Repair method of pulse synchronization signal

Abstract

The present invention belongs to the field of data processing technology, and in particular relates to a method for repairing a pulse synchronization signal. The method comprises the following steps: S1: obtaining a pulse synchronization signal sequence, preprocessing the pulse synchronization signal sequence, and obtaining a synchronization point time curve of the pulse synchronization signal sequence; S2: performing a first-order differential on the synchronization point time curve to obtain a first-order differential curve; S3: obtaining a trained repair network based on a BP neural network, and using the repair network to repair the first-order differential curve to obtain a repaired first-order differential curve; S4: performing reverse repair on the repaired first-order differential curve to obtain a repaired synchronization point time curve; S5: performing reverse repair on the repaired synchronization point time curve to obtain a repaired pulse synchronization signal sequence. The present invention proposes repairing second-level pulse synchronization signals based on a BP neural network, which has positive significance for the utilization of electromagnetic detection data.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to a method for repairing a pulse synchronization signal. Background Art

[0002] The transient electromagnetic method is a detection method through electromagnetic induction. Under the excitation of the primary electromagnetic field, an induced current is generated inside the underground conductor. During the gap of the primary pulse electromagnetic field, the secondary field generated by the eddy current will not disappear immediately with the disappearance of the primary field, that is, there is a transient process. The secondary field is observed using a coil or a grounding electrode, and its relationship with time is studied to determine the electrical distribution structure and spatial form of the underground conductor. According to the application method, it is divided into ground transient electromagnetic, semi-aeronautical transient electromagnetic, aeronautical transient electromagnetic, etc. Wireless synchronization is the preferred synchronization method for semi-airborne electromagnetic detection systems. It is not constrained by synchronization cables. The transmitter of the semi-airborne electromagnetic detection system is located on the ground, the receiver is located on the semi-airborne platform, and the receiver is in real-time mobile state. The second pulse signal is often used as the synchronization signal for wireless synchronization between the transmitter and receiver of the semi-airborne electromagnetic detection system. The transmitter and receiver of the semi-airborne electromagnetic system use the agreed synchronization signal edge as the trigger edge to perform excitation and reception operations respectively. The receiver mostly adopts a continuous acquisition method. After receiving the acquisition instruction, the receiver continuously acquires the synchronization signal and the signal of the receiving coil. Later, the synchronization signal trigger edge is used to intercept and superimpose the received data and perform data inversion. Therefore, the synchronization signal is particularly important.

[0003] However, in an environment with a weak pulse-per-second signal or affected by weather and geographical location, the synchronization signal may be lost, resulting in discontinuity in the superimposed data and failure to effectively reduce random noise. In addition, the loss of the synchronization signal will also cause data waste and reduce data utilization. Summary of the Invention

[0004] In view of this, the present invention aims to provide a method for repairing a pulse synchronization signal to solve the problems of data waste and reduced data utilization caused by the missing second-level pulse synchronization signal, and also to solve the problem that the repair results of traditional repair methods deviate from the actual problem on a differential layer. The present invention proposes to repair the second-level pulse synchronization signal based on the BP neural network, which has positive significance for the utilization of electromagnetic detection data. The present invention can more accurately restore the missing second-level pulse synchronization signal.

[0005] To achieve the above object, the technical solution created by the present invention is implemented as follows:

[0006] A method for repairing a pulse synchronization signal specifically comprises the following steps:

[0007] S1: Acquire a pulse synchronization signal sequence, pre-process the pulse synchronization signal sequence, and obtain a synchronization point time curve of the pulse synchronization signal sequence;

[0008] S2: Perform first-level difference on the synchronization point time curve to obtain a first-level difference curve;

[0009] S3: Obtain a trained repair network based on a BP neural network, and use the repair network to repair the first-order difference curve to obtain a repaired first-order difference curve;

[0010] S4: Perform reverse repair on the repaired first-order differential curve to obtain a repaired synchronization point time curve;

[0011] S5: Perform reverse repair on the repaired synchronization point time curve to obtain a repaired pulse synchronization signal sequence.

