A Deep Learning-Based Method and System for Matching Multi-Band Gravitational Wave Signals

Abstract

This invention discloses a deep learning-based method and system for matching multi-band gravitational wave signals, relating to the field of gravitational wave detection and signal processing technology. The method includes the following steps: establishing a unified time anchor chain across frequency bands; extracting the instantaneous frequency change trend and phase transition amplitude of the gravitational wave signal during the frequency sweep phase from the time anchor chain; and constructing a mutation feature expression sequence as a mutation fingerprint band. This invention characterizes the temporal changes during the frequency sweep phase through a unified time anchor chain and mutation fingerprint band, avoiding overall signal shift and event splitting across multiple bands. Furthermore, by combining the unlock trigger point and the breathing-style time anchor lateral shift scheduling mechanism, the alignment granularity is dynamically adjusted to maintain a single, continuous output of gravitational wave events during complex and stable evolution phases, thereby improving the reliability of subsequent astrophysical parameter analysis.

Description

Technical Field

[0001] This invention relates to the field of gravitational wave detection and signal processing technology, specifically to a method and system for matching multi-band gravitational wave signals based on deep learning. Background Technology

[0002] Deep learning-based multi-band gravitational wave signal matching refers to the process of automatically learning and representing the time-domain, frequency-domain, and time-frequency joint features of gravitational wave signals from different frequency ranges, detection stages, or detection channels during gravitational wave detection. This allows for consistent identification and establishment of correspondences between signals across different bands, even in complex noise environments. The process trains a neural network to extract stable and physically correlated feature representations from multi-band data, enabling the system to determine whether signals appearing in different bands originate from the same gravitational wave event and to correlate and match their evolution processes. This improves the reliability of weak gravitational wave signal identification and the overall detection accuracy under multi-band observation conditions.

[0003] Existing technologies suffer from the following shortcomings: In current technologies, during gravitational wave signal processing and multi-band matching, deep learning models typically rely on cross-band correlation features formed within historical time windows to perform signal alignment and event identification. However, when a gravitational wave source enters a rapid frequency sweep phase, the gravitational wave signal exhibits significant nonlinear evolution characteristics within an extremely short timescale, causing abrupt changes in the phase correspondence, temporal sequence, and frequency mapping relationships between multi-band signals. Existing deep learning models struggle to perceive and adapt to these abrupt changes in a timely manner, continuing to use correlation features established during the previous stable phase for matching. This leads to an overall shift in the temporal alignment of multi-band signals. This shift causes multi-band signals belonging to the same gravitational wave physical event to be incorrectly split along the time axis, resulting in multiple independent false event identification results. Ultimately, this severely interferes with the inversion of astrophysical parameters such as the mass, spin, and distance of the gravitational wave source, reducing the reliability of gravitational wave event determination and physical interpretation.

[0004] Is this a genuine technical problem existing in current technology? The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a deep learning-based method and system for matching multi-band gravitational wave signals, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a deep learning-based method for matching multi-band gravitational wave signals, comprising the following steps: Establish a unified time anchor chain across frequency bands, extract the instantaneous frequency change trend and phase transition amplitude of the gravitational wave signal during the frequency sweep phase on the time anchor chain, and construct a mutation feature expression sequence as a mutation fingerprint. Based on the nonlinear evolution segment in the time anchor chain of the mutation fingerprint band identification, within the nonlinear evolution segment, the band signal sequence is reconstructed according to the phase continuity of the multi-band signal to generate a cross-band consistent alignment correction trajectory. Drift monitoring nodes are deployed along the alignment correction trajectory to record the time residuals of signals in each frequency band on the alignment correction trajectory, and the location of the lockout trigger point is determined based on the sudden increase in residuals. A dual-time anchor traction zone is constructed at both ends of the alignment correction trajectory where the unlock trigger point is located. One end is anchored to the early stable evolution segment of the gravitational wave signal, and the other end is anchored to the boundary of the merging stage of the gravitational wave signal, forming an alignment constraint region with dynamic contraction and release capabilities. Based on the alignment constraint region, a breathing time anchor side-shift scheduling mechanism is initiated. The alignment window shrinks in the dense region of the unlock trigger point and expands in the sparse region of the unlock trigger point, dynamically adjusting the alignment granularity and maintaining a single continuous output sequence of gravitational wave events.

[0007] Preferably, the steps for establishing a unified time anchor chain across frequency bands and constructing mutation fingerprint bands are as follows: Time reference integration processing is performed on gravitational wave signals from different frequency ranges, mapping the timing information of each frequency band signal on the original acquisition time axis to the same continuous time reference frame, forming a unified time anchor chain across frequency bands; Under the constraint of the time anchor chain, the frequency distribution of each frequency band signal at adjacent time nodes is extracted along the time anchor chain around the evolution interval after the gravitational wave signal enters the frequency sweeping stage, forming an instantaneous frequency change trend that reflects the dynamic evolution characteristics of the frequency sweeping stage. Based on the instantaneous frequency change trend, the phase transition amplitude of the gravitational wave signal between adjacent time nodes is extracted along the time anchor chain, so that the phase transition amplitude and the instantaneous frequency change trend maintain a corresponding relationship on the time anchor chain. By combining and arranging the instantaneous frequency change trends and phase transition amplitudes corresponding to each time node in the time anchor chain in chronological order, a mutation feature expression sequence covering the frequency sweep stage is constructed, forming a mutation fingerprint band used to characterize mutation behavior.

