Noise reduction method and device for current pulse generator

Abstract

The invention discloses a current pulse generator noise reduction method and device, and relates to the technical field of current pulse generator noise reduction, and the method specifically comprises the following steps: removing a main pulse waveform from a filter response sequence, extracting filter residual response segments of a plurality of periods, carrying out the time alignment and overlapping processing of the residual response segments of the plurality of periods, and carrying out the noise reduction of the current pulse generator. Constructing a residual energy superposition trend line; based on the residual energy superposition trend line, a Savitzky-Golay sliding window polynomial fitting derivative algorithm is adopted to calculate a trend offset rate parameter, and a wavelet packet decomposition and Shannon entropy calculation algorithm is adopted to generate a disturbance frequency domain diffusivity parameter so as to reflect whether the residual energy superposition behavior of the filter is evolved into self-excitation background noise. According to the method, the problem that residual energy superposition of the filter is difficult to recognize under continuous narrow pulses is solved, precise recognition and dynamic intervention control of self-excitation background noise are realized, and the system stability is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of noise reduction for current pulse generators, and particularly to a noise reduction method and device for current pulse generators. Background Art

[0002] A current pulse generator is an electronic device used to generate current pulse signals with specific amplitude, frequency, duty cycle, and waveform characteristics, and is widely applied in technical fields such as radar systems, communication systems, power electronics control, electromagnetic compatibility testing, and pulsed laser driving. In these applications, the accuracy and stability of the pulse signal have a crucial impact on the system performance. However, due to the high-speed switching characteristics of current pulses, it is extremely easy to introduce noise problems such as electromagnetic interference (EMI), ground noise, voltage spikes, and echo perturbations. These noises not only affect the signal integrity of the system itself but may also interfere with other sensitive circuits or communication modules, resulting in an increase in the bit error rate, system instability, and even damage to components. Therefore, to ensure the reliable operation of the system and signal quality, it is necessary to effectively reduce the noise during the operation of the current pulse generator, suppress the generation of noise from the source, and reduce its propagation and influence through control strategies, filtering mechanisms, or structural optimizations, thereby enhancing the anti-interference ability and working efficiency of the overall system.

[0003] The existing noise reduction technologies for current pulse generators mainly start from three aspects: circuit structure optimization, signal processing technology, and electromagnetic compatibility design, to mitigate or suppress the noise interference caused by the high-speed switching of pulse signals. At the circuit level, components such as low-pass filters, common-mode inductors, and RC absorption circuits are often added to filter out high-frequency noise components and suppress voltage spikes; meanwhile, in the layout design, the propagation path of electromagnetic interference is reduced by optimizing the wiring, reasonably arranging the grounding layer, and power supply decoupling. In terms of signal processing, some systems introduce digital control algorithms or PWM modulation strategies to smooth the current change by adjusting the pulse edge slope, limiting the mutation rate, etc., thereby reducing the interference caused by high-speed switching. At the same time, in the electromagnetic compatibility design link, the release and coupling effect of electromagnetic radiation noise are further weakened through shielding structures, electromagnetic absorption materials, and reasonable housing grounding designs. The entire noise reduction process usually includes three core links: noise source analysis, interference path identification and control, and protection of the sensitive receiving end, aiming to construct an all-round noise suppression mechanism from the three levels of "source-path-terminal" to enhance the stability and reliability of the current pulse generator in complex systems.

[0004] The existing technologies have the following deficiencies: When the current pulse generator outputs continuous narrow pulses with an extremely short pulse width, especially when the pulse period is shorter than the natural energy dissipation time of the filter itself, energy residues will be generated inside the filter. Since these residual energies have not been fully released before the next pulse arrives, they will be superimposed multiple times, forming equivalent parasitic harmonics or low-amplitude self-exciting interferences. Since these background noises do not manifest as obvious spikes, distortions, or edge aberrations, but rather as a periodic increase in the baseline noise, it is difficult to be identified by traditional noise reduction methods based on fixed-frequency band filtering, amplitude limiting clamping, or edge amplitude monitoring. The existing noise reduction technologies for current pulse generators cannot identify the formation process of this self-exciting background noise based on the energy superposition behavior of the filter residues in the case of continuous narrow pulse output, which will cause blind spots in the identification of the noise reduction method in the dynamic operating state, and then lead to hidden interferences in the pulse output superposition. This may not only cause misjudgment of the reference threshold by the subsequent circuit, resulting in false triggering or timing disorders, but also accumulate to form broad-spectrum electromagnetic interference, affecting the anti-interference performance and stability of the overall system.

[0005] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, so it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0006] The object of the present invention is to provide a noise reduction method and device for a current pulse generator to solve the problems in the above background art.

[0007] To achieve the above object, the present invention provides the following technical solution: A noise reduction method for a current pulse generator, specifically including the following steps: S01. During the continuous narrow pulse output of the current pulse generator, collect the pulse width, pulse interval, and edge slope parameters of each pulse, obtain the output voltage of the filter within the corresponding period, and form a filter response sequence for multiple periods; S02. Remove the main pulse waveform from the filter response sequence, extract the filter residual response segments for multiple periods, perform time alignment and overlapping processing on the residual response segments for multiple periods, and construct a residual energy superposition trend line; S03. Based on the residual energy superposition trend line, use the Savitzky-Golay sliding window polynomial fitting derivative algorithm to calculate the trend offset rate parameter, and use the wavelet packet decomposition and Shannon entropy calculation algorithm to generate the perturbation frequency domain diffusivity parameter to reflect whether the filter residual energy superposition behavior evolves into self-exciting background noise; S04. Map the trend offset rate parameter and the perturbation frequency domain diffusivity parameter into a two-dimensional trajectory path, use the similar conical path mapping method to extract the trajectory deformation characteristics, and generate a state factor representing the stage of the formation process of self-exciting background noise; S05. Perform stage determination based on the generated state factors, and execute intervention measures according to the determination results, including adjusting the pulse output rhythm, inserting skip interference pulses, or applying rhythm disruption control, so as to block the continuous superposition of the residual energy of the filter.

[0008] Preferably, in S01, it specifically includes the following steps: During the continuous narrow pulse output of the current pulse generator, collect the start time and end time of each pulse based on the trigger clock, calculate the pulse width through the start and end times, calculate the pulse interval through the start time difference between adjacent pulses, and calculate the edge slope through the start edge fitting curve; Taking the start time of each pulse as the reference point, obtain the voltage signal waveform at the output end of the filter within a fixed sampling window, including the main pulse and the response section after the main pulse; Repeat the parameter collection and voltage acquisition operations for multiple pulse periods, and record the data of each period according to the time stamp number; Arrange the parameters and response voltage data of multiple periods in the preset period order to form a filter response behavior sequence on the continuous time axis for subsequent residual energy analysis.

[0009] Preferably, in S02, it specifically includes the following steps: Determine the time interval corresponding to the main pulse in each period, calculate the start and end boundaries of the main pulse based on the pulse start time and pulse width, and remove the filter voltage data within the time interval corresponding to the main pulse in each period from the original response sequence; In each period, set a fixed-length residual observation window after the main pulse, extract the voltage sampling points retained within this window as the residual response section, and uniformly record their relative time coordinates relative to the pulse start point; Align the residual response sections extracted from all periods according to their start times, and overlap and stack them on the unified time axis to generate a cross-period residual response distribution set; Average the voltage data corresponding to each time point in the set of residual response distributions to obtain a residual energy evolution curve covering multiple cycles, and construct a residual energy superposition trend line based on this curve. The specific construction process is as follows: Perform a sliding window process on the residual energy evolution curve covering multiple cycles, fit the data points within each window to a local linear segment, and obtain the local slope distribution within a continuous time interval; Perform a smoothing filter process on the slope sequences of multiple local linear segments to eliminate the high-frequency noise offset caused by minor perturbations and generate a continuous trend guiding signal; Sample the trend guiding signal at a fixed time interval, and use a third-order polynomial to fit the data intervals before and after each sampling point to form a continuous interpolation expression of the trend curve; Connect all the interpolation points to form a continuous change trajectory covering the entire residual response time axis, and use this trajectory as the superposition trend line of the filter residual energy.

[0010] Preferably, in S03, the process of calculating the trend offset rate parameter based on the residual energy superposition trend line includes the following steps: Perform a Savitzky-Golay sliding window process on the residual energy superposition trend line, perform a third-order polynomial fit within each fixed-length window, and generate a local smoothing curve; At the central position of each sliding window, extract the first derivative value of this curve to form a derivative sequence of the residual trend on the entire time axis; Perform a boundary extension process on the derivative sequence to avoid the interference of edge window fitting distortion on the overall rate continuity; Perform a moving average smoothing process on the derivative sequence to suppress the interference of local mutations on the rate trend; Extract the current derivative at each fixed sampling time point as the trend offset rate parameter, which is used to characterize the change amplitude and direction of the residual energy superposition trend.

[0011] Preferably, in S03, the process of generating the perturbation frequency domain diffusivity parameter based on the residual energy superposition trend line includes the following steps: Perform a multi-level wavelet packet decomposition on the residual energy superposition trend line, and use the db4 wavelet basis function to split the original signal into multiple sub-band signals; Calculate the energy proportion of the corresponding sub-signal on each frequency band to form a complete frequency band energy distribution vector; Apply the Shannon entropy calculation method to the frequency band energy distribution vector to obtain the dispersion index of the energy distribution; Normalize the Shannon entropy value to a standard value range for horizontal comparison between different cycles; Use the normalized entropy value as the perturbation frequency domain diffusivity parameter to reflect the diffusion degree of the residual energy in the frequency domain.

