Pelvic floor muscle training system based on mobile terminal

Abstract

The invention relates to the technical field of intelligent health management, in particular to a pelvic floor muscle training system based on a mobile terminal, which comprises a signal collaborative acquisition module, a multi-parameter linkage recognition module, a three-section action decomposition module, a time sequence voice guide module and an execution deviation evaluation module. According to the method, through synchronous time sequence analysis of pelvic floor electromyographic signals and heart rate signals in the training process, collaborative changes of signal mutation nodes are mined, multi-source physiological parameter linkage recognition and induction of action stage morphological characteristics are combined, training actions are disassembled into multi-stage processes, and time sequences of signal nodes and voice guidance are dynamically mapped; precise butt joint of step-by-step voice instructions and action states is achieved, performance of each stage of the training process is quantified into data features, training performance differences are visually evaluated through the time sequence deviation degree, scientificity and continuity of rehabilitation guidance are enhanced through a step-by-step feedback mechanism, and data basic support is provided for individual rehabilitation path adjustment and progress tracking.

Description

Technical Field

[0001] This invention relates to the field of intelligent health management technology, and in particular to a pelvic floor muscle training system based on a mobile terminal. Background Technology

[0002] Intelligent health management involves using wearable devices, mobile terminals, sensor networks, and health informatics to continuously monitor, collect, and analyze data on an individual's physiological parameters, behavioral habits, and lifestyles. This enables multi-level health management services, including disease prevention, health promotion, and rehabilitation guidance, and is widely applied in areas such as chronic disease management, sports rehabilitation, elderly care, and psychological adjustment. Traditional pelvic floor muscle training systems, on the other hand, refer to offline rehabilitation training services offered by hospital physical therapy departments. These services are primarily guided by rehabilitation therapists, who instruct patients to perform muscle contraction exercises. Visual cues, verbal guidance, and electrical stimulation devices are often used to repeatedly train the pelvic floor muscles. These services typically rely on fixed locations and dedicated personnel.

[0003] Existing technologies rely on manual observation and traditional instruments for pelvic floor training, which cannot automatically collect and link multiple physiological signals. The process of training movements cannot be broken down in detail, key nodes are difficult to identify accurately, and the performance of movements mainly relies on perception or static data review. Training performance lacks data-driven characteristics, differences between stages are difficult to quantify, the process feedback mechanism is fragmented, and rehabilitation training lacks continuous step-by-step assistance and scientific guidance. Data traceability and progress assessment are not consistent, making it difficult for users to obtain timely, detailed, and individualized rehabilitation adjustment support. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a pelvic floor muscle training system based on a mobile terminal.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a pelvic floor muscle training system based on a mobile terminal, the system comprising:

[0006] The signal collaborative acquisition module analyzes pelvic floor electromyography and heart rate signals based on the training process, determines the synchronous fluctuations of the two on the same time axis, compares the order of mutation occurrence, identifies synchronous mutation nodes and counts their number, and obtains the characteristics of synchronous mutation nodes.

[0007] Based on the characteristics of the synchronous mutation nodes, the multi-parameter linkage identification module filters signal synchronization nodes, compares the acquisition time of electromyography, photoelectric heart rate and triaxial displacement signals, analyzes the correspondence, judges the node linkage performance, and obtains a three-parameter linkage feature group.

[0008] The three-segment motion decomposition module, based on the three-parameter linkage feature group, determines the rising, maintaining, and falling phases of the displacement signal. Through continuous analysis, it compares the signal changes in the different phases, summarizes the differences in each phase, and obtains the motion segmentation morphological characteristics.

[0009] Based on the action segmentation morphological features, the timing-based voice guidance module compares signal nodes with voice prompt triggering points, analyzes key nodes with voice command response timing, filters nodes with the best response differences, establishes corresponding relationships, and obtains voice guidance timing features.

[0010] The present invention is improved in that the synchronous mutation node features include the node occurrence time, synchronous distribution type, and associated signal number; the three-parameter linkage feature group includes multi-signal coupling type, event discrimination category, and linkage amplitude identifier; the action segmentation morphology features include action structure parameters, segmentation morphology category, and action cycle marker; and the voice guidance timing features include trigger signal type, segmented voice nodes, and guidance timing annotation.

[0011] The present invention is improved in that the signal cooperative acquisition module includes:

[0012] The temporal fluctuation analysis submodule analyzes the collected pelvic floor electromyography and heart rate signals based on the training process, identifies the continuous change segments of each signal during the training process, calculates the rate of change of each signal segment, judges the trend change of the signal on the time axis, optimizes the location of signal inflection points, and obtains the temporal change trend quantity.

