A method, medium, and device for filtering seismic signals

Abstract

This invention relates to the field of signal filtering technology, and in particular to a method, medium, and device for filtering seismic signals. By performing modal decomposition on the target signal, complex mixed seismic signals are broken down into sub-mode signals with single characteristics. A clear sub-mode screening criterion is established based on the average instantaneous energy of the sub-mode signals within a preset time window, accurately distinguishing between non-noise and noise modes. Dynamic control of the iteration process is achieved through iterative updates and iteration stopping conditions based on cross-correlation, avoiding over-iteration or under-iteration caused by a fixed number of iterations, thus ensuring the quality of non-noise sub-mode signals. By superimposing all non-noise sub-mode signals, effective signal components from different iteration levels are aggregated, ensuring that the output effective seismic signal fully retains the effective characteristics of each level while significantly reducing noise interference, thereby improving the purity of the effective seismic signal.

Description

Technical Field

[0001] This invention relates to the field of signal filtering technology, and in particular to a method, medium, and device for filtering seismic signals. Background Technology

[0002] Seismic signal filtering is a core step in seismic exploration data processing. Its goal is to separate surface interference, instrument noise, surface waves, and other noise from noisy seismic signals, as well as effective signals such as reflected and refracted waves, to provide high-quality data for subsequent reservoir prediction and structural interpretation.

[0003] Existing seismic signal filtering methods typically rely on fixed-parameter mode decomposition techniques, such as empirical mode decomposition (EMD) and ensemble empirical mode decomposition (ESD). When dealing with shallow, high-frequency seismic signals (frequency > 50 Hz), the fixed decomposition step size leads to excessive fragmentation of the effective signal into numerous sub-modes, resulting in the loss of some signal features and reducing the accuracy of subsequent seismic signal analysis. Furthermore, for deep, low signal-to-noise ratio (SNR) seismic signals (SNR < 3), a fixed initial decomposition layer is insufficient to uncover weak effective signals embedded in noise, making them unrecognizable in subsequent screening stages and severely impacting the assessment of deep geological structures. In addition, existing decomposition techniques lack effective separation mechanisms specifically designed for jump interference such as sudden surface vibrations and instrument pulses in seismic signals. This allows these jump interferences to easily infiltrate effective sub-modes, reducing the purity of the decomposed sub-modes and interfering with subsequent assessment of the effective seismic signal.

[0004] Therefore, how to adjust the mode decomposition method and parameters according to the dynamic characteristics of seismic signals, avoid over-decomposition of effective signals or burying of weak effective signals, and at the same time, directionally separate skipping interference to improve the purity of the filtered sub-mode signals has become an urgent problem to be solved. Summary of the Invention

[0005] To address the aforementioned technical problems, the present invention provides a seismic signal filtering method, which includes the following steps:

[0006] S1, initialize i=1, and determine the initial seismic signal as the i-th level target signal.

[0007] S2, for any i-th level target signal, perform mode decomposition processing on the current i-th level target signal to obtain several (i+1)-th level sub-mode signals corresponding to the current i-th level target signal, wherein the sub-mode signals are signal components with amplitude modulation and frequency modulation characteristics.

[0008] S3. Based on the average instantaneous energy of each (i+1)th level sub-mode signal within each preset time window, obtain the signal type corresponding to each (i+1)th level sub-mode signal, where the signal type is either a noise mode or a non-noise mode.

[0009] S4, each (i+1)th level submode signal of the signal type non-noise mode is determined as the (i+1)th level target signal, and the cross-correlation degree between each (i+1)th level target signal and the (i)th level target signal that generated the (i+1)th level target signal is obtained.

[0010] S5. If the cross-correlation degree of all target signals of level i+1 is greater than the preset degree threshold, then all sub-mode signals of non-noise mode are obtained and step S6 is executed. Otherwise, update i=i+1 and repeat step S2.

[0011] S6 superimposes all submode signals of the non-noise mode to obtain the filtered effective seismic signal.

[0012] The present invention also provides a non-transitory computer-readable storage medium storing at least one instruction or at least one program, wherein the at least one instruction or at least one program is loaded and executed by a processor to implement the above-described seismic signal filtering method.

[0013] The present invention also provides an electronic device, including a processor and the aforementioned non-transitory computer-readable storage medium.

[0014] This invention has at least the following beneficial effects: By performing modal decomposition on the i-th level target signal, the complex mixed seismic signal is decomposed into sub-mode signals with single characteristics, overcoming the problem of difficult separation of traditional mixed signals, while retaining the core amplitude and frequency modulation characteristics of the effective signal, avoiding the effective signal being masked by noise due to lack of decomposition; by establishing a clear sub-mode screening criterion through the average instantaneous energy of the sub-mode signals within a preset time window, local energy analysis is achieved with the help of the preset time window, avoiding misjudgment of weak effective signals caused by overall energy averaging, so as to accurately distinguish between non-noise modes and noise modes; by calculating the degree of cross-correlation, the similarity between upper and lower level target signals is quantified, providing data for subsequent judgment on whether the decomposition is reasonable and whether effective features are retained. Based on this, to prevent the breakage of effective signal features caused by excessive decomposition; through iterative updates and iterative stopping conditions based on cross-correlation, dynamic control of the iterative process is achieved, avoiding over-iteration or under-iteration caused by a fixed number of iterations, ensuring that the finally obtained non-noise mode sub-signals all have stable and effective features, without any insufficiently purified signal residues, thus guaranteeing the quality of subsequent reconstruction; by superimposing all non-noise mode sub-mode signals, the filtered effective seismic signal is obtained, and the effective signal components of different iteration levels are aggregated to achieve a closed loop from sub-mode decomposition to complete effective signal restoration. The final output effective seismic signal not only fully retains the effective features of each level, but also significantly reduces noise interference, improving the purity of the effective seismic signal. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of an earthquake signal filtering method provided in Embodiment 1 of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It is understood that, where appropriate, the terms used to distinguish similar objects can be interchanged so that the invention can also be implemented in other embodiments besides the illustrated or described embodiments. Furthermore, the terms "including," "having," and any variations are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices.

[0019] Example 1

[0020] This embodiment provides a method for filtering seismic signals, such as... Figure 1 As shown, the seismic signal filtering method includes the following steps:

[0021] S1, initialize i=1, and determine the initial seismic signal as the i-th level target signal.