[0012] Preferably, in S03, the trend offset rate parameter and the perturbation frequency domain diffusion parameter are used to reflect whether the filter residual energy superposition behavior evolves into self-excited background noise. Specifically: On the same time axis, the trend offset rate parameter and the perturbation frequency domain diffusion parameter are combined into a two-dimensional time series vector sequence; Perform sliding window clustering analysis on this vector sequence to identify segments where the trend growth rate and the frequency domain diffusion degree increase simultaneously within a continuous time interval, and mark this segment as a suspicious self-excited section, indicating that the residual energy has a continuous enhancement and shows a tendency of frequency domain diffusion; Perform duration and amplitude statistics on each suspicious self-excited section to determine whether it exceeds the preset continuous threshold and energy diffusion threshold; If within any suspicious self-excited section, the continuous sampling time length is not less than the set minimum time length threshold, and the average value of the trend offset rate parameter in this section is greater than the trend rate threshold, and at the same time the average value of the perturbation frequency domain diffusion parameter is greater than the diffusion threshold, it is determined that the filter is in the process of self-excited background noise evolution, and subsequent classification and intervention operations are triggered.

[0013] Preferably, in S04, it specifically includes the following steps: Taking the trend offset rate parameter corresponding to each moment as the abscissa and the perturbation frequency domain diffusion parameter as the ordinate in chronological order, and sequentially constructing a two-dimensional trajectory path to form a set of trajectory segments reflecting the parameter change process; In the two-dimensional trajectory path, extract the geometric shape features of each trajectory segment with a fixed-length sliding window, including local curvature, path bending rate, angle rotation trend, and change rate; Project the extracted geometric shape features into the conical path mapping space, and by analyzing the contour shape of each trajectory segment in this space and performing clustering, divide it into multiple continuously evolving state regions; According to the evolution order, deformation amplitude, and distribution characteristics of each state region in the trajectory, generate a state factor representing the stage of the self-excited background noise formation process, and use it as the basis for subsequent classification and intervention.

[0014] Preferably, in S05, it specifically includes: Based on the distribution position of the generated state factor in the multi-dimensional state space, determine the stage at which the current filter residual energy superposition behavior is located, and divide the state factor into an initial fluctuation stage, an evolution critical stage, and a continuous self-excitation stage; When the state factor is in the initial fluctuation stage, by introducing a fixed-amplitude time interval perturbation in each continuous pulse period, change the rhythm arrangement of the continuous pulse output to prevent the filter from accumulating energy under a constant excitation; When the state factor is in the evolution critical stage, skip interference pulses with a set interval are inserted into the continuous pulse output sequence to change the pulse trigger timing and interrupt the residual energy superposition link in the filter; When the state factor is in the continuous self-excitation stage, a rhythm disturbance control operation is performed, including setting the rhythm change rule within multiple consecutive pulse output periods, and clearing the accumulated energy inside the filter by extending the pulse interval and inserting blank periods to block the continuous formation of self-excitation background noise.

[0015] Preferably, a current pulse generator noise reduction device includes a pulse characteristic analysis module, a residual energy trend modeling module, a dynamic evolution parameter extraction module, a state factor generation module, and a rhythm intervention and regulation module; The pulse characteristic analysis module collects the pulse width, pulse interval, and edge slope parameters of each pulse during the continuous narrow pulse output of the current pulse generator, obtains the output voltage of the filter corresponding to each period, and forms a filter response sequence for multiple periods; The residual energy trend modeling module removes the main pulse waveform from the filter response sequence, extracts the residual response segments of the filter for multiple periods, performs time alignment and overlapping processing on the residual response segments for multiple periods, and constructs a residual energy superposition trend line; The dynamic evolution parameter extraction module calculates the trend offset rate parameter using the Savitzky-Golay sliding window polynomial fitting derivative algorithm based on the residual energy superposition trend line, and generates the perturbation frequency domain diffusion degree parameter using the wavelet packet decomposition and Shannon entropy calculation algorithm to reflect whether the residual energy superposition behavior of the filter evolves into self-excitation background noise; The state factor generation module maps the trend offset rate parameter and the perturbation frequency domain diffusion degree parameter into a two-dimensional trajectory path, extracts the trajectory deformation characteristics using the similar conical path mapping method, and generates a state factor representing the stage of the self-excitation background noise formation process; The rhythm intervention and regulation module performs stage determination based on the generated state factor, and executes intervention measures according to the determination result, including adjusting the pulse output rhythm, inserting skip interference pulses, or applying rhythm disturbance control to block the continuous superposition of the filter residual energy.

[0016] In the above technical solution, the technical effects and advantages provided by the present invention: 1. The noise reduction method for the current pulse generator provided by the present invention first collects pulse characteristic parameters and filter response data in real time during the output of continuous narrow pulses, and processes the main pulse waveform to obtain a filter residual response behavior sequence across multiple cycles. This structured data processing method effectively breaks through the limitations of traditional methods that rely on amplitude spikes or fixed threshold recognition, enabling the accurate characterization and extraction of background energy accumulation behaviors that were previously difficult to detect, establishing a dynamic observation basis based on the evolution behavior of the residual response, and providing sufficient support for subsequent trend modeling and behavior evolution recognition. 2. By introducing an analysis mechanism that combines Savitzky-Golay derivative fitting and wavelet packet decomposition, the present invention can finely characterize the change rate and diffusion characteristics of the residual energy in both the time domain and the frequency domain, and further map these two parameters into a two-dimensional trajectory path. By using the conical path mapping method to extract the trajectory deformation characteristics, the present invention realizes the process modeling and state partitioning of the residual energy from accumulation to evolution into self-exciting interference. This processing path does not rely on traditional frequency domain filtering or fixed amplitude limiting rules, has stronger robustness and adaptability, and is particularly suitable for the dynamic recognition and monitoring of concealed background interference in nonlinear systems. 3. After the system identifies that the filter is in a specific evolution stage, it can trigger different intervention strategies according to the generated state factors, such as adjusting the pulse rhythm, inserting gap pulses, or applying disturbance control, and actively intervene in the continuous superposition link of energy, thereby blocking the continuous formation of self-exciting noise. This closed-loop control method from recognition to regulation not only greatly improves the noise reduction ability of the current pulse system during dynamic operation, but also significantly enhances the anti-interference performance and output stability of the system, ensuring the accuracy and safety of signal transmission and event triggering in high-frequency and high-speed circuit environments, and having outstanding technical advantages and engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0018] Figure 1 It is a schematic flowchart of a noise reduction method and device for a current pulse generator according to the present invention.

[0019] Figure 2 It is a schematic block diagram of a noise reduction method and device for a current pulse generator according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art.

[0021] The present invention provides a method for reducing noise of a current pulse generator as Figure 1 shown, which specifically includes the following steps: S01. During the continuous narrow pulse output of the current pulse generator, collect the pulse width, pulse interval, and edge slope parameters of each pulse, obtain the output voltage of the filter within the corresponding period, and form a filter response sequence for multiple periods; In this embodiment, in S01, it specifically includes the following steps: During the continuous narrow pulse output of the current pulse generator, collect the start time and end time of each pulse based on the trigger clock, calculate the pulse width through the start and end times, calculate the pulse interval through the difference between the start times of adjacent pulses, and calculate the edge slope through curve fitting of the start edge; During the continuous narrow pulse output of the current pulse generator, collect the start time and end time of each pulse based on the trigger clock. The timestamp recording of the pulse edge can be realized by a high-precision time-to-digital converter (TDC). The TDC has a picosecond-level resolution and is suitable for nanosecond-level pulse measurement. The calculation method of the pulse width is as follows: record the rising edge time point t_rise and the falling edge time point t_fall of the same pulse, and the difference t_fall - t_rise is the pulse width. For example, if the rising edge of a certain pulse appears at 10.000 ns and the falling edge appears at 10.900 ns, the pulse width is 0.900 ns. The pulse interval is obtained by the difference between the rising edge time points of two adjacent pulses. For example, if the previous pulse starts at 10.000 ns and the next pulse starts at 11.200 ns, the interval is 1.200 ns. The edge slope parameter can be digitally fitted from the oscilloscope waveform data. For example, a voltage-time fitting model is established for 3 - 5 sampling points before and after the rising edge using the least squares linear fitting method, and the slope is the slope value of the fitted line. For example, if the voltage rises from 0 V to 3 V between 0 ns and 0.3 ns, the slope is 3 / 0.3 = 10 V / ns. The above method can realize edge capture and waveform readback through an FPGA equipped with a TDC and a high-bandwidth ADC, providing key parameter support for the subsequent identification of the filter residual response.

[0022] Taking the start time of each pulse as a reference point, obtain the voltage signal waveform at the output end of the filter within a fixed sampling window, including the main pulse and the response segment after the main pulse; It can be achieved by a high-speed analog-to-digital converter (ADC) cooperating with external trigger control logic. The specific approach is to set an external pulse trigger source in the system (such as the trigger flag signal output by the FPGA). Whenever the starting time point of a pulse is detected, the data acquisition task of the ADC is started, and a sampling time window with a fixed length is defined. For example, the waveform data with a length of 1.5 ns is collected each time. The starting point of the sampling window is aligned with the starting moment of the pulse, and the sampling interval is determined by the ADC sampling rate. For example, if an ADC with a sampling rate of 10 GSa / s (10 billion samples per second) is used, a data point is obtained every 0.1 ns, and a total of 15 points are collected. The sampling window should be sufficient to cover the complete main pulse and the natural response section of the filter after the main pulse. Usually, the duration of the main pulse is 0.9 ns, and the filter decay section may continue for 0.4 - 0.6 ns. After data acquisition, this section of the waveform is temporarily stored in the buffer and packed together with the pulse time mark to form the complete voltage response of the pulse event. This method realizes the accurate tracking and capture of the "main pulse + tail response" at the output end of the filter, facilitating the subsequent separation and modeling of residual disturbances. To improve real-time performance, a circular buffer structure can be used in combination with the DMA method to quickly transfer the acquired waveform to the backend processing unit for further analysis.