[0013] The mutation relationship judgment submodule, based on the time-series change trend, filters key nodes with obvious signal jumps, determines the order of occurrence of various signal mutations, analyzes the primary and secondary relationships of mutation occurrences, and calculates the time difference between nodes to obtain mutation priority information.

[0014] Based on the mutation priority information, the synchronization node identification submodule filters nodes with similar time differences, calculates the number and distribution of synchronous mutation nodes, compares the synchronization changes of each signal type, analyzes the coupling characteristics of signals in each time period, and obtains the characteristics of synchronous mutation nodes.

[0015] The present invention is improved in that the multi-parameter linkage recognition module includes:

[0016] Based on the characteristics of the synchronous mutation nodes, the signal synchronization extraction submodule analyzes the collected electromyographic signals, heart rate signals, and displacement signals, determines the synchronous change trend of each signal on the time axis, and filters the signal combinations that change synchronously within the same time period to obtain the set of synchronous nodes.

[0017] The node matching and discrimination submodule, based on the set of synchronous nodes, compares the direction of change and fluctuation of each signal within the recording period, judges the consistency of the performance of the different signals on the same node, analyzes the correspondence between time series and trend, and obtains the node coupling distribution group.

[0018] The linkage feature generation submodule, based on the node coupling distribution group, judges the combination of changes in each signal, analyzes the synchronicity and linkage features of the signals at key nodes, filters out the related change types, and obtains a three-parameter linkage feature group.

[0019] The present invention is improved in that the three-segment action decomposition module includes:

[0020] The stage identification submodule analyzes the coordinated performance of each signal based on the three-parameter linkage feature group, determines the rising, stable and falling trends of the displacement signal curve in different time periods, optimizes the start and end nodes of stage division, filters effective intervals, and obtains the stage boundary time series.

[0021] The signal comparison submodule compares the displacement signal and electromyographic signal trajectory at each stage based on the stage boundary time series, analyzes the fluctuation amplitude and consistency of the two types of signal trajectories at each stage, calculates the fluctuation difference between the trajectories at each stage, and obtains the stage signal difference amplitude by collecting the fluctuation differences.

[0022] Based on the difference amplitude of the stage signals, the morphological induction submodule determines the key stage number, analyzes the change process of the displacement signal and the electromyographic signal response of the stage, summarizes the difference characteristics of the three stages, and obtains the morphological characteristics of the action segmentation.

[0023] The present invention is improved in that the time-series voice guidance module includes:

[0024] The voice node alignment submodule, based on the action segmentation morphological features and the corresponding time point of the voice prompt, compares the sequence of each stage signal node and the voice prompt in terms of timing, and selects signal nodes with similar times to pair with the voice prompt to obtain a node alignment time interval group.

[0025] The response difference calculation submodule compares the time interval between each pair of signal nodes and the voice prompt based on the node alignment time interval group, calculates the deviation of the synchronous change of signal and voice, and obtains the average response difference amplitude.

[0026] Based on the average response difference amplitude, the guidance matching construction submodule optimizes the pairing results of signal nodes and voice prompts at each stage, selects the node combination with the best pairing effect, establishes a one-to-one correspondence between signal nodes and voice prompts, and obtains the voice guidance timing features.

[0027] The present invention has an improvement, wherein the system further includes:

[0028] The execution deviation assessment module calculates the difference between the voice guidance timing characteristics and the start and end times of each stage of the user's actual actions, analyzes the execution time, weights the stage differences, judges the consistency with the target, and obtains the execution deviation index.

[0029] The execution deviation indicators include the time sequence deviation level, action phase evaluation, and execution consistency grading.

[0030] The present invention is improved in that the deviation evaluation module includes:

[0031] The training period extraction submodule, based on the voice guidance timing features, matches the instruction time points marked by the voice prompts with the start and end nodes of the user's actions one by one. By judging the degree of overlap of nodes and the degree of temporal closeness, it extracts the start and end times of the corresponding actions in each stage to obtain the stage execution time period sequence.

[0032] The stage difference calculation submodule compares the stage execution time period sequence with the preset action timing arrangement, focuses on the distribution and differences of the start and end nodes of each stage action in the time sequence, and determines the timing error type of each stage by sorting out the offset performance of the nodes before and after, and obtains the time matching offset set.

[0033] The deviation degree summary submodule, based on the time matching offset set, determines the relationship between rhythmic features in the action structure parameters, analyzes the phenomenon of advance, lag and rhythm asynchrony in the actual action at the difference stage, and integrates the offset performance of all stages to obtain the execution deviation degree index.