[0012] Furthermore, each pulse synchronization signal included in the pulse synchronization signal sequence is a periodic positive pulse signal.

[0013] Furthermore, step S1 specifically includes:

[0014] S11: Calculating the median amplitude of each pulse synchronization signal included in the pulse synchronization signal sequence, where the median amplitude is the average of the maximum synchronization edge amplitude and the minimum synchronization edge amplitude of the pulse synchronization signal;

[0015] S12: If the sampling point data of the i-th pulse synchronization signal includes the median amplitude, the sampling time corresponding to the median amplitude is used as the synchronization point time; otherwise, two sampling point data with the smallest amplitude difference from the median amplitude are searched on the synchronization edge of the current pulse synchronization signal, and the two sampling point data are interpolated to obtain the sampling time corresponding to the median amplitude, and the sampling time corresponding to the median amplitude is used as the synchronization point time;

[0016] S13: Replace the i-th pulse synchronization signal with the i+1-th pulse synchronization signal, repeat step S12, obtain the synchronization point time of all pulse synchronization signals, and plot all synchronization point times in the order of sampling time to obtain the synchronization point time curve of the pulse synchronization signal sequence.

[0017] Furthermore, in step S2, the time difference between two adjacent synchronization points in the sampling time sequence is taken, and all the differences are plotted in the sampling time sequence to obtain a first-order difference curve;

[0018] The difference between two adjacent synchronization point times is the value obtained by subtracting the synchronization point time before the sampling time from the synchronization point time after the sampling time.

[0019] Furthermore, in step S3, the trained BP neural network-based restoration network includes a peak restoration network, a trough restoration network, and a transition recovery network.

[0020] Furthermore, the steps of training the peak recovery network and the trough recovery network include:

[0021] Extract the peak data and corresponding peak index array of the first-level differential curve to be processed, use 0 to fill the missing peaks, remove the peak placeholder data and corresponding peak index of each peak, and use the remaining peak data and corresponding peak index array to create a peak training set; use the peak training set to train the BP neural network to obtain the peak recovery network;

[0022] The trough data and the corresponding trough index array of the first-level differential curve to be processed are extracted, and the missing troughs are filled with 0. The trough placeholder data and the corresponding trough index of each trough are eliminated, and the remaining trough data and the corresponding trough index array are used to prepare a trough training set; the trough training set is used to train the BP neural network to obtain a trough recovery network.

[0023] Furthermore, the transition segment recovery network includes n groups of transition segment recovery networks, where n is the maximum total number of data points included in each transition segment in the valid data segment minus one;

[0024] The steps of training n transition recovery networks include:

[0025] SA1: Extract the transition segments of the valid data segments from the first-level differential curve to construct a transition segment array, and retain the index of each transition segment of the valid data segment based on the sampling time sequence;

[0026] The transition section is the set of sampling points between adjacent peaks and troughs;

[0027] SA2: Subtract the adjacent data points in each transition segment by the rule of subtracting the data point with the sampling time before from the data point with the sampling time after, and obtain the transition segment difference data corresponding to each transition segment;

[0028] SA3: extracting the first difference data point in each transition segment difference data, and the index value of the first difference data point in each transition segment difference data is the same as the index value of the corresponding transition segment;

[0029] SA4: Create the first set of transition data training sets using the first set of difference data points extracted from the difference data of each transition segment and the corresponding indexes;

[0030] SA5: Use the first set of transition data training sets to train the BP neural network to obtain the first set of transition segment recovery networks;

[0031] SA6: Replace the first difference data point with the next difference data point and repeat steps SA3-SA5 until n sets of transition segment recovery networks are obtained.

[0032] Furthermore, the process of repairing the first-order difference curve using the repair network is as follows:

[0033] The peak index corresponding to each peak occupancy data is input into the peak recovery network to realize peak data repair;

[0034] The trough index corresponding to each trough placeholder data is input into the trough recovery network to realize trough data repair;

[0035] The index value of the transition segment with missing data points is input into the transition segment recovery network, and the transition segment corresponding to the difference data of the current transition segment is reversed based on the recovery result. The current transition segment is replaced by the transition segment of the next missing data point and the above operation is repeated until all transition segments are repaired;

[0036] Repair the first-level differential curve.

