Device status monitoring method based on time-synchronous compression mechanism

Abstract

This invention belongs to the field of data analysis and processing technology, specifically relating to a device status monitoring method based on a time-series synchronous compression mechanism. The method includes: Step 1: The monitoring terminal collects vibration, acoustic, and acceleration signals on a unified time baseline of the BeiDou timing signal, and registers the second and frame numbers in a unified segment index table; Step 2: The original time-series data packet is input, and a phase-locked loop constructs a phase grid frame based on a reference phase rail, extracting path-level flag event bits frame by frame and fusing them into frame-level flag event bits; Step 3: Within the recognition window, based solely on the coupling symbols, mode commands, and abnormal segment markers in the synchronously compressed time-series packet and the unified segment index table, the device status label is output according to a mutual exclusion priority rule. This method significantly outperforms existing technologies in terms of time alignment accuracy, pattern recognition determinism, data compression efficiency, and reliability of health status determination.

Description

Technical Field

[0001] This invention belongs to the field of data analysis and processing technology, specifically relating to a device status monitoring method based on a time-series synchronous compression mechanism. Background Technology

[0002] With the development of sensor technology, embedded systems, and smart terminals, equipment condition monitoring has become an important research direction in the fields of industrial equipment, wearable terminals, and health monitoring. Existing monitoring systems generally rely on multi-source signals (such as vibration signals, acoustic signals, and acceleration signals) to determine the operating status of equipment or the health status of humans. However, these signals often originate from different sampling channels and independent clock references, leading to significant timing inconsistencies during actual acquisition and analysis. Especially under motion or complex operating conditions, signal phase drift, sampling offset, and time reference errors can cause deviations in feature extraction and status recognition, affecting the stability and accuracy of monitoring results.

[0003] Currently common time synchronization methods mainly include: late alignment based on software timestamps, dynamic correction based on sampling clock drift compensation, and unified time synchronization using external timing signals (such as GPS, BeiDou, etc.). Among these, software timestamp methods are easily affected by system latency and multi-task scheduling, and cannot achieve millisecond-level synchronization; sampling drift compensation relies on model estimation, which is complex and has poor adaptability to non-stationary signals; while synchronization methods based on timing signals can provide a unified time reference, they still lack alignment mechanisms and anomaly management strategies in the signal analysis and data fusion stages, resulting in limited accuracy of multi-source signal fusion. On the other hand, existing pattern recognition and data compression mechanisms mostly focus on extracting time-domain or frequency-domain features of independent signals, such as short-time Fourier transform, wavelet packet decomposition, or principal component analysis. These methods are effective in identifying single-channel anomalies, but they cannot effectively characterize the coupling characteristics between different signal sources, and it is difficult to achieve efficient online processing with limited computing resources. In addition, traditional compression algorithms such as LZW, Huffman or differential coding are mainly designed for general data sequences and lack adaptive structures for temporal coupling characteristics. Therefore, they often have problems such as insufficient compression ratio or high decompression delay in real-time monitoring systems. Summary of the Invention

[0004] Therefore, the main objective of this invention is to provide a device status monitoring method based on a time-series synchronous compression mechanism. This method achieves high-precision time alignment by resampling, phase-locking, and cyclically shifting multi-source signals such as vibration, acoustics, and acceleration on a unified time baseline of the BeiDou timing signal. Furthermore, it utilizes morphological symbolization and entropy coupling analysis to construct multi-dimensional coupling features, combined with pattern command execution synchronous compression and index consistency management, thereby achieving efficient signal fusion and stable identification in complex environments. This method significantly outperforms existing technologies in terms of time alignment accuracy, pattern recognition determinism, data compression efficiency, and reliability of health status determination, effectively improving the real-time performance, stability, and accuracy of monitoring terminals. It provides an efficient and scalable solution for multi-source signal fusion-based health and device status monitoring.

[0005] The technical solution adopted in this invention is as follows:

[0006] A device status monitoring method based on a time-synchronization compression mechanism, the method comprising:

[0007] Step 1: The monitoring terminal collects vibration, acoustic and acceleration signals on the unified time baseline of the Beidou timing signal, resamples them according to the preset phase grid frame and sample grid, generates alignment tags and original time data packets, records abnormal segment markers, and registers the second sequence number and frame sequence number in the unified segment index table.

[0008] Step 2: Input the original time-series data packet. The phase-locked loop constructs a phase grid frame based on the reference phase track. Path-level flag event bits are extracted frame by frame and fused into frame-level flag event bits. The intra-frame sequence is cyclically shifted according to early / late determination to align the frame-level flag event bits to the reference sample. Combined with cross-frame continuity verification and capture restart, a phase-locked loop alignment sequence is obtained. Subsequently, a morphological symbol is generated for the phase-locked loop alignment sequence according to a fixed determination. The entropy coupler generates coupling symbols based on recognition window statistics and intra-second symmetrical frame mapping. Coupling symbols include synchronous rare coupling, complementary mirror coupling, energy transfer coupling, and steady-state rhythm coupling. The second sequence number, frame sequence number, morphological symbol, and coupling symbol are written into a unified segment index table. For completely repeated subsequences in the coupling symbols, synchronous compression is performed using mode instructions to form a synchronous compressed time-series packet containing only the frame sequence number, coupling symbol, mode instructions, and abnormal segment markers. For invalid phase-locked seconds, freezing, backtracking, and overwriting are triggered to maintain index consistency.

[0009] Step 3: Within the identification window, based solely on the coupling symbols, mode instructions, and abnormal fragment markers in the synchronous compressed timing packet and the unified fragment index table, output the device status label according to the mutual exclusion priority rule.

[0010] Furthermore, the BeiDou timing signal is measured in seconds; each BeiDou second is divided into 64 phase grid frames, numbered from 1 to 64; within each phase grid frame, each signal is uniformly resampled into 16 samples, numbered from 1 to 16; the phase-locked reference position is fixed at sample number 8 within the frame; if the maximum and minimum values ​​of the 16 samples of the same signal within a frame are equal, then that frame is marked as an abnormal segment for that signal; the set of morphological symbols is fixed as rising spikes, falling spikes, narrow peaks, wide peaks, flat segments, and gaps; the identification window is fixed as 8 consecutive BeiDou seconds, 64 frames per second; within the same second, two frames whose frame numbers sum to sixty-five are defined as a pair of symmetrical frames; each record item in the unified segment index table is set as second number, frame number, signal source, morphological symbol, coupling symbol, mode command, and abnormal segment marker; the frame quadrants are defined as 1 to 16, 17 to 32, 33 to 48, and 49 to 64.

[0011] Furthermore, in step 2, the process of extracting the flag event bit includes: In each frame, for each signal, a judgment correction is performed according to the following three judgment rules: First judgment rule: If a single peak exists, and strictly decreasing sequences appear on both sides of the peak sample point, with the length of the left decreasing sequence not less than 2 and the length of the right decreasing sequence not less than 2, and the peak value minus the minimum value of the frame is not less than 3 original count units, then the peak sample point is taken as the flag event bit; Second judgment rule: If there is no single peak that satisfies the first judgment rule, but there is a rising surge pattern candidate: that is, there is one continuous rise with a continuous sample number in the range of 3 to 6. Furthermore, if the value of the last sample point minus the value of the first sample point is not less than 3 original count units, then the last sample point of the rising segment is used as the flag event bit; the third judgment rule: if there is a gap pattern candidate: that is, there is a local valley value, and there are strictly increasing sequences on both sides of the valley value, the length of both increasing sequences is not less than 2, and the maximum value of the frame minus the valley value is not less than 3 original count units, then the valley value sample point is used as the flag event bit; if multiple judgment rules are satisfied at the same time, the corresponding judgment rule is selected for judgment correction according to the priority order of the first judgment rule taking precedence over the second judgment rule, and the second judgment rule taking precedence over the third judgment rule.

[0012] Furthermore, if none of the three judgment rules are met, the sample point with sample point number 8 in the frame is used as the flag event bit; there are a maximum of 3 flag event bits in the same frame, and the selection method is as follows: first, select the flag event bit that is not marked as an abnormal segment; if there are 2 or more, select the one closest to sample point number 8; if the distance is the same, select according to a fixed priority: vibration type is higher than acoustic type, acoustic type is higher than acceleration type, and the frame-level flag event bit of the frame is obtained.

[0013] Furthermore, if the frame-level flag event bit is less than 8, it is defined as early arrival, and the frame sequence is cyclically shifted to the right within the frame, with the shift step number being 8 minus the current position; if it is greater than 8, it is defined as late arrival, and a cyclic shift is performed to the left, with the shift step number being the current position minus 8; after the shift, if the frame-level flag event bit is equal to 8, it is considered a successful phase lock; if the required shift step number is greater than 4, it is judged as a phase lock failure; if five consecutive frames fail to lock phase, a capture restart is triggered: the shift of these 5 frames is canceled, and the original order is maintained; if the number of successfully locked phase frames within the same second is not less than 48, that second is counted as a valid phase lock second, otherwise it is counted as an invalid phase lock second and the whole sequence is marked as a traceable segment.