[0008] Preferably, the steps for generating a cross-frequency band consistent alignment correction trajectory based on the mutation fingerprint band are as follows: Based on the established time anchor chain, the mutation fingerprint band is continuously unfolded along the direction of the time anchor chain. By comparing the changes in the mutation feature expression unit corresponding to adjacent time nodes, candidate segments in which the evolution state of the gravitational wave signal changes are identified. By combining the continuous distribution characteristics of mutation fingerprint bands on the time anchor chain within the candidate segment, the arrangement density and variation amplitude of mutation feature expression units are analyzed to determine the nonlinear evolution segment where the gravitational wave signal enters the nonlinear evolution state. Within the nonlinear evolution region, the phase change process of multi-band gravitational wave signals is unfolded using a time anchor chain as a unified time reference, and the multi-band signal segments are sequentially reconstructed based on the degree of phase continuity. The arrangement of the multi-band signal segments after sequential reconstruction on the time anchor chain is updated to ensure that the signal arrangement within the nonlinear evolution segment is consistent with the time connection of the preceding and following evolution segments, forming a consistent alignment correction trajectory across frequency bands.

[0009] Preferably, during the sequential reconstruction process within the nonlinear evolution segment, the arrangement order of multi-frequency signal segments in the time anchor chain is adjusted by maintaining the continuous connection relationship of the phase change process on the time anchor chain, so that the reconstructed signal arrangement conforms to the overall evolution trend of the gravitational wave signal, and is used to constrain the continuous extension of the alignment correction trajectory within the nonlinear evolution segment.

[0010] Preferably, the steps for outputting the unlock trigger point based on the alignment correction trajectory are as follows: Based on the generated alignment correction trajectory, multiple monitoring positions are set along the time extension direction of the alignment correction trajectory, so that the alignment correction trajectory forms a continuous and traceable reference sequence in the time dimension. The actual time position of the multi-band gravitational wave signal in the time anchor chain is recorded at each monitoring location and compared with the reference time position corresponding to the alignment correction trajectory to form a time residual evolution sequence arranged along the alignment correction trajectory. The temporal residual changes between adjacent monitoring locations are compared along the alignment correction trajectory to identify candidate locations where the temporal residual changes from slow to rapid increase. Candidate locations where the time residuals synchronously increase in multiple frequency bands are identified as the unlock trigger points, and the time position of the unlock trigger point in the time anchor chain is output.

[0011] Preferably, the monitoring positions are continuously distributed along the alignment correction trajectory according to the time anchor chain. The time residual is obtained by comparing the actual time position of the multi-band gravitational wave signal in the time anchor chain with the reference time position in the alignment correction trajectory. The unlock trigger point is defined as the position where the time residual synchronously deviates from the previous change trend in terms of time continuity, so that the unlock trigger point maintains a continuous and associative time positioning attribute in the time anchor chain.

[0012] Preferably, the steps for forming the alignment constraint region based on the unlock trigger point are as follows: Using the temporal position of the unlock trigger point in the alignment correction trajectory as the central reference, the time anchor chain is extended along the forward temporal direction of the alignment correction trajectory to bind the time anchor chain to the early stable evolution segment of the gravitational wave signal, thus constructing a forward time anchor traction relationship. Based on the completion of the forward time anchor traction relationship, the time backward direction along the alignment correction trajectory is expanded so that the time anchor chain is correspondingly bound to the boundary of the merging stage of the gravitational wave signal, thus constructing the backward time anchor traction relationship. The time interval between the forward time anchor traction relationship and the backward time anchor traction relationship is defined as the alignment constraint region, so that the alignment correction trajectory is simultaneously subject to the time constraints of the early stable evolution segment and the merger stage boundary within this interval. Based on the time offset changes of the alignment correction trajectory within the alignment constraint region, the alignment constraint region is guided to shrink or release on the time anchor chain, so that the alignment correction trajectory maintains time continuity and consistency near the unlock trigger point.

[0013] Preferably, within the alignment constraint region, the adjustment of the alignment correction trajectory on the time anchor chain is jointly constrained by the forward time anchor traction relationship and the backward time anchor traction relationship, so that the time position change of the alignment correction trajectory near the unlock trigger point region is less than the time position change of the region far from the unlock trigger point.

[0014] Preferably, the steps for initiating the breathing-type time anchor lateral shift scheduling mechanism based on the alignment constraint region are as follows: Based on the established alignment constraint area, a time reference arrangement that can be moved laterally is established within the alignment constraint area, so that the time anchor chain has adjustment space to move back and forth along the time axis within the alignment constraint area. Based on the lateral displacement capability of the time anchor chain, according to the distribution density of the unlocking trigger points on the time anchor chain within the alignment constraint area, the dense region and sparse region of the unlocking trigger points are identified, and the alignment window is guided to shrink in the dense region and expand in the sparse region. When the alignment window shrinks or expands, the multi-band gravitational wave signal is guided to be remapped onto the time axis according to the updated time anchor chain arrangement, thereby achieving dynamic adjustment of the alignment granularity. By connecting the updated time anchor chain mapping results within the alignment constraint region with the alignment correction trajectory, the multi-band gravitational wave signals can form a single continuous gravitational wave event output sequence on the time axis.