[0022] The initial seismic signal is the seismic signal collected by the seismic exploration equipment. Its waveform includes effective seismic signals such as reflected waves and refracted waves, as well as various noise and jump interference components such as surface interference, instrument noise, and surface waves. It is the original input data of the filtering method.

[0023] Seismic signal filtering requires multiple iterations to gradually separate noise from effective signals. Each iteration must clearly define the object of decomposition, i.e., the target signal. The initial seismic signal is defined as the first-level target signal. In subsequent iterations, the non-noise modes determined to require further decomposition become the next-level target signals, forming a multi-level decomposition system that ensures the iterative process is orderly and traceable.

[0024] S2, for any i-th level target signal, perform mode decomposition processing on the current i-th level target signal to obtain several (i+1)-th level sub-mode signals corresponding to the current i-th level target signal, wherein the sub-mode signals are signal components with amplitude modulation and frequency modulation characteristics.

[0025] Mode decomposition, based on non-stationary signal analysis theory, uses a pre-defined decomposition algorithm (such as jump-enhanced amplitude-frequency modulation mode decomposition) to break down a complex i-th-order target signal into several sub-mode signals with non-overlapping frequencies and unique characteristics. In this embodiment, mode decomposition must prioritize the separation of sub-modes with amplitude-frequency modulation characteristics.

[0026] The (i+1)th level sub-mode signal is the next level sub-mode signal generated after the i-th level target signal is decomposed. Each sub-mode signal has independent signal characteristics, such as frequency range and amplitude variation. Each level sub-mode signal is a denoising optimization of the previous level target signal. The effective signal ratio of the (i+1)th level sub-mode signal is higher than that of the i-th level target signal. Through multiple rounds of decomposition iteration, a high degree of purification of the effective signal is finally achieved.

[0027] Amplitude-frequency modulation (AM / FM) characteristics combine the two core dynamic features of seismic signals: amplitude modulation (AM) and frequency modulation (FM). Amplitude modulation refers to the periodic or regular change in the instantaneous amplitude of a signal over time. For example, the amplitude of a seismic reflection wave fluctuates periodically from strong to weak to strong, varying with the reflection coefficient of geological interfaces. Frequency modulation refers to the dynamic adjustment of the instantaneous frequency of a signal over time. For instance, as a seismic signal moves from shallow to deep layers, the dominant frequency gradually decreases from 50Hz to 5Hz due to differences in geological layer density, exhibiting a stable frequency decrease trend. Noise, on the other hand, lacks stable amplitude or frequency modulation patterns. The combined AM / FM characteristics are the defining attribute that distinguishes effective seismic signals from noise. During modal decomposition, the presence of AM / FM characteristics is used as the core screening criterion to ensure that the (i+1)th level sub-mode signals generated are all effective signal candidates. This reduces noise interference at the source and preserves the core waveform information of the effective signal. When superimposing these sub-modes, the waveform characteristics of the initial seismic signal can be restored to the greatest extent possible, avoiding signal distortion after reconstruction.

[0028] As described above, by performing mode decomposition processing on the i-th level target signal, the complex non-stationary seismic signal is decomposed into several sub-mode signals with single characteristics. This transforms the separation of effective signals and noise from the overall separation of the mixed signal to the individual separation of sub-modes, significantly reducing the separation difficulty and improving the separation accuracy. By limiting the sub-mode signals to signal components with amplitude modulation and frequency modulation characteristics, noise modes without amplitude modulation and frequency modulation characteristics are eliminated from the decomposition stage. This ensures that the generated i+1-th level sub-mode signals are all effective signal candidates, reducing invalid operations in subsequent screening stages and improving filtering efficiency.

[0029] In one specific embodiment, the mode decomposition process is a skip-enhanced amplitude-frequency modulation mode decomposition process, and S2 includes the following steps:

[0030] S21, extract the key prior features of the current i-th level target signal, where the key prior features include the dominant frequency range and the signal-to-noise ratio.

[0031] S22, determine the iteration step size and initial decomposition level N in the skip-enhanced amplitude-frequency modulation mode decomposition process based on key prior features.

[0032] S23, perform jump-enhanced amplitude-frequency modulation mode decomposition processing according to the iteration step size and the initial decomposition level N to obtain several i+1 level sub-mode signals and jump interference components. The jump interference components are determined based on the ratio of the peak amplitude to the average amplitude of the current i-th level target signal being greater than a preset ratio threshold, and the duration of the peak being less than a preset time threshold.

[0033] The dominant frequency range refers to the main frequency distribution interval of the effective signal in the i-th level target signal. It can be calculated through power spectrum analysis or wavelet transform, and is expressed as the range from the minimum to the maximum dominant frequency. For example, the dominant frequency range of shallow signals is 50Hz to 80Hz, and the dominant frequency range of deep signals is 8Hz to 20Hz. It is a core indicator reflecting the frequency characteristics of the effective signal. The dominant frequency range is used to guide the setting of the iteration step size. By extracting the dominant frequency range features, the decomposition method can be automatically adapted to shallow and deep exploration scenarios, improving its versatility.

[0034] Signal-to-noise ratio (SNR) is the ratio of the effective signal power to the noise signal power in the i-th level target signal, usually expressed in decibels (dB). SNR = 10lg(Pi) valid / P noise ), where P valid For effective signal power, P noise This represents noise power, reflecting the proportion of effective components in the signal. In low signal-to-noise ratio (SNR) signals, skipping interference is easily confused with weak effective signals, so the preset ratio threshold needs to be reduced (e.g., from 5 to 4) to improve the interference recognition rate. In high SNR signals, interference characteristics are more obvious, so a higher ratio threshold can be maintained to reduce false positives of effective signals.