[0023] Repeat the parameter acquisition and voltage acquisition operations for multiple pulse cycles, and record the data of each cycle according to the time stamp number; It can be achieved by the FPGA control logic cooperating with the cyclic sampling scheduling mechanism. The specific method is that during the continuous output of narrow pulses by the pulse generator, the FPGA receives the starting trigger signal of each pulse in real time, and starts a data acquisition process at each trigger. This process includes reading the starting time and ending time of the current pulse from the TDC, calculating parameters such as pulse width, interval, and edge slope, and at the same time controlling the high-speed ADC to read the output voltage waveform of the filter according to the preset sampling window. After the acquisition of each cycle is completed, the pulse parameters and the corresponding voltage waveform data are formed into a set of data packets, and the trigger moment of this pulse is added as a time stamp identifier, for example, recorded as "cycle number #N, starting time Tn, parameter group Pn, voltage sequence Wn". To ensure data order and real-time performance, an FIFO buffer can be set inside the FPGA to write data packets in the order of trigger sequence, or a cyclic write structure can be used to automatically increment the number, and the batch data can be written into the high-speed buffer (such as DDR) through the DMA interface for subsequent processing. This method can stably collect pulse data for hundreds or even thousands of consecutive cycles, providing high-time-resolution dynamic data support for the subsequent construction of residual response trend lines and the identification of background noise evolution behaviors.

[0024] Arrange the parameters and response voltage data of multiple cycles in the preset cycle order to form a sequence of filter response behaviors on the continuous time axis for subsequent residual energy analysis.

[0025] It can be achieved through the method of structured data splicing and index management. Specifically, after collecting the parameters of each pulse cycle and obtaining the voltage waveform, the system constructs the data of this cycle into a data unit containing "timestamp, pulse parameter vector, voltage sampling sequence", and writes these data units into an ordered buffer structure in sequence according to the order of trigger time, such as a two-dimensional array, a circular buffer queue or a chained structure. Each time when writing, the data units are automatically sorted according to the timestamp (if the acquisition trigger order is the same, they are directly stored in sequence), ensuring that the data maintains strict time continuity logically. In addition, the system can set a unified reference clock to remap the voltage sampling points in each data unit to a global time axis. For example, the starting position of the voltage sampling points in the Nth cycle is calibrated as T0 + N×ΔT, where ΔT is the pulse period, so as to construct a continuous filter response behavior matrix across cycles and samples. This sequence not only reflects the transient response of the filter to the pulse excitation in each cycle, but also completely retains the response continuation and superposition between cycles, providing a structured and time-order consistent data basis for extracting the residual energy change trend subsequently. This processing process can be automatically completed through the preset data scheduling logic in the embedded processor or FPGA, and is applicable to the management of high-density and high-timeliness pulse data sequences.

[0026] During the continuous narrow pulse output of the current pulse generator, collecting the pulse width, pulse interval and edge slope parameters of each pulse, and obtaining the output voltage of the filter in the corresponding period, to form the filter response sequence of multiple periods, is to analyze the dynamic evolution characteristics of the filter response behavior under high-speed excitation conditions. Since the continuous narrow pulse has the switching characteristics of high density and high steepness, the filter may not fully recover to the steady state after each pulse excitation, resulting in the superposition of energy between adjacent periods and generating potential residual disturbances. By systematically collecting these parameters, the mapping relationship between the pulse excitation and the filter response can be accurately restored, especially to identify whether the residual energy cannot be fully released due to too fast rhythm or too steep edge. The pulse width and interval reflect the pulse rhythm structure, the edge slope is related to the excitation intensity, and the filter output voltage reflects its transient response and remaining influence within each period. Organizing this information into a multi-period response sequence can provide accurate and continuous dynamic input for extracting the superposition trend of the filter residual energy and constructing the background noise evolution model subsequently, which is the basic premise for realizing high-reliability noise reduction and identification.

[0027] S02. Remove the main pulse waveform from the filter response sequence, extract the filter residual response segments of multiple periods, perform time alignment and overlapping processing on the residual response segments of multiple periods, and construct a residual energy superposition trend line; In this embodiment, in S02, it specifically includes the following steps: Determine the time interval corresponding to the main pulse in each cycle, calculate the start and end boundaries of the main pulse based on the pulse start time and pulse width, and remove the filter voltage data in the time interval corresponding to the main pulse in each cycle from the original response sequence; "Determining the time interval corresponding to the main pulse in each cycle, calculating the main pulse start and end boundaries based on the pulse start time and pulse width, and removing the filter voltage data within this time period from the original response sequence" can be achieved through pulse trigger timing control and a data window marking mechanism. Specifically, during the operation of the current pulse generator, the system records the start time T_start and pulse width W of each pulse event in real time, thereby determining the end time of the main pulse as T_end = T_start + W. After obtaining this time interval, all voltage data with sampling times within the interval [T_start, T_end] in the acquired filter output voltage sequence are marked as the main pulse segment using the timestamp index. Subsequently, data cleaning processes remove all sampling points within this time period or replace them with zero values to maintain timeline integrity. A common implementation structure includes a timestamp-based voltage sequence array. Before extracting the residual response segments, the start and end time indexes are used in each cycle array to perform segment stripping. For example, if the main pulse start time of a cycle is 10ns and the pulse width is 0.9ns, the main pulse interval is [10ns, 10.9ns], and the system deletes all voltage data within this interval from the response sequence of the current cycle. The core purpose of this operation is to accurately distinguish the main excitation portion from the subsequent residual disturbance portion. This ensures that subsequent trend analysis is based only on the natural release behavior of the filter rather than the results of active excitation, improving the accuracy of identifying residual energy evolution.

[0028] In each cycle, a fixed-length residual observation window is set after the main pulse ends, and the voltage sampling points retained in the window are extracted as the residual response segment, and their relative time coordinates relative to the pulse starting point are uniformly recorded; "In each cycle, after the main pulse ends, a residual observation window of a fixed length is set, and the voltage sampling points retained within this window are extracted as the residual response segment, and their relative time coordinates with respect to the pulse start point are uniformly recorded." This can be achieved by means of time window truncation and unified time normalization processing. Specifically, after the system completes the identification and elimination of the main pulse time interval [T_start, T_end], immediately in each cycle, starting from T_end, a time window ΔT of a set length is extended backward. For example, if it is set to 20 ns, then the observation window interval is [T_end, T_end + 20 ns]. The system will extract all voltage data whose sampling times fall within this time period in the filter voltage response sequence as the residual response segment. At the same time, for the convenience of unified superposition processing of data between different cycles, the sampling points in each residual response segment will convert their original absolute time stamps (such as 13.5 ns) into relative times (such as 3.5 ns) with respect to the current pulse start time T_start, thus achieving time normalization. This normalization operation enables the residual response segments to be calculated based on a unified relative time reference in subsequent alignment and overlap analysis even though the actual start times of each pulse are different. For example, in the first cycle, the pulse starts at 10 ns and the main pulse ends at 10.9 ns, then the extraction window is [10.9 ns, 30.9 ns], and the relative time of any sampling point time t is t - 10 ns. This method can accurately capture the residual response behavior of the filter during the natural release process after the main excitation, and is the key data basis for identifying energy accumulation and trend evolution.

[0029] Align the residual response segments extracted in all cycles according to their start times, and perform overlapping superposition on a unified time axis to generate a set of cross-cycle residual response distributions; It can be achieved through methods such as relative time normalization + array stacking + unified interpolation alignment. The specific implementation method is as follows: In the previous stage, the time axes of all residual response segments have been converted into relative time coordinates relative to their respective pulse start points. Therefore, the residual response segments of each cycle can be regarded as voltage sequences under the same time reference (for example, taking the pulse start point as 0 ns). The system can construct all cycle residual segments into a two-dimensional matrix structure, where each row represents the residual response of one cycle, and the column index represents unified relative time points (for example, a total of 200 points sampled at intervals of 0.1 ns between 0 ns and 20 ns). If the actual sampling points of different cycles are not completely aligned, linear interpolation or high-order spline interpolation methods can be used to resample the data within each cycle to fill in the corresponding time points, so as to achieve alignment consistency. Subsequently, by stacking column by column, that is, converging the voltage values corresponding to multiple cycles at the unified time point, a "set matrix of residual response distributions across cycles" can be formed. This matrix can not only be used for subsequent trend extraction, but also reflect the response volatility and stability at each time point. For example, the 50th column of the matrix aggregates the residual response values of all cycles at the 5 ns position, which can be used to calculate the average response trend at this time point and also to statistically analyze its variance change to support the identification of the dynamic evolution process of energy superposition. This processing method can fuse the non-steady interference information under multiple cycles within a unified time framework, which is the core basis for constructing a reliable trend recognition model.