[0037] Furthermore, step S4 specifically includes:

[0038] S41: adding the first data value of the restored segment of the repaired first-level differential curve to the synchronization point time immediately preceding the first missing synchronization point time of the missing segment of the synchronization point time curve to obtain the first synchronization point time restored value of the missing segment of the synchronization point time curve;

[0039] S42: adding the second data value of the restored segment of the repaired first-level differential curve to the first synchronization point time recovery value of the missing segment of the synchronization point time curve to obtain the second synchronization point time recovery value of the missing segment of the synchronization point time curve;

[0040] S43: Repeat step S42 using the next data value of the restored segment of the repaired first-level differential curve and the current synchronization point time restoration value of the missing segment of the synchronization point time curve until the repaired synchronization point time curve is obtained.

[0041] Furthermore, in step S5, the pulse synchronization signal is inserted into the pulse synchronization signal sequence in the order of sampling time, and the synchronization edge time of the inserted pulse synchronization signal is the same as the corresponding synchronization point time recovery value, so as to obtain a repaired pulse synchronization signal sequence.

[0042] Compared with the prior art, the present invention can achieve the following beneficial effects:

[0043] The present invention creates the aforementioned method for repairing a pulse synchronization signal, extracting a second-level pulse synchronization signal synchronization point time curve from all second-level pulse synchronization signals, and further performing a first-level differential on the second-level pulse synchronization signal synchronization point time curve to obtain a first-level differential curve of the second-level pulse synchronization signal synchronization point time, and then recovering data from the first-level differential curve level, thereby obtaining recovery data that is closer to reality and solving the problem of missing second-level pulse synchronization signals. The present invention also solves the problem that the repair results of traditional repair methods deviate from reality at the first-level differential level: the present invention extracts peak data and trough data from the first-level differential curve to construct a training set of a peak recovery network and a trough recovery network based on a BP neural network, and extracts the transition segment between the peak and the trough to construct a training set of a transition segment recovery network based on a BP neural network. By adopting a method based on a BP neural network rather than traditional interpolation methods, data with nonlinear relationships can be repaired more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings, which constitute part of the present invention, are intended to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are intended to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings:

[0045] Figure 1 1 is a flow chart of a method for repairing a pulse synchronization signal according to an embodiment of the present invention;

[0046] Figure 2a The waveform of the original missing second-level pulse synchronization signal provided by an embodiment of the present invention;

[0047] Figure 2b A synchronization point time curve of an original missing second-level pulse synchronization signal provided by an embodiment of the present invention;

[0048] Figure 2c is a first-level differential curve of an original missing second-level pulse synchronization signal provided by an embodiment of the present invention;

[0049] Figure 3a The waveform of a second-level pulse synchronization signal based on interpolation repair provided by an embodiment of the present invention;

[0050] Figure 3b This is a synchronization point time curve of a second-level pulse synchronization signal based on interpolation repair provided by an embodiment of the present invention;

[0051] Figure 3c It is a first-level differential curve of a second-level pulse synchronization signal based on interpolation repair provided by an embodiment of the present invention;

[0052] Figure 4aThe waveform of a second-level pulse synchronization signal based on BP neural network repair provided by an embodiment of the present invention;

[0053] Figure 4b The synchronization point time curve of the second-level pulse synchronization signal based on BP neural network repair provided by an embodiment of the present invention;

[0054] Figure 4c It is a first-order differential curve of a second-level pulse synchronization signal repaired based on a BP neural network provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and do not constitute a limitation of the present invention.

[0056] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other.

[0057] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second" and the like are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, features defined as "first", "second" and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0058] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "installed," "connected," and "connected" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to internal connections between two components. Those skilled in the art can understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0059] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0060] like Figure 1 As shown, the present invention proposes a method for repairing a pulse synchronization signal, which specifically includes the following steps: S1: obtaining a pulse synchronization signal sequence, preprocessing the pulse synchronization signal sequence, and obtaining a synchronization point time curve of the pulse synchronization signal sequence; S2: performing a first-level difference on the synchronization point time curve to obtain a first-level difference curve; S3: obtaining a trained repair network based on a BP neural network, and using the repair network to repair the first-level difference curve to obtain a repaired first-level difference curve; S4: performing reverse repair on the repaired first-level difference curve to obtain a repaired synchronization point time curve; S5: performing reverse repair on the repaired synchronization point time curve to obtain a repaired pulse synchronization signal sequence.