[0035] The iteration step size determines the fitting accuracy of the sub-mode signals after decomposition; a smaller iteration step size results in a finer fit but lower efficiency. The initial decomposition layer number N determines the decomposition depth of the seismic signal; a larger initial decomposition layer number allows for better separation of deeply nested signals but also increases computational cost. By establishing quantitative mapping relationships between the dominant frequency range and iteration step size, and between the signal-to-noise ratio and the initial decomposition layer number, dynamic adaptation of decomposition parameters can be achieved. These quantitative mapping relationships are established by experts through experimental data and theoretical derivation, specifically:

[0036] The higher the master frequency, the greater the rate of frequency change of the effective signal, requiring a smaller iteration step size to capture high-frequency details. For example, high master frequency signals (e.g., >50Hz) change frequency rapidly, requiring a smaller iteration step size (e.g., 0.005) to avoid over-decomposition of the effective signal; low master frequency signals (e.g., <20Hz) change frequency gradually, allowing for a larger iteration step size (e.g., 0.01) to improve decomposition efficiency. The iteration step size for master frequency signals within the range of [20Hz, 50Hz] can be set to 0.008 to balance the degree of signal decomposition.

[0037] The lower the signal-to-noise ratio (SNR), the deeper the effective signal is masked by noise, requiring more initial decomposition layers to fully remove the noise. For example, in low SNR signals (SNR < 3dB), the effective signal is deeply masked by noise, necessitating an increased initial decomposition layer count (e.g., 5 layers) to effectively separate nested weak effective signals. In high SNR signals (SNR > 10dB), the effective signal accounts for a high proportion, allowing for a reduction in the initial decomposition layer count (e.g., 3 layers) to avoid redundant computation. For SNR signals in the range of [3dB, 10dB], the initial decomposition layer count can be set to 4 to balance the degree of decomposition and computational efficiency.

[0038] Jump interference components are interference signal segments in seismic signals generated by sudden factors such as vehicles passing by on the ground or instruments colliding. They do not have amplitude modulation or frequency modulation characteristics and have short duration (usually <0.01 seconds) and high amplitude (usually 3-8 times the average amplitude of the signal). They are the main sudden interference sources affecting the quality of effective signals and need to be separated and removed separately.

[0039] The key difference between Jump Plus AM-FM Mode Decomposition (JMD) and traditional AM-FM mode decomposition lies in its simultaneous extraction of effective AM-FM submodes and separation of jump interference. Jump interference in seismic signals has unique characteristics of short duration and high amplitude, which contrasts sharply with the long-term stationary AM-FM characteristics of effective signals. It can be accurately identified using a dual threshold of amplitude ratio and duration.

[0040] During the execution of JMD, firstly, based on the iteration step size and the initial decomposition level N, amplitude-modulated and frequency-modulated submode fitting is performed on the i-th level target signal (e.g., fitting the instantaneous amplitude and frequency function using the least squares method) to generate preliminary submodes; then, for the peak segments in the preliminary submodes, skipping interference components are screened and separated by the dual conditions of peak amplitude / average amplitude > preset ratio threshold (determining high amplitude) and peak duration < preset time threshold (determining short duration); finally, a clean and effective amplitude-modulated and frequency-modulated submode (i.e., the i+1-th level submode signal) and a separate skipping interference component are output, achieving complete separation of effective signals and interference classification.

[0041] The preset ratio threshold is the amplitude judgment standard for determining whether a peak value is a jump interference. It is usually an integer in the range of 3-8, preferably 5. The calculation method is the amplitude of the jump interference peak value / the average amplitude of the current i-th level target signal. When this ratio exceeds the preset ratio threshold, the peak value is judged to have high amplitude characteristics, which is one of the necessary conditions for jump interference.

[0042] The preset time threshold is the standard for determining whether a peak value is a jump interference. It is a short time value, usually in the range of 0.001-0.01 seconds, preferably 0.005 seconds. It is calculated as the time during which the peak amplitude continuously exceeds 1.5 times the average amplitude. When this time is less than the preset time threshold, the peak value is determined to have a short duration characteristic, which is another necessary condition for jump interference.

[0043] As described above, determining the iteration step size by the main frequency range ensures that the fitting accuracy of the decomposed sub-modes matches the frequency variation characteristics of the effective signal. This avoids fitting errors caused by excessively large step sizes for high-frequency signals and inefficiencies caused by excessively small step sizes for low-frequency signals. Determining the initial decomposition layer N by the signal-to-noise ratio ensures that the decomposition depth matches the degree of masking of the effective signal in the signal. Low signal-to-noise ratio signals can be sufficiently separated by a larger N, while high signal-to-noise ratio signals can be decomposed on demand by controlling costs with a smaller N. Performing decomposition by using the iteration step size and the initial decomposition layer N ensures that the fitting accuracy and decomposition depth of the effective amplitude-modulated and frequency-modulated sub-modes are adapted to the signal characteristics, avoiding over-decomposition or under-decomposition and providing a high-quality signal foundation for subsequent signal reconstruction.

[0044] In one specific embodiment, S23 includes the following steps:

[0045] S231, initialize the core decomposition parameters based on the current i-th level target signal, wherein the core decomposition parameters include the decomposition window length, the mode convergence threshold, and the jump interference detection window.

[0046] S232, based on the initial decomposition level N and the core decomposition parameters, perform N-level cascaded decomposition on the current i-th level target signal to obtain one preliminary sub-mode signal output from each decomposition level and the residual component output from the Nth level decomposition level.

[0047] S233, compare the peak amplitude and peak duration of each preliminary submode signal and residual component with a preset ratio threshold and a preset time threshold, and extract the jump interference component from all preliminary submode signals and residual components.

[0048] S234, remove signal segments that overlap with the time of the jump interference component from the preliminary sub-mode signal and the residual component, and obtain several (i+1)th level sub-mode signals corresponding to the current i-th level target signal.

[0049] The decomposition window length is the time window used to extract the target signal and perform local amplitude-frequency modulation (AM / FM) mode fitting during the decomposition process. It is measured in units of sampling points and needs to be dynamically calculated based on the dominant frequency and sampling rate of the current i-th level target signal. This ensures the window covers at least two complete effective signal cycles, avoiding signal fragmentation fitting errors due to an excessively short window or a decrease in time resolution due to an excessively long window, thus guaranteeing local fitting accuracy. For example, with a dominant frequency of 50Hz and a sampling rate of 1000Hz, the window length is 40 sampling points, corresponding to 0.04 seconds, covering two 50Hz signal cycles.