[0014] Furthermore, in step 2, for each frame and each signal after lock-lock alignment, output one morphological symbol according to the following rules: Narrow peak: There is a unique peak sample; the peak value minus the median value of the frame is not less than 3 original count units; half-width is defined as the first sample position searched from the peak to the left and right that is not higher than half the difference between the peak value and the median value of the frame, and the number of samples between the two positions is not greater than 4; Wide peak: There is a unique peak sample; the peak value minus the median value of the frame is not less than 3 original count units; the peak width is calculated using the same half-width method, and the number of samples is not less than 8; Rising peak: There is no single peak. However, there is a single continuous rise, with a continuous number of sample points between 3 and 6 (including the boundary), and the end value minus the starting value is not less than 3 original count units; Drop spike: There is no single peak or rise spike, but there is a single continuous drop, with a continuous number of sample points between 3 and 6, and the starting value minus the ending value is not less than 3 original count units; Gap: There is a unique valley value sample point; the maximum value of the frame minus the valley value is not less than 3 original count units; Half width is calculated as half the difference between the valley value and the maximum value of the frame, and the number of sample points is not greater than 6; Flat segment: The difference between the maximum value and the minimum value within the frame is not greater than 2 original count units.

[0015] Furthermore, within the 512-frame recognition window, if a certain morphological symbol appears 1 time in a certain path, then that morphological symbol is recorded as a rare symbol; the mirror symbol pair mapping includes: rising peaks and falling peaks forming a pair; narrow peaks and gaps forming a pair; and wide peaks and flat segments forming a pair.

[0016] Furthermore, within the same frame, at least two signals simultaneously exhibit rare symbols, generating synchronous rare coupling; within a pair of symmetrical frames in the same second, if a pair of mirror symbols appears from the same signal source, complementary mirror coupling is generated; within four consecutive frames, the following rules are met to generate energy transfer coupling at the fourth frame position: the acceleration type of the first frame is a rising spike or a broad peak; the vibration type of the second frame is a narrow peak or a broad peak; the acoustic type of the third frame is a gap; the acoustic type of the fourth frame is a flat segment; statistically, for each BeiDou second: if within that second, at least two of the three signals are marked as flat segments in 48 or more of the 64 frames, and the total number of narrow peaks used for commutation within that second is no more than 1, then steady-state rhythm coupling is generated in the last frame of that second.

[0017] Furthermore, in step 2, within the identification window, a completely identical subsequence with a length of not less than 8 frames and an occurrence frequency of not less than 2 is searched from the coupling mark sequence; each repetition is replaced with a mode instruction, and the mapping relationship between the mode instruction and the start and end frames of the original sequence is written into the unified fragment index table; the output is a synchronous compressed timing packet, including frame number, coupling mark, mode instruction and abnormal fragment marker.

[0018] Furthermore, step 3 specifically includes: within each identification window, the following mutually exclusive priority determination is performed based solely on the coupling symbols, mode instructions, and abnormal segment markers in the synchronous compressed timing packet and the unified segment index table: if the effective phase-locked seconds within the identification window are less than 6 or the number of frames with abnormal segment markers is greater than 64, the signal quality is deemed insufficient; if the occurrence of rare synchronous couplings within the identification window is not less than 2 times and covers at least 2 frame quadrants, an abnormality is indicated; if there is repeated occurrence of complementary mirror couplings within the identification window, and the frame interval between two adjacent occurrences is between 7 and 9 frames, a tendency for resonance instability is determined; if there is at least one energy transfer coupling within the identification window, and the occurrence of rare synchronous couplings within that window is equal to 0 times, the operating load is determined; if steady-state rhythmic couplings are generated for each BeiDou second within the identification window, the standby is determined to be stable.

[0019] By adopting the above technical solution, the present invention has produced the following beneficial effects:

[0020] Achieving high-precision timing alignment of multi-source signals: This invention constructs a unified time reference based on BeiDou timing signals. Through a phase-locked loop and cyclic shift mechanism, different types of signals (vibration, acoustic, and acceleration signals) achieve phase consistency alignment at the same time. This design effectively eliminates sampling offset and time delay differences between multi-channel signals, significantly improving the time consistency of data fusion and state determination, and providing a stable foundation for subsequent feature recognition and health assessment.

[0021] Improving the determinism and robustness of complex pattern recognition: This invention establishes a structured temporal pattern recognition system through frame-level flag event bit extraction, morphological symbolization, and cross-frame statistics. The entropy coupler utilizes multi-dimensional features such as synchronous rare coupling, complementary mirror coupling, energy transfer coupling, and steady-state rhythm coupling to achieve fine characterization of the correlation between signals. This mechanism can accurately identify key state changes even in the presence of noise interference or motion artifacts, improving the robustness and interpretability of the recognition.

[0022] Significantly improving data compression efficiency and index consistency: This invention introduces a pattern instruction mechanism at the coupling symbol level, performing synchronous compression on recurring coupled subsequences, retaining only necessary frame numbers, coupling symbols, and anomaly marker information. This design significantly reduces storage and transmission redundancy while ensuring data integrity, achieving efficient management of long-term monitoring data. Simultaneously, the second number, frame number, morphological symbol, and pattern instruction recorded in the unified fragment index table maintain a one-to-one correspondence, ensuring consistency and determinism in subsequent data backtracking and retrieval.

[0023] Enhancing the stability and responsiveness of real-time monitoring terminals: By dynamically determining the effective second of phase-locked loop and employing a traceable segment mechanism, this invention can perform freeze and backtracking operations when signals are abnormal or phase-locking fails, ensuring the stable operation of the monitoring system in complex environments. This method supports low-latency state updates, significantly improving the terminal's real-time performance and system recovery capabilities compared to traditional solutions. Attached Figure Description

[0024] Figure 1 A schematic diagram of the method flow for a device status monitoring method based on a timing synchronization compression mechanism provided in an embodiment of the present invention;

[0025] Figure 2 This is a schematic diagram of the experimental curves for the core technology of BeiDou time synchronization compression provided in an embodiment of the present invention;

[0026] Figure 3 This is a schematic diagram illustrating the experimental accuracy of health status classification based on the mutual exclusion priority rule provided in an embodiment of the present invention. Detailed Implementation

[0027] All features disclosed in this specification, or all steps in all disclosed methods or processes, may be combined in any way, except for mutually exclusive features and / or steps.

[0028] Any feature disclosed in this specification (including any appended claims and abstract) may be replaced by other equivalent or similar features, unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is merely one example of a series of equivalent or similar features.

[0029] refer to Figure 1 A device status monitoring method based on a time-series synchronization compression mechanism, the method comprising:

[0030] Step 1: The monitoring terminal collects vibration, acoustic, and acceleration signals on the unified time baseline of the BeiDou timing signal. It resamples the signals according to a preset phase grid frame and sample grid, generating alignment tags and original time-series data packets. Abnormal segment markers are recorded, and the second and frame numbers are registered in the unified segment index table. Vibration signals can be heart rate signals, and acoustic signals can be blood oxygenation signals.

[0031] In the specific implementation, the whole-second pulses of the BeiDou timing signal are used as time boundaries to form a continuous second sequence. After the device is powered on, the first available whole-second pulse is used as the first second sequence number, and then the second sequence number is incremented by one for each subsequent whole-second pulse. This process fixes the three signals to the same time starting point, avoiding the problem of misalignment caused by the internal clock drift of each signal. Within each second sequence number range, the second is divided into 64 phase grid frames, with frame numbers from 1 to 64. The equal division of the phase grid frames ensures that the three signals obtain the same phase reference point within the second level, and any subsequent cross-channel comparisons are carried out under the same frame number, reducing the sources of phase error. Within each phase grid frame, a sampling grid is constructed, with each frame containing 16 samples, with sample numbers from 1 to 16. The uniform sampling grid provides a uniform value position within the frame, facilitating the use of the same time coordinate when performing resampling within the frame. The alignment label is formed by combining the second sequence number and the frame sequence number. Alignment tags are attached to each of the 16 samples within the frame, enabling precise time-location tracking of each sample in subsequent steps. This explicit tagging method helps maintain index stability during frame drops or insertions, without relying on implicit counts.