[0061] It should be noted that the present invention can solve the problem of missing second-level (not limited to second-level) pulse synchronization signals, and can also solve the problem that the repair results of traditional repair methods deviate from the actual problem on the first differential plane. Taking the second-level pulse synchronization signal as an example, the processing process of the present invention is as follows: pre-processing each second-level pulse synchronization signal, extracting the synchronization point time of all second-level pulse synchronization signals, and obtaining the synchronization point time curve of each second-level pulse synchronization signal; performing a first-level differential on the synchronization point time curve of each second-level pulse synchronization signal to obtain the first-level differential curve of each second-level pulse synchronization signal; constructing a recovery network based on a BP neural network to repair the first-level differential curve of each second-level pulse synchronization signal; using the repaired first-level differential curve to reversely repair and obtain a repaired synchronization point time curve; using the repaired synchronization point time curve to reversely repair and obtain a repaired second-level pulse synchronization signal.

[0062] In some embodiments, each pulse synchronization signal included in the pulse synchronization signal sequence is a periodic positive pulse signal.

[0063] It should be noted that the pulse synchronization signal sequence is a digital signal collected by a data acquisition board, the horizontal axis is time, and the vertical axis is amplitude.

[0064] Furthermore, in the present invention, the storage period of the synchronization signal acquisition file is set to be consistent with the period of the pulse synchronization signal (since there are errors in the pulse synchronization signal period and the storage period of the synchronization signal acquisition file, the two periods must not be completely consistent in fact, and there will be slight deviations). Each sampling point of each pulse synchronization signal corresponds to a synchronization signal acquisition file, and whether the sampling point data is missing is determined based on the content of the synchronization signal acquisition file.

[0065] In some embodiments, step S1 specifically includes:

[0066] S11: Calculating the median amplitude of each pulse synchronization signal included in the pulse synchronization signal sequence, where the median amplitude is the average of the maximum synchronization edge amplitude and the minimum synchronization edge amplitude of the pulse synchronization signal;

[0067] S12: If the sampling point data of the i-th pulse synchronization signal includes the median amplitude, the sampling time corresponding to the median amplitude is used as the synchronization point time; otherwise, two sampling point data with the smallest amplitude difference from the median amplitude are searched on the synchronization edge of the current pulse synchronization signal, and the two sampling point data are interpolated to obtain the sampling time corresponding to the median amplitude, and the sampling time corresponding to the median amplitude is used as the synchronization point time;

[0068] S13: Replace the i-th pulse synchronization signal with the i+1-th pulse synchronization signal, repeat step S12, obtain the synchronization point time of all pulse synchronization signals, and plot all synchronization point times in the order of sampling time to obtain the synchronization point time curve of the pulse synchronization signal sequence.

[0069] It should be noted that the strategy for extracting the synchronization point time of each second-level pulse synchronization signal is consistent. The maximum and minimum values ​​of the synchronization edge amplitude of the second-level pulse synchronization signal are searched, and the average of the maximum and minimum values ​​is taken as the median amplitude. If the sampling point data happens to contain the median amplitude, the time of this data is used as the synchronization point time. If the sampling data does not contain the median data amplitude data, the signal points near the median amplitude of the synchronization edge are searched, and the time corresponding to the median amplitude is obtained by interpolation as the synchronization point time of the second-level pulse synchronization signal. Interpolation is used to estimate unknown data points between known data points. The core idea is to construct a function or curve through known discrete data points so that the value of the function at these points is consistent with the known value, thereby predicting or estimating the values ​​at other positions. The synchronization edge is the change process of the signal from low to high (or high to low) (for example, from 0 to 100), and the median amplitude (such as 50) is the "middle point" of this change process. To find the "two sampling points with the smallest difference from the median amplitude", the two points must meet the following conditions: the amplitude of one point is less than the median amplitude (for example, 49), and the amplitude of the other point is greater than the median amplitude (for example, 51). This ensures that the median amplitude (50) falls exactly between the two points, and the subsequent interpolation calculation is meaningful (estimating the time corresponding to the intermediate value through the linear relationship between the two points). Therefore, when searching for the two sampling point data with the smallest difference from the median amplitude on the synchronization edge of the pulse synchronization signal, the amplitude of one sampling point data must be less than the median amplitude (for example, 49), and the amplitude of the other sampling point data must be greater than the median amplitude (for example, 51).