[0050] Modal convergence threshold is a quantitative standard for determining whether the fitting of AM-FM submodes has reached a stable state. It is represented by the root mean square error of the submode signals generated in two adjacent iterations, and its value ranges from

[10] . -4 10 -3 ], preferably 10 -4 When the root mean square error is less than the convergence threshold of the mode, the sub-mode fitting is considered to have converged, and the iterative fitting of the current level is stopped.

[0051] The skipping interference detection window is a time window used for sliding detection of short-duration, high-amplitude skipping interference in a signal. It is measured in seconds or sampling points, and its length is usually 0.001-0.01 seconds, preferably 0.005 seconds. The sliding step size is one sampling point, which ensures that no sudden interference is missed and avoids misjudging the effective signal peak as interference due to an excessively long window, or causing the interference to be divided into multiple segments due to an excessively short window. This allows for accurate location of the interference in the time domain and improves interference detection efficiency.

[0052] The current i-th level target signal is a mixture of effective AM-FM signal, noise, and skipping interference. A single decomposition is insufficient to completely separate the deeply nested effective components. Through N-level cascaded decomposition, signal components are peeled off layer by layer in descending frequency (or descending energy). Correspondingly, the first layer extracts the highest frequency (or strongest energy) AM-FM sub-mode, the second layer extracts the second highest frequency (or second strongest energy) AM-FM sub-mode from the residual signal after the first layer decomposition, and so on, until the Nth layer, ultimately forming a decomposition result of N preliminary sub-modes + 1 residual component. In this process, the initial decomposition layer number N determines the peeling depth, and the core decomposition parameters ensure consistent accuracy at each layer. This hierarchical operation avoids the problem of effective signals being masked by noise in a single decomposition, gradually separating the effective AM-FM sub-modes from the mixed signal, providing more characteristic preliminary sub-modes for subsequent skipping interference extraction.

[0053] Those skilled in the art will recognize that the prior art of performing N-level cascaded decomposition of a target signal using the JMD method falls within the protection scope of this invention, and will not be elaborated further here.

[0054] The above-mentioned method, by initializing the decomposition window length, modal convergence threshold, and jump interference detection window, and serving N-level cascaded decomposition, improves the separation rate of deeply nested low-frequency effective signals, avoids the problem of weak effective signals being missed due to high-frequency noise masking in a single decomposition, and improves the fitting accuracy and efficiency of sub-signal modes. By comparing the dual thresholds of amplitude and duration, the false positive rate is reduced and the signal separation rate is improved, ensuring the output of high-purity effective sub-modes and providing high-quality data support for subsequent signal screening and reconstruction.

[0055] S3. Based on the average instantaneous energy of each (i+1)th level sub-mode signal within each preset time window, obtain the signal type corresponding to each (i+1)th level sub-mode signal, where the signal type is either a noise mode or a non-noise mode.

[0056] The preset time window is a fixed time segment set for analyzing the average instantaneous energy characteristics of the sub-mode signal. This avoids the averaging of energy characteristics caused by directly analyzing the entire sub-mode signal. By using time windowing, the signal is divided into multiple local segments, and the energy characteristics of each segment are easier to distinguish.

[0057] The average instantaneous energy is the local energy value of the i+1th level submode signal within a preset time window. It reflects the energy intensity of the submode signal within a local time segment and is the core quantitative indicator for distinguishing between noise and non-noise modes.

[0058] Noise mode is a sub-mode unit in the (i+1)th level sub-mode signal whose average instantaneous energy exhibits irregular random fluctuations and does not possess the characteristics of an effective AM-FM signal. It mainly includes surface wave interference, instrument background noise, and low-frequency interference from the ground surface.

[0059] The non-noise mode is a sub-mode unit in the i+1 level sub-mode signal whose average instantaneous energy exhibits a stable periodic change and has typical AM-FM effective signal characteristics. It mainly includes effective signal components such as seismic reflection waves, refracted waves, and first waves.

[0060] As described above, by setting a preset time window that matches the dominant frequency of the sub-mode, the average instantaneous energy analysis focuses on a local segment containing a complete effective signal period, avoiding the masking of noise features due to an excessively long time window or the incomplete analysis of effective signal energy due to an excessively short time window. By distinguishing between noise modes and non-noise modes by the average instantaneous energy features corresponding to the preset time window, the stable energy features of weak effective signals can be accurately identified, reducing the misclassification rate of signal types.

[0061] In one specific embodiment, S3 includes the following steps:

[0062] S31. For any (i+1)th level submode signal, perform a Hilbert transform on the current (i+1)th level submode signal to obtain the Hilbert spectrum corresponding to the current (i+1)th level submode signal.

[0063] S32, based on the Hilbert spectrum corresponding to the current (i+1)th level submode signal, obtain the average instantaneous energy of the current (i+1)th level submode signal within each preset time window, wherein the length of the preset time window is determined according to the dominant frequency of the current (i+1)th level submode signal.

[0064] S33. Based on all the average instantaneous energies corresponding to the current (i+1)th level submode signal, obtain the kurtosis coefficient, instantaneous frequency standard deviation, and energy entropy value corresponding to the current (i+1)th level submode signal.

[0065] S34, obtain the preset peak coefficient threshold, preset standard deviation threshold and preset entropy threshold corresponding to the current i+1 level submode signal.

[0066] S35, if the kurtosis coefficient of the current (i+1)th level submode signal is greater than or equal to the preset kurtosis coefficient threshold, the instantaneous frequency standard deviation is less than the preset standard deviation threshold, and the energy entropy value is less than the preset entropy value threshold, then the current (i+1)th level submode signal is determined to be a non-noise mode.

[0067] The Hilbert transform is a mathematical operation that converts a real signal into an analytic signal. It overcomes the limitations of traditional frequency domain analysis, accurately extracting the instantaneous time-frequency characteristics of non-stationary submode signals (such as the trend of instantaneous frequency change over time), providing time-frequency data support for subsequent multi-dimensional judgments. The Hilbert spectrum is a three-dimensional spectrum with time on the horizontal axis, instantaneous frequency on the vertical axis, and instantaneous amplitude as the magnitude. Its function is to visualize the time-frequency characteristics of submode signals, avoiding the limitations of single time-domain or frequency-domain analysis.