[0032] The raw timestamps and raw values ​​of heart rate, blood oxygen, and acceleration signals are collected, ensuring that the three signals do not interfere with each other but are uniformly grouped under the same second sequence number. This allows for consistent time archiving without constraining the original sampling rhythm of each signal. For each second sequence number, each frame sequence number, and each sample point sequence number, the target time on the actual time axis is determined. The target time is determined as follows: between the start and end integer pulses of the second, 64 equidistant frame boundaries are taken, and within each frame, 16 equidistant sample point positions are taken. This ensures that all samples are evenly distributed within the second, avoiding comparison deviations caused by uneven time steps. For each signal, deterministic value selection is performed for each target time: if an original sample point exists within the frame containing the target time, the original sample point that is closest to the target time in time is selected as the value of that sample point; if two equidistant original sample points exist, the one with the earlier timestamp is selected. This strategy avoids weighting, preventing implementation differences due to varying weight choices, and closely approximates the true waveform at high sampling densities. If no original sampling point exists within the frame containing the target time, but an original sampling point exists in an adjacent frame within the same second, the closest original sampling point within that second is selected as the value for that sample. If two equally spaced original sampling points exist, the earlier timestamp is chosen. This fills in sporadic missing data within a second, maintaining data continuity within that second. If no usable original sampling points exist within the same second (e.g., a short-term disconnection occurs within that second), a placeholder value is filled for that sample point across all frames within that second, and a missing note is added to the unified segment index table, preventing cross-second historical propagation during resampling. This limits the missing data to the current second, and subsequent quality control is performed through abnormal segment marking. After resampling, all 16 samples corresponding to each frame number within each second are assigned values. The three signals form a one-to-one correspondence under the same frame number and sample number, so that subsequent cross-channel comparison and alignment do not require time registration, reducing calculation steps and error accumulation.

[0033] Within each second and frame number, a consistency check is performed on 16 samples of heart rate, blood oxygen, and acceleration signals, respectively. If the maximum and minimum values ​​of the 16 samples in a frame are equal, the frame is marked as an abnormal segment for that signal. This condition corresponds to "flat segments" caused by device saturation, silence, or severe jumps. The advantage of early marking is that it delineates the quality problem at its source and time position, making it easier for subsequent steps to bypass or reduce its weight, while not changing the original values, for traceability. For each frame marked with an abnormal segment, the second number, frame number, and signal source of the abnormality are recorded, maintaining consistency with the alignment label, for rapid location during subsequent cross-frame or cross-channel analysis.

[0034] The original time-series data packets are organized as follows: all frames are arranged in ascending order of second number, and within each frame, 16 samples are arranged in ascending order of sample number. Within the same frame, the three signals are arranged sequentially as heart rate signal, blood oxygenation signal, and acceleration signal. This fixed arrangement order ensures data sequence consistency between different implementations, avoiding parsing errors caused by different signal sequences. Alignment tags are attached to each frame-level unit, corresponding to one second number and one frame number. Since the sample grid is fixed within a frame, the parser can directly locate the specific sample using the sample number within the frame. This preserves frame-level alignment and eliminates the need to repeatedly store redundant tags for the samples. The second number and frame number are registered in a unified fragment index table. The status of abnormal fragments within the frame, containing all three signals, is also registered for subsequent retrieval. The unified fragment index table serves as a reference directory, allowing subsequent steps to retrieve the corresponding frame in the original time-series data packets using only the index, improving retrieval efficiency and reducing memory usage.

[0035] Second-level integrity check: For each second sequence number, check if it contains 64 frame sequence numbers and 16 samples per frame. If the quantities are inconsistent, locate the missing frame sequence number or sample number, triggering the nearest neighbor value or placeholder filling strategy in this step until the structure of that second is complete. This ensures that subsequent steps face inputs with completely consistent structure, eliminating the need to write additional branches for structural differences. Cross-second boundary stability: The start and end of the phase grid frame strictly match two adjacent whole-second pulses of the BeiDou timing signal, and samples are not shared between seconds. This rule avoids the risk of cross-second data being counted twice, while ensuring that the alignment label naturally increments at the second boundary, facilitating synchronization maintenance during long-term operation. Testing and calibration recommendations: After the device is used for the first time or after changing the environment, it can be run for several seconds in a static state to check whether the variation range of the three signals within each frame meets expectations; if a large number of frames are marked as abnormal segments, the alignment of the sampling rhythm with the BeiDou timing signal should be checked. Through this self-check, problems such as loose connections, power supply noise, or weak timing reception can be detected during the deployment phase, reducing on-site backtracking costs.

[0036] Employing equally divided phase grid frames and sample grids allows for a stable grid structure to be fixed within a second, with all calculations performed on this stable grid, avoiding feature drift caused by dynamic time steps. The use of a temporally closest strategy for sample point selection eliminates reliance on weighting and adjustable parameters, ensuring consistent output across different devices and firmware versions. Furthermore, this strategy naturally suppresses short-term jitter: when original sample points fluctuate around the target time, they are always snapped to the same grid position by the nearest strategy, reducing intra-frame morphological errors caused by jitter. Abnormal segment marking uses "the maximum and minimum values ​​of the frame are equal" as the sole criterion, which is simple, clear, and repeatable. This criterion accurately locates frames with flat values ​​caused by saturation, silence, and continuous sample loss, allowing subsequent steps to handle these frames specifically without requiring further quality assessment during the identification phase. Without altering the sequence numbering system of seconds, frames, and samples, the nearest neighbor value can be replaced by the median value within the frame: when a sample location corresponds to multiple original sampling points within the frame, the median of these sampling points is taken as the value of that sample; when no original sampling point exists, the strategy reverts to the nearest sampling point within the second. This method is more robust to spike noise and is suitable for scenarios with strong electromagnetic interference. Without changing the abnormal segment marking conditions, a cause code field can be added to the abnormal segment marker in the unified segment index table to record information returned by the device self-test, such as weak reception, incomplete sensor self-calibration, or storage write failure. This additional information does not participate in the identification logic and is only used for fault location and maintenance.

[0037] Step 2: Input the original time-series data packet. The phase-locked loop constructs a phase grid frame based on the reference phase track. Path-level flag event bits are extracted frame by frame and fused into frame-level flag event bits. The intra-frame sequence is cyclically shifted according to early / late determination to align the frame-level flag event bits to the reference sample point. Combined with cross-frame continuity verification and capture restart, a phase-locked loop alignment sequence is obtained. Subsequently, a morphological symbol is generated for the phase-locked loop alignment sequence according to a fixed determination. The entropy coupler generates coupling symbols based on recognition window statistics and intra-second symmetrical frame mapping. Coupling symbols include synchronous rare coupling, complementary mirror coupling, energy transfer coupling, and steady-state rhythm coupling. The second sequence number, frame sequence number, morphological symbol, and coupling symbol are written into a unified segment index table. For completely repeated subsequences in the coupling symbols, synchronous compression is performed using mode instructions to form a synchronous compressed time-series packet containing only the frame sequence number, coupling symbol, mode instructions, and abnormal segment markers. For invalid phase-locked seconds, freezing, backtracking, and overwriting are triggered to maintain index consistency.

[0038] Specifically, the alignment tags formed in step 1 are read and traversed incrementally by second number. Each second number is divided into 64 phase grid frames, with frame numbers ranging from 1 to 64. The reference phase track is constructed sequentially from these equally spaced phase grid frames. Using an equally spaced reference phase track ensures that subsequent early / late determinations are limited to a unified time scale, avoiding relative drift caused by internal clock jitter. Within each phase grid frame, 16 sample point numbers are retained, ranging from 1 to 16. Fixed sample point numbers provide a unified reference within the frame, ensuring that alignment and comparison across paths and frames are performed in the same coordinate system.

[0039] Under each frame number, three deterministic judgment rules are applied to 16 samples of heart rate, blood oxygen, and acceleration signals to generate road-level marker event bits: First judgment rule: If a single peak exists, and both sides of the peak sample point are strictly decreasing sequences with a left decreasing sequence length of at least 2 and a right decreasing sequence length of at least 2, and the peak value minus the minimum value of the frame is at least 3 original count units, then the sample number of the peak sample point is recorded as the road-level marker event bit. This prioritizes the location of significant and stable instantaneous change points, facilitating the alignment of the most representative event to the reference position after cyclic shifting. Second judgment rule: If the first judgment rule is not met, but there is a continuous rise with a continuous sample number between 3 and 6, and the value of the last sample point minus the value of the first sample point is at least 3 original count units, then the sample number of the last sample point of the rising segment is recorded as the road-level marker event bit. This allows the capture of the end position of rapid energy accumulation even in the absence of isolated peaks, and subsequent alignment can ensure that this end position is aligned between different roads. The third determination rule: If a local valley value exists, with strictly increasing sequences on both sides of the valley value, and the length of each increasing sequence is not less than 2, and the maximum value of the frame minus the valley value is not less than 3 original count units, then the sample number of the valley value sample is recorded as the road-level flag event bit. This allows the center position of the instantaneous release or dilution behavior to be used as an alignment anchor point, facilitating correspondence with rising or peak events of other roads. If none of the three determination rules are satisfied and the road is not recorded as an abnormal segment marker in the frame, then sample number 8 is used as the road-level flag event bit. Setting sample number 8 as a fallback option ensures that each frame has an alignable anchor point, thus enabling continuous operation of the phase-locked loop.