[0034] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0035] In this invention, by synchronous temporal analysis of pelvic floor electromyography signals and heart rate signals during training, the coordinated changes of signal mutation nodes are explored. Combined with the linkage recognition of multi-source physiological parameters and the summarization of morphological characteristics of movement stages, the training movement is decomposed into a multi-stage process. The time sequence of signal nodes and voice guidance is dynamically mapped to achieve precise docking between step-by-step voice commands and movement states. The performance of each stage of the training process is quantified as data features. The differences in training performance are intuitively evaluated through temporal deviation. The step-by-step feedback mechanism enhances the scientific nature and continuity of rehabilitation guidance, and provides data foundation support for individual rehabilitation path adjustment and progress tracking. Attached Figure Description

[0036] Figure 1 This is a system flowchart of the present invention;

[0037] Figure 2 This is a flowchart of the signal collaborative acquisition module in this invention;

[0038] Figure 3 This is a flowchart of the multi-parameter linkage recognition module in this invention;

[0039] Figure 4 This is a flowchart of the three-stage action decomposition module in this invention;

[0040] Figure 5 This is a flowchart of the timing-based voice guidance module in this invention;

[0041] Figure 6 This is a flowchart of the deviation evaluation module in this invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0043] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0044] For examples, please refer to Figure 1 The present invention provides a technical solution: a pelvic floor muscle training system based on a mobile terminal includes:

[0045] The signal collaborative acquisition module analyzes the acquired pelvic floor electromyography and heart rate signals based on the training action process, judges their temporal fluctuations on the same time axis, compares the order of abrupt changes in the acquired signals, identifies nodes where abrupt changes occur simultaneously, and counts the number of synchronous abrupt change nodes to obtain the characteristics of synchronous abrupt change nodes.

[0046] The multi-parameter linkage identification module, based on the characteristics of synchronous mutation nodes, filters signal synchronization nodes, compares the acquisition time of Bluetooth synchronous electromyography acquisition device, photoelectric heart rate belt and triaxial displacement sensor, analyzes the correspondence of each type of signal under each node, judges the signal linkage performance between nodes, and obtains three-parameter linkage feature group.

[0047] The three-segment motion decomposition module is based on a three-parameter linkage feature group to determine the rising, maintaining, and falling phases of the displacement signal. By performing continuous analysis on the signal of each phase, comparing the signal change characteristics of each phase, and summarizing the performance of the different phases, the motion segmentation morphological characteristics are obtained.

[0048] The temporal voice guidance module compares the triggering time of each stage signal node with the triggering time of the voice prompt based on the action segmentation morphological features, analyzes the timing of each stage key node and the voice command response, and establishes a correspondence by selecting the node with the best difference between the signal and command timing response to obtain the voice guidance timing features.

[0049] The execution deviation assessment module calculates the difference between the voice-guided timing characteristics and the start and end times of each stage of the user's actual training actions, analyzes the actual execution time of each stage, and judges the consistency between the action execution and the target by weighting the differences of each stage, thus obtaining the execution deviation index.

[0050] Synchronous mutation node features include node occurrence time, synchronization distribution type, and associated signal number; three-parameter linkage feature group includes multi-signal coupling type, event discrimination category, and linkage amplitude identifier; action segmentation morphology features include action structure parameters, segmentation morphology category, and action cycle marker; voice guidance timing features include trigger signal type, segmented voice nodes, and guidance timing annotation; execution deviation index includes timing deviation level, action stage evaluation, and execution consistency grading.

[0051] In the signal collaborative acquisition module, pelvic floor electromyography (EMG) signals refer to the electrical activity signals of the pelvic floor muscles acquired in real time through surface EMG sensors, reflecting the physiological processes of contraction and relaxation of the pelvic floor muscles; heart rate signals refer to the heart rate signals acquired using ECG or photoelectric sensors, reflecting the activity state of the autonomic nervous system; temporal fluctuations refer to the dynamic process of changes in the intensity or value of each signal on the time axis, that is, the fluctuation of the signal over time; the sequence of mutations refers to which signal mutation occurs first in the signal acquisition, and by comparison, the primary and secondary causal relationships between different physiological events can be determined; the node of simultaneous mutation refers to the node where multiple signals mutate at the same time point or within a very short period of time (such as the synchronous change of muscle contraction and heart rate), and this node is used to mark the moment of key physiological events.