[0068] The duration of the preset time window can be dynamically determined based on the dominant frequency of the (i+1)th level submode signal, using the formula T. win =k / f sub , among which, T win f is the duration of the preset time window. sub To correspond to the dominant frequency of the sub-mode signal, k is an empirical coefficient, typically set to 2-5, to ensure that the time window contains 2-5 complete valid signal cycles. For example, for a sub-mode signal with a dominant frequency of 20Hz, k=2, and the preset time window length can be 0.1 seconds, containing two 20Hz cycles. When the sampling rate is 1000Hz, the time window length is 100 sampling points.

[0069] First, extract the instantaneous amplitude at each moment from the Hilbert spectrum. Then, calculate the instantaneous energy at each moment based on the relationship that instantaneous energy equals the square of instantaneous amplitude. The average instantaneous energy of the submode signal within each preset time window is the average of the instantaneous energies at all moments within the time window.

[0070] The kurtosis coefficient reflects the steepness of the instantaneous energy distribution, the instantaneous frequency standard deviation reflects the stability of the instantaneous frequency, and the energy entropy reflects the uniformity of the instantaneous energy distribution within the time window. By calculating the characteristics of these three dimensions, the submode signal characteristics can be comprehensively characterized from the energy distribution steepness, frequency stability, and energy distribution uniformity, avoiding the limitations of judging by a single instantaneous energy.

[0071] Those skilled in the art will recognize that the calculation methods for kurtosis coefficient, instantaneous frequency standard deviation, and energy entropy in the prior art fall within the protection scope of this invention, and will not be elaborated upon here.

[0072] Since meeting only the kurtosis coefficient standard might indicate low-energy noise (such as low-frequency background noise), meeting only the frequency standard might indicate frequency-stable interference (such as power frequency interference), and meeting only the entropy value standard might indicate concentrated burst noise, a non-noise mode must simultaneously possess the characteristics of low-energy distribution (kurtosis coefficient ≥ preset kurtosis coefficient threshold), stable frequency (instantaneous frequency standard deviation < preset standard deviation threshold), and concentrated energy (energy entropy value < preset entropy value threshold). All three are indispensable to ensure that the sub-mode signal possesses the complete characteristics of an effective AM-FM signal, ultimately outputting a non-noise mode. If any condition is not met, it is determined to be a noise mode and must be removed in subsequent processes.

[0073] As described above, by calculating the kurtosis coefficient, instantaneous frequency standard deviation, and energy entropy value, the determination dimension of the sub-mode signal is expanded from a single instantaneous energy to three-dimensional features. Through the determination logic that satisfies the three conditions simultaneously, the determination accuracy of non-noise modes is improved.

[0074] In one specific embodiment, S34 includes the following steps:

[0075] S341, collect a set of known effective earthquake signal samples and noise signal samples of the same type and region as the initial seismic signal.

[0076] S342, calculate the first kurtosis coefficient of each valid signal sample and the second kurtosis coefficient of each noise signal sample respectively, and obtain the minimum value of the first kurtosis coefficient and the maximum value of the second kurtosis coefficient.

[0077] S343, statistically analyze the kurtosis coefficients of all (i+1)th level submode signals to obtain the (i+1)th level kurtosis coefficient distribution.

[0078] S344: Based on the distribution of the kurtosis coefficients of the (i+1)th level, the minimum value of the first kurtosis coefficient, and the maximum value of the second kurtosis coefficient, obtain the preset kurtosis coefficient threshold corresponding to the current (i+1)th level submode signal.

[0079] The known effective earthquake signal sample set contains at least 50 geologically verified pure effective signal segments, and the noise signal sample set contains at least 50 labeled pure noise signal segments. The duration of each sample is consistent with the duration of the (i+1)th level submode signal, and the sampling rate is the same as the current signal.

[0080] The effective and noise components of seismic signals exhibit significant regional dependence. Within the same region and exploration scenario, the kurtosis coefficient ranges of both the effective and noise signals remain relatively stable, while cross-regional signal characteristics differ considerably. By acquiring a known sample set from the same region and of the same type as the initial seismic signal, we can ensure a high degree of consistency between the sample's feature distribution and the current processed signal's feature distribution. This provides a regionally adapted anchor point for subsequent threshold calculations, preventing threshold failure due to a disconnect between the sample and the current signal.

[0081] The preset standard deviation threshold and preset entropy threshold can also be obtained by referring to the method for obtaining the preset kurtosis coefficient threshold, which is obtained through statistical analysis of the standard deviation characteristics and entropy characteristics of known effective seismic signal sample sets and noise signal sample sets, and will not be elaborated here.

[0082] The above-mentioned progressive operation of sample set collection, feature statistics, current signal analysis, and threshold determination refines the generation process of the preset kurtosis coefficient threshold, solves the problem of poor adaptability of traditional fixed thresholds, and ensures that the threshold is based on historical experience and fits the current sub-mode signal characteristics, thereby improving the accuracy of non-noise mode determination.

[0083] In one specific embodiment, S344 includes the following steps:

[0084] S3441, the distribution range is determined based on the minimum value of the first kurtosis coefficient.

[0085] S3442, obtain the percentage of overlap between the peak coefficient distribution and the distribution range of the (i+1)th level submode signal.

[0086] S3443, if the overlap range ratio is greater than the preset ratio threshold, then the product of the average of the minimum value of the first kurtosis coefficient and the maximum value of the second kurtosis coefficient and the first preset coefficient is determined as the preset kurtosis coefficient threshold.

[0087] S3444, if the overlap ratio is less than or equal to the preset ratio threshold, then the product of the minimum value of the first kurtosis coefficient and the maximum value of the second kurtosis coefficient with the second preset coefficient is determined as the preset kurtosis coefficient threshold.

[0088] In one specific implementation, the second preset coefficient is greater than the first preset coefficient.

[0089] Specifically, the distribution range of the kurtosis coefficients of the effective signal must cover more than 90% of the known effective signal samples and have no overlap with the noise signal range, combined with the minimum value C of the first kurtosis coefficient. min The distribution range is set to [C] min C min +2×σ], where σ is the standard deviation of the kurtosis coefficient of the effective signal sample set, and 2×σ is an empirical value based on the 95% confidence interval of the normal distribution, ensuring that this range can cover most effective signals.

