units and the noise amplitude is 0.2 units, and the signal-to-noise ratio is 4 ÷ 0.2 = 20. When calculating the frequency change amount, the frequency of the previous second is 15 million times / s and the frequency of the next second is 15.01 million times / s, and the change amount is 10,000 times / s.

[0078] S13, based on the signal-to-noise ratio and frequency change amount of all frequency bands, dynamically adjusting the filter characteristic parameters through a preset three-level tuning structure, and filtering the satellite communication signals of different frequency bands based on the dynamically adjusted filter characteristic parameters to obtain a filtered signal.

[0079] Wherein, the three-level tuning structure is a combination of adjusting filter settings in three parts; the filter characteristic parameter refers to the setting affecting the filter effect (such as filtering range, center position, amplification degree); dynamic adjustment refers to the change of the parameter with the signal characteristics; the filtered signal is the signal after filtering processing; the generated result of this step is the filtered signal of each frequency band.

[0080] In the embodiments of the present application, a three-level tuning structure is constructed based on the signal-to-noise ratio and the frequency variation: the first level calculates the filter bandwidth adjustment parameter according to the signal-to-noise ratio, such as 20 million times / s corresponding to the bandwidth adjustment parameter 200 million times / s; the second level calculates the center frequency adjustment parameter according to the frequency variation, such as a variation of 200,000 times / s corresponding to an adjustment parameter of 2 million times / s; the third level integrates the signal-to-noise ratio and the frequency variation through weighted calculation to obtain a comprehensive evaluation parameter to determine the signal gain adjustment parameter, such as a comprehensive evaluation parameter of 12.4 corresponding to a gain of 1.2 times. After integrating the three-level structure, the filter characteristic parameter is adjusted, and then the signal is filtered.

[0081] S14, transmit the filtered signal to the federated learning framework, and calculate the multi-frequency band signal entropy value based on the federated learning framework and the multi-dimensional feature group.

[0082] Wherein, the federated learning framework is a system for multiple devices to cooperatively process data; the multi-frequency band signal entropy value is a numerical value reflecting the complexity of the signal, and the larger the value, the more information it may contain; the generated result of this step is the signal entropy value of each frequency band.

[0083] In the embodiments of the present application, the filtered signal is transmitted to the federated learning framework, and the entropy value is calculated in combination with the multi-dimensional feature group, the formula being H = -∑(p_i x log(p_i)), H being the entropy value, and p_i being the feature occurrence ratio.

[0084] S15, based on the multi-frequency band signal entropy value, optimize the adaptive filter parameter, and perform global aggregation on the optimized adaptive filter parameter through a differential privacy protection rule to generate an aggregated feature map group.

[0085] Wherein, the adaptive filter parameter is a filter setting that can automatically adjust with the signal; the differential privacy protection rule protects data privacy by adding random values; global aggregation is a process of integrating the processing results of multiple devices; the aggregated feature map group is a set of images containing multi-frequency band features after integration; the generated result of this step is the aggregated feature map group.

[0086] In the embodiments of the present application, the adaptive filter parameter is optimized based on the entropy value, and the higher the entropy value, the more precise the parameter; a random value within a preset range is added according to the rule, such as 20.3 being obtained by adding 0.3 to the optimized parameter 20; the parameter is transmitted to the aggregation node, and the parameters of multiple nodes are integrated according to a preset proportion (such as 50% for each node); and finally, the aggregated feature map group is obtained.

[0087] S16, based on the channel and space attention mechanism, the aggregated feature map group is weighted to enhance the effective information of the high entropy value frequency band and suppress the noise interference of the low entropy value frequency band, and adaptive filtering of the pulse signal based on the multi-frequency band signal entropy value is completed.

[0088] In the channel and space attention mechanism, the channel refers to the signal source, and the space refers to the signal position; weighting refers to giving different importance to different parts; the high entropy value frequency band refers to the frequency band with high entropy value and possibly containing more effective information; the low entropy value frequency band refers to the frequency band with low entropy value and possibly containing more noise; and the result of this step is to complete the adaptive filtering of the pulse signal.

[0089] In the embodiments of the present application, the channel features and spatial features of each frequency band are extracted from the aggregated feature map group; the weighting coefficients are determined based on the entropy values (high entropy value coefficients are high, such as 0.8 for an entropy value of 1.20 and 0.2 for an entropy value of 0.85); and the high entropy value frequency band features are enhanced and the low entropy value frequency band features are weakened according to the coefficients, and finally the filtering is completed.

[0090] For example, when processing multi-frequency band satellite pulse signals, first, A, B, and C frequency band signals are received, and the phase offset, amplitude value, and frequency value of each frequency band are extracted. Then, a multi-dimensional feature group of each frequency band is generated, and the signal-to-noise ratio (20, 15, and 25) and the frequency change (10, 20, and 0.5 million times per second) of the A, B, and C frequency bands are calculated. Based on these indicators, a three-level tuning structure is constructed: the bandwidth adjustment parameter is calculated according to the signal-to-noise ratio, the center frequency adjustment parameter is calculated according to the frequency change, the comprehensive evaluation parameter is integrated, and the gain adjustment parameter is determined, and the filtering characteristic parameter is adjusted and filtered after integration. The filtered signal is transmitted into the federated learning framework, and the entropy values of each frequency band (1.03, 0.85, and 1.20) are calculated according to H = -∑ (p_i × log (p_i)) based on the feature group. Based on the entropy value, the filtering parameter is optimized, a random value (such as A frequency band parameter 20 plus 0.3 to 20.3) is added, and the parameter is transmitted to the aggregation node and integrated with other node parameters to generate an aggregated feature map group. Finally, the weighting coefficients are determined according to the entropy values (C = 0.8, B = 0.2), the high entropy value frequency band features are enhanced, and the low entropy value frequency band features are suppressed, and the adaptive filtering is completed.

[0091] By performing S11-S16, the embodiments of the present application achieve accurate filtering of multi-frequency band satellite pulse signals through step-by-step processing: from obtaining basic signal information to converting it into key indicators, then preliminary filtering through a dynamic tuning structure, calculating signal complexity in combination with a federated learning framework, integrating multi-source features while protecting privacy, and finally specifically enhancing effective information and suppressing noise through an attention mechanism. The synergistic effect of each step enables the filtering process to adapt to changes in signal characteristics, balances information retention and noise suppression, and improves the stability and accuracy of pulse signal transmission.

[0092] In a possible embodiment, S13 dynamically adjusts the filter characteristic parameters through a preset three-level tuning structure based on the signal-to-noise ratios and the frequency variation amounts of all frequency bands, including:

[0093] Step 131: Based on the signal-to-noise ratios of all frequency bands, a filter bandwidth adjustment parameter of a corresponding frequency band is calculated according to a preset inverse proportion correspondence, so as to construct a first-level tuning structure.

[0094] The signal-to-noise ratio is the ratio of an effective signal to noise, and the larger the ratio is, the clearer the signal is; the filter bandwidth adjustment parameter is used to adjust a signal frequency range allowed to pass; the first-level tuning structure is a part used to adjust a filter range and composed of filter bandwidth adjustment parameters of all frequency bands; and the first-level tuning structure is generated in this step.

[0095] In the embodiment of the present application, the signal-to-noise ratios of all frequency bands are collected, and parameters of all frequency bands are calculated according to a preset inverse proportion correspondence (the larger the signal-to-noise ratio is, the smaller the filter bandwidth adjustment parameter is), for example, a preset filter bandwidth adjustment parameter = 1000 ÷ signal-to-noise ratio, and the signal-to-noise ratios of A, B and C frequency bands are 20, 15 and 25 respectively, so A frequency band parameter = 1000 ÷ 20 = 50, B frequency band = 1000 ÷ 15 ≈ 67, and C frequency band = 1000 ÷ 25 = 40; and then the parameters are used to construct the first-level tuning structure.

[0096] Step 132: Based on the frequency variation amounts of all frequency bands, a center frequency adjustment parameter of a corresponding frequency band is calculated according to a preset direct proportion correspondence, so as to construct a second-level tuning structure.

[0097] The frequency variation amount is a change degree of a signal frequency in a period of time; the center frequency adjustment parameter is used to adjust a center position of a filter range; the second-level tuning structure is a part used to adjust the center position of the filter range and composed of center frequency adjustment parameters of all frequency bands; and the second-level tuning structure is generated in this step.

[0098] In the embodiment of the present application, the frequency variation amounts of all frequency bands are collected, and parameters of all frequency bands are calculated according to a preset direct proportion correspondence (the larger the frequency variation amount is, the larger the center frequency adjustment parameter is), for example, a preset center frequency adjustment parameter = frequency variation amount × 20, and the frequency variation amounts of A, B and C frequency bands are 10,000 times / s, 20,000 times / s and 5,000 times / s respectively, so A frequency band parameter = 1 × 20 = 20, B frequency band = 2 × 20 = 40, and C frequency band = 0.5 × 20 = 10; and then the parameters are used to construct the second-level tuning structure.