[0040] Within the same frame, there can be a maximum of three path-level flag event bits. To obtain a unique frame-level flag event bit, the following fusion order is performed: path-level flag event bits recorded as abnormal segments are first removed, and selection is made only from the remaining valid path-level flag event bits. The distance from each valid path-level flag event bit to sample number 8 is calculated, and the one with the smallest distance is selected as the frame-level flag event bit. Selecting the event bit closest to the reference position minimizes the number of subsequent cyclic shift steps and reduces perturbation to the waveform morphology. If there are ties in distance, a fixed priority is used for selection, with the priority order being heart rate signals over blood oxygen signals, and blood oxygen signals over acceleration signals. The fixed priority ensures that different implementations produce the same results in parallel situations, avoiding irreproducible random differences.

[0041] In each frame, the frame-level flag event bit is compared with sample number 8: if it is less than 8, it is early and needs to be cyclically shifted to the right; if it is greater than 8, it is late and needs to be cyclically shifted to the left; if it is equal to 8, no shift is needed. The number of cyclic shift steps is equal to the absolute value of the difference between the frame-level flag event bit and sample number 8. When the number of cyclic shift steps is not greater than 4, the cyclic shift is performed and the shifted frame is counted as a successful phase lock; when the number of cyclic shift steps is greater than 4, it is counted as a phase lock failure. Limiting the allowed number of shift steps to no more than 4 ensures that the alignment action does not cross most of the samples within the frame, maintaining the stability of the morphological symbols within the frame. If phase lock fails for 5 consecutive frames, a capture restart is triggered. The capture restart operation is to undo any temporary shifts in these 5 frames, maintain the original order, and re-enter the path-level flag event bit extraction and fusion process from the next second. Setting 5 consecutive frames as the trigger condition can distinguish between occasional noise and continuous loss of lock, avoiding overreaction to short-term disturbances. At the end of each second, a cross-frame continuity check is performed: if the number of successfully phase-locked frames within that second is not less than 48, then that second is recorded as a valid phase-locked second; otherwise, that second is recorded as an invalid phase-locked second and marked as a traceable segment. The determination of valid phase-locked seconds provides a quality boundary for subsequent entropy coupling statistics, avoiding the use of a large amount of lost-lock data in coupling symbol generation.

[0042] For all frames counted as successfully phase-locked, the shifted sample sequence is output to form the phase-locked alignment sequence. For frames that fail to phase-lock, the original sample sequence is retained and the phase-locking failure information is registered in a unified segment index table for subsequent backtracking segment processing and identification window quality judgment. The phase-locked alignment sequence concentrates key events around sample number 8 within each frame. The direct benefit of this is that it aligns key events from different paths and different seconds on a unified coordinate system, allowing subsequent morphological symbol generation to be completed near a fixed position, thereby improving judgment consistency.

[0043] On the locked alignment sequence, morphological symbols are generated frame by frame and path by path. The candidate set of morphological symbols includes rising spikes, falling spikes, narrow peaks, wide peaks, flat segments, and gaps. Specific rules are as follows: Narrow Peak: A unique peak sample exists; the peak value minus the median value of the frame is not less than 3 original count units; the half-width is calculated by searching from the peak to the left and right until the first position is no higher than half the difference between the peak value and the median value of the frame, and the number of samples between the two positions is not greater than 4. Narrow peaks are used to represent short-term concentrated changes and can stably fall near sample number 8 after alignment. Wide Peak: A unique peak sample exists; the peak value minus the median value of the frame is not less than 3 original count units; calculated using the same half-width method as narrow peaks, the number of samples is not less than 8. Wide peaks characterize slow and significant accumulation processes, providing stable support for the detection of energy transfer coupling. Rising Spike: No single peak exists, but there is a continuous rise, the number of continuous samples is between 3 and 6, and the final value minus the initial value is not less than 3 original count units. Rising spikes reflect the termination position of rapid ascents and are often associated with motion triggering of acceleration signals. Falling spikes: There are no single peaks or rising spikes, but there is a continuous decline with a continuous sample count between 3 and 6, and the difference between the initial value and the final value is not less than 3 original count units. Falling spikes reflect the termination position of rapid decay and are often associated with recovery or deceleration. Gaps: There is a unique valley sample; the difference between the maximum value and the valley value in this frame is not less than 3 original count units; the half-width is calculated as half the difference between the valley value and the maximum value in this frame, and the sample count is not greater than 6. Gaps correspond to short-term exhaustion or occlusion, facilitating mirroring with narrow peaks. Flat segments: The difference between the maximum and minimum values ​​within this frame is not greater than 2 original count units. Flat segments are used to represent steady-state or unchanging segments. The determination order and rules for morphological symbols are fixed according to the above items, and all implementations are executed in the same order to ensure consistent reproducibility across devices and versions.

[0044] The identification window is a continuous 512 frames. Statistics and detection are performed only within this window, without looking back at historical windows. Limiting statistics to a fixed-length identification window ensures timeliness and traceability, preventing outdated information from being included in current judgments. First, the occurrence frequency of each morphological symbol for each channel is counted within the identification window. If a certain morphological symbol for a particular channel occurs only once within the identification window, it is marked as a rare symbol. Marking uniquely occurring patterns as rare symbols helps highlight non-repeating events and reduces the masking of anomalous behavior by common patterns. Mirror symbol pair mapping is established: rising spikes and falling spikes form a pair, narrow peaks and gaps form a pair, and wide peaks and flat segments form a pair. Mirror mapping establishes pairs of opposite morphologies, facilitating pairing detection on symmetrical frames. Coupled symbols are generated frame-by-frame and pair-by-pair: Synchronous rare coupling: If at least two signals simultaneously exhibit rare symbols within the same frame, synchronous rare coupling is generated in that frame. This symbol highlights simultaneous and rare event combinations, enabling rapid exposure of multiple simultaneous anomalies. Complementary mirror coupling: Within the same second, two frames whose frame numbers sum to 65 are taken as a pair of symmetrical frames. If a pair of mirror symbols appears in each of these symmetrical frames from the same signal source, then complementary mirror coupling is generated on the latter frame. Mirror matching using symmetrical frames can link reverse changes within half the cycle, revealing the rhythmic structure. Energy transfer coupling: In four consecutive frames, if the acceleration signal in the first frame is a rising spike or a broad peak, the heart rate signal in the second frame is a narrow peak or a broad peak, the blood oxygen signal in the third frame is a gap, and the blood oxygen signal in the fourth frame is a flat segment, then energy transfer coupling is generated in the fourth frame. This sequence links mechanical triggering, cardiac response, and a short-term drop and recovery of blood oxygen into a fixed order, facilitating the identification of the physiological response chain induced by exercise. Steady-state rhythm coupling: Statistics are compiled for each second. If at least two of the three signals within that second are determined to be flat segments in 48 or more of the 64 frames, and the total number of narrow peaks used for commutation within that second is no more than one, then steady-state rhythm coupling is generated in the last frame of that second. This symbol is used to identify slight commutation under a stable rhythm, easily distinguishing it from abnormal coupling. Each generated coupling symbol, along with its corresponding second and frame number, is registered in a unified segment index table. This registration allows subsequent retrieval to locate the coupling trigger position without recalculation and establishes a one-to-one correspondence between the coupling symbol and the phase-locked alignment sequence. When a second is counted as an invalid phase-locked second, the coupling symbols generated within that second are not cleared but marked as a traceable segment for later processing. After a valid phase-locked second appears in a subsequent second, the traceable segment can be reviewed, its morphological symbol and coupling symbol recalculated, and the old record overwritten. This approach limits the judgment affected by loss of lock to local conditions and corrects it promptly, reducing the spillover of false triggers.When a capture restart is triggered after 5 consecutive frames, the starting point of the identification window remains unchanged. When the frame sequence number within the window advances, the undone shift records are skipped, and only the latest mark in the unified fragment index table is used to ensure that the index and content are consistent.

[0045] Within the identification window, coupling markers and corresponding abnormal segment markers are read incrementally by frame number, forming a continuous frame-level sequence. Each frame can be retrieved for its second and frame numbers, facilitating a one-to-one correspondence in a unified segment index table. Frames belonging to invalid phase-locked seconds are initially excluded from the final replacement decision, marked as frozen, and used only for placeholder and boundary checks, to be processed uniformly in the backtracking phase. The scanning order adopts a deterministic strategy from left to right and from long to short: at each starting frame number, the longest possible length is first attempted, then gradually shortened until the length is 8 frames. At the same length, each starting frame number is compared to ensure that different implementations obtain the same candidate order.