[0090] The overlap ratio between the kurtosis coefficient distribution of the (i+1)th level submode signal and the effective signal distribution range is essentially the proportion of submodes in the current submode whose kurtosis coefficients conform to the characteristics of an effective signal. A high overlap ratio indicates that the coefficients of most submodes in the current submode are close to the characteristics of an effective signal. Therefore, the threshold is lowered by multiplying the average value by a first preset coefficient (first preset coefficient < 1, e.g., 0.8) to ensure that weak effective signals can be correctly identified as non-noise modes, while also ensuring that strong noise with coefficients far exceeding the threshold can still be eliminated. A low overlap ratio indicates that the coefficients of most submodes in the current submode deviate from the characteristics of an effective signal. Therefore, the threshold needs to be increased by multiplying the average value by a second preset coefficient (second preset coefficient > 1, e.g., 1.2) to strengthen the noise removal capability, ensuring that only submodes with coefficients close to the characteristics of an effective signal are retained, reducing the accumulation of noise in subsequent iterations.

[0091] The specific value of the preset percentage threshold can be set by the implementer according to the actual situation, such as 60%.

[0092] The above calculation of the overlap range ratio converts the correlation between the current sub-mode and the effective signal into a quantitative value, providing a clear quantitative basis for subsequent threshold adjustment and avoiding adjustment deviations caused by subjective judgment. By differentiating the second preset coefficient from the first preset coefficient, it achieves accurate adaptation between the two scenarios of effective mode and noise mode, solving the problem that traditional fixed thresholds or unidirectional threshold adjustments cannot adapt to the two scenarios, thereby improving the accuracy of non-noise mode determination.

[0093] S4, each (i+1)th level submode signal of the signal type non-noise mode is determined as the (i+1)th level target signal, and the cross-correlation degree between each (i+1)th level target signal and the (i)th level target signal that generated the (i+1)th level target signal is obtained.

[0094] In this process, the non-noise mode is identified as the target signal of the (i+1)th level of the next decomposition, ensuring that the iteration process can continue. That is, only the high-purity non-noise mode has the value of further decomposition, from which the weak effective signals of deeper nesting can be separated, while the noise mode is eliminated to avoid the accumulation of noise in the iteration.

[0095] The (i+1)th level target signal is generated by decomposing the i-th level target signal. Both should retain the core characteristics of the effective signal, such as the main frequency and phase. However, excessive decomposition may cause the effective signal characteristics to break, making the (i+1)th level target signal and the i-th level target signal too different and losing the effective signal attributes.

[0096] Cross-correlation is a quantitative indicator of the similarity between two signals, with a value range of [-1, 1]. By calculating the cross-correlation, we can quantitatively assess whether the (i+1)th level target signal inherits the effective features of the ith level target signal: a high cross-correlation indicates that the effective features of the two signals are consistent and the decomposition is reasonable; a low cross-correlation indicates that the decomposition may introduce false signals or lose effective features, and subsequent iteration termination logic needs to determine whether to stop the decomposition.

[0097] S5. If the cross-correlation degree of all target signals of level i+1 is greater than the preset degree threshold, then all sub-mode signals of non-noise mode are obtained and step S6 is executed. Otherwise, update i=i+1 and repeat step S2.

[0098] Specifically, the current iteration level number i is updated to i+1, so that the processing object of the next iteration changes from the target signal of level i to the target signal of level i+1, ensuring that the iteration process progresses in the order of 1→2→3→…→n.

[0099] The termination criterion is that the cross-correlation degree of all (i+1)th level target signals is greater than a preset threshold. This determines whether the effective signal features of the current non-noise mode have stabilized. If the cross-correlation degree of any (i+1)th level target signal is less than or equal to the preset threshold, it indicates that the signal still contains strippable noise and needs to continue iterating. If the cross-correlation degree of all (i+1)th level target signals is greater than the preset threshold, it indicates that the effective signal features of all non-noise modes have stabilized and there is no more strippable noise. Continuing to iterate is meaningless. At this point, the iteration is terminated, and all non-noise mode sub-mode signals are summarized to provide high-purity material for subsequent reconstruction.

[0100] The specific value of the preset threshold can be set by the implementer according to the actual situation. For example, it can be set based on a large amount of experimental data, with a value range of 0.7-0.9, preferably 0.8, so as to ensure that the purity of the effective signal meets the standard while avoiding excessive iteration.

[0101] All submode signals of the non-noise mode are the set of submode signals that are determined to be non-noise modes from all iterations from the first iteration to the termination iteration.

[0102] As described above, hierarchical progressive purification is achieved through iterative updates of i=i+1, which gradually improves the purity of the effective signal of the non-noise mode with each iteration. By using the termination condition that the cross-correlation degree of all target signals at level i+1 is greater than a preset threshold, over-iteration or under-iteration caused by a fixed number of iterations is avoided, thereby improving the decomposition rate and accuracy of the effective signal.

[0103] S6 superimposes all submode signals of the non-noise mode to obtain the filtered effective seismic signal.

[0104] Among them, all sub-mode signals of the non-noise mode are high-purity effective signals, and each corresponds to an effective component of different frequencies and energies in the original seismic signal. These sub-mode signals are essentially characteristic fragments of the original effective signal after decomposition. Their phase and frequency characteristics are consistent with the original effective signal. Through linear superposition, they can be re-aggregated into a complete effective signal waveform, ultimately outputting a high-fidelity, low-noise filtered effective seismic signal, providing core data support for subsequent geological interpretation tasks such as reservoir prediction and fault identification.

[0105] In one specific embodiment, S6 includes the following steps:

[0106] S61, for each non-noise mode submode signal, the effective signal confidence level corresponding to the current submode signal is obtained based on the kurtosis coefficient, instantaneous frequency standard deviation and energy entropy value corresponding to the current submode signal. The effective signal confidence level ranges from [0, 1]. The absolute value of the difference between the kurtosis coefficient and the mean of the first kurtosis coefficient, the instantaneous frequency standard deviation and the energy entropy value are all negatively correlated with the effective signal confidence level.

[0107] S62, based on the effective signal confidence levels corresponding to the sub-mode signals of all non-noise modes, determine the superposition weight of each non-noise mode sub-mode signal, wherein the superposition weight is positively correlated with the effective signal confidence level, and the sum of the superposition weights corresponding to the sub-mode signals of all non-noise modes is 1.