[0099] Step 133: The signal-to-noise ratios and the frequency variation amounts of all frequency bands are integrated, and a signal comprehensive evaluation parameter is obtained through a preset weighted calculation mode based on the integration result.

[0100] The integration is a process of combining the signal-to-noise ratio and the frequency variation amount; the weighted calculation mode is a mode of calculating after assigning different weights to the two; the signal comprehensive evaluation parameter is a value comprehensively reflecting the two, and is used for comprehensively evaluating the signal; and the signal comprehensive evaluation parameter of each frequency band is generated in this step.

[0101] In the embodiment of the present application, the weights of the signal-to-noise ratio and the frequency variation amount are determined (for example, the signal-to-noise ratio is 60%, and the frequency variation amount is 40%), the comprehensive evaluation parameter is calculated for each frequency band, for example, the A frequency band signal-to-noise ratio is 20, the frequency variation amount is 10,000 times / s, and the parameter = 20*60%+1*40%=12+0.4=12.4; the B frequency band = 15*60%+2*40%=9+0.8=9.8; and the C frequency band = 25*60%+0.5*40%=15+0.2=15.2.

[0102] In the embodiment of the present application, the weights of the signal-to-noise ratio and the frequency variation amount are determined (for example, the signal-to-noise ratio is 60%, and the frequency variation amount is 40%), the comprehensive evaluation parameter is calculated for each frequency band, for example, the A frequency band signal-to-noise ratio is 20, the frequency variation amount is 10,000 times / s, and the parameter = 20*60%+1*40%=12+0.4=12.4; the B frequency band = 15*60%+2*40%=9+0.8=9.8; and the C frequency band = 25*60%+0.5*40%=15+0.2=15.2.

[0103] The signal gain adjustment parameter is used for adjusting the signal amplification degree; the third-level tuning structure is a part used for adjusting the amplification degree and composed of the signal gain adjustment parameters of each frequency band; and the third-level tuning structure is generated in this step.

[0104] In the embodiment of the present application, the corresponding relationship between the preset comprehensive evaluation parameter and the gain adjustment parameter is determined (for example, the parameter increases by 1, and the gain increases by 0.1, and the initial value is 1), the parameters of each frequency band are calculated, for example, the A frequency band comprehensive evaluation parameter is 12.4, and the gain parameter = 1+12.4*0.1=2.24; the B frequency band = 1+9.8*0.1=1.98; and the C frequency band = 1+15.2*0.1=2.52, and then the third-level tuning structure is constructed by using these parameters.

[0105] In the embodiment of the present application, the first-level tuning structure, the second-level tuning structure, and the third-level tuning structure are integrated to obtain the preset three-level tuning structure.

[0106] The first-level, second-level, and third-level tuning structures respectively adjust the filtering range, the center position, and the amplification degree; the preset three-level tuning structure is a complete adjustment system of the three; and the preset three-level tuning structure is generated in this step.

[0107] In the embodiment of the present application, the first-level, second-level, and third-level tuning parameters of each frequency band are combined in the order of the filtering range, the center position, and the amplification degree, for example, the A frequency band is combined as (50, 20, 2.24), the B frequency band is (67, 40, 1.98), and the C frequency band is (40, 10, 2.52), and the three-level tuning structure is formed by integrating these combinations.

[0108] In step 136, based on the three-level tuning structure, the filter bandwidth adjustment parameter, the center frequency adjustment parameter and the signal gain adjustment parameter are mapped into the filter characteristics of the corresponding frequency band to dynamically adjust the filter characteristic parameters, and the dynamically adjusted filter characteristic parameters are obtained.

[0109] Wherein, the mapping is the process of applying the adjustment parameters to the filter characteristics; the filter characteristics include the range, the center position, the amplification degree, etc.; the dynamically adjusted filter characteristic parameters are the parameters that are adjusted to be more in line with the signal characteristics; this step generates the dynamically adjusted filter characteristic parameters of each frequency band.

[0110] In the embodiments of the present application, the parameters in the three-level tuning structure are applied to the corresponding filter characteristics, for example, the original filter range of the A frequency band is 1-2 million times / s, and after applying the bandwidth parameter 50, it becomes 1-2.5 million times / s; the original center position is 1.5 million times / s, and after applying the center parameter 20, it becomes 1.7 million times / s; the original amplification factor is 1.0, and after applying the gain parameter 2.24, it becomes 2.24, and the adjusted filter characteristic parameters are obtained.

[0111] For example, when a certain system processes satellite signals of A, B and C frequency bands, the signal-to-noise ratios of A, B and C frequency bands are known to be 20, 15 and 25, respectively. The filter bandwidth adjustment parameters of each frequency band are calculated as 50, 67 and 40 respectively according to the formula filter bandwidth adjustment parameter = 1000 ÷ signal-to-noise ratio, and these parameters are integrated to form the first-level tuning structure. The frequency variation amounts of the three frequency bands are known to be 10, 20 and 0.5 thousand times / s respectively, and the center frequency adjustment parameters are calculated as 20, 40 and 10 respectively according to the formula center frequency adjustment parameter = frequency variation amount × 20, and these parameters are integrated to form the second-level tuning structure. The signal comprehensive evaluation parameters of each frequency band are calculated according to the weights of signal-to-noise ratio 60% and frequency variation amount 40%, and the signal comprehensive evaluation parameters of A, B and C frequency bands are 12.4, 9.8 and 15.2 respectively. The gain adjustment parameters of each frequency band are calculated as 2.24, 1.98 and 2.52 respectively according to the formula signal gain adjustment parameter = 1 + comprehensive evaluation parameter × 0.1, and these parameters are integrated to form the third-level tuning structure. The first, second and third level tuning parameters of the three frequency bands are combined as (50, 20, 2.24), (67, 40, 1.98) and (40, 10, 2.52) respectively, and these combinations are integrated to form the preset three-level tuning structure. Finally, these parameters are mapped into the filter characteristics of the corresponding frequency band, for example, the filter range of the A frequency band is adjusted from the original 1-2 million times / s to 1-2.5 million times / s, the center position is adjusted from 1.5 million times / s to 1.7 million times / s, and the amplification factor is adjusted from 1.0 to 2.24, and the dynamically adjusted filter characteristic parameters of B and C frequency bands are obtained in the same way.

[0112] By performing steps 131-136, the embodiment of the application realizes dynamic adjustment of the filter characteristic parameters by constructing and integrating the tuning structure in three stages. From adjusting the filter range based on the signal-to-noise ratio, to adjusting the center position based on the frequency variation, to adjusting the amplification degree after comprehensive evaluation, each link is closely connected, so that the filter characteristic can fully adapt to the clarity, dynamic change and overall performance of the signal. Finally, the filter characteristic parameters obtained by mapping are more targeted and can better adapt to the characteristics of signals in different frequency bands, improving the adaptability and accuracy of the filter effect.

[0113] In a possible embodiment, step 136 maps the filter bandwidth adjustment parameter, the center frequency adjustment parameter and the signal gain adjustment parameter into the filter characteristic of the corresponding frequency band based on the three-stage tuning structure, to dynamically adjust the filter characteristic parameter, to obtain the dynamically adjusted filter characteristic parameter, including:

[0114] a1, determining a basic parameter set of each frequency band, the basic parameter set containing a basic bandwidth parameter, a basic center frequency parameter and a basic gain parameter.

[0115] The basic parameter set is a set of initial settings for each frequency band filter, containing a basic bandwidth parameter (the size of the initial allowed signal frequency range), a basic center frequency parameter (the center position of the initial filter range), and a basic gain parameter (the initial signal amplification multiple). This step generates the basic parameter set of each frequency band.

[0116] In the embodiment of the application, the basic bandwidth, basic center frequency and basic gain parameters of each frequency band are determined according to the signal characteristics of each frequency band and the conventional filter requirements, for example, the A frequency band basic bandwidth is 1 million times / second (allowing signals within 1 million times / second to pass), the basic center frequency is 1.5 million times / second (the center of the filter range is at 1.5 million times / second), and the basic gain is 1.0 times (no amplification); the B frequency band basic bandwidth is 1.5 million times / second, the basic center frequency is 2.5 million times / second, and the basic gain is 1.0 times; the C frequency band basic bandwidth is 2 million times / second, the basic center frequency is 3.5 million times / second, and the basic gain is 1.0 times. These parameters are sorted to form the basic parameter set of each frequency band.