[0070] In some embodiments, in step S2, the time difference between two adjacent synchronization points in the sampling time sequence is obtained, and all the differences are plotted in the sampling time sequence to obtain a first-order difference curve;

[0071] The difference between two adjacent synchronization point times is the value obtained by subtracting the synchronization point time before the sampling time from the synchronization point time after the sampling time.

[0072] It should be noted that the first-level difference is the value obtained by subtracting the previous point data from the later point data in the synchronization point time curve.

[0073] In some embodiments, in step S3, the trained BP neural network-based restoration network includes a peak restoration network, a trough restoration network, and a transition recovery network.

[0074] In some embodiments, the steps of training the peak recovery network and the trough recovery network include:

[0075] Extract the peak data and corresponding peak index array of the first-level differential curve to be processed, use 0 to fill the missing peaks, remove the peak placeholder data and corresponding peak index of each peak, and use the remaining peak data and corresponding peak index array to create a peak training set; use the peak training set to train the BP neural network to obtain the peak recovery network;

[0076] The trough data and the corresponding trough index array of the first-level differential curve to be processed are extracted, and the missing troughs are filled with 0. The trough placeholder data and the corresponding trough index of each trough are eliminated, and the remaining trough data and the corresponding trough index array are used to prepare a trough training set; the trough training set is used to train the BP neural network to obtain a trough recovery network.

[0077] It should be noted that the peak data and trough data of the first-level differential curve are extracted, and the number of peaks and troughs of each missing data segment is calculated according to the data amount of each missing data segment in the first-level differential curve and the frequency of occurrence of peaks and troughs of valid data in the first-level differential curve. In the extracted peak data, 0 is used to replace the missing peaks, and in the extracted trough data, 0 is used to replace the missing troughs. A one-dimensional peak array and a one-dimensional trough array are generated respectively, and a peak index array and a trough index array are generated in sequence; the extracted peak data are used to construct a training set, the peak data are arranged in sequence as the correct output, and the peaks with a placeholder of 0 are removed as input, and this is used as the training set to train the peak recovery network based on the BP neural network; the extracted trough data are used to construct a training set, the trough data are arranged in sequence as the correct output, and the troughs with a placeholder of 0 are removed as input, and this is used as the training set to train the trough recovery network based on the BP neural network.

[0078] The transition segment recovery network includes n groups of transition segment recovery networks, where n is the maximum value of the total number of data points included in each transition segment in the valid data segment minus one;

[0079] The steps of training n transition recovery networks include:

[0080] SA1: Mark the index of each transition segment in the first-level differential curve according to the sampling time sequence, extract the transition segments of the valid data segments in the first-level differential curve to construct a transition segment array, and retain the index of each transition segment of the valid data segment based on the sampling time sequence;

[0081] The transition section is the set of sampling points between adjacent peaks and troughs;

[0082] Assume that there are five transition segments in the first-order difference curve. The five transition segments are indexed in the order of sampling time. The index numbers 1, 3, and 5 are transition segments without missing segments (transition segments without missing segments belong to valid data segments). That is, there are three transition segments with valid data segments. The indexes of the transition segments of the valid data segments are retained based on the order of sampling time, that is, the index numbers are 1, 3, and 5.

[0083] SA2: Subtract the adjacent data points in each transition segment by the rule of subtracting the data point with the sampling time before from the data point with the sampling time after, and obtain the transition segment difference data corresponding to each transition segment;

[0084] Perform differential processing on the transition segments with index numbers 1, 3, and 5 to obtain the transition segment difference data corresponding to each transition segment.