[0108] S63, based on the superposition weights of the sub-mode signals of all non-noise modes, superimpose the sub-mode signals of all non-noise modes to obtain the filtered effective seismic signal.

[0109] Although all non-noise modes satisfy the conditions that the kurtosis coefficient is greater than or equal to the preset kurtosis coefficient threshold, the instantaneous frequency standard deviation is less than the preset standard deviation threshold, and the energy entropy value is less than the preset entropy value threshold, the effective signal purity of different sub-modes still varies. For example, some sub-modes are close to the threshold boundary and contain a small amount of noise, while some sub-modes are far from the threshold and have extremely high purity.

[0110] By using the kurtosis coefficient, instantaneous frequency standard deviation, and energy entropy value, a negative correlation mapping between feature deviation and the confidence level of the effective signal can be established. Correspondingly, the smaller the absolute value of the difference between the kurtosis coefficient and the mean kurtosis coefficient of the effective signal sample set, the closer the kurtosis characteristics of the submode are to the effective signal, and the higher the effective purity. The smaller the instantaneous frequency standard deviation, the more stable the submode frequency is, and the higher the effective purity. The smaller the energy entropy value, the more concentrated the energy distribution of the submode is, and the higher the effective purity.

[0111] The kurtosis deviation, instantaneous frequency standard deviation, and energy entropy deviation are obtained by normalizing the absolute value of the difference between the kurtosis coefficient and the mean of the first kurtosis coefficient, the instantaneous frequency standard deviation, and the energy entropy deviation, respectively. Then, a weighted average inverse operation is used to convert the three deviations into the confidence level of the effective signal. The weights corresponding to the absolute value of the difference between the kurtosis coefficient and the mean of the first kurtosis coefficient, the instantaneous frequency standard deviation, and the energy entropy value can be set to 0.4, 0.3, and 0.3, respectively.

[0112] Furthermore, when superimposing the submode signals of all non-noise modes, since each submode signal comes from decomposition at different iteration levels, there may be a slight time offset. Using the time axis of the initial seismic signal as a reference, the time starting point of all submode signals is adjusted, and each submode signal is aligned with the initial seismic signal using the maximum cross-correlation method. For example, the cross-correlation coefficient between the submode signal and the initial signal is calculated, the time offset corresponding to the maximum value of the coefficient is found, and the submode signal is shifted to be consistent with the time starting point, ensuring that the effective signal components at the same time point correspond accurately during superposition.

[0113] Since the amplitude ranges of different sub-mode signals vary greatly, such as the shallow sub-mode amplitude of 0.5V-2V and the deep sub-mode amplitude of 0.1V-0.5V, direct superposition will cause the weak effective signal in the deep layer to be masked by the strong signal in the shallow layer. Therefore, amplitude normalization processing can be used to uniformly map the amplitude range of all sub-mode signals to [0, 2V] to ensure that sub-mode signals with different amplitudes can contribute effective features when superimposed.

[0114] The above-mentioned progressive operations of effective signal confidence quantification, dynamic weight allocation, and weighted superposition reconstruction, and the accurate weighted superposition driven by confidence, solve the pain point of low-purity sub-modes and high-purity sub-modes contributing equally in traditional average superposition, thereby improving the fidelity and signal-to-noise ratio of the filtered effective signal.

[0115] As described above, by performing modal decomposition on the i-th level target signal, the complex mixed seismic signal is decomposed into sub-mode signals with single characteristics, overcoming the problem of difficult separation of traditional mixed signals. Simultaneously, the core amplitude-modulation and frequency-modulation characteristics of the effective signal are preserved, preventing the effective signal from being masked by noise due to lack of decomposition. A clear sub-mode screening criterion is established by using the average instantaneous energy of the sub-mode signals within a preset time window. Local energy analysis is achieved using the preset time window, avoiding misjudgment of weak effective signals caused by overall energy averaging, thus accurately distinguishing between non-noise and noise modes. The similarity between upper and lower level target signals is quantified by calculating the cross-correlation degree, providing data basis for subsequent judgments on the rationality of the decomposition and whether effective features are preserved, preventing... Excessive decomposition leads to the fragmentation of effective signal features. By iteratively updating and using iteration stopping conditions based on cross-correlation, dynamic control of the iteration process is achieved, avoiding over-iteration or under-iteration caused by a fixed number of iterations. This ensures that the final acquired non-noise modal sub-signals all possess stable and effective features, with no unpurified signal residue, thus guaranteeing the quality of subsequent reconstruction. By superimposing all non-noise modal sub-signals, filtered effective seismic signals are obtained. Effective signal components from different iteration levels are aggregated, achieving a closed loop from sub-mode decomposition to complete effective signal restoration. The final output effective seismic signal not only fully retains the effective features of each level but also significantly reduces noise interference, improving the purity of the effective seismic signal.

[0116] Example 2

[0117] Embodiment 2 of the present invention provides a non-transient computer-readable storage medium, which can be disposed in an electronic device to store at least one instruction or at least one program related to implementing a method in the method embodiment. The at least one instruction or at least one program is loaded and executed by the processor to implement the seismic signal filtering method provided in the above embodiment.

[0118] Example 3

[0119] Embodiment 3 of the present invention provides an electronic device, which includes a processor and the non-transitory computer-readable storage medium of Embodiment 2 of the present invention.