[0015] 10. A deep learning-based gravitational wave multi-band signal matching system, used to implement the deep learning-based gravitational wave multi-band signal matching method according to any one of claims 9, characterized in that it includes a time anchor construction module, a nonlinear reconstruction module, a drift monitoring module, a dual-anchor traction module, and an adaptive scheduling module: The time anchor construction module establishes a unified time anchor chain across frequency bands, extracts the instantaneous frequency change trend and phase transition amplitude of the gravitational wave signal during the frequency sweep phase on the time anchor chain, and constructs a mutation feature expression sequence as a mutation fingerprint band. The nonlinear reconstruction module identifies the nonlinear evolution segment in the time anchor chain based on the mutation fingerprint band. Within the nonlinear evolution segment, it reconstructs the band signal sequence according to the phase continuity of the multi-band signal to generate a consistent alignment correction trajectory across the frequency band. The drift monitoring module deploys drift monitoring nodes along the alignment correction trajectory, records the time residual of each frequency band signal on the alignment correction trajectory, and determines the location of the lockout trigger point based on the sudden increase in residual. The dual-anchor traction module constructs dual time-anchor traction belts at both ends of the alignment correction trajectory where the unlock trigger point is located. One end is anchored to the early stable evolution segment of the gravitational wave signal, and the other end is anchored to the boundary of the merging stage of the gravitational wave signal, forming an alignment constraint region with dynamic contraction and release capabilities. The adaptive scheduling module initiates a breathing-style time anchor lateral shift scheduling mechanism based on the alignment constraint region. It shrinks the alignment window in dense regions of the unlock trigger point and expands the alignment window in sparse regions of the unlock trigger point, dynamically adjusting the alignment granularity and maintaining a single continuous output sequence of gravitational wave events.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention constructs a unified time anchor chain across frequency bands and introduces abrupt change fingerprints during the frequency sweep phase to characterize signal evolution, ensuring that multi-band gravitational wave signals always have a consistent reference basis on the time axis. When the signal enters a phase of rapid frequency change, it can promptly capture the temporal changes brought about by frequency change trends and phase transitions. Through the synergistic effect of alignment correction trajectories and alignment constraint regions, it avoids overall shifts in the time dimension of multi-band signals, thereby effectively preventing the same gravitational wave event from being split and output, and improving the continuity and consistency of multi-band signal matching.

[0017] This invention introduces a release trigger point and a breathing-style time anchor lateral shift scheduling mechanism to enable dynamic adjustment of the time alignment process, allowing the alignment granularity to adaptively adjust according to the complexity of the gravitational wave signal evolution. It strengthens time constraints in densely populated release trigger point regions and releases alignment windows in sparsely populated regions, ensuring stable correlation of multi-band signals during complex evolution stages and avoiding over-constraint during stable stages. This maintains a single, continuous output of gravitational wave events throughout the entire evolution process, which is beneficial for improving subsequent astrophysical parameter analysis. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart of the deep learning-based gravitational wave multi-band signal matching method of the present invention.

[0020] Figure 2 This is a schematic diagram of the module of the deep learning-based gravitational wave multi-band signal matching system of the present invention. Detailed Implementation

[0021] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0022] This invention provides, for example Figure 1 The deep learning-based multi-band gravitational wave signal matching method shown includes the following steps: Establish a unified time anchor chain across frequency bands, extract the instantaneous frequency change trend and phase transition amplitude of the gravitational wave signal during the frequency sweep phase on the time anchor chain, and construct a mutation feature expression sequence as a mutation fingerprint. To obtain a unified, stable, and physically consistent time reference basis for multi-band gravitational wave signals during rapid frequency sweeping, the following steps are taken to establish a unified cross-band time anchor chain, and to construct a mutation fingerprint band on this time anchor chain to characterize mutation behavior. The specific steps are as follows: Gravitational wave signals from different frequency ranges undergo time-reference integration processing, mapping the temporal information of each frequency band signal on its original acquisition time axis to the same continuous time reference frame, thus forming a unified cross-frequency band time anchor chain. Specifically, the absolute time identifiers corresponding to each frequency band signal during acquisition are introduced into a unified time series arrangement rule. By standardizing the start time, duration interval, and time sampling interval of each frequency band signal, signals from different frequency bands form a set of sequentially arranged continuous time nodes on a unified time axis. This time anchor chain spans the entire process of gravitational wave signals from the early evolution stage to the frequency sweep stage, providing a stable time-carrying structure for subsequent feature extraction, thereby avoiding the accumulation of misalignments caused by differences in the original time identifiers of different frequency band signals.

[0023] Under the constraint of a unified time anchor chain, the state of the gravitational wave signal at each time point is continuously unfolded within the evolutionary range after the gravitational wave signal enters the frequency sweeping phase. Instantaneous frequency change trends reflecting the dynamic evolutionary characteristics of the frequency sweeping phase are extracted from the time anchor chain. Specifically, the frequency distribution of each frequency band at adjacent time points is read sequentially along the time anchor chain, mapping the frequency change process over time as a continuous trajectory. By comparing the direction and magnitude of the shift in the center of gravity of the frequency distribution between consecutive time points, a trend expression describing the frequency evolution trend is formed. This frequency change trend is arranged using the time anchor chain as an index, allowing the frequency change processes of different frequency bands to be presented and listed under the same time reference, thus laying the foundation for characterizing the overall evolutionary form of the frequency sweeping phase.