[0117] a2, combining the filter bandwidth adjustment parameter in the three-stage tuning structure with the basic bandwidth parameter of the corresponding frequency band based on a preset bandwidth mapping rule, to obtain a dynamic bandwidth parameter.

[0118] The preset bandwidth mapping rule is a rule for combining the filter bandwidth adjustment parameter with the basic bandwidth parameter, and the dynamic bandwidth parameter is a new signal frequency passing range size obtained after combination. This step generates the dynamic bandwidth parameter of each frequency band.

[0119] In the embodiments of the present application, the preset bandwidth mapping rule (such as dynamic bandwidth parameter = basic bandwidth parameter + filter bandwidth adjustment parameter) is used to combine the filter bandwidth adjustment parameter in the three-level tuning structure with the basic bandwidth parameter of the corresponding frequency band, for example, the basic bandwidth of A frequency band is 1 million times / s, the filter bandwidth adjustment parameter is 0.5 million times / s, the dynamic bandwidth = 1 million + 0.5 million = 1.5 million times / s; the basic bandwidth of B frequency band is 1.5 million times / s, the adjustment parameter is 0.67 million times / s, the dynamic bandwidth = 1.5 million + 0.67 million = 2.17 million times / s; the basic bandwidth of C frequency band is 2 million times / s, the adjustment parameter is 0.4 million times / s, the dynamic bandwidth = 2 million + 0.4 million = 2.4 million times / s.

[0120] a3, based on the preset frequency mapping rule, combine the center frequency adjustment parameter in the three-level tuning structure with the basic center frequency parameter of the corresponding frequency band to obtain a dynamic center frequency parameter.

[0121] The preset frequency mapping rule is a rule of combining the center frequency adjustment parameter with the basic center frequency parameter, and the dynamic center frequency parameter is a new filter range center position obtained after the combination. The step generates the dynamic center frequency parameter of each frequency band.

[0122] In the embodiments of the present application, the preset frequency mapping rule (such as dynamic center frequency parameter = basic center frequency parameter + center frequency adjustment parameter) is used to combine the center frequency adjustment parameter in the three-level tuning structure with the basic center frequency parameter of the corresponding frequency band, for example, the basic center frequency of A frequency band is 1.5 million times / s, the center frequency adjustment parameter is 0.2 million times / s, the dynamic center frequency = 1.5 million + 0.2 million = 1.7 million times / s; the basic center frequency of B frequency band is 2.5 million times / s, the adjustment parameter is 0.4 million times / s, the dynamic center frequency = 2.5 million + 0.4 million = 2.9 million times / s; the basic center frequency of C frequency band is 3.5 million times / s, the adjustment parameter is 0.1 million times / s, the dynamic center frequency = 3.5 million + 0.1 million = 3.6 million times / s.

[0123] a4, based on the preset gain mapping rule, combine the signal gain adjustment parameter in the three-level tuning structure with the basic gain parameter of the corresponding frequency band to obtain a dynamic gain parameter.

[0124] The preset gain mapping rule is a rule of combining the signal gain adjustment parameter with the basic gain parameter, and the dynamic gain parameter is a new signal amplification multiple obtained after the combination. The step generates the dynamic gain parameter of each frequency band.

[0125] In the embodiments of the present application, the gain mapping rule (e.g. dynamic gain parameter = basic gain parameter x signal gain adjustment parameter) is explicitly preset, and the signal gain adjustment parameter in the three-stage tuning structure and the basic gain parameter of the corresponding frequency band are combined according to the rule, for example, the A frequency band basic gain is 1.0 times, the signal gain adjustment parameter is 2.24, the dynamic gain = 1.0 x 2.24 = 2.24 times; the B frequency band basic gain is 1.0 times, the adjustment parameter is 1.98, the dynamic gain = 1.0 x 1.98 = 1.98 times; the C frequency band basic gain is 1.0 times, the adjustment parameter is 2.52, the dynamic gain = 1.0 x 2.52 = 2.52 times.

[0126] a5, mapping the dynamic bandwidth parameter, the dynamic center frequency parameter and the dynamic gain parameter to the filter characteristics of the corresponding frequency band to generate the dynamically adjusted filter characteristic parameters of the corresponding frequency band.

[0127] Wherein, the dynamic bandwidth parameter, the dynamic center frequency parameter and the dynamic gain parameter are the adjusted filter range size, center position and amplification multiple respectively, mapping to the filter characteristics is to apply these parameters to the actual filter setting, and the dynamically adjusted filter characteristic parameter is the final filter setting after integration, and the step generates the dynamically adjusted filter characteristic parameters of each frequency band.

[0128] In the embodiments of the present application, the corresponding relationship between the dynamic parameters and the filter characteristics (dynamic bandwidth corresponding to the range size, dynamic center frequency corresponding to the center position, and dynamic gain corresponding to the amplification multiple) is determined, and the three are integrated into the filter setting of each frequency band, for example, the A frequency band takes the dynamic center frequency of 1.7 million times / s as the center, the dynamic bandwidth of 1.5 million times / s (the range is 1.7 million ± 0.75 million times / s), and the amplification is 2.24 times; the B frequency band takes 2.9 million times / s as the center, the bandwidth is 2.17 million times / s (the range is 2.9 million ± 0.1085 million times / s), and the amplification is 1.98 times; the C frequency band takes 3.6 million times / s as the center, the bandwidth is 2.4 million times / s (the range is 3.6 million ± 1.2 million times / s), and the amplification is 2.52 times, to form the final filter characteristic parameters.

[0129] For example, when a system processes satellite signals of three frequency bands A, B and C, the basic parameter sets of each frequency band are first determined: the basic bandwidth of A frequency band is 1 million times / s, the basic center frequency is 1.5 million times / s, and the basic gain is 1.0 times; the basic bandwidth of B frequency band is 1.5 million times / s, the basic center frequency is 2.5 million times / s, and the basic gain is 1.0 times; the basic bandwidth of C frequency band is 2 million times / s, the basic center frequency is 3.5 million times / s, and the basic gain is 1.0 times. Then, according to the bandwidth mapping rule (dynamic bandwidth = basic bandwidth + filter bandwidth adjustment parameter), combined with the adjustment parameters 50, 67 and 40 million times / s in the three-level tuning structure, the dynamic bandwidths of A frequency band 1 million + 50 million = 150 million times / s, B frequency band 1.5 million + 67 million = 217 million times / s, and C frequency band 2 million + 40 million = 240 million times / s are calculated. According to the frequency mapping rule (dynamic center frequency = basic center frequency + center frequency adjustment parameter), combined with the adjustment parameters 20, 40 and 10 million times / s, the dynamic center frequencies of A frequency band 1.5 million + 20 million = 1.7 million times / s, B frequency band 2.5 million + 40 million = 2.9 million times / s, and C frequency band 3.5 million + 10 million = 3.6 million times / s are calculated. Then, according to the gain mapping rule (dynamic gain = basic gain x signal gain adjustment parameter), combined with the adjustment parameters 2.24, 1.98 and 2.52, the dynamic gains of A frequency band 1.0 x 2.24 = 2.24 times, B frequency band 1.0 x 1.98 = 1.98 times, and C frequency band 1.0 x 2.52 = 2.52 times are calculated. Finally, these parameters are mapped to the filter characteristics, and the parameters of A frequency band are formed as centering on 1.7 million times / s, with a bandwidth of 1.5 million times / s and an amplification of 2.24 times; the parameters of B frequency band are formed as centering on 2.9 million times / s, with a bandwidth of 217 million times / s and an amplification of 1.98 times; and the parameters of C frequency band are formed as centering on 3.6 million times / s, with a bandwidth of 240 million times / s and an amplification of 2.52 times.

[0130] By performing a1-a5, the embodiments of the present application realize the complete conversion of the filter characteristics from the basic setting to the dynamic adaptation from determining the initial filter parameters to adjusting the parameters combined with the signal characteristics, and then integrating into the final filter settings. Each step combines the basic parameters and the signal characteristic adjustment parameters through explicit mapping rules, so that the filter range, center position and amplification can fully match the signal characteristics of different frequency bands. The final filter parameters are more targeted and can better adapt to signal changes, improving the accuracy and adaptability of filter processing.

[0131] In a possible embodiment, S14, the filtered signal is transmitted to a federated learning framework, and the multi-frequency band signal entropy value is calculated based on the federated learning framework and the multi-dimensional feature group, including:

[0132] Step 141, based on the preset frequency band division standard, the filtered signal is divided into a plurality of sub-signals, and the plurality of sub-signals are transmitted to the nodes corresponding to the frequency bands in the federated learning framework respectively, so that after each node receives the sub-signal of the corresponding frequency band, the sub-signal is associated with the multi-dimensional feature group, and based on the association result, the feature parameter calculation is performed on the associated data of the sub-signal and the multi-dimensional feature group according to the preset local calculation rule, and the signal feature distribution parameter of the corresponding frequency band is obtained.