[0085] SA3: extracting the first difference data point in each transition segment difference data, and the index value of the first difference data point in each transition segment difference data is the same as the index value of the corresponding transition segment;

[0086] Next, the first difference data point is extracted from the transition segment difference data corresponding to the transition segments with index numbers 1, 3, and 5, and the index of the first difference data point extracted from the transition segment difference data corresponding to the transition segment with index number 1 is marked as 1, the index of the first difference data point extracted from the transition segment difference data corresponding to the transition segment with index number 3 is marked as 3, and the index of the first difference data point extracted from the transition segment difference data corresponding to the transition segment with index number 5 is marked as 5.

[0087] SA4: Create the first set of transition data training sets using the first set of difference data points extracted from the difference data of each transition segment and the index of the corresponding mark;

[0088] SA5: Use the first set of transition data training sets to train the BP neural network to obtain the first set of transition segment recovery networks;

[0089] SA6: Replace the first difference data point with the next difference data point and repeat steps SA3-SA5 until n sets of transition segment recovery networks are obtained.

[0090] It should be noted that the valid data segment emphasizes the validity of the data, that is, it is noise-free, anomaly-free, missing, and break-free and can be used for subsequent processing. The transition segment between the peak and the trough in the intact (valid) data segment in the first-level differential curve is extracted. By comparing the number of data points in each transition segment in the valid data segment, the maximum total number of data in the transition segment, M, is obtained; at the same time, the number of data points in each transition segment in the missing data is compared to obtain the maximum total number of data in the transition segment in the missing data, N. The present invention is applicable to the case where M≥N. The two data points before and after each transition segment are subtracted to obtain the corresponding transition segment difference data. The first number in the difference data of each transition segment is extracted, and a training set is constructed based on the index of the transition segment. The BP neural network is trained using the training set to obtain n groups of transition segment recovery networks.

[0091] In some embodiments, the process of repairing the first-order difference curve using the repair network is as follows:

[0092] The peak index corresponding to each peak occupancy data is input into the peak recovery network to realize peak data repair;

[0093] The trough index corresponding to each trough placeholder data is input into the trough recovery network to realize trough data repair;

[0094] The index value of the transition segment with missing data points is input into the transition segment recovery network, and the transition segment corresponding to the difference data of the current transition segment is reversed based on the recovery result. The current transition segment is replaced by the transition segment of the next missing data point and the above operation is repeated until all transition segments are repaired;

[0095] Repair the first-level differential curve.

[0096] It should be noted that the indexes corresponding to the transition segments with missing data are input into the first to nth transition segment recovery networks respectively, and the transition segment difference data corresponding to each group of transition segment recovery networks are obtained. Based on the transition segment difference data corresponding to each group of transition segment recovery networks, a transition segment difference array is constructed, and all transition segments are repaired by inferring the transition segment difference data contained in the transition segment difference array.

[0097] Back-calculation means: assuming that the first data point of the transition segment is a1 (if there is an initial reference value, it can be used directly; if not, it needs to be estimated based on the context);

[0098] The second point a2=a1+the first number in the difference array;

[0099] The third point a3=a2+the second number in the difference array;

[0100] …

[0101] The mth point +The m-1th number in the difference array.

[0102] Through this accumulation, the complete "transition segment" can be inferred from the transition segment difference data, thereby completing the repair of the missing transition segment.

[0103] In some embodiments, step S4 specifically includes:

[0104] S41: adding the first data value of the restored segment of the repaired first-level differential curve to the synchronization point time immediately preceding the first missing synchronization point time of the missing segment of the synchronization point time curve to obtain the first synchronization point time restored value of the missing segment of the synchronization point time curve;

[0105] S42: adding the second data value of the restored segment of the repaired first-level differential curve to the first synchronization point time recovery value of the missing segment of the synchronization point time curve to obtain the second synchronization point time recovery value of the missing segment of the synchronization point time curve;

[0106] S43: using the next data value of the restored segment of the repaired first-level differential curve and the current synchronization point time restoration value of the missing segment of the synchronization point time curve, repeat step S42 until the repaired synchronization point time curve is obtained.