[0120] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method of filtering seismic signals, characterized by, The seismic signal filtering method comprises the following steps: S1, initializing i=1, and determining an initial seismic signal as an ith level target signal; S2, for any ith level target signal, performing modal decomposition processing on the current ith level target signal to obtain a plurality of (i+1)th level sub-modal signals corresponding to the current ith level target signal, wherein the sub-modal signal is a signal component with amplitude modulation and frequency modulation characteristics; S3, according to the average instantaneous energy of each (i+1)th level sub-modal signal in each preset time window, obtaining the signal type corresponding to each (i+1)th level sub-modal signal, wherein the signal type is a noise mode or a non-noise mode, and S3 comprises the following steps: S31, for any (i+1)th level sub-modal signal, performing Hilbert transform on the current (i+1)th level sub-modal signal to obtain the Hilbert spectrum corresponding to the current (i+1)th level sub-modal signal; S32, according to the Hilbert spectrum corresponding to the current (i+1)th level sub-modal signal, obtaining the average instantaneous energy of the current (i+1)th level sub-modal signal in each preset time window, wherein the length of the preset time window is determined according to the main frequency of the current (i+1)th level sub-modal signal; S33, according to all the average instantaneous energies corresponding to the current (i+1)th level sub-modal signal, obtaining the peakiness coefficient, the instantaneous frequency standard deviation and the energy entropy value corresponding to the current (i+1)th level sub-modal signal; S34, obtaining the preset peakiness coefficient threshold, the preset standard deviation threshold and the preset entropy value threshold corresponding to the current (i+1)th level sub-modal signal, wherein S34 comprises the following steps: S341, collecting a known seismic effective signal sample set and a noise signal sample set of the same region and type as the initial seismic signal; S342, calculating a first peakiness coefficient of each effective signal sample and a second peakiness coefficient of each noise signal sample respectively, and obtaining the minimum value of the first peakiness coefficient and the maximum value of the second peakiness coefficient; S343, statistically processing the peakiness coefficients of all (i+1)th level sub-modal signals to obtain an (i+1)th level peakiness coefficient distribution; S344, according to the (i+1)th level peakiness coefficient distribution, the minimum value of the first peakiness coefficient and the maximum value of the second peakiness coefficient, obtaining the preset peakiness coefficient threshold corresponding to the current (i+1)th level sub-modal signal; S35, if the peakiness coefficient of the current (i+1)th level sub-modal signal is greater than or equal to the preset peakiness coefficient threshold, the instantaneous frequency standard deviation is less than the preset standard deviation threshold, and the energy entropy value is less than the preset entropy value threshold, it is determined that the current (i+1)th level sub-modal signal is a non-noise mode; S4, determining each (i+1)th level sub-modal signal with the signal type of the non-noise mode as an (i+1)th level target signal, and obtaining the cross-correlation degree between each (i+1)th level target signal and the ith level target signal generating the (i+1)th level target signal; S5, if the cross-correlation degrees corresponding to all (i+1)th level target signals are greater than a preset degree threshold, obtaining all sub-modal signals with the signal type of the non-noise mode, and performing step S6, otherwise, updating i=i+1, and repeating step S2. S6, superimpose all sub-modal signals of the signal type of non-noise modal to obtain filtered seismic effective signals.

2. The seismic signal filtering method of claim 1, wherein, The modal decomposition processing is a jump-enhanced amplitude-frequency modulation modal decomposition processing, and S2 includes the following steps. S21, extract key prior features of the current ith level target signal, wherein the key prior features include a main frequency range and a signal-to-noise ratio; S22, determine an iteration step and an initial decomposition layer number N in the jump-enhanced amplitude-frequency modulation modal decomposition processing according to the key prior features; S23, perform the jump-enhanced amplitude-frequency modulation modal decomposition processing according to the iteration step and the initial decomposition layer number N to obtain a plurality of (i+1)th level sub-modal signals and a jump interference component, wherein the jump interference component is determined according to that a peak amplitude to average amplitude ratio of the current ith level target signal is greater than a preset ratio threshold value and a peak duration is less than a preset time threshold value.

3. The seismic signal filtering method of claim 2, wherein, S23 includes the following steps. S231, initialize core decomposition parameters according to the current ith level target signal, wherein the core decomposition parameters include a decomposition window length, a modal convergence threshold value and a jump interference detection window; S232, perform N-layer cascaded decomposition on the current ith level target signal according to the initial decomposition layer number N and the core decomposition parameters to obtain one preliminary sub-modal signal of each layer decomposition output and a residual component of the Nth layer decomposition output; S233, compare a peak amplitude and a peak duration of each preliminary sub-modal signal and the residual component with the preset ratio threshold value and the preset time threshold value to extract the jump interference component from all the preliminary sub-modal signals and the residual component; S234, remove signal segments that are time overlapped with the jump interference component from the preliminary sub-modal signals and the residual component to obtain a plurality of (i+1)th level sub-modal signals corresponding to the current ith level target signal.

4. The seismic signal filtering method of claim 1, wherein, S344 includes the following steps. S3441, determine a distribution range according to a minimum value of the first peakiness coefficient; S3442, obtain an overlap range proportion of a peakiness coefficient distribution of the (i+1)th level sub-modal signal and the distribution range; S3443, if the overlap range proportion is greater than a preset proportion threshold value, determine a product of an average value of the minimum value of the first peakiness coefficient and the maximum value of the second peakiness coefficient and the first preset coefficient as the preset peakiness coefficient threshold value; S3444, if the overlap range proportion is less than or equal to the preset proportion threshold value, determine a product of the average value of the minimum value of the first peakiness coefficient and the maximum value of the second peakiness coefficient and the second preset coefficient as the preset peakiness coefficient threshold value.

5. The seismic signal filtering method of claim 4, wherein, The second preset coefficient is greater than the first preset coefficient.

6. The method of filtering seismic signals according to claim 1, wherein, S6 includes the following steps. S61, for each sub-modal signal of the non-noise modal, obtain an effective signal confidence corresponding to the current sub-modal signal according to a peakiness coefficient, an instantaneous frequency standard deviation and an energy entropy value corresponding to the current sub-modal signal, wherein the effective signal confidence has a value range of [0, 1], an absolute value of a difference between the peakiness coefficient and an average value of the first peakiness coefficient, the instantaneous frequency standard deviation and the energy entropy value are all negatively correlated with the effective signal confidence; S62, determining a superposition weight of each submodal signal of the non-noise modal according to the effective signal confidence corresponding to the submodal signal of the non-noise modal, wherein the superposition weight is in a positive correlation with the effective signal confidence, and a sum of the superposition weights corresponding to all the submodal signals of the non-noise modal is 1; S63, superimposing all the submodal signals of the non-noise modal according to the superposition weights of the submodal signals of the non-noise modal to obtain a filtered seismic effective signal. 7.A non-transitory computer-readable storage medium having stored therein at least one instruction or at least one piece of program, characterized in that, The at least one instruction or the at least one program is loaded and executed by the processor to implement the seismic signal filtering method as claimed in any one of claims 1-6.

8. An electronic device, comprising: The non-transitory computer readable storage medium as claimed in claim 7 is included. The non-transitory computer readable storage medium as claimed in claim 7 is included.

Images