[0133] Wherein, the preset frequency band division standard is a provision for splitting the filtered signal according to different frequency ranges; the filtered signal is a signal after preliminary processing; the sub-signal is a small signal corresponding to each frequency band after splitting; the federated learning framework is a system in which a plurality of processing units work cooperatively; the node is a processing unit corresponding to each frequency band in the system; the multi-dimensional feature group is a set containing signal phase, amplitude, frequency and other information; the association is to correspond the sub-signal with the multi-dimensional feature group information of the same frequency band; the local calculation rule is a method for each node to independently calculate the feature parameter; the associated data is the associated sub-signal and multi-dimensional feature group information; the feature parameter calculation is to calculate the parameter describing the signal feature; the signal feature distribution parameter is a parameter reflecting the distribution of the feature in the signal; this step generates the signal feature distribution parameter of each frequency band.

[0134] In the embodiments of the present application, first, the filtered signal is split into A, B and C sub-signals according to the preset frequency band division standard (for example, 1000-2000 million times / s is A frequency band, 2000-3000 million times / s is B frequency band, and 3000-4000 million times / s is C frequency band); then the sub-signals are transmitted to the nodes corresponding to the frequency bands in the federated learning framework (A to node 1, B to node 2, and C to node 3); after each node receives the sub-signal, it is associated with the multi-dimensional feature group of the corresponding frequency band (for example, the A frequency band contains 25 degrees, 4 units and other information); then the associated data is calculated according to the local calculation rule (for example, the proportion of the occurrence times of the feature in the three intervals), and the signal feature distribution parameter is obtained, for example, node 1 calculates that the A sub-signal feature appears 30 times in the first interval, 50 times in the second interval, and 20 times in the third interval (total 100 times), and the distribution parameter is 30, 50 and 20.

[0135] Step 142, based on the federated learning framework and the preset cross-node cooperation rule, the signal feature distribution parameters of each node corresponding to the frequency band are integrated to form a global feature distribution parameter.

[0136] Wherein, the cross-node cooperation rule is a provision for integrating the parameters of each node in the federated learning framework; the signal feature distribution parameter is the feature distribution parameter calculated by each node; the global feature distribution parameter is the parameter reflecting the overall feature distribution after integration; this step generates the global feature distribution parameter.

[0137] In the embodiment of the present application, first, the cross-node coordination rule is determined (such as assigning weights according to the importance of the frequency bands: A accounts for 40%, B accounts for 30%, and C accounts for 30%); then the signal feature distribution parameters of each node are collected (such as node 1: 30, 50, 20; node 2: 20, 60, 20; node 3: 40, 40, 20); and the parameters are integrated according to the rule, for example, the global parameter of the first interval = 30*40% + 20*30% + 40*30% = 30, the second interval = 50*40% + 60*30% + 40*30% = 50, and the third interval = 20*40% + 20*30% + 20*30% = 20, to obtain the global feature distribution parameters 30, 50, and 20.

[0138] Step 143, calculating the global feature distribution parameters according to the preset entropy value calculation rule to obtain the multi-band signal entropy value.

[0139] In the embodiment of the present application, first, the cross-node coordination rule is determined (such as assigning weights according to the importance of the frequency bands: A accounts for 40%, B accounts for 30%, and C accounts for 30%); then the signal feature distribution parameters of each node are collected (such as node 1: 30, 50, 20; node 2: 20, 60, 20; node 3: 40, 40, 20); and the parameters are integrated according to the rule, for example, the global parameter of the first interval = 30*40% + 20*30% + 40*30% = 30, the second interval = 50*40% + 60*30% + 40*30% = 50, and the third interval = 20*40% + 20*30% + 20*30% = 20, to obtain the global feature distribution parameters 30, 50, and 20.

[0140] In the embodiment of the present application, first, the cross-node coordination rule is determined (such as assigning weights according to the importance of the frequency bands: A accounts for 40%, B accounts for 30%, and C accounts for 30%); then the signal feature distribution parameters of each node are collected (such as node 1: 30, 50, 20; node 2: 20, 60, 20; node 3: 40, 40, 20); and the parameters are integrated according to the rule, for example, the global parameter of the first interval = 30*40% + 20*30% + 40*30% = 30, the second interval = 50*40% + 60*30% + 40*30% = 50, and the third interval = 20*40% + 20*30% + 20*30% = 20, to obtain the global feature distribution parameters 30, 50, and 20.

[0141] For example, when a system processes filter signals of A, B, and C frequency bands, the filter signals are first split into A, B, and C sub-signals according to the preset standard (1000-2000 million times / s for A, 2000-3000 million times / s for B, and 3000-4000 million times / s for C), and then transmitted to nodes 1, 2, and 3 of the federated learning framework, respectively; each node associates the sub-signals with the multi-dimensional feature groups corresponding to the frequency bands, calculates according to the local rule (the proportion of the occurrence times of the statistical features in the three intervals), obtains the A frequency band distribution parameters 30, 50, and 20, the B frequency band 20, 60, and 20, and the C frequency band 40, 40, and 20; then integrates according to the cross-node rule (A accounts for 40%, B accounts for 30%, and C accounts for 30%), the global parameter of the first interval = 30*0.4 + 20*0.3 + 40*0.3 = 30, the second interval = 50*0.4 + 60*0.3 + 40*0.3 = 50, and the third interval = 20*0.4 + 20*0.3 + 20*0.3 = 20, to form the global parameters 30, 50, and 20; finally, the parameters are converted to 0.3, 0.5, and 0.2 according to the entropy formula, and the multi-band signal entropy value is calculated (H = -(0.3*log0.3 + 0.5*log0.5 + 0.2*log0.2) = 1.0295).

[0142] By performing steps 141-143, the embodiments of the present application achieve accurate calculation of the characteristics of each frequency band by splitting the signal by frequency band and processing by the corresponding node; the local parameters are integrated by means of the cross-node collaborative rule to form global parameters reflecting the overall signal characteristics; and finally, the multi-band signal entropy value calculated based on the global parameters can fully reflect the complexity of the signal. The steps are closely linked, which not only preserves the uniqueness of each frequency band, but also realizes the fusion of global information, providing a reliable basis for subsequent optimization of filter parameters and improving the rationality and effectiveness of the overall processing.

[0143] In one possible embodiment, step 141, based on the association result, performs feature parameter calculation on the associated data in the multi-dimensional feature group and the sub-signal according to a preset local calculation rule to obtain signal feature distribution parameters of the corresponding frequency band, including:

[0144] b1, compare the phase offset of the associated data in the multi-dimensional feature group with a preset reference phase value to generate a phase difference value sequence.

[0145] Wherein, the phase offset of the associated data in the multi-dimensional feature group is a series of values reflecting the phase offset from the reference position in the associated signal; the preset reference phase value is a standard phase value preset as a reference; the phase difference value sequence is a sequence composed of a series of difference values in order after comparing the phase offset with the reference phase value; this step generates the phase difference value sequence.

[0146] In the embodiments of the present application, first, a series of values of the phase offset are extracted from the associated data, for example, the phase offset of the A frequency band is 25 degrees, 28 degrees, 23 degrees, and 30 degrees in turn; then each value is compared with the preset reference phase value (for example, 0 degrees), and the difference value is obtained by subtracting the reference phase value from the phase offset, that is, 25 degrees minus 0 degrees equals 25 degrees, 28 degrees minus 0 degrees equals 28 degrees, 23 degrees minus 0 degrees equals 23 degrees, and 30 degrees minus 0 degrees equals 30 degrees; finally, these difference values are arranged in order to form the phase difference value sequence 25 degrees, 28 degrees, 23 degrees, and 30 degrees.

[0147] b2, extract the maximum difference value and the minimum difference value from the phase difference value sequence, and calculate the phase fluctuation range parameter based on the maximum difference value and the minimum difference value.

[0148] Wherein, the phase difference value sequence is a series of phase difference values generated in step b1; the maximum difference value is the largest value in the sequence; the minimum difference value is the smallest value in the sequence; the phase fluctuation range parameter is a value reflecting the overall phase fluctuation amplitude obtained by subtracting the minimum difference value from the maximum difference value; this step generates the phase fluctuation range parameter.

[0149] In the embodiments of the present application, first, the maximum and minimum values in the phase difference value sequence are found out, for example, the phase difference value sequence of the A frequency band in b1 is 25 degrees, 28 degrees, 23 degrees and 30 degrees, wherein the maximum difference value is 30 degrees and the minimum difference value is 23 degrees; then the maximum difference value is subtracted from the minimum difference value, that is, 30 degrees minus 23 degrees equals 7 degrees, to obtain the phase fluctuation range parameter.