[0046] Complete duplication is determined by comparing the coupling tokens frame-by-frame. The comparison does not involve anomalous segment markers to avoid interference from quality markers in repeatability identification; anomalous segment markers are imported into the pattern instruction record and unified segment index table after successful replacement, without loss. A subsequence of a certain length is considered a completely duplicated subsequence if it appears at least twice at different starting frame numbers and its content is completely identical. If multiple feasible lengths exist at the same position, the longer subsequence is preferred; if the lengths are the same and the number of repeated groups is more than two, the group with the smaller starting frame number in the second occurrence is preferred, ensuring the replacement result is unique and reproducible. To avoid overlap conflicts, frames covered by longer subsequences are no longer used as starting frames for other subsequences in subsequent matching. This maximizes the replacement span, reduces the number of pattern instructions, and maintains the continuity of frame numbers.

[0047] For each defined completely repeating subsequence, a pattern instruction is generated. The pattern instruction is represented as a single record in the synchronous compression timing packet, with the record position taken from the starting frame number of the subsequence. The replacement rule is full-bit replacement: the original frame-by-frame coupling mark records within the subsequence are no longer listed separately, but are represented by a single pattern instruction record; abnormal segment markers are not lost and are retained in two places: the pattern instruction record in the synchronous compression timing packet carries the set of frame numbers containing abnormal segment markers within the subsequence, used to quickly determine the quality status of the area covered by the instruction; the unified segment index table retains the abnormal segment markers and corresponding frame numbers for each frame within the subsequence, facilitating the retrieval of the original frame-level quality information. Frames not covered by any pattern instructions are still retained as single-frame records in the synchronous compression timing packet, recording the frame number, coupling mark, and abnormal segment marker. The unified segment index table registers the mapping relationship between each pattern instruction and the starting and ending frames of the original sequence. Registration uses a combination of second and frame numbers to ensure direct location during cross-window retrieval, without relying on additional context.

[0048] The synchronous compression timing packet and the unified fragment index table must be completely consistent in terms of frame sequence number: the start and end frame numbers of any mode instruction record must be consistent with the mapping relationship between mode instructions and the start and end frames of the original sequence in the unified fragment index table. The frame number of any single frame record in the synchronous compression timing packet must be able to locate the same frame-level record in the unified fragment index table. When a change in the adjacency relationship across subsequences occurs after a replacement, only the adjacent references of the affected records are updated; the confirmed start and end frame numbers are not changed to avoid cascading rewrites causing index jitter.

[0049] Freeze: When an invalid phase-locked second occurs within the identification window, frames within that second are marked as frozen. Frozen frames do not participate in the final confirmation and mode instruction replacement of completely repeating subsequences, but are still retained in the frame-level sequence for length and boundary determination. This avoids misleading replacement results due to lost-lock data and does not disrupt the continuity of frame sequence numbers. Backtracking Trigger: When a valid phase-locked second subsequently occurs, backtracking is immediately performed on the sequence number of the second involved in the most recent freeze. The backtracking range covers all frames within that second and adjacent, potentially affected consecutive frames, with the upper limit not exceeding the identification window boundary. Backtracking Recalculation: The morphological symbols and coupling symbols of the frames within the backtracking range are recalculated, and the complete repeating subsequence discovery and replacement process is re-executed. If the new result differs from that before freezing, overwriting is initiated. Overwriting: Based on the result obtained from the latest backtracking recalculation, the record at the corresponding position in the synchronous compressed timing packet is replaced, and the mapping relationship between mode instructions and the start and end frames of the original sequence is updated in the unified fragment index table. The overwriting time and overwriting reason fields are retained for subsequent auditing. Overlay only affects the backtracking range and does not change records outside the window, maintaining overall index consistency. The advantages are that freezing explicitly isolates unreliable data segments, backtracking limits updates to the minimum necessary range, and overlay ensures index and content synchronization, preventing inconsistencies between the frame range pointed to by the pattern command and the index.

[0050] When two completely repeating subsequences partially overlap and have different lengths, the longer subsequence is replaced first; if the lengths are the same, the one with the smaller starting frame number is selected first. This rule allows different implementations to achieve the same compression result. When a completely repeating subsequence appears multiple times, and the number of occurrences exceeds two, all occurrences are replaced with the same pattern instruction, and a unified fragment index table records multiple start and end frame intervals corresponding to this pattern instruction. This allows multiple occurrences of the same pattern to be covered in a single registration, reducing the number of records and facilitating batch retrieval. When a subsequence crosses the boundary of the recognition window, it is not considered a valid candidate to avoid partial matching caused by window sliding; the subsequence only participates in the comparison when it completely falls within the recognition window. Performing replacement in subsequences with abnormal fragment markers does not change the abnormal fragment marker status of any single frame. The pattern instruction record only integrates and presents these frame numbers, and the recognition process can still use them for quality judgment.

[0051] The frame number recorded in the mode command can be the starting frame number of the subsequence or a fixed reference position within the subsequence, such as the middle frame. As long as the mapping relationship between the mode commands in the unified fragment index table and the starting and ending frames of the original sequence is complete, the retrieval and backtracking functions are equivalent. The triggering time for backtracking recalculation can be set at the last frame of the recognition window or immediately after detecting a valid phase-locked second. The two methods are consistent in the final synchronized compressed timing packet; the only difference is the output delay.

[0052] Step 3: Within the recognition window, based solely on the coupling symbols, mode instructions, and abnormal fragment markers in the synchronous compressed timing packet and the unified fragment index table, output the health status label according to the mutual exclusion priority rule.

[0053] Specifically, the synchronously compressed timing packets within the identification window are read, yielding two types of records: single-frame records and pattern instruction records. Single-frame records contain frame numbers, coupling symbols, and aberration markers. Pattern instruction records provide the starting frame number and coverage area, and establish a one-to-one mapping with the start and end frames of the original sequence in the unified fragment index table. To perform statistics on a continuous coordinate system, the frames within the identification window are rearranged according to their second and frame numbers in window order, starting from frame number 1 at the first second and proceeding sequentially to frame number 64, then moving to the next second until the identification window is covered. This approach ensures that all statistics and interval calculations are performed on a continuous sequence, avoiding duplication or omissions across seconds. For pattern instruction records, the mapping relationship provided by the unified fragment index table is used to restore the covered frame number set according to the coverage area, and these are treated equivalently to single-frame records during statistics. This maintains the storage advantages of compression while ensuring that the compressed coupling symbols and aberration markers are not missed during statistics.

[0054] **Phase-locked Effective Seconds Statistics:** The valid or invalid phase-locked seconds identifier corresponding to each second number within the identification window is read from the unified segment index table. The number of valid phase-locked seconds is accumulated for determining insufficient signal quality. **Abnormal Segment Marking Statistics:** All frame numbers are traversed in the synchronous compressed timing packet. For a single frame record, the abnormal segment mark status of that frame is directly read and accumulated. For mode command records, abnormal segment marks for each covered frame are extracted and accumulated according to the mapping relationship. The accumulated result is used for determining insufficient signal quality. **Coupled Mark Occurrence Location Set:** Four sets are constructed to store the frame numbers of synchronous rare coupling, complementary mirror coupling, energy transfer coupling, and steady-state rhythm coupling within the window, respectively. The frame numbers in the sets adopt window order for easy subsequent interval and coverage statistics. **Frame Quadrant Coverage Set:** For each synchronous rare coupling occurrence frame, the frame number within its second is obtained and mapped to a frame quadrant. Frame quadrants are defined as 1 to 16, 17 to 32, 33 to 48, and 49 to 64. The mapping result is added to the coverage set for determining abnormality indication. Complementary Mirror Coupling Interval Sequence: Arrange the window sequence frame numbers where complementary mirror coupling occurs in ascending order, calculate the sequence difference between two adjacent occurrences to form an interval sequence, used for determining the tendency of resonance instability. To avoid ambiguity caused by spanning seconds, the interval calculation is performed on the window sequence, without needing to handle second boundaries separately. Energy Transfer Coupling Existence: Check if the energy transfer coupling set is empty; simultaneously check if the occurrence count of rare synchronous couplings within the identification window is equal to 0, used for determining operating load. Steady-State Rhythmic Coupling Coverage: Check if each second number included in the identification window generates steady-state rhythmic coupling in its last frame. To this end, check second by second whether the last frame of that second records steady-state rhythmic coupling. If the identification window spans multiple second numbers, all second numbers are required to meet this condition.