[0024] Based on the obtained frequency change trend, the phase transition amplitude information of the gravitational wave signal during the frequency sweeping phase is further extracted along the time anchor chain, and the phase change process is correspondingly linked to the aforementioned frequency change trend. Specifically, at each consecutive time node of the time anchor chain, the change in the phase state of the gravitational wave signal between adjacent time nodes is recorded. By quantifying the jump amplitude during the continuous phase evolution process, an amplitude sequence describing the phase transition behavior is formed. This phase transition amplitude sequence maintains a one-to-one correspondence with the instantaneous frequency change trend on the time anchor chain, ensuring that each time node simultaneously carries both frequency change information and phase transition information. This allows for the construction of a joint feature foundation reflecting the multidimensional evolutionary characteristics of the frequency sweeping phase on a unified time anchor chain.

[0025] Based on the joint expression of frequency change trends and phase transition amplitudes formed on a unified time anchor chain, the mutation behavior during the frequency sweep stage is centrally characterized, and a mutation feature expression sequence is constructed as a mutation fingerprint band. Specifically, the frequency change trends and phase transition amplitudes corresponding to consecutive time nodes in the time anchor chain are combined and arranged in chronological order, so that the joint features corresponding to each time node constitute a component unit in the mutation feature expression sequence. This mutation feature expression sequence extends continuously along the time anchor chain, completely covering the dynamic change process of gravitational wave signals during the frequency sweep stage, thus forming a mutation fingerprint band used to characterize the mutation characteristics of the frequency sweep stage. Through the construction of the mutation fingerprint band, the frequency evolution acceleration and phase change concentration behaviors exhibited by gravitational wave signals in different frequency bands during the frequency sweep stage are uniformly mapped to the same time reference, providing a stable, continuous, and physically consistent feature basis for subsequent identification and processing of nonlinear evolution segments.

[0026] Based on the nonlinear evolution segment in the time anchor chain of the mutation fingerprint band identification, within the nonlinear evolution segment, the band signal sequence is reconstructed according to the phase continuity of the multi-band signal to generate a cross-band consistent alignment correction trajectory. When gravitational wave signals undergo rapid nonlinear evolution during the frequency sweep phase, it is crucial to accurately identify key segments where the time relationships of multi-frequency band signals reverse, and to restore the physical order of each frequency band signal under a unified time reference within these segments. The following steps are employed to analyze the time anchor chain based on abrupt change fingerprints and generate consistent alignment correction trajectories across frequency bands. The specific steps are as follows: Based on the previously constructed time anchor chain, the mutation fingerprint band is continuously unfolded along the time anchor chain. By comparing the changes in the mutation fingerprint band between adjacent time nodes segment by segment, the locations of intervals in the time anchor chain where the evolution state of the gravitational wave signal changes rapidly are identified. Specifically, the mutation feature expression units corresponding to each time node in the mutation fingerprint band are read sequentially along the time anchor chain. The overall trend of frequency change and phase transition amplitude in terms of temporal continuity is observed. When the mutation feature expression unit shows a transition from smooth change to rapid change between adjacent time nodes, the corresponding time node range is marked as a candidate segment where the evolutionary state has changed. In this way, a continuous time range containing the complex evolutionary behavior of the frequency sweep stage is initially delineated on the time anchor chain, providing a clear temporal positioning basis for subsequent refined processing.

[0027] Within the identified candidate segments, further refined analysis is conducted by combining the continuous distribution characteristics of the mutation fingerprint bands on the time anchor chain, thereby identifying the specific segment boundaries where the gravitational wave signal enters the nonlinear evolution state. Specifically, by continuously unfolding the internal change structure of the mutation fingerprint bands within the candidate segments, attention is paid to the arrangement density and change amplitude distribution of the mutation feature expression units in the time sequence. When the mutation features exhibit a concentrated distribution on the time anchor chain and form a continuous high-change interval, this continuous interval is identified as the nonlinear evolution segment. This nonlinear evolution segment maintains continuity in the time anchor chain and fully covers the key periods during the frequency sweep phase when the multi-band signal experiences time relationship reversal and phase relationship rearrangement, thus laying the time range foundation for subsequent reconstruction of the signal sequence within this segment.

[0028] Within the nonlinear evolution region, using a time anchor chain as a unified reference, the phase change process of gravitational wave signals in each frequency band is described, and the sequence of multi-frequency signal segments is reconstructed based on the degree of phase continuity. Specifically, the signals in each frequency band within the nonlinear evolution region are divided into multiple continuous signal segments along the time anchor chain, and the phase change state of each signal segment at its corresponding position on the time anchor chain is recorded. Subsequently, the smoothness of phase transition between adjacent signal segments is used as the basis for sequence reconstruction. Signal segments with continuous phase evolution relationships are preferentially arranged in adjacent positions, thereby gradually adjusting the arrangement order of different frequency band signal segments on the time anchor chain. Through this sequence adjustment based on phase continuity, the multi-frequency signal segments, whose order was originally disordered due to nonlinear evolution, are restored to a temporal arrangement relationship consistent with the physical evolution of gravitational waves.