[0030] Perform an average calculation on the voltage data corresponding to each time point in the set of residual response distributions to obtain a residual energy evolution curve covering multiple cycles, and construct a residual energy superposition trend line based on this curve. The specific construction process is as follows: Perform a sliding window process on the residual energy evolution curve covering multiple cycles, fit the data points within each window into local linear segments, and obtain the local slope distribution within the continuous time interval; perform a smoothing filter process on the slope sequence of multiple local linear segments to eliminate the high-frequency noise offset caused by small perturbations and generate a continuous trend guiding signal; sample the trend guiding signal at a fixed time interval, and use a third-order polynomial to fit the data intervals before and after each sampling point to form a continuous interpolation expression of the trend curve; connect all the interpolation points to form a continuous change trajectory covering the entire residual response time axis, and use this trajectory as the superposition trend line of the filter residual energy.

[0031] "Perform an average calculation on the voltage data corresponding to each time point in the set of residual response distributions to obtain a residual energy evolution curve covering multiple cycles, and construct a residual energy superposition trend line based on this curve", which can be achieved by combining methods such as averaging point by point + sliding window local modeling + slope sequence filtering + polynomial fitting interpolation. The steps have high data processing feasibility and evolution trend modeling capabilities. Specifically, first, in the set of residual response distributions, calculate the mean value of the voltage values at each unified relative time point (such as 0.1 ns, 0.2 ns...) for all cycles, forming a single one-dimensional time series, representing the average evolution process of residual energy over multiple cycles. Next, apply a sliding window process to this one-dimensional evolution curve. The length of each window can be set to several sampling points (such as 7 points). Within each window, fit a local linear curve through the least squares method and extract its slope as the trend rate within this interval. In this way, a set of "local slope sequences" that change with time can be obtained, reflecting the accelerating or stable sections of the residual energy evolution. Since these slope values may introduce high-frequency fluctuations due to minor perturbations or discrete sampling, further smooth filtering is performed on the entire slope sequence (such as using moving average filtering, wavelet denoising, or exponential weighting method) to remove non-structural high-frequency noise and extract a stable trend guiding signal. Then, select key sampling points at fixed time intervals (such as sampling once every 1 ns) on this smoothed trend signal. Take a certain number of data points on both sides of each sampling point to construct local data segments, and use a third-order polynomial fitting algorithm to fit these data to obtain the trend expression function for each key point. Continuously splice all the fitting segments to form a continuous change trajectory covering the entire residual response time axis. This trajectory is the superposition trend line of the filter residual energy. This trend line not only retains the dominant change characteristics of the energy evolution but also effectively suppresses background perturbations, and can truly reflect the dynamic accumulation behavior of the filter under continuous narrow pulse excitation, providing a structural criterion for subsequent background noise identification and intervention strategy selection.

[0032] Removing the main pulse waveform from the filter response sequence, extracting the filter residual response segments of multiple periods, and performing time alignment and overlapping processing on them to construct the residual energy superposition trend line is to accurately identify the potential energy accumulation behavior inside the filter under continuous narrow pulse excitation. Since the current pulse generator is in the high-frequency output mode, the main pulse itself has a high amplitude and a steep edge, which will trigger an obvious transient response in the filter, and the energy release process requires a certain time window. However, if the next pulse arrives before the residual energy of the previous pulse has been completely dissipated, the residual part will be periodically superimposed, and then evolve into a weak-amplitude but persistent background disturbance. Such interference does not manifest as significant pulse distortion and is often ignored by traditional filtering or edge discrimination methods. Therefore, the main pulse excitation itself must be removed, and the focus should be on analyzing the natural release segment of the filter after the main pulse. Through time alignment and overlapping processing after multi-period sampling, sufficient statistical samples can be accumulated to eliminate the influence of individual pulse jitters, and thus the cross-period energy evolution trend can be extracted. This trend line reveals whether there is a cumulative increase in residual energy in the filter under continuous excitation, which is the key basis for judging whether it enters the potential self-excited state. Therefore, the purpose of setting the whole step is to strip the main excitation interference, extract the real release behavior, and provide the input basis for subsequent intervention judgment in the form of trend modeling.

[0033] S03. Based on the residual energy superposition trend line, use the Savitzky-Golay sliding window polynomial fitting derivative algorithm to calculate the trend offset rate parameter, and use the wavelet packet decomposition and Shannon entropy calculation algorithm to generate the disturbance frequency domain diffusion parameter to reflect whether the residual energy superposition behavior of the filter evolves into self-excited background noise; In this embodiment, in S03, the process of calculating the trend offset rate parameter based on the residual energy superposition trend line includes the following steps: Perform Savitzky-Golay sliding window processing on the residual energy superposition trend line, perform third-order polynomial fitting within each fixed-length window to generate a local smooth curve; Perform Savitzky-Golay sliding window processing on the residual energy superposition trend line, specifically referring to: on the time axis of the entire residual energy trend curve, successively intercept adjacent data segments by sliding a window of a preset fixed length, and within each data segment, with the window center point as a reference, use the Savitzky-Golay filtering algorithm to perform third-order polynomial fitting on the data points of this segment. The Savitzky-Golay sliding window is a local smoothing algorithm that eliminates high-frequency fluctuation noise while maintaining the data trend characteristics by fitting a low-order polynomial within the sliding window. Different from simple moving average filtering, its advantage lies in being able to better retain the local slope, peak, and bending structure of the original signal. In this application, through the polynomial curves generated by fitting each window, the smooth change form of the trend in this interval can be extracted, thereby constructing multiple local smooth segments. These segments are spliced to form a continuous smooth expression of the entire residual energy trend, providing basic support for subsequent calculation of the derivative sequence and analysis of the energy evolution rate. At the same time, this method has high sensitivity to weak energy changes under continuous narrow pulse excitation and can effectively capture the potential cumulative growth trend in the residual response of the filter.

[0034] At the central position of each sliding window, extract the first derivative value of this curve to form a derivative sequence of the residual trend on the entire time axis; At the central position of each sliding window, extracting the first derivative value of this curve means: for the smooth curve generated by fitting within the sliding window, calculate the rate of change of the current trend at its central point position. The essence of this operation is to judge whether the residual energy at the current moment is rising, falling, or remaining stable. Specifically, when implementing, through numerical differentiation or difference methods, take the data points adjacent to the left and right of the central point within the window, and estimate the rate of change at this position with their voltage change and time interval. As the window continuously slides on the entire trend line, the rate of change of all central points is extracted and recorded one by one, and finally a complete "derivative sequence" is formed, that is, the rate of change trajectory of the trend. This trajectory reflects the cumulative or dissipated state of the filter residual energy under continuous pulse excitation and is the key judgment basis for identifying whether there is continuous energy superposition and evolving into background noise. In this way, the "shape" of the energy change can be transformed into an identifiable dynamic change "trend", laying a foundation for subsequent background noise analysis.

[0035] Perform boundary extension processing on the derivative sequence to avoid the interference of fitting distortion of the edge window on the overall rate continuity; Boundary extension processing of the derivative sequence means that when dealing with both ends (start and end) of the trend curve, in order to avoid problems such as inaccurate fitting or sudden changes in derivatives caused by insufficient data in the edge region, a certain data compensation strategy is adopted to expand and smooth the edge region. Specific implementation methods include mirror extension, constant value extension, or polynomial extrapolation, etc. For example, at the starting point of the curve, the first few data points can be "mirror copied" forward according to their changing trend to simulate their trend at an earlier position on the time axis, so as to provide sufficient data support for the sliding window; similarly, at the end point, the simulated values can be extended backward to ensure that the end window still has complete data input. The role of boundary extension is to improve the accuracy of sliding fitting at the start and end positions of the time axis, ensuring that the derivative sequence does not have abnormal jumps due to missing fitting data at the beginning and end positions. In this way, a more continuous, smooth, and stable trend rate trajectory can be constructed, enabling the entire derivative sequence to have a unified physical meaning in the full time domain, providing a reliable basis for subsequent trend analysis and clustering judgment.

[0036] Perform moving average smoothing on the derivative sequence to suppress the interference of local mutations on the rate trend; Performing moving average smoothing on the derivative sequence means that on the entire trend rate trajectory, the local mean value of the derivative value at each moment is calculated through a sliding window, thereby weakening the influence of sudden jitters or individual outliers on the overall trend judgment. The specific implementation method is as follows: in the derivative sequence, taking a certain point as the center, taking several adjacent derivative values before and after it, calculating the arithmetic mean of these values, and taking the result as the new derivative value of this center point. Then the window slides sequentially and repeats this process to finally obtain a smoother derivative curve. Moving average smoothing is a commonly used data denoising method. It can effectively suppress sharp changes caused by electrical transients, measurement errors, or fitting edge effects, making the trend change more coherent and analyzable. In the noise reduction and identification process of the current pulse generator, using this smoothing method can help filter out rate mutations that do not have physical significance, thus ensuring that the process of judging the residual energy superposition trend is more stable and reliable.

[0037] Extract the current derivative at each fixed sampling time point as the trend offset rate parameter, which is used to characterize the change amplitude and direction of the residual energy superposition trend.

[0038] Extracting the current derivative as the trend offset rate parameter at each fixed sampling time point means sampling the derivative sequence after smoothing processing at a preset time interval and extracting the derivative value at each sampling point to quantify the change rate and direction of the superposition trend of the filter residual energy within this time period. In specific implementation, the time axis can be equally divided into several sampling points. For example, sampling is performed every 10 microseconds, and the derivative value corresponding to each position is recorded at each sampling point to form a discretized rate parameter sequence. The reason for doing this is that the accumulation or release of residual energy is not continuous and uniform, but dynamically fluctuates with factors such as pulse rhythm and filter recovery state. By extracting the rate information at these representative time points, not only can the microscopic evolution characteristics of the trend be captured, but also the interference of invalid high-frequency noise can be eliminated, thereby providing a stable and accurate basis for subsequent identification of phenomena such as continuous energy accumulation and self-excited evolution. This parameter ultimately serves as the key input for judging the evolution trend of background noise and is an important part of the entire noise reduction evaluation mechanism.