[0150] b3, based on the time domain distribution statistical rule in the preset local calculation rule, calculating the amplitude value of the associated data to obtain the amplitude change rate parameter.

[0151] The time domain distribution statistical rule in the preset local calculation rule is a rule based on time change statistical data (such as calculating the change amount in a unit of time); the amplitude value of the associated data is a series of values reflecting the strength in the associated signal; the amplitude change rate parameter is a value reflecting the speed of change of the amplitude value in a unit of time; and the amplitude change rate parameter is generated in this step.

[0152] In the embodiments of the present application, first, the time domain distribution statistical rule is determined, for example, the amplitude difference between two adjacent time points is calculated and then divided by the time interval; then the amplitude values at different time points are extracted from the associated data, for example, the amplitude of the A frequency band is 4 units at t1 and 4.5 units at t2, and the time interval is 1 second; according to the rule, the amplitude value at the latter time point is subtracted from the amplitude value at the former time point, and then divided by the time interval, that is, (4.5 units minus 4 units) divided by 1 second equals 0.5 units / second, to obtain the amplitude change rate parameter.

[0153] b4, calculating the difference between the frequency value in the associated data and the preset reference frequency value to obtain the frequency offset parameter.

[0154] The frequency value in the associated data is a series of values reflecting the fluctuation times per second in the associated signal; the preset reference frequency value is a standard frequency value preset as a reference; the frequency offset parameter is the difference between the frequency value and the reference frequency value, reflecting the degree of frequency deviation from the reference; and the frequency offset parameter is generated in this step.

[0155] In the embodiments of the present application, first, the frequency value is extracted from the associated data, for example, the frequency value of the A frequency band at a certain time is 15.01 million times / second; then the frequency value is subtracted from the preset reference frequency value (such as 15 million times / second), that is, 15.01 million times / second minus 15 million times / second equals 10,000 times / second, to obtain the frequency offset parameter.

[0156] b5, constructing a feature parameter set according to the phase fluctuation range parameter, the amplitude change rate parameter and the frequency offset parameter.

[0157] The phase fluctuation range parameter, the amplitude change rate parameter, and the frequency offset parameter are respectively parameters reflecting phase fluctuation amplitude, amplitude change speed, and frequency deviation degree; the feature parameter set is a group of data formed by integrating the three parameters, and is used for comprehensively describing signal features; and the feature parameter set is generated in this step.

[0158] In the embodiment of the present application, the phase fluctuation range parameter obtained by b2, the amplitude change rate parameter obtained by b3, and the frequency offset parameter obtained by b4 are collected first, for example, 7 degrees, 0.5 units / s, and 10,000 times / s for the A frequency band; then the three parameters are integrated in sequence to form the feature parameter set (7 degrees, 0.5 units / s, and 10,000 times / s).

[0159] b6, performing distribution characteristic analysis on the feature parameter set, and generating signal feature distribution parameters of the corresponding frequency band based on the result of the distribution characteristic analysis.

[0160] The feature parameter set is the comprehensive parameter group formed in the step b5; the distribution characteristic analysis is the analysis of the proportion of the number of times that the values of the parameters in the parameter set fall in different intervals; the result of the distribution characteristic analysis is the proportion data of each interval obtained by the analysis; the signal feature distribution parameter is a value reflecting the distribution of the feature parameters in different intervals; and the signal feature distribution parameters of the corresponding frequency band are generated in this step.

[0161] In the embodiment of the present application, the method of the distribution characteristic analysis is determined first, for example, the value range of each parameter is divided into 3 intervals, and the proportion of the number of times that falls in each interval is counted; then each parameter in the feature parameter set is analyzed, for example, the phase fluctuation range of 7 degrees in the A frequency band falls in the 5-10 degree interval (the proportion is 100%), the amplitude change rate of 0.5 units / s falls in the 0-1 unit / s interval (the proportion is 100%), and the frequency offset of 10,000 times / s falls in the 0-20,000 times / s interval (the proportion is 100%); and finally, the interval proportions of the parameters are integrated to generate the signal feature distribution parameters (100%, 100%, and 100%).

[0162] For example, when a system processes the associated data of A frequency band, it first extracts the phase offset values 25 degrees, 28 degrees, 23 degrees and 30 degrees from the associated data, compares them with the reference phase value 0 degrees, calculates the difference values 25-0=25 degrees, 28-0=28 degrees, 23-0=23 degrees and 30-0=30 degrees, and generates a phase difference value sequence; then finds the maximum difference value 30 degrees and the minimum difference value 23 degrees from the sequence, calculates the phase fluctuation range parameter 30-23=7 degrees; extracts the amplitude value 4 units at t1 and the amplitude value 4.5 units at t2 (interval 1 second), calculates the amplitude change rate parameter (4.5-4) ÷ 1=0.5 units / second according to the rule; then extracts the frequency value 1501 million times / second, compares it with the reference frequency 1500 million times / second, and obtains the frequency offset parameter 1501 million-1500 million=10,000 times / second; integrates the three parameters into a feature parameter set (7 degrees, 0.5 units / second, 10,000 times / second); and finally determines that each parameter interval accounts for 100% according to the distribution analysis method, and generates the signal feature distribution parameter (100%, 100%, 100%).

[0163] By performing b1-b6, the embodiments of the present application quantize the signal features from the three dimensions of phase, amplitude and frequency through step-by-step processing: first, the phase fluctuation range is obtained by comparing the phase with the reference value, reflecting the phase stability; then the amplitude change rate is analyzed in the time dimension, reflecting the signal strength dynamics; then the offset parameter is obtained by comparing the frequency with the reference value, reflecting the fluctuation stability; finally, these parameters are integrated and the distribution is analyzed to form comprehensive feature distribution parameters. The steps are closely connected, from local feature extraction to global distribution analysis, providing multi-dimensional and quantifiable basis for subsequent signal processing, and improving the comprehensiveness and accuracy of signal feature description.

[0164] In a possible embodiment, S15, based on the multi-band signal entropy value, optimizes the adaptive filtering parameter, performs global aggregation on the optimized adaptive filtering parameter through the differential privacy protection rule, and generates an aggregated feature map group, including:

[0165] Step 151, according to the size of the multi-band signal entropy value, determine the adjustment weight of the adaptive filtering parameter.

[0166] Wherein, the multi-band signal entropy value is a value reflecting the complexity of multiple frequency band signals, the larger the entropy value, the more information the signal may contain; the adaptive filtering parameter is a filtering setting parameter that can be automatically adjusted according to the signal feature; the adjustment weight is a value reflecting the importance of adjusting the filtering parameter of different frequency bands, the larger the weight, the higher the adjustment priority; this step generates the adjustment weight of the adaptive filtering parameter of each frequency band.

[0167] In this embodiment, the entropy values ​​of multi-band signals in each frequency band are first collected, for example, band A is 1.03, band B is 0.85, and band C is 1.20. Then, the adjustment weights are determined according to the entropy values, and the total weights are distributed proportionally with 100% weight. The total entropy value is 3.08. The weights for band C are 1.20 ÷ 3.08 × 100% ≈ 39%, band A is 1.03 ÷ 3.08 × 100% ≈ 33%, and band B is 0.85 ÷ 3.08 × 100% ≈ 28%.

[0168] Step 152: Based on the adjusted weights, the initial adaptive filter parameters are optimized step by step to obtain the optimized adaptive filter parameters.

[0169] Among them, the adjustment weight is the importance of each frequency band determined in step 151; the initial adaptive filter parameters are the unoptimized basic filter parameters; the step-by-step optimization is to adjust the parameters from high to low according to the weight, with higher weights resulting in larger adjustment ranges; the optimized adaptive filter parameters are the adjusted parameters that better fit the signal characteristics; this step generates the optimized adaptive filter parameters.

[0170] In this embodiment, the optimization order is first determined (C→A→B); initial parameters are set (A: 2.0, B: 1.8, C: 2.2); the adjustment range is determined according to the weight (each 10% weight corresponds to an adjustment of 0.1). For the C band, 39% corresponds to 0.4, and after optimization, the range is 2.2 + 0.4 = 2.6; for the A band, 33% corresponds to 0.3, and after optimization, the range is 2.0 + 0.3 = 2.3; for the B band, 28% corresponds to 0.3, and after optimization, the range is 1.8 + 0.3 = 2.1.

[0171] Step 153: Based on the differential privacy protection rules, add privacy perturbation values ​​that meet the preset threshold conditions to the optimized adaptive filtering parameters to obtain the target adaptive filtering parameters.