[0029] After sequentially reconstructing multi-band signal segments within the nonlinear evolution region, the reconstruction results are re-bound to the time anchor chain, forming an alignment correction trajectory that extends continuously throughout the nonlinear evolution region. Specifically, the arrangement of the sequentially reconstructed multi-band signal segments on the time anchor chain is updated as a whole, ensuring temporal consistency with the preceding and following stable evolution regions within the nonlinear evolution region. This results in a continuous and smooth alignment correction path on a unified time anchor chain. This alignment correction trajectory covers the complete evolution process of multiple frequency band signals on the time anchor chain, enabling different frequency band gravitational wave signals to maintain their correspondence under a unified time reference even after undergoing nonlinear evolution during the frequency sweep stage. This provides a stable time alignment basis for subsequent tracking and scheduling processing based on changes in time residuals.

[0030] Drift monitoring nodes are deployed along the alignment correction trajectory to record the time residuals of signals in each frequency band on the alignment correction trajectory, and the location of the lockout trigger point is determined based on the sudden increase in residuals. After the multi-band gravitational wave signal undergoes nonlinear evolution and completes sequential reconstruction, the cross-band time alignment state is continuously tracked, and the location where the time alignment relationship becomes unstable is identified in a timely manner. The following steps are used to deploy drift monitoring nodes along the alignment correction trajectory, record the time residuals, and output the unlock trigger point. The specific implementation steps are as follows: Based on the generated alignment correction trajectory, multiple monitoring positions with time positioning capabilities are set along the time extension direction of the alignment correction trajectory, forming a continuous and traceable monitoring reference sequence in the time dimension. Specifically, the alignment correction trajectory is refined according to the continuous time nodes of the time anchor chain, and a corresponding monitoring position is established at each division position, so that the alignment correction trajectory has a clear reference point in different time periods. These monitoring positions are evenly distributed along the alignment correction trajectory, covering the complete time range from the start position to the end position of the nonlinear evolution segment, so that the alignment state of multi-band signals within this time range can be continuously observed and recorded, providing a stable time-bearing structure for subsequent acquisition of time residuals.

[0031] At the established monitoring locations, the actual arrival status of gravitational wave signals in each frequency band at their corresponding time positions is recorded based on the alignment correction trajectory. This data forms a time residual record of the multi-frequency band signals relative to the alignment correction trajectory. Specifically, at each monitoring location, the actual time position of each frequency band signal within the time anchor chain is compared with the reference time position given by the alignment correction trajectory. By recording the time offset between the two, a time residual reflecting the degree of alignment deviation of the frequency band signal at that monitoring location is formed. These time residuals are arranged sequentially along the alignment correction trajectory, creating a complete time residual evolution sequence of the time offset states of different frequency band signals at multiple monitoring locations. This reflects the overall stability change process of cross-frequency band signals along the alignment correction trajectory.

[0032] After obtaining the temporal residual evolution sequence, the changes in temporal residuals between adjacent monitoring positions are analyzed along the alignment correction trajectory to identify regions where the temporal residuals exhibit abrupt increases. Specifically, the temporal residual changes of each frequency band signal at consecutive monitoring positions are compared. When the temporal residual changes from a slow change to a rapid increase between adjacent monitoring positions, the corresponding monitoring position is identified as a candidate position where the temporal alignment relationship has changed abnormally. Simultaneously, by combining the consistency of temporal residual changes of multiple frequency band signals at the same monitoring position, the candidate positions are further focused, highlighting locations where the temporal residuals show concentrated increases across multiple frequency bands. In this way, a set of key positions reflecting changes in the stability of temporal alignment is formed on the alignment correction trajectory.

[0033] Based on the marked key locations, the points where the temporal residuals experience concentrated surges are identified as lock-out trigger points, which serve as the time reference output for subsequent alignment constraint adjustments. Specifically, when the temporal residual change at a certain monitoring location exhibits a significant deviation from the preceding evolutionary state in terms of temporal continuity, and this deviation occurs synchronously in multiple frequency bands, the temporal coordinate of that monitoring location in the alignment correction trajectory is determined as the lock-out trigger point. The lock-out trigger point maintains a clear temporal positioning attribute within the time anchor chain and forms a continuous correlation with the preceding and following time positions in the alignment correction trajectory, thus providing a clear time reference basis for subsequently constructing a time anchor traction relationship around the lock-out trigger point and adjusting the alignment constraint interval. Through the output of the lock-out trigger point, the temporal alignment instability behavior of multi-frequency gravitational wave signals in complex evolutionary stages can be accurately captured, laying the foundation for maintaining the continuous representation of gravitational wave events in the time dimension.