[0039] In this embodiment, in S03, the process of generating the perturbation frequency domain diffusion degree parameter based on the residual energy superposition trend line includes the following steps: Perform multi-level wavelet packet decomposition on the residual energy superposition trend line, and use the db4 wavelet basis function to split the original signal into multiple sub-band signals; Performing multi-level wavelet packet decomposition on the residual energy superposition trend line means recursively decomposing the trend signal at multiple scales so that its frequency domain characteristics can be finely expressed at different resolutions. In implementation, first select the db4 wavelet basis function as the decomposition basis. It is a basis function in the Daubechies wavelet family with good smoothness and compact support and is suitable for processing signals of medium complexity. Then perform the first-level wavelet packet decomposition on the original trend curve, dividing it into equally wide low-frequency and high-frequency sub-signals; on this basis, continue to perform the second-level or even third-level decomposition on each sub-signal until the signal is finally split into multiple sub-frequency bands with uniform frequency band coverage and gradually increasing resolution. Different from traditional wavelet transform, wavelet packet decomposition not only refines high-frequency signals but also equivalently processes low-frequency parts, so it is more suitable for analyzing trend data with uncertain energy distribution density changes. In this way, the energy proportion of the sub-signals on each frequency band can be analyzed independently, comprehensively revealing the diffusion of the filter residual energy in the frequency domain and laying a foundation for subsequent entropy value calculation and perturbation characteristic identification. The key to this process is to realize the conversion from time series trend to frequency structure, thereby capturing hidden interference evolution signs that are not easily visually identified in the time domain.

[0040] Calculate the energy proportion of the corresponding sub-signals on each frequency band to form a complete frequency band energy distribution vector; Calculating the energy proportion of the corresponding sub-signal in each frequency band means calculating the energy occupancy ratio of each frequency band sub-signal obtained after multi-level wavelet packet decomposition in the entire trend curve, so as to construct a set of vectors reflecting the characteristics of the frequency-domain energy distribution. In specific implementation, first square the sub-signal of each frequency band to obtain the sum of the squared amplitudes of all sampling points as the energy value of this frequency band; then sum the energy values of all frequency bands to obtain the total energy; then divide the energy value of each frequency band by the total energy to obtain the energy proportion of this frequency band. For example, if 8 frequency bands, namely A1 - A8, are obtained after three-level wavelet packet decomposition, and their corresponding energies are E1 - E8 respectively, and the total energy is E_total, then the energy proportion of each frequency band is E1 / E_total, E2 / E_total... E8 / E_total, finally forming a frequency band energy distribution vector with a length of 8. This vector reflects the distribution structure of the residual energy of the filter in different frequency bands. If the energy proportion of some high-frequency bands continues to increase, it may mean that the energy diffuses in the frequency domain, implicitly containing the trend of noise evolution. In this way, not only can the diffusion degree of the frequency-domain perturbation be quantified, but also a reliable data basis is provided for the subsequent entropy value calculation.

[0041] Apply the Shannon entropy calculation method to the frequency band energy distribution vector to obtain the dispersion index of the energy distribution; Applying the Shannon entropy calculation method to the frequency band energy distribution vector aims to quantify the "dispersion degree" of the distribution of the residual energy of the filter in different frequency bands, that is, whether the energy is concentrated in a few frequency bands or widely dispersed. In specific implementation, first use each item in the frequency band energy distribution vector as a component in a probability distribution, that is, regard the energy proportion of each frequency band as its "occurrence probability"; then substitute each proportion value into the Shannon entropy formula for calculation. The core of this method is to perform the calculation of "-p×log(p)" on all probability components and sum the results. The Shannon entropy is essentially used to measure the uncertainty of information or the degree of uniformity of the distribution. When the energy is concentrated in a few frequency bands, the entropy value is small; on the contrary, if the energy is evenly distributed in multiple frequency bands, the entropy value will be large. For example, if in a certain decomposition, the energy of the frequency band is almost concentrated in two frequency bands, the entropy value is low, indicating that the perturbation structure is clear; if the energy is evenly distributed in all frequency bands, the entropy value increases, indicating that the frequency-domain energy diffuses and has a high degree of uncertainty. This entropy value, as a parameter of the diffusion degree of the perturbed frequency domain, directly reflects the possibility that the residual energy of the filter is changing from regular perturbation to broad-spectrum noise, and is one of the key reference bases for self-excited background noise recognition.

[0042] Normalize the Shannon entropy value to a standard value range for horizontal comparison between different periods; Normalizing the Shannon entropy value to a standard value range is to eliminate the absolute entropy value differences caused by different signal lengths, numbers of frequency bands, or energy distribution structures between different periods or data segments, so that the entropy values between different periods are comparable, facilitating horizontal comparative analysis of their disturbance diffusion trends. In specific implementation, first determine the theoretical minimum and maximum values of the Shannon entropy. The minimum value is usually 0 (indicating that the energy is completely concentrated in one frequency band), and the maximum value is log(N), where N is the number of frequency bands (for example, when there are 8 frequency bands, the maximum entropy value is log(8) ≈ 2.08). Then substitute the entropy value H of the current period into the normalization formula: H_norm = H / log(N), so that the result is restricted within the standard value range of 0 to 1. The closer the normalized entropy value is to 1, the more uniform the energy distribution is, and the more the disturbance tends to be broad-spectrum; the closer it is to 0, the more concentrated the energy is, and the disturbance has a structure. Through this normalization process, not only can a unified evaluation scale be ensured when analyzing the residual behavior of multiple periods, but it can also be used as a reliable indicator to participate in subsequent state classification and intervention determination, improving the dynamic adaptability and algorithm stability of the overall noise reduction method.

[0043] Take this normalized entropy value as the disturbance frequency-domain diffusion degree parameter to reflect the diffusion degree of the residual energy in the frequency domain.

[0044] Taking the normalized entropy value as the disturbance frequency-domain diffusion degree parameter is to quantitatively reflect the diffusion degree of the filter residual energy in the frequency domain and use it as one of the core bases for determining whether there is the evolution of self-excited background noise. The specific implementation method is: record the Shannon entropy value calculated and normalized for each period as the disturbance frequency-domain diffusion degree parameter corresponding to that period, forming a sequence of diffusion index evolving with time, which is used to dynamically monitor the change trend of the frequency-domain energy distribution. The reason for doing this is that in the special pulse working state where the filter is not fully recovered and energy accumulates, self-excited interference does not appear in the form of obvious spikes, but is manifested as "background noise" with spectrum expansion. Traditional methods cannot identify this change, while this parameter can reveal the evolution process of energy from concentration to diffusion, and then judge whether the system is entering the stage of potential instability or crosstalk risk. By tracking the trend of this parameter, early identification and intervention deployment of hidden interference can be achieved, improving the intelligence and predictability of the noise reduction process of the entire current pulse generator.

[0045] In this embodiment, in S03, the trend offset rate parameter and the disturbance frequency-domain diffusion degree parameter are used to reflect whether the superposition behavior of the filter residual energy evolves into self-excited background noise. Specifically: Combine the trend offset rate parameter and the disturbance frequency-domain diffusion degree parameter into a two-dimensional time series vector sequence on the same time axis; Combining the trend shift rate parameter and the perturbation frequency domain diffusion parameter into a two-dimensional time series vector sequence on the same time axis is intended to integrate the time-domain evolution trend and frequency-domain diffusion characteristics of the filter residual energy into a dynamic analysis framework, thereby more comprehensively identifying the formation trajectory of self-excited background noise. Specifically, a time index is first established for each fixed time sampling point (e.g., every 10 microseconds). Then, at that time point, the corresponding trend shift rate parameter value (indicating the speed and direction of residual energy change) and the perturbation frequency domain diffusion parameter value (indicating the degree of frequency-domain energy diffusion) are extracted, and these two values are combined to form a two-dimensional vector (v1, v2). Over time, the two-dimensional vectors corresponding to all sampling points are sequentially arranged to form a two-dimensional time series vector sequence, i.e., each moment contains a coordinate pair representing "energy change - frequency domain diffusion." This sequence can not only be visualized as a two-dimensional trajectory plot, intuitively reflecting the parameter evolution path, but also serves as the input for subsequent trajectory morphology analysis and state classification. In this way, the system can dynamically capture the linkage changes of residual energy superposition behavior in two dimensions, make up for the limitations of single indicator judgment, and improve the recognition accuracy of complex interference evolution processes.

[0046] A sliding window cluster analysis is performed on the vector sequence to identify segments where both the trend growth rate and the frequency domain diffusion degree increase simultaneously within a continuous time interval. Such segments are marked as suspected self-excitation segments, indicating that the residual energy is continuously increasing and showing a tendency of frequency domain diffusion. Sliding window cluster analysis is performed on this vector sequence to identify continuous segments on the dynamic time axis where both the trend excursion rate and frequency domain diffusion increase simultaneously. These segments are known as "suspicious self-excitation segments" where self-excited background noise may be evolving. Specifically, a fixed-length sliding window (e.g., containing 5 to 10 consecutive sampling points) is set. During each sliding window, a set of two-dimensional vectors within the window is extracted. A clustering algorithm (such as DBSCAN or K-means) is then applied to the vector data within the window, dividing them into several clusters based on their distribution characteristics in two-dimensional space. Clusters of vectors that are concentrated in intervals with high trend excursion rate and high diffusion values are then identified, and their presence in multiple sliding windows is determined. If a vector cluster maintains this "rapid growth combined with frequency domain diffusion" pattern over a continuous period of time, the segment is labeled a "suspicious self-excitation segment." "Sliding window cluster analysis" involves repeatedly clustering the data within a sliding window on the time axis, dynamically tracking the evolution of the clustering structure. This method is more adaptable than static threshold judgment, can more accurately identify subtle signs of background noise evolution, and achieve early identification of nonlinear and self-driven interference trends.