[0172] Among them, the differential privacy protection rule is a rule that protects data by adding random values; the optimized adaptive filtering parameters are the adjusted parameters in step 152; the privacy perturbation value is a random value within the range of ±0.5; the target adaptive filtering parameter is the final parameter after superimposing the perturbation value; this step generates the target adaptive filtering parameter.

[0173] In the embodiments of this application, the rules are first defined (adding random values ​​within ±0.5); in order to optimize the parameter superposition disturbance value, A band 2.3+0.2=2.5, B band 2.1-0.3=1.8, C band 2.6+0.4=3.0.

[0174] Step 154: Based on the preset parameter fusion ratio, the target adaptive filtering parameters are input into the aggregation node in the federated learning framework and then globally integrated to generate global parameters.

[0175] The preset parameter fusion ratio is the proportion of each node parameter in the global integration (for example, 50% for each of two nodes); the target adaptive filtering parameter is the final parameter of step 153; the aggregation node is a processing unit for integrating parameters of each node; the global parameter is a parameter reflecting the overall situation after integration; and the global parameter is generated in this step.

[0176] In the embodiments of the present application, the fusion ratio (50% for each of two nodes) is first determined; the parameters of node 1 (A: 2.5, B: 1.8, C: 3.0) and node 2 (A: 2.6, B: 1.7, C: 3.1) are input into the aggregation node; and the parameters are integrated according to the ratio, A band 2.5*50%+2.6*50%=2.55, B band 1.8*50%+1.7*50%=1.75, and C band 3.0*50%+3.1*50%=3.05.

[0177] Step 155, reorganizing the global parameter according to a preset feature dimension to generate an aggregated feature map group.

[0178] The global parameter is the integrated parameter of step 154; the preset feature dimension is the classification angle of the parameter (such as the filtering range, the center position, etc.); the aggregated feature map group is a parameter chart set arranged according to the dimension; and the aggregated feature map group is generated in this step.

[0179] In the embodiments of the present application, the feature dimension (filtering range, center position, and amplification degree) is first determined; the global parameter is reorganized according to the dimension, A band 2.55 corresponds to the filtering range, B band 1.75 corresponds to the center position, and C band 3.05 corresponds to the amplification degree; and the parameters are arranged into charts to form the aggregated feature map group.

[0180] For example, when a system processes A, B, and C frequency band signals, the entropy values of each frequency band (A: 1.03, B: 0.85, C: 1.20) are first collected, the total entropy value is 3.08, and the adjustment weight is determined according to the ratio (C: 39%, A: 33%, B: 28%). Based on the weight, the initial parameters (A: 2.0, B: 1.8, C: 2.2) are optimized, the C band is adjusted by 0.4 to be 2.6, the A band is adjusted by 0.3 to be 2.3, and the B band is adjusted by 0.3 to be 2.1. According to the differential privacy rule, the perturbation value is superimposed, A: 2.3+0.2=2.5, B: 2.1-0.3=1.8, and C: 2.6+0.4=3.0. The parameters of two nodes (node 1: A: 2.5, B: 1.8, C: 3.0; node 2: A: 2.6, B: 1.7, C: 3.1) are integrated according to the ratio of 50% for each, and the global parameters A: 2.55, B: 1.75, and C: 3.05 are obtained. Finally, the parameters are reorganized according to the filtering range, the center position, and the amplification degree dimension to generate the aggregated feature map group.

[0181] By performing steps 151-155, the embodiments of the present application form a complete parameter processing system from determining the adjustment priority according to the signal complexity, to optimizing the parameters by weight, to protecting data security by privacy protection rules, and finally integrating multi-source parameters and arranging them by feature dimensions. Each link is closely connected, which not only ensures that the filtering parameters accurately adapt to the signal characteristics, but also takes into account data security, and finally generates an intuitive and clear aggregated feature map group, which provides a comprehensive and reliable basis for subsequent filtering processing, and improves the rationality and security of the overall processing.

[0182] In one possible embodiment, S16, based on the channel and space attention mechanism, weights the aggregated feature map group to enhance the effective information of high entropy value frequency bands and suppress the noise interference of low entropy value frequency bands, and completes the adaptive filtering of the pulse signal based on the multi-frequency band signal entropy value, including:

[0183] Step 161, extracts the initial channel features and initial space features corresponding to each frequency band from the aggregated feature map group.

[0184] Wherein, the aggregated feature map group is a parameter chart set arranged by feature dimensions; the initial channel features are raw data extracted from the chart group, reflecting the characteristics of signals in different transmission channels; the initial space features are raw data extracted from the chart group, reflecting the characteristics of signals in spatial distribution; this step generates initial channel features and initial space features corresponding to each frequency band.

[0185] In the embodiments of the present application, first, determine the charts corresponding to each frequency band in the aggregated feature map group, for example, the filter range chart corresponding to the A frequency band, the center position chart corresponding to the B frequency band, and the amplification degree chart corresponding to the C frequency band; then extract data reflecting channel characteristics from these charts as initial channel features, such as 5 for the A frequency band, 4 for the B frequency band, and 6 for the C frequency band; at the same time, extract data reflecting spatial distribution as initial space features, such as 3 for the A frequency band, 2 for the B frequency band, and 5 for the C frequency band.

[0186] Step 162, based on the multi-frequency band signal entropy value of each frequency band and the attention mechanism, determines the first weighting coefficient and the second weighting coefficient for the initial channel features and the initial space features, respectively.

[0187] Wherein, the multi-frequency band signal entropy value is a numerical value reflecting the complexity of signals in each frequency band; the attention mechanism is a method of highlighting important features and weakening secondary features; the first weighting coefficient is a numerical value for adjusting the importance of the initial channel features; the second weighting coefficient is a numerical value for adjusting the importance of the initial space features; this step generates the first and second weighting coefficients.

[0188] In the embodiment of the present application, first, the multi-band signal entropy values of each frequency band are collected, for example, A frequency band 1.03, B frequency band 0.85, C frequency band 1.20; then, the attention mechanism is combined, the principle of the greater the entropy value, the higher the coefficient is determined, the total coefficient sum is 1, the total entropy value is 1.03+0.85+1.20=3.08, the first weighting coefficient is C frequency band 1.20÷3.08≈0.4, A frequency band 1.03÷3.08≈0.3, B frequency band 0.85÷3.08≈0.3; the second weighting coefficient is the same, C frequency band 0.5, A frequency band 0.3, B frequency band 0.2.

[0189] Step 163, based on the first weighting coefficient, the initial channel feature is enhanced or attenuated to obtain the target channel feature.

[0190] Wherein, the first weighting coefficient is the coefficient determined in step 162 for adjusting the initial channel feature; the enhancement or attenuation processing is to amplify or reduce the initial channel feature according to the coefficient; the target channel feature is the channel feature that can better reflect the importance after processing; the target channel feature is generated in this step.

[0191] In the embodiment of the present application, first, the processing rule is determined as the initial channel feature multiplied by the first weighting coefficient; then, each frequency band is processed, for example, the initial channel feature of A frequency band 5 is multiplied by 0.3 to obtain 5×0.3=1.5; the initial channel feature of B frequency band 4 is multiplied by 0.3 to obtain 4×0.3=1.2; the initial channel feature of C frequency band 6 is multiplied by 0.4 to obtain 6×0.4=2.4, these results are the target channel feature.

[0192] Step 164, based on the second weighting coefficient, the initial spatial feature is regionally weighted to obtain the target spatial feature.

[0193] Wherein, the second weighting coefficient is the coefficient determined in step 162 for adjusting the initial spatial feature; the regional weight distribution is to distribute the importance of the initial spatial feature according to the coefficient; the target spatial feature is the feature that can better reflect the importance of the space after distribution; the target spatial feature is generated in this step.

[0194] In the embodiment of the present application, first, the processing rule is determined as the initial spatial feature multiplied by the second weighting coefficient; then, each frequency band is processed, for example, the initial spatial feature of A frequency band 3 is multiplied by 0.3 to obtain 3×0.3=0.9; the initial spatial feature of B frequency band 2 is multiplied by 0.2 to obtain 2×0.2=0.4; the initial spatial feature of C frequency band 5 is multiplied by 0.5 to obtain 5×0.5=2.5, these results are the target spatial feature.

[0195] Step 165, the target channel feature and the target spatial feature are superimposed and fused to generate a weighted aggregated feature map group.

[0196] Wherein, the target channel feature is the adjusted channel feature obtained in step 163; the target space feature is the adjusted space feature obtained in step 164; the superimposed fusion is to add and integrate the two features; the weighted aggregated feature map group is a chart set formed after the fusion, which reflects the importance of the features; and the step generates the weighted aggregated feature map group.