[0034] A dual-time anchor traction zone is constructed at both ends of the alignment correction trajectory where the unlock trigger point is located. One end is anchored to the early stable evolution segment of the gravitational wave signal, and the other end is anchored to the boundary of the merging stage of the gravitational wave signal, forming an alignment constraint region with dynamic contraction and release capabilities. After time alignment instability occurs in a multi-band gravitational wave signal, the temporal relationship before and after the instability segment is effectively constrained, and the propagation of alignment error is limited while maintaining the continuity of the overall event. The following steps are used to construct dual time anchor traction bands at both ends of the alignment correction trajectory where the loss-of-lock trigger point is located, thereby forming an adjustable alignment constraint region. The specific implementation steps are as follows: Using the determined time position of the unlock trigger point within the alignment correction trajectory as a central reference, the time anchor chain is traced backward along the temporal extension direction of the alignment correction trajectory to guide its corresponding binding with the early stable evolution segment of the gravitational wave signal, thus constructing a first time anchor traction relationship. Specifically, starting from the unlock trigger point and gradually expanding along the alignment correction trajectory towards an earlier time direction, the continuous time interval corresponding to the stable evolution stage of the gravitational wave signal within the time anchor chain is selected as the traction reference segment, fixing the arrangement of multi-frequency band signals on the time anchor chain within this segment as a stable reference. By establishing a traction association between the portion of the alignment correction trajectory before the unlock trigger point and the early stable evolution segment, the alignment correction trajectory obtains a stable temporal constraint basis in the forward temporal direction, thereby preventing alignment offset from spreading into the early evolution segment.

[0035] Building upon the established forward traction relationship, a second time anchor traction relationship is constructed by extending backward along the time extension direction of the alignment correction trajectory, centered on the same unlock trigger point. This guides the time anchor chain to be correspondingly bound to the time boundary of the gravitational wave signal merging phase. Specifically, extending the alignment correction trajectory from the unlock trigger point towards a later time, the continuous time boundaries corresponding to the entry of the gravitational wave signal into the merging phase are used as anchoring references, creating a clear time constraint position for these time boundaries within the alignment correction trajectory. This method constrains the alignment correction trajectory in the backward time direction to the merging phase boundary, preventing time alignment drift caused by the complex evolution of the frequency sweep phase from further intruding into the merging phase and ensuring the temporal integrity of the gravitational wave event during key physical phases.

[0036] After establishing the traction relationships of both the preceding and following time anchors, the time interval between the anchor position of the early stable evolution stage and the boundary anchor position of the merging stage in the alignment correction trajectory is defined as the alignment constraint zone, and an adjustable temporal traction tension distribution is introduced within this zone. Specifically, the position of the alignment correction trajectory on the time anchor chain within the alignment constraint zone is no longer completely fixed, but forms a restricted adjustment space under the combined action of forward and backward time anchor traction. This ensures that the temporal alignment relationship within this zone is constrained by both the traction of the early stable evolution stage and the traction of the merging stage boundary. Through this bidirectional traction method, the alignment constraint zone forms a closed but adjustable constraint band in the time dimension, ensuring that the adjustment of the alignment correction trajectory near the unlocking trigger point is always limited to a controllable range.

[0037] After the alignment constraint region is formed, the alignment constraint region is guided to dynamically contract and release on the time anchor chain based on the time offset changes of the alignment correction trajectory within the alignment constraint region. Specifically, when the alignment correction trajectory shows a tendency for time alignment to concentrate within the alignment constraint region, the coverage area of ​​the alignment constraint region on the time anchor chain contracts towards the unlock trigger point, making the time traction constraint more concentrated; when the alignment correction trajectory shows a tendency for time alignment to disperse within the alignment constraint region, the coverage area of ​​the alignment constraint region on the time anchor chain is appropriately released in the forward and backward anchoring directions, allowing the alignment correction trajectory to obtain the necessary adjustment space. Through this dynamic contraction and release alignment constraint method, the time alignment adjustment near the unlock trigger point is always constrained and controlled by the dual time anchor traction band, thereby maintaining the overall continuity and time consistency of the alignment correction trajectory when multi-band gravitational wave signals undergo complex evolution stages, laying a stable time constraint foundation for maintaining a single continuous output of gravitational wave events.

[0038] Based on the alignment constraint region, a breathing time anchor side-shift scheduling mechanism is initiated. The alignment window shrinks in the dense region of the unlock trigger point and expands in the sparse region of the unlock trigger point, dynamically adjusting the alignment granularity and maintaining a single continuous output sequence of gravitational wave events. Within the established alignment constraint region, the time alignment relationship of multi-band gravitational wave signals is finely controlled so that the time alignment adjustment can adapt to the spatial distribution changes of the unlock trigger point and avoid the gravitational wave event being split and output on the time axis. The following steps are used to initiate a breathing-type time anchor lateral shift scheduling mechanism based on the alignment constraint region: Based on the established alignment constraint zone, a laterally movable time reference arrangement is established within the alignment constraint zone, allowing the time anchor chain to no longer maintain a static, fixed position. Specifically, the coverage area of ​​the alignment constraint zone on the time anchor chain is used as the scheduling boundary. Each time node within this range is set as a set of reference nodes that can be moved as a whole, giving the time anchor chain space to move slightly forward and backward along the time axis within the alignment constraint zone. This lateral movement does not change the time boundary positions defined by the dual time anchor traction bands at both ends of the alignment constraint zone, but rather continuously adjusts the relative spacing of the internal time nodes under boundary constraints, thus providing an operable time reference basis for subsequent dynamic changes in alignment granularity.