[0055] First, determining insufficient signal quality allows for the timely exclusion of windows marked with invalid phase-locked loops or abnormal segments, reducing false detections caused by subsequent coupling markers. Binding anomaly alerts to frame quadrant coverage enables rapid identification of rare synchronization events occurring in different phase intervals, providing strong localization and facilitating subsequent tracking. The requirement for fixed frame interval repetition in complementary mirror coupling helps to isolate approximately periodic alternating behaviors from random pairings. Simultaneously requiring energy transfer coupling and a rare synchronization coupling count of 0 for operational loads eliminates abnormal interference and isolates motion-induced coherent responses. Standby stability requires the generation of steady-state rhythmic coupling every BeiDou second, ensuring window-level stability rather than local stability; standby stability characterizes resting stability.

[0056] The statistical order can be implemented using a single traversal: during a single window scan, the effective seconds of phase-locked coupling, the number of abnormal segment markers, the location of various coupling symbols, and frame quadrant coverage are simultaneously accumulated, and the results are determined according to mutual exclusion priority after the scan. This method yields the same results as multiple traversals, but with lower execution overhead. For the interval calculation of complementary mirror coupling, a binary sorting of second and frame numbers can be used, then mapped back to the window order for difference calculation; both methods yield consistent interval results. When the identification window is precisely at the boundary of the synchronous compressed timing packet, if part of the frame interval covered by the mode instruction falls outside the window, only the set of frames falling within the window is taken during statistics. This avoids false duplications or omissions caused by boundary truncation. After outputting the health status label, if the identification window slides to the next set of frames, the registration content of the previous window remains unchanged and is not modified repeatedly in subsequent windows. This ensures that the judgment and registration of each window are independent, facilitating playback and auditing.

[0057] The identification window covers eight consecutive BeiDou seconds, numbered from 1000 to 1007. Each BeiDou second contains 64 phase grid frames, numbered from 1 to 64; each phase grid frame contains 16 samples, numbered from 1 to 16. For ease of description, a time positioning function is defined. ;in Indicates the second sequence number. Indicates the frame number. This indicates the sample point number. This function is only used to specify the absolute position of the sample point on a unified time baseline; all acquisition and judgment are based on the alignment label. and For accuracy. The counting baseline for the three signals is set as follows: heart rate signal baseline. baseline of blood oxygenation signals Acceleration-type signal baseline .here This represents the typical stationary value under the three raw counting units. In each... and Below, the 16 target times of the sample grid are given by alignment labels. Resampling is performed using the "closest to date principle": for each target time, the original sample point with the smallest time distance within that second is found and its count value is taken; if the two points are equidistant, the earlier timestamp is taken.

[0058] To avoid misclassifying "flat segments" as anomalous segments, within frames where no events occur, the three signals are analyzed in a deterministic sequence. Small fluctuations of no more than 2 occur nearby. For example, in a frame without events, heart rate signals can be analyzed by sample number (odd or even). Alternating between them; blood oxygenation signals and acceleration signals similarly alternate near their respective baselines, thus satisfying the "flat segment" criterion and not triggering the abnormal segment marking condition of "maximum and minimum values ​​being equal". In the unified segment index table, for each Register the alignment tags (second number and frame number) and the abnormal segment marking status of the three signals. Upon inspection, the number of abnormal segment marked frames in this example is 0, which meets the requirements for subsequent quality assessment.

[0059] Between frame numbers 20 and 23 (second number 1003), a set of cross-path events is constructed to trigger energy transfer coupling; other frames maintain a "flat segment" or normal minor fluctuations. Second number 1003, frame number 20 (accelerometer-type signal "sudden rise"); the 16 sample points of the acceleration-type signal are... A continuous increase occurred between sample point numbers 3 and 6, with a total of 4 sample points. The difference between the final sample point value and the initial sample point value was... According to the road-level determination rules for "progression" candidates, the road-level flag event bit is 6. Compared with reference sample point 8, it belongs to "early arrival," and the number of cyclic shift steps is... After shifting two steps to the right, phase locking is considered successful. Second sequence number 1003, frame sequence number 21 (heart rate signal "narrow peak"), the 16 samples of the heart rate signal are taken as follows: Calculate the median value of this frame. After sorting the 16 samples, the 8th and 9th values ​​were 522 and 523 respectively. The peak sample point number is 10, and the peak value is... The check should strictly decrease on both sides: 10 on the left side is... Gradually decreasing, to the right of 10 Gradually decreasing, all requirements are met. The half-amplitude threshold is... .

[0060] Searching from the peak outwards, the first sample point not exceeding 526.25 is found. The left side is sample point 9 (value 526), ​​and the right side is sample point 11 (value 526). The number of sample points between these two positions is 3, not exceeding 4, satisfying the "narrow peak" condition. Therefore, the road-level marker event bit is 10. Compared to reference sample point 8, this is considered "late," and the cyclic shift step is... After shifting left by 2 steps, phase locking was successful. Second number 1003, frame number 22 (blood oxygenation signal "gap"), the 16 samples of the blood oxygenation signal are... The valley value sample point number is 7, and the valley value is... The maximum value of this frame The check is strictly incremental on both sides: 7 is on the left. Gradually increasing, to the right of 7 is Gradually increasing. The half-amplitude threshold is... .

[0061] Starting from the valley value, search outwards for sample points with a minimum value of 407.5. On the left, sample point 6 (value 407, less than 407.5, ignored) and sample point 5 (value 408, greater than or equal to 407.5, used as the left boundary) are used. On the right, sample point 8 (value 407, less than 407.5, ignored) and sample point 9 (value 408, greater than or equal to 407.5, used as the right boundary) are used. The number of sample points corresponding to half the width is defined by the left and right boundaries. The value is no greater than 6, satisfying the "gap" condition. Therefore, the road-level marker event bit is 7. Compared to reference point 8, it is considered "early," and the cyclic shift step count is... Shift 1 step to the right in a circular fashion, phase locking successful. Second number 1003, frame number 23 (blood oxygenation signal "flat segment"): Blood oxygenation signal surrounds within this frame. Make fluctuations with an amplitude no greater than 2, for example If the difference between the maximum and minimum values ​​is no greater than 2, it is judged as a "flat segment". When none of the three paths have a path-level flag event bit in the frame, the sample number 8 is used as a fallback, and the frame-level flag event bit is 8, without the need for cyclic shifting.

[0062] For each second number within the identification window, the number of successfully phase-locked frames is counted. In this example, the number of successfully phase-locked frames for each of the 7 second numbers is not less than 60, counted as 7 valid phase-locked seconds; the number of successfully phase-locked frames for the other second number is 46, counted as invalid phase-locked seconds and marked as a traceable segment. Since the number of valid phase-locked seconds is 7, which is not less than 6, the quality access requirement of step 3 is met. On the phase-locked alignment sequence after the cyclic shift is completed, the morphological symbol is determined frame by frame and path by path. The above 4 key frames are respectively: the acceleration signal of frame 20 is "rising spike", the heart rate signal of frame 21 is "narrow peak", the blood oxygen signal of frame 22 is "gap", and the blood oxygen signal of frame 23 is "flat segment". Other frames are judged as "flat segments" based on the fluctuation of the difference between the maximum and minimum values ​​within the sample points not exceeding 2.

[0063] Within the recognition window, the occurrence frequency of each morphological symbol for each channel is counted. In this example, each channel's "narrow peak, gap, and rising spike" are arranged to appear at least twice, so there are no morphological symbols that appear only once, thus preventing the triggering of "synchronization rare coupling" in any frame. A fixed sequence is checked at frame numbers 20 to 23 of second number 1003: the first frame shows an acceleration signal as a "rising spike," the second frame shows a heart rate signal as a "narrow peak," the third frame shows a blood oxygen signal as a "gap," and the fourth frame shows a blood oxygen signal as a "flat segment," satisfying the sequence condition for energy transfer coupling, and generating "energy transfer coupling" at frame number 23. "Steady-state rhythm coupling" is counted at the end of each second: in this example, 6 second numbers satisfy the condition that "at least 2 channels are judged as flat segments in 48 or more of the 64 frames, and the narrow peak used for commutation is no more than once," and "steady-state rhythm coupling" is generated at the end of these seconds; the other 2 second numbers do not meet this condition. Checking "complementary mirror coupling" on symmetrical frames: This example only constructs one occurrence and does not form more than two near-equal interval repetitions, so it does not meet the interval requirement of "resonance instability tendency". Resonance instability tendency characterizes breathing instability tendency.

[0064] To illustrate synchronous compression, a specific repetition is introduced: in frames 1 to 8 of second sequence 1000 and frames 1 to 8 of second sequence 1001, the coupling symbol distribution is identical, with no coupling symbols in the first 7 frames and "steady-state rhythmic coupling" in the last frame. Assume these two subsequences are 8 frames long and repeat twice, satisfying the condition of "length not less than 8 and occurrence not less than 2". This subsequence is replaced by a single mode instruction: in the synchronous compression timing packet, the mode instruction is recorded at the position of second sequence 1000, frame sequence 1, and the mapping between this mode instruction and the start and end frames of the two original sequences is registered in the unified fragment index table. arrive and arrive The mode command also carries a set of abnormal segment marker frame numbers within these 16 frames (the set is empty in this example) for quality backtracking. Frames not covered by any mode command still appear as single-frame records in the synchronous compression timing packet.