[0197] In the embodiment of the present application, first, the target channel features of each frequency band are superimposed with the target space features; for example, 1.5 of A frequency band plus 0.9, obtaining 1.5+0.9=2.4; 1.2 of B frequency band plus 0.4, obtaining 1.2+0.4=1.6; 2.4 of C frequency band plus 2.5, obtaining 2.4+2.5=4.9; then these results are arranged into a chart to form the weighted aggregated feature map group.

[0198] Step 166, using the weighted aggregated feature map group to enhance the effective information of the high-entropy-value frequency band and suppress the noise interference of the low-entropy-value frequency band, to complete the pulse signal adaptive filtering based on the multi-frequency band signal entropy value.

[0199] Wherein, the weighted aggregated feature map group is the integrated feature chart obtained in step 165; the high-entropy-value frequency band is a frequency band with large entropy value and much effective information; the low-entropy-value frequency band is a frequency band with small entropy value and much noise; the pulse signal adaptive filtering is a process of automatically adjusting to retain effective information and reduce noise; and the step completes the adaptive filtering.

[0200] In the embodiment of the present application, first, the high-entropy-value and low-entropy-value frequency bands are distinguished from the weighted aggregated feature map group, for example, C frequency band 4.9 corresponds to high-entropy-value, and B frequency band 1.6 corresponds to low-entropy-value; then the effective information of the high-entropy-value frequency band is enhanced according to the chart information, such as amplifying the C frequency band signal; and at the same time, the noise of the low-entropy-value frequency band is suppressed, such as reducing the B frequency band signal, to complete the adaptive filtering.

[0201] For example, when a system processes A, B, and C three-band signals, the initial channel features 5 and initial spatial features 3 of the A-band, the initial channel features 4 and initial spatial features 2 of the B-band, and the initial channel features 6 and initial spatial features 5 of the C-band are extracted from the aggregated feature map group. According to the multi-band signal entropy values of each band (A: 1.03, B: 0.85, and C: 1.20), the total entropy value 3.08 is calculated by combining the attention mechanism, the first weighting coefficient (A: 1.03 ÷ 3.08 ≈ 0.3, B: 0.85 ÷ 3.08 ≈ 0.3, and C: 1.20 ÷ 3.08 ≈ 0.4) and the second weighting coefficient (A: 0.3, B: 0.2, and C: 0.5) are determined. The initial channel features are processed according to the first weighting coefficient, A-band 5 × 0.3 = 1.5, B-band 4 × 0.3 = 1.2, and C-band 6 × 0.4 = 2.4, to obtain the target channel features. The initial spatial features are processed according to the second weighting coefficient, A-band 3 × 0.3 = 0.9, B-band 2 × 0.2 = 0.4, and C-band 5 × 0.5 = 2.5, to obtain the target spatial features. The target channel features and the target spatial features are superimposed, A-band 1.5 + 0.9 = 2.4, B-band 1.2 + 0.4 = 1.6, and C-band 2.4 + 2.5 = 4.9, to generate the weighted aggregated feature map group. Finally, the effective information of the C-band (high entropy value) is enhanced and the noise interference of the B-band (low entropy value) is suppressed through the map group, to complete the adaptive filtering of the pulse signal.

[0202] By performing steps 161-166, the embodiment of the application extracts the basic features from the aggregated feature map group, determines the feature weight by combining the signal complexity and the attention mechanism, highlights the important features by weighting adjustment, fuses the channel and spatial features to form a comprehensive chart, and finally enhances the effective information of the high-entropy value band and suppresses the noise of the low-entropy value band. The steps are closely connected, so that the filtering process can accurately adapt to the signal features of different bands, effectively improves the clarity of the effective information in the signal, reduces irrelevant interference, and makes the filtering result more in line with actual needs.

[0203] Figure 2 A structural diagram of a pulse signal adaptive filtering system based on multi-band signal entropy values provided by the embodiment of the application is shown in FIG. 1. Figure 2 As shown in FIG. 1, the system comprises:

[0204] The receiving module 21 is configured to receive satellite communication signals of multiple different bands, and extract the phase offset, amplitude value, and frequency value in the satellite communication signals.

[0205] The first generating module 22 is configured to generate a multi-dimensional feature group corresponding to each band based on the phase offset, amplitude value, and frequency value, and extract the signal-to-noise ratio and frequency change of the corresponding band from each multi-dimensional feature group.

[0206] The adjusting module 23 dynamically adjusts the filter characteristic parameters through a preset three-level tuning structure based on the signal-to-noise ratios and frequency variation amounts of all frequency bands, filters the satellite communication signals of different frequency bands based on the dynamically adjusted filter characteristic parameters, and obtains a filtered signal.

[0207] The transmission module 24 transmits the filtered signal to a federated learning framework, and calculates a multi-frequency signal entropy value based on the federated learning framework and the multi-dimensional feature group.

[0208] The second generation module 25 optimizes adaptive filter parameters based on the multi-frequency signal entropy value, globally aggregates the optimized adaptive filter parameters through a differential privacy protection rule, and generates an aggregated feature map group.

[0209] The weighting module 26 weights the aggregated feature map group based on a channel and spatial attention mechanism, so as to enhance the effective information of a high-entropy value frequency band and suppress noise interference of a low-entropy value frequency band, and complete the multi-frequency signal entropy value-based adaptive filtering of the pulse signal.

[0210] Figure 2 The multi-frequency signal entropy value-based adaptive filtering system of the pulse signal can perform Figure 1 The multi-frequency signal entropy value-based adaptive filtering method of the pulse signal of the embodiment described above will not be described in detail. The specific manner in which each module, unit of the multi-frequency signal entropy value-based adaptive filtering system of the pulse signal in the above embodiment performs operations has been described in detail in the embodiment related to the method, and will not be described in detail here.

[0211] In one possible design, Figure 2 The multi-frequency signal entropy value-based adaptive filtering system of the pulse signal of the embodiment described above can be implemented as a computing device, such as Figure 3 As shown, the computing device can include a storage component 31 and a processing component 32.

[0212] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32.

[0213] The processing component 32 is configured to perform the following processes: receiving satellite communication signals of multiple different frequency bands, extracting phase offsets, amplitude values and frequency values in the satellite communication signals; generating a plurality of multi-dimensional feature groups corresponding to the respective frequency bands based on the phase offsets, the amplitude values and the frequency values, extracting signal-to-noise ratios and frequency variation amounts of the corresponding frequency bands from the multi-dimensional feature groups; dynamically adjusting filter characteristic parameters based on the signal-to-noise ratios and the frequency variation amounts of all the frequency bands through a preset three-level tuning structure, filtering the satellite communication signals of different frequency bands based on the dynamically adjusted filter characteristic parameters to obtain filtered signals; transmitting the filtered signals to a federated learning framework, calculating multi-band signal entropy values based on the federated learning framework and the multi-dimensional feature groups; optimizing adaptive filtering parameters based on the multi-band signal entropy values, performing global aggregation on the optimized adaptive filtering parameters through a differential privacy protection rule to generate an aggregated feature map group; and weighting the aggregated feature map group based on a channel and spatial attention mechanism to enhance effective information of high-entropy value frequency bands and suppress noise interference of low-entropy value frequency bands, thereby completing adaptive filtering of the pulse signals based on the multi-band signal entropy values.

[0214] The processing component 32 can include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component can also be an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a controller, a microcontroller, a microprocessor, or other electronic elements for executing the above method.

[0215] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component can be implemented by any type of volatile or nonvolatile storage devices or a combination thereof, such as a random access memory (RAM), a static random access memory (SRAM), an erasable programmable read only memory (EPROM), an electrically erasable programmable read only memory (EEPROM), a programmable read only memory (PROM), a read only memory (ROM), a magnetic storage, a flash memory, a magnetic disk, or a compact disk.

[0216] Of course, the computing device can also include other components, such as an input / output interface, a display component, a communication component, etc.

[0217] The input / output interface provides an interface between the processing component and peripheral interface modules, which can be output devices, input devices, etc.

[0218] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0219] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform, and the computing device can be a cloud server, and the processing component, the storage component, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0220] The embodiments of the present application also provide a computer storage medium storing a computer program, and the computer program can implement the above-mentioned Figure 1 The pulse signal adaptive filtering method based on multi-band signal entropy values of the embodiments shown.