[0039] Assuming the time anchor chain has lateral shift capability, the distribution of unlocking trigger points within the alignment constraint region on the time anchor chain is described in detail, and the shrinking or expanding of the alignment window is guided based on the distribution density of the unlocking trigger points in the time dimension. Specifically, all unlocking trigger points within the alignment constraint region are arranged according to the time anchor chain sequence. When the unlocking trigger points exhibit a concentrated distribution within a certain continuous time range, this time range is identified as a dense region of unlocking trigger points, and the time anchor chain is guided to compress the spacing between adjacent time nodes within this region, causing the alignment window to shrink in the time dimension. When the unlocking trigger points exhibit a scattered distribution within a certain continuous time range, this time range is identified as a sparse region of unlocking trigger points, and the time anchor chain is guided to stretch the spacing between adjacent time nodes within this region, causing the alignment window to expand in the time dimension. Through this lateral shifting of the time anchor chain and spacing adjustment based on the distribution characteristics of the unlocking trigger points, the alignment window can adaptively adjust to changes in the complexity of time evolution.

[0040] As the alignment window contracts or expands with the distribution of the unlock trigger points, the multi-band gravitational wave signals are guided to be remapped onto the time axis within the alignment constraint region according to the updated time anchor chain arrangement, thereby achieving dynamic adjustment of the alignment granularity. Specifically, when the alignment window contracts, the mapping positions of the multi-band signals on the time anchor chain become more concentrated, making the time alignment granularity finer and thus strengthening the time constraint on complex evolutionary behavior within densely populated unlock trigger point regions. When the alignment window expands, the mapping positions of the multi-band signals on the time anchor chain become more dispersed, making the time alignment granularity coarser and thus reducing the risk of over-adjusting time positions in relatively stable evolutionary regions. By synchronously adjusting the alignment granularity with the changes in the alignment window, the multi-band gravitational wave signals maintain a consistent mapping relationship with the time anchor chain throughout the entire alignment constraint region, avoiding time alignment instability caused by a fixed alignment scale.

[0041] After completing the shrinking and expanding of the alignment window and the dynamic adjustment of the alignment granularity, the updated time anchor chain mapping results within the alignment constraint region are integrated with the alignment correction trajectory, enabling the multi-band gravitational wave signals to form a single, continuous output sequence on the time axis. Specifically, by maintaining the temporal continuity between the boundaries before and after the alignment constraint region and the boundaries of the early stable evolution segments and merging stages, the time mapping results within the alignment constraint region, after being scheduled through a breathing-like time anchor shift, can naturally transition to the segments before and after the alignment correction trajectory, thereby avoiding discontinuous or repetitive event outputs on the time axis. Through this method, even during complex evolutionary stages with densely distributed unlock trigger points, the multi-band gravitational wave signals can still be continuously expressed as a single gravitational wave event output sequence under a unified time anchor chain constraint, providing a stable and coherent temporal basis for the overall determination of gravitational wave events and subsequent physical interpretation.

[0042] This invention constructs a unified time anchor chain across frequency bands and introduces abrupt change fingerprints during the frequency sweep phase to characterize signal evolution, ensuring that multi-band gravitational wave signals always have a consistent reference basis on the time axis. When the signal enters a phase of rapid frequency change, it can promptly capture the temporal changes brought about by frequency change trends and phase transitions. Through the synergistic effect of alignment correction trajectories and alignment constraint regions, it avoids overall shifts in the time dimension of multi-band signals, thereby effectively preventing the same gravitational wave event from being split and output, and improving the continuity and consistency of multi-band signal matching.

[0043] This invention introduces a release trigger point and a breathing-style time anchor lateral shift scheduling mechanism to enable dynamic adjustment of the time alignment process, allowing the alignment granularity to adaptively adjust according to the complexity of the gravitational wave signal evolution. It strengthens time constraints in densely populated release trigger point regions and releases alignment windows in sparsely populated regions, ensuring stable correlation of multi-band signals during complex evolution stages and avoiding over-constraint during stable stages. This maintains a single, continuous output of gravitational wave events throughout the entire evolution process, which is beneficial for improving subsequent astrophysical parameter analysis.

[0044] This invention provides, for example Figure 2 The deep learning-based gravitational wave multi-band signal matching system shown includes a time anchor construction module, a nonlinear reconstruction module, a drift monitoring module, a dual-anchor traction module, and an adaptive scheduling module. The time anchor construction module establishes a unified time anchor chain across frequency bands, extracts the instantaneous frequency change trend and phase transition amplitude of the gravitational wave signal during the frequency sweep phase on the time anchor chain, and constructs a mutation feature expression sequence as a mutation fingerprint band. The nonlinear reconstruction module identifies the nonlinear evolution segment in the time anchor chain based on the mutation fingerprint band. Within the nonlinear evolution segment, it reconstructs the band signal sequence according to the phase continuity of the multi-band signal to generate a consistent alignment correction trajectory across the frequency band. The drift monitoring module deploys drift monitoring nodes along the alignment correction trajectory, records the time residual of each frequency band signal on the alignment correction trajectory, and determines the location of the lockout trigger point based on the sudden increase in residual. The dual-anchor traction module constructs dual time-anchor traction belts at both ends of the alignment correction trajectory where the unlock trigger point is located. One end is anchored to the early stable evolution segment of the gravitational wave signal, and the other end is anchored to the boundary of the merging stage of the gravitational wave signal, forming an alignment constraint region with dynamic contraction and release capabilities. The adaptive scheduling module initiates a breathing-style time anchor lateral shift scheduling mechanism based on the alignment constraint region. It shrinks the alignment window in dense regions of the unlock trigger point and expands the alignment window in sparse regions of the unlock trigger point, dynamically adjusting the alignment granularity and maintaining a single continuous output sequence of gravitational wave events.