[0107] It should be noted that the first point data of the missing segment in the synchronization point time curve of the pulse synchronization signal is added to the first point data of the recovery segment of the first-level differential curve to restore the first point data of the missing segment in the synchronization point time curve, and the second point data of the missing segment in the synchronization point time curve is restored by adding the first point data recovered in the synchronization point time curve of the second-level pulse synchronization signal to the second point data of the recovery segment of the first-level differential curve, and so on, to obtain the complete repaired second-level pulse synchronization signal synchronization point time curve.

[0108] Furthermore, the number of recovery segments of the first-order differential curve is greater than or equal to 1, and a single recovery segment is a set of corresponding recovery data. The number of missing segments of the synchronization point time curve is greater than or equal to 1, and a single missing segment is a set of corresponding missing data.

[0109] In some embodiments, in step S5, in the pulse synchronization signal sequence, the pulse synchronization signal is inserted in the sampling time order, and the synchronization edge time of the inserted pulse synchronization signal is the same as the corresponding synchronization point time recovery value to obtain a repaired pulse synchronization signal sequence.

[0110] It should be noted that, by using the synchronization point time of each pulse synchronization signal repaired in the complete synchronization point time curve, the pulse synchronization signals consistent with the repaired synchronization point time are inserted in sequence into the original missing pulse synchronization signal to obtain a complete repaired pulse synchronization signal sequence.

[0111] based on Figure 2a 、 Figure 2b and Figure 2c Various curves of the second-level pulse synchronization signal with missing Figure 3a 、 Figure 3b and Figure 4a 、 Figure 4b In contrast, the traditional repair method and the repair method based on BP neural network do not easily show obvious differences in the original data and synchronization point time curve levels, but further Figure 3c and Figure 4c It can be seen from the comparison that at the first-order differential curve level of the synchronous time point, the first-order differential curve of the traditional repair method deviates significantly from the actual one, while the repair result of the present invention is closer to the actual one. Figure 2c 、 Figure 3c and Figure 4c The horizontal axis is the serial number of the time difference between two adjacent synchronization points arranged in the order of sampling time.

[0112] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved. This is not limited herein.

[0113] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A method for repairing a pulse synchronization signal, characterized in that: The specific steps include: S1: Acquire a pulse synchronization signal sequence, pre-process the pulse synchronization signal sequence, and obtain a synchronization point time curve of the pulse synchronization signal sequence; S2: performing a first-order difference on the synchronization point time curve to obtain a first-order difference curve; S3: Obtaining a trained repair network based on a BP neural network, and using the repair network to repair the first-order difference curve to obtain a repaired first-order difference curve; In step S3, the trained BP neural network-based repair network includes a peak recovery network, a trough recovery network, and a transition recovery network; The steps for training the peak recovery network and the trough recovery network include: Extract the peak data and corresponding peak index array of the first-level differential curve to be processed, use 0 to fill the missing peaks, remove the peak placeholder data and corresponding peak index of each peak, and use the remaining peak data and corresponding peak index array to create a peak training set; use the peak training set to train the BP neural network to obtain the peak recovery network; Extract the trough data and the corresponding trough index array of the first-level differential curve to be processed, use 0 to fill the missing troughs, remove the trough placeholder data and the corresponding trough index, and use the remaining trough data and the corresponding trough index array to create a trough training set; use the trough training set to train the BP neural network to obtain the trough recovery network; The transition segment recovery network includes n groups of transition segment recovery networks, where n is the maximum value of the total number of data points included in each transition segment in the valid data segment minus one; The steps of training n transition recovery networks include: SA1: Extract the transition segments of the valid data segments from the first-level difference curve to construct a transition segment array, and retain the index of each transition segment of the valid data segment based on the sampling time sequence; the transition segment is the set of sampling points between adjacent peaks and troughs; SA2: Subtract the adjacent data points in each transition segment by the rule of subtracting the data point with the sampling time before from the data point with the sampling time after, and obtain the transition segment difference data corresponding to each transition segment; SA3: extracting the first difference data point in each transition segment difference data, and the index value of the first difference data point in each transition segment difference data is the same as the index value of the corresponding transition segment; SA4: Create the first set of transition data training sets using the first set of difference data points extracted from the difference data of each transition segment and the index of the corresponding mark; SA5: Use the first set of transition data training sets to train the BP neural network to obtain the first set of transition segment recovery networks; SA6: Replace the first difference data point with the next difference data point and repeat steps SA3-SA5 until n sets of transition recovery networks are obtained; The process of repairing the first-order difference curve using the repair network is as follows: The peak index corresponding to each peak occupancy data is input into the peak recovery network to realize peak data repair; The trough index corresponding to each trough placeholder data is input into the trough recovery network to realize trough data repair; The index value of the transition segment with missing data points is input into the transition segment recovery network, and the transition segment corresponding to the difference data of the current transition segment is reversed based on the recovery result. The current transition segment is replaced by the transition segment of the next missing data point and the above operation is repeated until all transition segments are repaired; Repairing the first-level differential curve; S4: Perform reverse repair on the repaired first-order differential curve to obtain a repaired synchronization point time curve; S5: Perform reverse repair on the repaired synchronization point time curve to obtain a repaired pulse synchronization signal sequence.