[0047] Statistically analyze the duration and amplitude of each suspicious self-excitation section to determine whether they exceed the preset duration threshold and energy diffusion threshold; Statistically analyzing the duration and amplitude of each suspicious self-excitation section is for further quantitative determination of the preliminarily marked suspicious areas, ensuring that the identified background noise evolution behaviors have sufficient persistence and intensity, and avoiding misjudgment caused by occasional perturbations. In specific implementation, first, count the number of consecutive sampling points that each suspicious section lasts on the time axis, and calculate the actual duration of this section in combination with the sampling period; then extract the maximum and average values of the trend offset rate parameter and the perturbation frequency domain diffusion degree parameter within this section to measure the energy change amplitude and frequency domain diffusion intensity within this section. Then compare these statistical results with the preset thresholds: if the duration exceeds the set minimum duration threshold (such as 50 μs), and at the same time the average diffusion degree parameter or the trend rate parameter exceeds the set energy diffusion threshold (such as the normalized value 0.7), then determine that this section is a "significant self-excitation segment". In this way, short-time and low-amplitude perturbations can be effectively excluded, the recognition accuracy of real and evolving background noise can be improved, and a more reliable data basis can be provided for subsequent classification responses and intervention strategies.

[0048] If within any suspicious self-excitation section, the continuous sampling time length is not less than the set minimum duration threshold, and the average value of the trend offset rate parameter in this section is greater than the trend rate threshold, and at the same time the average value of the perturbation frequency domain diffusion degree parameter is greater than the diffusion degree threshold, then it is determined that the filter is in the process of self-excited background noise evolution, and subsequent classification and intervention operations are triggered.

[0049] To accurately determine whether the filter is in the process of self-excited background noise evolution, it is necessary to quantitatively analyze each "suspicious self-excited section" and combine three key thresholds for decision-making logic judgment. First, count the continuous sampling time length within this section, that is, the time range from the start to the end of this section, and compare it with the set minimum time length threshold, which is usually determined according to the average recovery time of the filter residual energy in the natural dissipation state, aiming to exclude misjudgments caused by short-term accidental disturbances. Then, calculate the average value of the trend offset rate parameter within this section and compare it with the trend rate threshold; this threshold reflects the significant degree of residual energy growth, and generally, the normal fluctuation range of the trend rate is measured through a large number of experiments, and then a certain safety margin is added. Next, calculate the average value of the disturbance frequency domain diffusivity parameter within the same section and compare it with the diffusivity threshold. This threshold represents whether the spectral energy begins to expand to higher-order frequency bands, indicating that the interference has propagation characteristics. Only when all three criteria are met simultaneously, that is, the time period is long enough, the energy trend continues to increase, and the frequency disturbance intensifies, can it be considered that the filter fails to complete natural attenuation, enters the excitation path of energy superposition, and may evolve into a stable background interference state. This determination method can not only effectively exclude atypical noise but also identify potential evolving interference in advance, ensuring that subsequent classification and intervention operations have an accurate trigger basis, and enhancing the robustness and practicality of the entire noise reduction scheme.

[0050] During the continuous narrow pulse output of the current pulse generator, if the pulse period is shorter than the natural dissipation time of the filter energy, the filter will be frequently excited by the incompletely attenuated energy, resulting in the accumulation of residual energy over multiple cycles, forming an imperceptible self-excited background noise. Since this type of background noise does not exhibit obvious spikes or distortions, its evolution process is concealed and progressive, and it is difficult for traditional noise reduction methods to effectively identify it. Therefore, it is necessary to conduct in-depth analysis based on the residual energy superposition trend line: on the one hand, the Savitzky-Golay sliding window polynomial fitting derivative algorithm is used to extract the trend offset rate parameter, which can capture the change rate and direction of the energy accumulation trend; on the other hand, the wavelet packet decomposition combined with the Shannon entropy calculation method is used to generate the disturbance frequency domain diffusivity parameter, which can describe whether the spectral energy diffuses to higher frequency bands and characterize the propagation trend of the interference. These two parameters quantify whether the residual energy superposition behavior of the filter has evolved into a persistent and diffusive background interference from two dimensions of time change and frequency domain disturbance, which is the key basis for achieving high-precision identification and triggering intervention operations, and also the core innovation point of this noise reduction method different from traditional technologies.

[0051] S04. Map the trend offset rate parameter and the disturbance frequency domain diffusivity parameter into a two-dimensional trajectory path, use the similar conical path mapping method to extract the trajectory deformation characteristics, and generate a state factor representing the stage of the self-excited background noise formation process; In this embodiment, in S04, the following steps are specifically included: Taking the trend offset rate parameter corresponding to each moment in chronological order as the abscissa and the disturbance frequency domain diffusion degree parameter as the ordinate, sequentially constructing a two-dimensional trajectory path, and forming a set of trajectory segments reflecting the parameter change process; This step constructs a two-dimensional trajectory path reflecting the superposition behavior of the filter residual energy evolving over time by pairing the trend offset rate parameter with the disturbance frequency domain diffusion degree parameter according to the timestamp. Specifically, in implementation, first, time index markings are made on the trend offset rate parameter and the disturbance frequency domain diffusion degree parameter calculated within each sampling period, and it is ensured that these two parameters have a one-to-one corresponding time order; then, taking the trend offset rate parameter as the horizontal axis coordinate value and the disturbance frequency domain diffusion degree parameter as the vertical axis coordinate value, connecting these coordinate points in chronological order on the two-dimensional plane to form a trajectory line with a time order. This trajectory essentially reflects the coupling relationship between the superposition trend (speed) of the filter residual energy and the frequency domain disturbance (complexity) in the dynamic process. For example, when the residual energy accumulates rapidly and is accompanied by an increase in frequency domain diffusion, the trajectory will shift towards the upper right; conversely, if the energy change slows down or the diffusion weakens, the trajectory may show a downward left convergence trend. After dividing these continuous trajectories into several segments, a set of trajectory segments can be formed, providing a basis for subsequent geometric feature extraction and stage state analysis. This path construction process not only retains the joint evolution information of the key parameters but also has strong visualization and pattern recognition potential.

[0052] In the two-dimensional trajectory path, geometric shape features of each trajectory segment are extracted by a sliding window with a fixed length, including local curvature, path bending rate, angle rotation trend, and change rate; In this step, by setting a sliding window of a fixed length on the two-dimensional trajectory path, the local geometric morphological features of the trajectory are extracted segment by segment to analyze the detailed change trend of parameter evolution. When specifically implemented, first set the length of the sliding window (for example, including 5 to 7 consecutive trajectory points), and slide this window along the trajectory path in chronological order, and extract the set of trajectory points within the current window each time. For each window, the following features are extracted through geometric calculation methods: The local curvature reflects the degree of bending of the trajectory within this interval, and can be estimated by fitting an arc or the second derivative; the path bending rate measures the frequency of trajectory turning back or turning, and usually calculates the standard deviation of the path direction change or the maximum angle change; the angle rotation trend evaluates the consistency or torsion trend of the trajectory turning, for example, using the angle trend between the line connecting the start and end points and the intermediate vector; the change rate quantifies the ratio of the forward distance of the trajectory within the window to time, and is used to reflect the acceleration or deceleration degree of the overall parameter change. After extracting these features, they can be used as the feature descriptors of the trajectory segment for subsequent similarity analysis and state recognition. This method can accurately describe the geometric characteristics of the self-excited background noise evolution trajectory without destroying the time sequence, providing a structured basis for subsequent state discrimination.

[0053] Project the extracted geometric morphological features onto the conical path mapping space. By analyzing the contour morphology of each trajectory segment in this space and performing clustering, it is divided into multiple continuous evolution state regions; In this step, the geometric morphological features (such as local curvature, path bending rate, angle rotation trend, and change rate) of each trajectory segment extracted in the previous stage are uniformly mapped into a high-dimensional geometric space called the conical path mapping space to enhance the distinguishability of different trajectory segment morphologies. The specific implementation method is as follows: First, standardize the feature vectors of each trajectory segment so that features with different dimensions have relatively balanced effects in the projection; then, embed these standardized vectors into the conical mapping space constructed based on the criterion of "continuous morphological evolution". In this space, trajectory segments close to the conical axis represent "stable evolution", and those far from the axis represent "drastic changes". In this space, the projection of each trajectory segment will show a certain contour morphology, such as spiral, turning back, bending, or spreading trajectories, etc. Subsequently, apply a clustering algorithm (such as density-based DBSCAN or continuous distribution K-means) to perform similarity clustering on the projection results of all trajectory segments, group trajectory segments with similar geometric morphological change trends into the same class, and finally form several continuous evolution state regions. Each region represents a potential energy superposition evolution stage. In this way, not only can a clear state transition structure be extracted from complex parameter changes, but also potential mutation points or pattern turning points can be identified, providing a structured basis for subsequent generation of state factors.

[0054] Generate a state factor characterizing the formation process stage of self-excited background noise based on the evolution order, deformation amplitude, and distribution characteristics of each state region in the trajectory, and use it as the basis for subsequent classification and intervention.