[0045] The deep learning-based gravitational wave multi-band signal matching method provided in this embodiment of the invention is implemented through the aforementioned deep learning-based gravitational wave multi-band signal matching system. For details of the specific methods and processes of the deep learning-based gravitational wave multi-band signal matching system, please refer to the embodiments of the deep learning-based gravitational wave multi-band signal matching method described above, which will not be repeated here.

[0046] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A deep learning-based method for matching multi-band gravitational wave signals, characterized in that, Includes the following steps: Establish a unified time anchor chain across frequency bands, extract the instantaneous frequency change trend and phase transition amplitude of the gravitational wave signal during the frequency sweep phase on the time anchor chain, and construct a mutation feature expression sequence as a mutation fingerprint. Based on the nonlinear evolution segment in the time anchor chain of the mutation fingerprint band identification, within the nonlinear evolution segment, the band signal sequence is reconstructed according to the phase continuity of the multi-band signal to generate a cross-band consistent alignment correction trajectory. Drift monitoring nodes are deployed along the alignment correction trajectory to record the time residuals of signals in each frequency band on the alignment correction trajectory, and the location of the lockout trigger point is determined based on the sudden increase in residuals. A dual-time anchor traction zone is constructed at both ends of the alignment correction trajectory where the unlock trigger point is located. One end is anchored to the early stable evolution segment of the gravitational wave signal, and the other end is anchored to the boundary of the merging stage of the gravitational wave signal, forming an alignment constraint region with dynamic contraction and release capabilities. Based on the alignment constraint region, a breathing time anchor side-shift scheduling mechanism is initiated. The alignment window shrinks in the dense region of the unlock trigger point and expands in the sparse region of the unlock trigger point, dynamically adjusting the alignment granularity and maintaining a single continuous output sequence of gravitational wave events.

2. The deep learning-based multi-band gravitational wave signal matching method according to claim 1, characterized in that, The steps for establishing a unified cross-frequency band time anchor chain and constructing mutation fingerprints are as follows: Time reference integration processing is performed on gravitational wave signals from different frequency ranges, mapping the timing information of each frequency band signal on the original acquisition time axis to the same continuous time reference frame, forming a unified time anchor chain across frequency bands; Under the constraint of the time anchor chain, around the evolution range after the gravitational wave signal enters the frequency sweeping stage, the frequency distribution state of each frequency band signal at adjacent time nodes is extracted along the time anchor chain to form the instantaneous frequency change trend reflecting the dynamic evolution characteristics of the frequency sweeping stage. Based on the instantaneous frequency change trend, the phase transition amplitude of the gravitational wave signal between adjacent time nodes is extracted along the time anchor chain, so that the phase transition amplitude and the instantaneous frequency change trend maintain a corresponding relationship on the time anchor chain. The instantaneous frequency change trend and phase transition amplitude corresponding to each time node in the time anchor chain are combined and arranged in chronological order to construct a mutation feature expression sequence covering the frequency sweep stage, forming a mutation fingerprint band.

3. A deep learning-based gravitational wave multi-band signal matching system, used to implement the deep learning-based gravitational wave multi-band signal matching method according to any one of claims-2, characterized in that, It includes a time anchor construction module, a nonlinear reconstruction module, a drift monitoring module, a dual-anchor traction module, and an adaptive scheduling module: The time anchor construction module establishes a unified time anchor chain across frequency bands, extracts the instantaneous frequency change trend and phase transition amplitude of the gravitational wave signal during the frequency sweep phase on the time anchor chain, and constructs a mutation feature expression sequence as a mutation fingerprint band. The nonlinear reconstruction module identifies the nonlinear evolution segment in the time anchor chain based on the mutation fingerprint band. Within the nonlinear evolution segment, it reconstructs the band signal sequence according to the phase continuity of the multi-band signal to generate a consistent alignment correction trajectory across the frequency band. The drift monitoring module deploys drift monitoring nodes along the alignment correction trajectory, records the time residual of each frequency band signal on the alignment correction trajectory, and determines the location of the lockout trigger point based on the sudden increase in residual. The dual-anchor traction module constructs dual time-anchor traction belts at both ends of the alignment correction trajectory where the unlock trigger point is located. One end is anchored to the early stable evolution segment of the gravitational wave signal, and the other end is anchored to the boundary of the merging stage of the gravitational wave signal, forming an alignment constraint region with dynamic contraction and release capabilities. The adaptive scheduling module initiates a breathing-style time anchor lateral shift scheduling mechanism based on the alignment constraint region. It shrinks the alignment window in dense regions of the unlock trigger point and expands the alignment window in sparse regions of the unlock trigger point, dynamically adjusting the alignment granularity and maintaining a single continuous output sequence of gravitational wave events.

Images