[0065] The number of effective seconds of phase-locked loop is .here This indicates the number of seconds counted as valid phase-locked within the identification window. The number of frames marked with abnormal segments is... .here This indicates the number of frames in which any of the three paths within the identification window is marked as an anomalous segment. The number of times energy transfer coupling occurs is... This occurred at second number 1003 and frame number 23. This represents the count of energy transfer couplings within the identification window. The number of times rare synchronization couplings occur is... .here This indicates the count of rare synchronous couplings within the identification window. Complementary mirror coupling occurs when the frame sequence set size is 1, and no interval sequence is formed between two adjacent frames, failing to satisfy the requirement that "the interval between two adjacent frames is between 7 and 9 frames". Steady-state rhythmic coupling covering all second sequences does not exist because two second sequences did not generate the symbol in the last frame.

[0066] Judge each item in a predetermined order: Insufficient signal quality: Judgment condition is... or Here , If not met, proceed to the next condition. Error message: Requires at least two rare coupling operations and coverage of at least two frame quadrants. [Here] This condition is not met. Resonant instability tendency: requires the interval between two adjacent occurrences of complementary mirror coupling to be between 7 and 9 frames. Here, it only occurs once, which is not met. Operating load: requires at least one energy transfer coupling and the number of synchronous rare coupling occurrences to be equal to 0. Here... , If the conditions are met, the output health status label is "Operating Load". Subsequent rules will not be checked. The registration and identification window in the unified fragment index table starts and ends at second numbers 1000 to 1007. The coupling mark frame number that triggered the judgment is second number 1003 and frame number 23. The start and end frames of the involved mode commands are second number 1000 and frames 1 to 8, and second number 1001 and frames 1 to 8. The abnormal fragment marker count is 0, and the effective phase-locked seconds count is 7. The output health status label is "Operating Load", where operating load represents the actual motion load.

[0067] Figure 2This is an experimental curve graph of the core technology of the BeiDou monitoring terminal based on the time-series synchronous compression mechanism. The graph accurately demonstrates the complete technical process of phase-locked loop alignment, morphological symbol recognition, and synchronous compression coding of this invention through 64 consecutive frames of data within a single BeiDou second. The horizontal axis represents the phase grid frame number, ranging from 1 to 64 frames, corresponding to a complete BeiDou second, with each frame containing 16 resampled points. The vertical axis represents the heart rate signal amplitude, measured in heartbeats per minute, with a test range of 50 to 110 bpm. The right vertical axis represents the cumulative number of morphological symbols recognized, ranging from 0 to 50. The graph contains four core curves, respectively showing the technical processing of the original heart rate signal, phase-aligned signal, cumulative number of morphological symbols, and data compression ratio. The experiment was strictly performed according to the unified time baseline of the BeiDou timing signal, dividing the 64 frames into four processing stages according to quadrants. Frames 1-16 constitute the first quadrant, where the original heart rate signal exhibits obvious phase shift and random noise characteristics, with the signal amplitude fluctuating within the range of 60-95 bpm. The phase-locked loop begins extracting the marker event bits. By executing three judgment rules frame by frame, the system identifies the flag event bits for single-peak mode, rising surge mode, and gap mode, and merges them into frame-level flag event bits. Frames 17-32 are the second quadrant, where the phase alignment algorithm plays a core role. When the frame-level flag event bit is less than the phase-locked reference position 8, the system performs a rightward cyclic shift; when it is greater than the reference position, it performs a leftward cyclic shift. The phase alignment signal curve shows that after cyclic shift correction, the phase jitter of the signal is significantly reduced, and the waveform regularity is significantly enhanced, proving the effectiveness of the phase-locked loop of this invention. Frames 33-48 are the third quadrant, where morphological symbol recognition enters the rapid accumulation stage. The system analyzes and identifies the phase-locked alignment sequence of each frame according to the judgment rules for six types of morphological symbols: narrow peak, wide peak, rising surge, falling surge, gap, and flat segment. The cumulative number of morphological symbols shows a linear growth trend, increasing from 12 symbols in frame 20 to 28 symbols in frame 40, with the recognition efficiency stabilizing at approximately 0.8 symbols per frame. Frames 49-64 are in the fourth quadrant, where synchronous compression coding achieves optimal results. The data compression ratio drops significantly from the initial 100% to around 15%, indicating that the algorithm successfully identified identical subsequences with a length of at least 8 frames and an occurrence frequency of at least 2, and performed replacement compression using pattern instructions. The sharp decrease in compression ratio verifies the technical advantages of this invention in data storage and transmission efficiency. Throughout the entire 64-frame processing, the phase-aligned signal consistently maintains higher stability and regularity than the original signal, the number of morphological symbols recognized shows a stable linear increase, and the data compression ratio achieves a significant optimization and improvement.

[0068] Figure 3This is an experimental curve showing the accuracy of health status classification based on the mutual exclusion priority rule. The curve verifies the classification performance and learning improvement capabilities of the four health status recognition algorithms of this invention through 20 consecutive tests. The horizontal axis represents the number of tests, corresponding to 20 independent recognition window numbers; the vertical axis represents the recognition accuracy, in percentage, with a test range of 65% to 100%. The curve contains four classification curves, showing the accuracy trends of four health statuses: stable standby, operating load, unstable breathing, and abnormal indication. The experiment strictly followed the mutual exclusion priority rule of this invention to determine the health status. The stable standby status showed the best recognition accuracy, achieving an initial accuracy of 92% based on the steady-state rhythm coupling mechanism, gradually improving to an excellent level of 98% as the tests progressed. This status requires steady-state rhythm coupling to be generated every second within the recognition window, meaning that at least two of the three signals are marked as flat segments in 48 or more of the 64 frames, and the total number of narrow peaks used for commutation does not exceed one. The judgment conditions are relatively clear, therefore the recognition accuracy is the highest and most stable. The accuracy of operational load identification showed a trend of first increasing and then stabilizing. Initially, the accuracy was around 85%, significantly improving during tests 6-15, and finally stabilizing at a good level of 93%. This state is based on energy transfer coupling judgment, requiring a specific sequence of acceleration spikes or broad peaks, heart rate narrow or broad peaks, blood oxygen gaps, and blood oxygen flat segments to appear sequentially within four consecutive frames. The complexity of pattern matching made initial identification difficult, but the accuracy steadily improved with algorithm learning and optimization. The accuracy of respiratory instability identification gradually improved from a baseline of 78% to 92%, demonstrating the adaptive learning capability of the algorithm. This state is based on a complementary mirror coupling recurring pattern, requiring the frame interval between two adjacent occurrences to be within the range of 7-9 frames. The timing requirements are strict, necessitating continuous optimization of parameter thresholds in practice, thus the accuracy improvement process was relatively long. The accuracy of anomaly alert identification showed the largest fluctuation, initially only 65%, but after algorithm training and parameter tuning, it finally reached a satisfactory level of 85%. This state is based on synchronous rare coupling determination, requiring that the occurrence frequency within the identification window be no less than 2 and cover at least 2 frame quadrants. The rarity and diversity of abnormal patterns make identification the most difficult. However, this invention sets it to the highest priority through a mutual exclusion priority rule, ensuring that abnormal situations can be identified and alerted in a timely manner. Experimental results show that the mutual exclusion priority rule of this invention can effectively distinguish four health states, and the identification accuracy of each state meets the requirements for practical application. The algorithm has good learning and improvement capabilities, verifying the technical advancement and application value of the health state identification method based on the time-series synchronous compression mechanism.

[0069] While specific embodiments of the present invention have been described above, those skilled in the art should understand that these specific embodiments are merely illustrative. Those skilled in the art can omit, substitute, and modify the details of the above methods and systems in various ways without departing from the principles and essence of the present invention. For example, combining the above method steps to perform substantially the same function and achieve substantially the same result according to substantially the same method falls within the scope of the present invention. Therefore, the scope of the present invention is defined only by the appended claims.