[0055] In this step, by analyzing the evolution order, deformation amplitude, and distribution characteristics of each state region in the two-dimensional trajectory path, a state factor is generated to characterize the formation process stage of self-excited background noise. The specific implementation method is as follows: First, arrange each clustering state region in the time order of the trajectory to construct a state sequence reflecting the energy evolution process; then, calculate the time length (duration) and deformation amplitude index (such as average curvature, maximum bending rate) occupied by each state region in the trajectory to quantify the stability and severity of the characteristics of this stage; then, normalize and combine these quantified characteristics to form a multi-dimensional state description vector. This description vector can be compressed into a low-dimensional state factor through dimensionality reduction techniques such as principal component analysis (PCA) or t-SNE as a compact representation of the overall behavior stage. Finally, map and match this state factor with multiple defined noise formation stages (such as "not triggered", "perturbation accumulation period", "critical turning period", "self-excited noise generation period", etc.) to complete the determination of the evolution state of the residual energy of the current filter. This state factor can not only be used for the judgment of subsequent classification labels but also as the basis for triggering intervention measures to achieve active control of the formation process of current pulse background noise.

[0056] The purpose of executing this step is to abstract the dynamic evolution process of the residual energy of the filter under continuous narrow pulse excitation from the original numerical parameter level into an interpretable and classifiable stage state. Since the trend offset rate parameter reflects the speed change of energy accumulation and the perturbation frequency domain diffusion parameter reflects the distribution breadth of energy in the frequency domain, after mapping these two parameters to a two-dimensional trajectory path, their co-evolution relationship in the time dimension can be intuitively captured. And by further extracting the geometric deformation characteristics of the trajectory through the conical path mapping method and dividing it into multiple continuous state regions, the key turning points and process stages of the residual energy of the filter evolving from the normal recovery stage to the self-excited noise stage can be effectively revealed. The finally generated state factor not only has the ability to judge the stage of complex dynamic behavior but also provides a clear target classification basis for subsequent intervention measures, which is the core bridge to break through the bottleneck of traditional noise reduction "fuzzy recognition" and achieve precise noise reduction control.

[0057] S05. Perform stage determination based on the generated state factor, and execute intervention measures according to the determination result, including adjusting the pulse output rhythm, inserting gap interference pulses, or applying rhythm disturbance control to block the continuous superposition of the residual energy of the filter.

[0058] In this embodiment, in S05, it specifically includes: Based on the distribution position of the generated state factor in the multi-dimensional state space, determine the stage of the current filter residual energy superposition behavior, and divide the state factor into an initial fluctuation stage, an evolution critical stage, and a continuous self-excitation stage; The stage of the filter residual energy superposition behavior can be determined based on the distribution position of the state factor in the multi-dimensional state space in the following way: First, the state factor is constructed from the deformation characteristics in the aforementioned trajectory path, and usually represents the dynamic changes of each trajectory segment with multiple dimensional values, such as the trajectory bending rate, rotation amplitude, change rate, etc. For the convenience of subsequent stage determination, project all state factors into a predefined multi-dimensional state space, where each dimension corresponds to a standardized representation of a feature quantity. In this space, three stage determination regions are predefined in advance: The initial fluctuation stage region covers state points with low feature amplitude and slow trajectory changes; the evolution critical stage region covers state points with medium feature amplitude and obvious bends in the trajectory path but the change rate has not continuously increased; while the continuous self-excitation stage region focuses on state clusters with high feature amplitude, severe trajectory twists, and continuous offsets. By matching the position of the current state factor in this multi-dimensional space and judging the nearest determination region, the stage of the current filter residual energy superposition evolution can be identified. This method can be implemented by means of a support vector machine (SVM) classifier, K-nearest neighbor (KNN) discrimination, or the determination of the nearest cluster belonging based on the Euclidean distance, effectively improving the resolution and anti-interference ability of state recognition. The stage determination completed in this way provides a basic basis for formulating differential intervention strategies in the subsequent stage.

[0059] When the state factor is in the initial fluctuation stage, by introducing a time interval perturbation with a fixed amplitude in each continuous pulse period, change the rhythm arrangement of the continuous pulse output to prevent the filter from accumulating energy under constant excitation; When the state factor is in the evolution critical stage, insert skip interference pulses with a set interval in the continuous pulse output sequence to change the pulse trigger timing and interrupt the residual energy superposition link in the filter; When the state factor is in the continuous self-excitation stage, perform a rhythm disruption control operation, including setting the rhythm change rules within multiple continuous pulse output periods, and adopting the methods of extending the pulse interval and inserting blank time periods to clear the accumulated energy inside the filter and block the continuous formation of self-excited background noise.

[0060] The following is a detailed description of the implementation methods and reasons for the intervention measures corresponding to these three state factors: When the state factor is in the initial fluctuation stage: The rhythm can be adjusted by means of dynamic perturbation pulse period. The specific method is as follows: On the premise of maintaining the stability of the overall pulse sequence, introduce a small time perturbation of ±ΔT level to each continuous pulse period. The perturbation amplitude can be generated by a pseudo-random sequence, such as based on the linear congruence method or M-sequence generator, so that the pulse interval is slightly offset within a reasonable range. This perturbation method can effectively break the resonance accumulation path caused by continuous equi-periodic pulses to the filter and avoid the formation of synchronous excitation of energy inside the filter. The reason for taking this measure is that a stable residual superposition chain usually has not been formed in the initial fluctuation stage, and only periodic responses appear under the condition of small perturbation. Therefore, the problem development can be effectively suppressed by slightly disturbing the rhythm.

[0061] When the state factor is in the evolution critical stage: The strategy of inserting skip interference pulses can be adopted for processing. The implementation method is as follows: In the pulse sequence, every N normal output periods, deliberately insert a pulse with a "skip" within one period, that is, no actual current output is carried out in this period, but only the trigger signal or a blank pulse signal is sent. This method can be realized by setting a skip beat table inside the drive controller and arranging the rhythm control bits in advance. Its purpose is to actively interrupt the superposition path formed inside the filter, similar to the interference method of "interrupting the current resonance chain", and prevent the residual energy from further superimposing and expanding along the time axis. This stage is the critical state before self-excitation formation, so effective intervention should be carried out through medium-strength rhythm reconstruction means.

[0062] When the state factor is in the continuous self-excitation stage: Rhythm disturbance control with stronger intervention should be executed. The specific implementation method is as follows: Set multiple continuous output periods as a perturbation rhythm unit. Within this unit, gradually extend the pulse interval (such as in a linear increase or exponential increase manner), and insert 1 - 2 complete blank periods in the middle or at the end, so that there is no input pulse at the output end of the filter, forming a complete energy release window. At the same time, combine adjustable PWM duty cycle control or digital delay line module to precisely manage the pulse timing, so that the disturbance operation is controllable and predictable. This method simulates the behavior of "actively emptying the filter" to clear the internal accumulated energy and restore its stable state. This intervention method is applicable to the scenario where self-excited background noise has been significantly formed and has a stable periodicity, and is the key measure to inhibit its continuous evolution.

[0063] Such as Figure 2 shown, a noise reduction device for a current pulse generator includes a pulse characteristic analysis module, a residual energy trend modeling module, a dynamic evolution parameter extraction module, a state factor generation module, and a rhythm intervention and regulation module; The pulse characteristic analysis module, during the continuous narrow pulse output of the current pulse generator, collects the pulse width, pulse interval, and edge slope parameters of each pulse, obtains the output voltage of the filter corresponding to each period, and forms a filter response sequence of multiple periods; The residual energy trend modeling module removes the main pulse waveform from the filter response sequence, extracts the filter residual response segments of multiple cycles, performs time alignment and overlapping processing on the residual response segments of multiple cycles, and constructs a residual energy superposition trend line; The dynamic evolution parameter extraction module, based on the residual energy superposition trend line, uses the Savitzky-Golay sliding window polynomial fitting derivative algorithm to calculate the trend offset rate parameter, and uses the wavelet packet decomposition and Shannon entropy calculation algorithm to generate the perturbation frequency domain diffusion parameter to reflect whether the filter residual energy superposition behavior evolves into self-excited background noise; The state factor generation module maps the trend offset rate parameter and the perturbation frequency domain diffusion parameter into a two-dimensional trajectory path, uses the similar conical path mapping method to extract the trajectory deformation characteristics, and generates a state factor representing the stage of the self-excited background noise formation process; The rhythm intervention and regulation module performs stage determination based on the generated state factor, and executes intervention measures according to the determination result, including adjusting the pulse output rhythm, inserting gap interference pulses or applying rhythm disturbance control to block the continuous superposition of the filter residual energy.

[0064] The above formulas are all dimensionless and take their numerical calculations. The formula is a formula obtained by software simulation by collecting a large amount of data to approximate the real situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0065] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired or wireless (such as infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that the computer can access, or a data storage device such as a server or data center that includes one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0066] It should be understood that in various embodiments of the present application, the magnitudes of the serial numbers of the above processes do not imply the order of execution, and the order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0067] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present application.