2. The method for repairing a pulse synchronization signal according to claim 1, wherein: Each pulse synchronization signal included in the pulse synchronization signal sequence is a periodic positive pulse signal.

3. The method for repairing a pulse synchronization signal according to claim 1, wherein: Step S1 specifically includes: S11: Calculating the median amplitude of each pulse synchronization signal included in the pulse synchronization signal sequence, wherein the median amplitude is the average value of the maximum synchronization edge amplitude and the minimum synchronization edge amplitude of the pulse synchronization signal; S12: If the sampling point data of the i-th pulse synchronization signal includes the median amplitude, the sampling time corresponding to the median amplitude is used as the synchronization point time; otherwise, two sampling point data with the smallest amplitude difference from the median amplitude are searched on the synchronization edge of the current pulse synchronization signal, and the two sampling point data are interpolated to obtain the sampling time corresponding to the median amplitude, and the sampling time corresponding to the median amplitude is used as the synchronization point time; S13: Replace the i-th pulse synchronization signal with the i+1-th pulse synchronization signal, repeat step S12, obtain the synchronization point time of all pulse synchronization signals, and plot all synchronization point times in the order of sampling time to obtain the synchronization point time curve of the pulse synchronization signal sequence.

4. The method for repairing a pulse synchronization signal according to claim 3, wherein: In step S2, the time difference between two adjacent synchronization points in the sampling time sequence is taken, and all the differences are plotted in the sampling time sequence to obtain a first-level difference curve; The difference between two adjacent synchronization point times is the value obtained by subtracting the synchronization point time before the sampling time from the synchronization point time after the sampling time.

5. The method for repairing a pulse synchronization signal according to claim 1, wherein: Step S4 specifically includes: S41: adding the first data value of the restored segment of the repaired first-level differential curve to the synchronization point time immediately preceding the first missing synchronization point time of the missing segment of the synchronization point time curve to obtain the first synchronization point time restored value of the missing segment of the synchronization point time curve; S42: adding the second data value of the restored segment of the repaired first-level differential curve to the first synchronization point time recovery value of the missing segment of the synchronization point time curve to obtain the second synchronization point time recovery value of the missing segment of the synchronization point time curve; S43: Repeat step S42 using the next data value of the restored segment of the repaired first-level differential curve and the current synchronization point time restoration value of the missing segment of the synchronization point time curve until the repaired synchronization point time curve is obtained.

6. The method for repairing a pulse synchronization signal according to claim 1, wherein: In step S5, the pulse synchronization signal is inserted into the pulse synchronization signal sequence in the sampling time sequence, and the synchronization edge time of the inserted pulse synchronization signal is the same as the corresponding synchronization point time recovery value, so as to obtain a repaired pulse synchronization signal sequence.

Images