Claims

1. A method for monitoring the state of a device based on a timing-synchronous compression mechanism, characterized in that, The method comprises: Step 1: monitoring the terminal to collect vibration, acoustic and acceleration signals on the unified time baseline of Beidou timing signals, resampling according to the preset phase grid frame and sample grid, generating alignment labels and original time sequence data packets, recording abnormal segment markers, and registering second sequence numbers and frame sequence numbers in the unified segment index table; Step 2: inputting the original time sequence data packet, constructing a phase grid frame according to a reference phase track by a phase-locked loop, extracting road-level marker event bits frame by frame and fusing them into frame-level marker event bits, performing cyclic shift on the in-frame sequence according to early or late judgment, aligning the frame-level marker event bits to the reference sample point, and obtaining the phase-locked alignment sequence by combining the cross-frame continuity check and capture restart; then generating a morphology symbol according to a fixed judgment on the phase-locked alignment sequence, and generating a coupling symbol based on the identification window statistics and the second in-symmetry frame mapping by an entropy coupler; the coupling symbol includes synchronous rare coupling, complementary mirror coupling, energy transfer coupling and steady-state rhythm coupling; the second sequence number, the frame sequence number, the morphology symbol and the coupling symbol are written into the unified segment index table; the completely repeated subsequence in the coupling symbol is executed by synchronous compression with a mode instruction, forming a synchronous compression time sequence packet containing only the frame sequence number, the coupling symbol, the mode instruction and the abnormal segment marker; the phase-locked invalid second triggers freezing, backtracking and covering to keep the index consistent; in the same frame, at least two signals simultaneously appear rare symbols, generating synchronous rare coupling; in a pair of symmetric frames in the same second, if a pair of mirror symbols appears on the same signal source, complementary mirror coupling is generated; in the next 4 frames, if the following rules are met, energy transfer coupling is generated at the 4th frame position: the acceleration class of the 1st frame is ascending or wide peak; the vibration class of the 2nd frame is narrow peak or wide peak; the acoustic class of the 3rd frame is notch; the acoustic class of the 4th frame is flat section; statistics are taken every Beidou second: if at least two of the three signals are marked as flat section in 48 frames or more in the 64 frames in the second, and the total number of narrow peaks used for commutation in the second is not more than 1, then a steady-state rhythm coupling is generated at the end frame of the second; Step 3: outputting the device state label according to the mutually exclusive priority rule in the identification window only according to the coupling symbol, the mode instruction and the abnormal segment marker in the synchronous compression time sequence packet and the unified segment index table.

2. The device status monitoring method based on time-series synchronous compression mechanism as described in claim 1, characterized in that, The Beidou timing signal is counted by seconds; each Beidou second is divided into 64 phase grid frames, and the frame numbers are 1 to 64; each phase grid frame is uniformly resampled into 16 samples for each signal, and the sample sequence numbers are 1 to 16; the phase-locked reference position is fixed as the sample sequence number 8 in the frame; if the maximum and minimum values of the same signal in the 16 samples in the frame are equal, the frame is marked as an abnormal segment marker for the signal; the morphology symbol set is fixed as ascending, descending, narrow peak, wide peak, flat section and notch; the identification window is fixed as 8 consecutive Beidou seconds, 64 frames per second; in the same second, the two frames with a frame number sum equal to sixty-five are defined as a pair of symmetric frames; each record item in the unified segment index table is set as a second sequence number, a frame sequence number, a signal source, a morphology symbol, a coupling symbol, a mode instruction and an abnormal segment marker; the frame quadrant is defined as 1 to 16, 17 to 32, 33 to 48 and 49 to 64.

3. The device status monitoring method based on time-series synchronization compression mechanism as described in claim 2, characterized in that, In step 2, the process of extracting the flag event bit includes: in each frame, each signal is determined and corrected according to the following three determination rules: the first determination rule: if there is a single peak, and a strictly decreasing sequence appears on both sides of the peak top sample, the left side decreasing sequence length is not less than 2, the right side decreasing sequence length is not less than 2, and the peak top value minus the minimum value of the frame is not less than 3 original count units, then the peak top sample is taken as the flag event bit; the second determination rule: if there is no single peak satisfying the first determination rule, but there is an ascending mode candidate: that is, there is a continuous rise for 1 time, the continuous sample point number is in the interval of 3 to 6, and the end sample point value minus the start sample point value is not less than 3 original count units, then the end sample point of the ascending segment is taken as the flag event bit; the third determination rule: if there is a notch mode candidate: that is, there is a local valley value, a strictly increasing sequence appears on both sides of the valley value, the bilateral increasing sequence length is not less than 2, and the maximum value of the frame minus the valley value is not less than 3 original count units, then the valley value sample is taken as the flag event bit; if multiple determination rules are satisfied at the same time, the corresponding determination rule is determined and corrected according to the priority order of the first determination rule being prior to the second determination rule, and the second determination rule being prior to the third determination rule.

4. The device status monitoring method based on time-series synchronization compression mechanism as described in claim 3, characterized in that, If the three determination rules are not satisfied, the sample at sample point number 8 of the frame is taken as the flag event bit; there are at most 3 flag event bits in the same frame, which are taken as follows: the flag event bit which is not marked as an abnormal segment is taken first; if there are 2 or more, the one closest to the sample point number 8 is selected; if the distances are the same, a fixed priority is selected: the vibration type is higher than the acoustic type, and the acoustic type is higher than the acceleration type, to obtain the frame-level flag event bit of the frame.

5. The device status monitoring method based on time-series synchronous compression mechanism as described in claim 4, characterized in that, If the frame-level flag event bit is less than 8, it is defined as early arrival, and the frame sequence is circularly shifted to the right in the frame, and the shift number is 8 minus the position; if it is greater than 8, it is defined as late arrival, and circularly shifted to the left, and the shift number is the position minus 8; after the shift, the frame-level flag event bit is equal to 8, which is counted as phase locking success; if the required shift number is greater than 4, it is determined as phase locking failure; if phase locking fails for 5 consecutive frames, it triggers capture restart: the shift of the 5 frames is cancelled, and the original order is maintained; if the number of frames with phase locking success in the same second is not less than 48, the second is counted as a valid phase locking second, otherwise it is counted as an invalid phase locking second and is marked as a traceable segment as a whole.

6. The method of claim 5, wherein the time-synchronous compression mechanism is based on a time-synchronous compression mechanism. Step 2: For each frame and each signal after phase-locked alignment, output one shape symbol according to the following rules: narrow peak: there is only one peak sample point; the peak value minus the median value of the frame is not less than 3 original count units; the half width is defined as the first sample position from the peak top to left and right search that is not higher than half of the difference between the peak value and the median value of the frame, and the number of sample points between the two positions is not more than 4; wide peak: there is only one peak sample point; the peak value minus the median value of the frame is not less than 3 original count units; the peak width is calculated by the same half width method, and the number of sample points is not less than 8; rising spike: there is no single peak, but there is a single continuous rising, the number of continuous sample points is 3 to 6, including boundary 3 and 6, and the end value minus the starting value is not less than 3 original count units; falling spike: there is no single peak or rising spike, but there is a single continuous falling, the number of continuous sample points is 3 to 6, and the starting value minus the end value is not less than 3 original count units; notch: there is only one valley sample point; the maximum value of the frame minus the valley value is not less than 3 original count units; the half width is calculated as half of the difference between the valley value and the maximum value of the frame, and the number of sample points is not more than 6; flat segment: the difference between the maximum value and the minimum value in the frame is not more than 2 original count units. In the 512-frame recognition window, if the number of occurrences of a certain shape symbol of a certain channel is equal to 1, the shape symbol is recorded as a rare symbol; mirror symbol pair mapping includes: rising spike and falling spike form a pair; narrow peak and notch form a pair; wide peak and flat segment form a pair.

7. The method of claim 6, wherein the time-synchronous compression mechanism is based on a time-synchronous compression mechanism. In step 2, in the recognition window, find a completely identical subsequence with a length of not less than 8 frames and a number of occurrences of not less than 2 from the coupled symbol sequence; replace each repetition with a mode instruction, and write the mapping relationship between the mode instruction and the original sequence start and end frames in the unified segment index table; output as a synchronous compression timing package, including frame number, coupled symbol, mode instruction and abnormal segment marker.

8. The method of claim 7, wherein the time-synchronous compression mechanism is based on a time-synchronous compression mechanism. Step 3 specifically includes: in each recognition window, only according to the coupled symbols, mode instructions and abnormal segment markers in the synchronous compression timing package and the unified segment index table, the following mutual exclusion priority judgment is performed: if the number of phase-locked valid seconds in the recognition window is less than 6 or the number of frames with abnormal segment markers is greater than 64, it is judged that the signal quality is insufficient; if the number of occurrences of synchronous rare coupling in the recognition window is not less than 2 and covers at least 2 frame quadrants, it is judged that there is an abnormal prompt; if there is a repeated occurrence of complementary mirror coupling in the recognition window, and the frame interval between the two adjacent occurrences is both within 7 to 9 frames, it is judged that there is a resonance instability tendency; if there is at least one energy transfer coupling in the recognition window, and the number of occurrences of synchronous rare coupling in the window is equal to 0, it is judged that the running load; if a stable rhythm coupling is generated for each Beidou second in the recognition window, it is judged that the standby is stable.

9. The device status monitoring method based on time-series synchronous compression mechanism as described in claim 8, characterized in that, ​

Images