[0068] In several embodiments provided by the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the above-described embodiments are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0069] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0070] In addition, the functional units in various embodiments of the present application can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0071] As described above, only the specific implementation manners of the present application are provided, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present application, and all should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A current pulse generator noise reduction method, characterized in that: The specific steps include: S01. During the continuous narrow pulse output process of the current pulse generator, the pulse width, pulse interval and edge slope parameters of each pulse are collected to obtain the output voltage of the filter in the corresponding cycle, thereby forming a filter response sequence of multiple cycles; S02. removing the main pulse waveform from the filter response sequence, extracting filter residual response segments of multiple cycles, performing time alignment and overlapping processing on the residual response segments of multiple cycles, and constructing a residual energy superposition trend line; S03. Based on the residual energy superposition trend line, the Savitzky-Golay sliding window polynomial fitting derivative algorithm is used to calculate the trend offset rate parameter, and the wavelet packet decomposition and Shannon entropy calculation algorithm are used to generate the disturbance frequency domain diffusion parameter to reflect whether the residual energy superposition behavior of the filter evolves into self-excited background noise; S04, mapping the trend offset rate parameter and the disturbance frequency domain diffusion parameter into a two-dimensional trajectory path, extracting trajectory deformation features using a similar cone path mapping method, and generating a state factor representing the stage of the self-excited background noise formation process; S05. Performing stage determination based on the generated state factor and executing intervention measures according to the determination result, including adjusting the pulse output rhythm, inserting gap interference pulses, or applying rhythm disturbance control to block the continuous superposition of the residual energy of the filter.

2. A method for reducing noise of a current pulse generator according to claim 1, characterized in that In S01, the following steps are specifically included: During the continuous narrow pulse output process of the current pulse generator, the start time and end time of each pulse are collected based on the trigger clock, the pulse width is calculated by the start and end time, the pulse interval is calculated by the start time difference of adjacent pulses, and the edge slope is calculated by the start edge fitting curve; Taking the start time of each pulse as the reference point, the voltage signal waveform at the output of the filter is obtained within a fixed sampling window, including the main pulse and the response segment after the main pulse; Repeat the parameter acquisition and voltage acquisition operations for multiple pulse cycles, and record the data of each cycle according to the timestamp number; The parameters and response voltage data of multiple cycles are arranged in a preset cycle order to form a filter response behavior sequence on a continuous time axis for subsequent residual energy analysis.

3. A noise reduction method for a current pulse generator according to claim 1, characterized in that, In S02, the following steps are specifically included: Determine the time interval corresponding to the main pulse in each cycle, calculate the start and end boundaries of the main pulse based on the pulse start time and pulse width, and remove the filter voltage data in the time interval corresponding to the main pulse in each cycle from the original response sequence; In each cycle, a fixed-length residual observation window is set after the main pulse ends, and the voltage sampling points retained in the window are extracted as the residual response segment, and their relative time coordinates relative to the pulse starting point are uniformly recorded; The residual response segments extracted from all cycles are aligned according to their start time and overlapped on a unified time axis to generate a cross-cycle residual response distribution set. The voltage data corresponding to each time point in the residual response distribution set are averaged to obtain a residual energy evolution curve covering multiple cycles, and the residual energy superposition trend line is constructed based on this curve. The specific construction process is as follows: sliding window processing is performed on the residual energy evolution curve covering multiple cycles, and the data points in each window are fitted into local linear segments to obtain the local slope distribution in the continuous time interval; the slope sequence of multiple local linear segments is smoothed and filtered to eliminate the high-frequency noise offset caused by small disturbances and generate a continuous trend-guiding signal; the trend-guiding signal is sampled at a fixed time interval, and a third-order polynomial is used to fit the data interval before and after each sampling point to form a continuous interpolation expression of the trend curve; all interpolation points are connected to form a continuous change trajectory covering the entire residual response time axis, and this trajectory is used as the superposition trend line of the filter residual energy.

4. A method for reducing noise of a current pulse generator according to claim 1, characterized in that, In S03, the process of calculating the trend shift rate parameter based on the residual energy superimposed trend line includes the following steps: The Savitzky-Golay sliding window process is performed on the residual energy superposition trend line, and a third-order polynomial fitting is performed within each fixed-length window to generate a local smooth curve; At the center of each sliding window, the first-order derivative value of the curve is extracted to form the derivative sequence of the residual trend on the entire time axis; The derivative sequence is processed by boundary extension to avoid the interference of edge window fitting distortion on the overall rate continuity; The derivative series is smoothed by sliding average to suppress the interference of local mutations on the rate trend; The current derivative is extracted at each fixed sampling time point as the trend shift rate parameter to characterize the change amplitude and direction of the residual energy superposition trend.

5. The current pulse generator noise reduction method according to claim 1, characterized in that: In S03, the process of generating the disturbance frequency domain diffusion parameter based on the residual energy superposition trend line includes the following steps: Multi-level wavelet packet decomposition is performed on the residual energy superposition trend line, and the db4 wavelet basis function is used to split the original signal into multiple sub-band signals; Calculate the energy proportion of the corresponding sub-signal in each frequency band to form the energy distribution vector of the complete frequency band; Applying Shannon entropy calculation method to the frequency band energy distribution vector, the energy distribution dispersion index is obtained; Normalize the Shannon entropy value to a standard value range to facilitate horizontal comparison between different cycles; The normalized entropy value is used as the disturbance frequency domain diffusion parameter to reflect the diffusion degree of residual energy in the frequency domain.

6. A method for reducing noise of a current pulse generator according to claim 1, characterized in that, In S03, the trend offset rate parameter and the disturbance frequency domain diffusion parameter are used to reflect whether the residual energy superposition behavior of the filter evolves into self-excited background noise, specifically: On the same time axis, the trend offset rate parameter and the disturbance frequency domain diffusion parameter are combined into a two-dimensional time series vector sequence; A sliding window cluster analysis is performed on the vector sequence to identify segments where both the trend growth rate and the frequency domain diffusion degree increase simultaneously within a continuous time interval. These segments are marked as suspected self-excitation segments, indicating that the residual energy is continuously increasing and showing a frequency domain diffusion tendency. The duration and amplitude of each suspected self-excitation segment are counted to determine whether it exceeds the preset duration threshold and energy diffusion threshold; If within any suspicious self-excitation section, the continuously existing sampling time length is not less than the set minimum time length threshold, and the average value of the trend offset rate parameter in this section is greater than the trend rate threshold, and at the same time the average value of the disturbance frequency domain diffusion parameter is greater than the diffusion threshold, then it is determined that the filter is in the process of self-excited background noise evolution, and subsequent classification and intervention operations are triggered.

7. A noise reduction method for a current pulse generator according to claim 1, characterized in that, In S04, it specifically includes the following steps: Taking the trend offset rate parameter corresponding to each moment in chronological order as the abscissa and the disturbance frequency domain diffusion parameter as the ordinate, sequentially constructing a two-dimensional trajectory path to form a set of trajectory segments reflecting the parameter change process; In the two-dimensional trajectory path, using a sliding window of fixed length to extract the geometric shape features of each trajectory segment, including local curvature, path bending rate, angle rotation trend, and change rate; Projecting the extracted geometric shape features into the conical path mapping space, analyzing the contour shape of each trajectory segment in this space, and performing clustering to divide it into multiple continuously evolving state regions; According to the evolution order, deformation amplitude, and distribution characteristics of each state region in the trajectory, generating a state factor characterizing the formation process stage of self-excited background noise, and using it as the basis for subsequent classification and intervention.

8. A method for reducing noise of a current pulse generator according to claim 1, characterized in that, In S05, it specifically includes: Based on the distribution position of the generated state factor in the multi-dimensional state space, determining the stage at which the current filter residual energy superposition behavior is located, and dividing the state factor into an initial fluctuation stage, an evolution critical stage, and a continuous self-excitation stage; When the state factor is in the initial fluctuation stage, by introducing a time interval disturbance of a fixed amplitude in each continuous pulse period, changing the rhythm arrangement of the continuous pulse output, and preventing the filter from accumulating energy under constant excitation; When the state factor is in the evolution critical stage, inserting skip interference pulses with a set interval in the continuous pulse output sequence, changing the pulse trigger timing, and interrupting the residual energy superposition link inside the filter; When the state factor is in the continuous self-excitation stage, performing a rhythm disruption control operation, including setting the rhythm change rule within multiple continuous pulse output periods, and using the method of extending the pulse interval and inserting blank periods to clear the accumulated energy inside the filter and block the continuous formation of self-excited background noise.

9. A noise reduction device for a current pulse generator, which is used to implement the noise reduction method for a current pulse generator described in any one of the above claims 1-8, and is characterized in that, Including a pulse characteristic analysis module, a residual energy trend modeling module, a dynamic evolution parameter extraction module, a state factor generation module, and a rhythm intervention and regulation module; The pulse characteristic analysis module, during the continuous narrow pulse output of the current pulse generator, collects the pulse width, pulse interval, and edge slope parameters of each pulse, obtains the output voltage of the filter corresponding to each period, and forms a filter response sequence of multiple periods; The residual energy trend modeling module removes the main pulse waveform from the filter response sequence, extracts the residual response segments of the filter for multiple periods, performs time alignment and overlapping processing on the residual response segments of multiple periods, and constructs a residual energy superposition trend line; The dynamic evolution parameter extraction module, based on the residual energy superposition trend line, uses the Savitzky-Golay sliding window polynomial fitting derivative algorithm to calculate the trend offset rate parameter, and uses the wavelet packet decomposition and Shannon entropy calculation algorithm to generate the perturbation frequency domain diffusion parameter to reflect whether the filter residual energy superposition behavior evolves into self-excited background noise; The state factor generation module maps the trend offset rate parameter and the perturbation frequency domain diffusion parameter into a two-dimensional trajectory path, uses the similar conical path mapping method to extract the trajectory deformation characteristics, and generates the state factor characterizing the stage of the self-excited background noise formation process; The rhythm intervention and regulation module performs stage determination based on the generated state factor, and executes intervention measures according to the determination result, including adjusting the pulse output rhythm, inserting a gap interference pulse or applying a rhythm disturbance control to block the continuous superposition of the filter residual energy.

Images