[0221] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0222] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purposes of the embodiments according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0223] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and necessary universal hardware platforms, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of software products, and the computer software products can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and include a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0224] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for adaptive filtering of a pulse signal based on entropy values of multi-band signals, characterized in that, The method comprises the following steps: receiving a plurality of different frequency band satellite communication signals, extracting the phase offset, amplitude value and frequency value in the satellite communication signal; based on the phase offset, amplitude value and frequency value, generating a plurality of multi-dimensional feature groups corresponding to each frequency band, and extracting the signal-to-noise ratio and frequency change of the corresponding frequency band from each multi-dimensional feature group; based on the signal-to-noise ratio and frequency change of all frequency bands, dynamically adjusting the filter characteristic parameters through a preset three-level tuning structure, filtering the satellite communication signals of different frequency bands based on the dynamically adjusted filter characteristic parameters to obtain a filtered signal; transmitting the filtered signal to a federated learning framework, calculating the multi-frequency signal entropy value based on the federated learning framework and the multi-dimensional feature group; based on the multi-frequency signal entropy value, optimizing the adaptive filtering parameters, and globally aggregating the optimized adaptive filtering parameters through a differential privacy protection rule to generate an aggregated feature map group; based on the channel and spatial attention mechanism, weighting the aggregated feature map group to enhance the effective information of high-entropy frequency bands and suppress the noise interference of low-entropy frequency bands, and completing the multi-frequency signal entropy value-based pulse signal adaptive filtering.

2. The method of claim 1, wherein, The method comprises the following steps: based on the signal-to-noise ratio and frequency change of all frequency bands, dynamically adjusting the filter characteristic parameters through a preset three-level tuning structure, comprising: based on the signal-to-noise ratio of all frequency bands, calculating the filter bandwidth adjustment parameter of the corresponding frequency band according to the preset inverse proportional corresponding relationship to construct the first-level tuning structure; based on the frequency change of all frequency bands, calculating the center frequency adjustment parameter of the corresponding frequency band according to the preset proportional corresponding relationship to construct the second-level tuning structure; integrating the signal-to-noise ratio and frequency change of all frequency bands, and obtaining a signal comprehensive evaluation parameter through a preset weighted calculation method based on the integration result; based on the signal comprehensive evaluation parameter, determining the signal gain adjustment parameter of the corresponding frequency band according to the preset corresponding relationship to construct the third-level tuning structure; integrating the first-level tuning structure, the second-level tuning structure and the third-level tuning structure to obtain the preset three-level tuning structure; 3. The method of claim 2, wherein, based on the three-level tuning structure, mapping the filter bandwidth adjustment parameter, the center frequency adjustment parameter and the signal gain adjustment parameter to the filter characteristic of the corresponding frequency band to dynamically adjust the filter characteristic parameter and obtain the dynamically adjusted filter characteristic parameter. The method comprises the following steps: determining the basic parameter set of each frequency band, wherein the basic parameter set includes a basic bandwidth parameter, a basic center frequency parameter and a basic gain parameter; based on a preset bandwidth mapping rule, combining the filter bandwidth adjustment parameter in the three-level tuning structure and the basic bandwidth parameter of the corresponding frequency band to obtain a dynamic bandwidth parameter; The center frequency adjustment parameter in the three-level tuning structure is combined with a basic center frequency parameter of a corresponding frequency band based on a preset frequency mapping rule to obtain a dynamic center frequency parameter; The signal gain adjustment parameter in the three-level tuning structure is combined with a basic gain parameter of a corresponding frequency band based on a preset gain mapping rule to obtain a dynamic gain parameter; The dynamic bandwidth parameter, the dynamic center frequency parameter, and the dynamic gain parameter are mapped to filter characteristics of the corresponding frequency band to generate a dynamically adjusted filter characteristic parameter of the corresponding frequency band.

4. The method of claim 1, wherein, The filter signal is transmitted to a federated learning framework, and a multi-frequency band signal entropy value is calculated based on the federated learning framework and the multi-dimensional feature group, including: The filter signal is divided into a plurality of sub-signals based on a preset frequency band division standard, and the plurality of sub-signals are respectively transmitted to nodes corresponding to each frequency band in the federated learning framework, so that after each node receives a sub-signal of a corresponding frequency band, the sub-signal is associated with the multi-dimensional feature group, and based on the association result, a feature parameter calculation is performed on the associated data in the sub-signal and the multi-dimensional feature group according to a preset local calculation rule to obtain a signal feature distribution parameter of the corresponding frequency band; Based on the federated learning framework and a preset cross-node cooperation rule, the signal feature distribution parameters of the corresponding frequency band of each node are integrated to form a global feature distribution parameter; The global feature distribution parameter is calculated according to a preset entropy value calculation rule to obtain a multi-frequency band signal entropy value.

5. The method of claim 4, wherein, The filter signal is transmitted to a federated learning framework, and a multi-frequency band signal entropy value is calculated based on the federated learning framework and the multi-dimensional feature group, including: The phase offset of the associated data in the multi-dimensional feature group is compared with a preset reference phase value to generate a phase difference value sequence; The maximum difference value and the minimum difference value are extracted from the phase difference value sequence, and a phase fluctuation range parameter is calculated based on the maximum difference value and the minimum difference value; The amplitude value of the associated data is calculated based on a time domain distribution statistical rule in the preset local calculation rule to obtain an amplitude change rate parameter; The frequency value in the associated data is calculated by difference with a preset reference frequency value to obtain a frequency offset parameter; According to the phase fluctuation range parameter, the amplitude change rate parameter, and the frequency offset parameter, a feature parameter set is constructed; The feature parameter set is analyzed for distribution characteristics, and based on the result of the distribution characteristic analysis, a signal feature distribution parameter of the corresponding frequency band is generated.

6. The method of claim 1, wherein, The filter signal is transmitted to a federated learning framework, and a multi-frequency band signal entropy value is calculated based on the federated learning framework and the multi-dimensional feature group, including: According to the size of the multi-frequency band signal entropy value, the adjustment weight of the adaptive filtering parameter is determined; Based on the adjustment weight, the initial adaptive filtering parameter is optimized step by step to obtain the optimized adaptive filtering parameter; Based on the differential privacy protection rule, a privacy disturbance value meeting the preset threshold condition is superimposed on the optimized adaptive filtering parameter to obtain a target adaptive filtering parameter; Based on a preset parameter fusion ratio, the target adaptive filtering parameter is input to an aggregation node in the federated learning framework for global integration to generate a global parameter; The global parameter is reorganized according to a preset feature dimension to generate an aggregated feature map group.

7. The method of claim 1, wherein, The channel and space based attention mechanism is used to weight the aggregated feature map group to enhance the effective information of the high entropy value frequency band and suppress the noise interference of the low entropy value frequency band, and complete the pulse signal adaptive filtering based on the multi-frequency band signal entropy value, including: Initial channel features and initial space features corresponding to each frequency band are extracted from the aggregated feature map group; Based on the multi-frequency band signal entropy value of each frequency band and the attention mechanism, first and second weighting coefficients are determined for the initial channel features and the initial space features, respectively; Based on the first weighting coefficient, the initial channel features are enhanced or attenuated to obtain target channel features; Based on the second weighting coefficient, the initial space features are regionally weighted to obtain target space features; The target channel features and the target space features are superimposed and fused to generate a weighted aggregated feature map group; The weighted aggregated feature map group is used to enhance the effective information of the high entropy value frequency band and suppress the noise interference of the low entropy value frequency band to complete the pulse signal adaptive filtering based on the multi-frequency band signal entropy value.

8. A multi-band signal entropy-based adaptive filtering system for pulse signals, characterized in that, It includes: A receiving module is configured to receive satellite communication signals of multiple different frequency bands, and extract phase offsets, amplitude values and frequency values in the satellite communication signals; A first generating module is configured to generate a multi-dimensional feature group corresponding to each frequency band based on the phase offsets, the amplitude values and the frequency values, and extract a signal-to-noise ratio and a frequency change of the corresponding frequency band from each multi-dimensional feature group; An adjusting module is configured to dynamically adjust filter characteristic parameters through a preset three-level tuning structure based on the signal-to-noise ratios and the frequency changes of all frequency bands, and filter the satellite communication signals of different frequency bands based on the dynamically adjusted filter characteristic parameters to obtain a filtered signal; A transmission module is configured to transmit the filtered signal to a federated learning framework, calculate a multi-frequency band signal entropy value based on the federated learning framework and the multi-dimensional feature group; A second generating module is configured to optimize adaptive filtering parameters based on the multi-frequency band signal entropy value, and generate an aggregated feature map group by globally aggregating the optimized adaptive filtering parameters through a differential privacy protection rule; A weighting module is configured to use a channel and space based attention mechanism to weight the aggregated feature map group to enhance the effective information of the high entropy value frequency band and suppress the noise interference of the low entropy value frequency band, and complete the pulse signal adaptive filtering based on the multi-frequency band signal entropy value.

9. A computing device, comprising: The application relates to a device for implementing a multi-band signal entropy value-based pulse signal adaptive filtering method, which comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to realize the multi-band signal entropy value-based pulse signal adaptive filtering method.

10. A computer storage medium, characterized in that, The application relates to a computer program, which is executed by a computer to realize the multi-band signal entropy value-based pulse signal adaptive filtering method.

Images