[0052] In the multi-parameter linkage recognition module, the Bluetooth synchronous electromyography (EMG) acquisition device refers to a dedicated device that collects EMG signals wirelessly via Bluetooth and transmits them synchronously to a mobile terminal; the photoelectric heart rate belt is a wearable device that uses photoelectric sensing technology (such as PPG) to measure heart rate, usually wrapped around the chest or wrist for real-time heart rate monitoring; the triaxial displacement sensor is a sensor used to detect displacement changes in three-dimensional space, reflecting physical motion information such as user movements and body positions; the mutual correspondence refers to the temporal correspondence and matching between various signals (EMG, heart rate, displacement) at the same event node; the signal linkage performance refers to whether multiple physiological signals change in synergy at the event node, and whether they show synchronization or a certain linkage pattern, used to judge the coupling characteristics between action and physiological state.

[0053] In the three-stage motion decomposition module, displacement signal refers to the physical signal collected by a triaxial displacement sensor that reflects changes in the user's body position or movement; signal change characteristics at each stage refer to the quantitative characteristics such as the continuity and rate of change of displacement or related physiological signals at different stages of motion execution (such as rising, maintaining, and falling); difference stage performance refers to summarizing the typical performance of each stage by comparing and analyzing the changes in signals at each stage, such as the difference between the force exertion stage and the relaxation stage.

[0054] In the temporal voice guidance module, each stage signal node refers to the physiological or displacement signal characteristic node at the start, end, and transition time points of each stage within the three-stage cycle of the action; the trigger time point refers to the specific time node at which the voice prompt module issues the prompt command, which must correspond to the signal node; the voice command response sequence refers to the actual temporal relationship of the user's action signal response after the voice prompt is issued, used to evaluate the timeliness of the guidance; the node with the optimal response difference refers to the key node with the smallest time difference and the closest coordination between the signal node and the voice prompt trigger time; establishing a correspondence means pairing the signal stage characteristics with the voice prompt time one by one to achieve the linkage between step-by-step voice guidance and action state; the voice guidance temporal characteristics are the time response data of the one-to-one pairing between "voice prompt" and "user's actual action node", used to achieve the quantitative evaluation of the accuracy, personalized adjustment, and training feedback of voice guidance, and to provide a basis for training data analysis and individualized rehabilitation path optimization.

[0055] In the deviation assessment module, the difference between the start and end times of each stage refers to the gap between the standard start and end times of the stage determined by the system and the time when the user actually completes the action stage; the actual execution time refers to the time taken by the user to actually complete each stage of the action, which can be calculated from the start and end nodes detected by the signal; weighted processing refers to the comprehensive analysis of the time differences of each stage according to a certain proportion, reflecting the degree of deviation in the overall action execution; target consistency refers to the system's evaluation of the consistency between the user's actual action and the preset target action in terms of time, rhythm, etc.

[0056] Please see Figure 2 The signal collaborative acquisition module includes:

[0057] The temporal fluctuation analysis submodule analyzes the collected pelvic floor electromyography and heart rate signals based on the training process, identifies the continuous change segments of each signal during the training process, calculates the rate of change of each signal segment, judges the trend change of the signal on the time axis, optimizes the location of signal inflection points, and obtains the temporal change trend quantity.

[0058] The system identifies the user's current training cycle, groups pelvic floor electromyography (EMG) signals and heart rate signals according to their sampling frequencies, and generates data segments corresponding to the time axis based on the start time of the movement. Each signal segment consists of multiple continuous time sampling points. After each signal sequence is input, the amplitude difference between adjacent sampling points is extracted, and the rate of change is calculated according to a fixed time interval. Subsequently, a rate sequence graph is built according to the time series, and the continuity of the change trend is analyzed. If three or more sampling points change in the same direction, i.e., the values ​​all increase or decrease, then the segment is marked as a continuous change segment. For example, in a Kegel exercise, if a user's EMG signal amplitude continuously increases between the 1.0 second and 1.6 second of training, and the amplitude of change at each sampling point is relatively stable, then this segment is considered a continuous change segment. The "rising segment" is then defined as follows: the range of changes in each segment is statistically analyzed. If the difference exceeds a specific threshold, it is marked as a "rapid change segment"; otherwise, it is marked as a "slow change segment." For example, if the total amplitude difference of the electromyographic signal in the rising segment reaches 1.2 and the duration is less than 0.6 seconds, then the segment is identified as a rapid rising segment. Subsequently, the trend signs of adjacent changing segments are compared. If the preceding segment rises and the following segment falls, and the difference between the end and the beginning of the two segments meets the inflection point determination requirements, then the sampling point at the boundary between the segments is marked as the signal inflection point. During implementation, if the electromyographic signal reaches its peak at 1.6 seconds and then immediately falls with a significant drop, then the inflection point is precisely located at 1.6 seconds. All identified changing segments and inflection point times are recorded in the database as the temporal trend output of the training process.

[0059] The mutation relationship judgment submodule filters key nodes with obvious signal jumps based on the time-series change trend, determines the order of occurrence of various signal mutations, analyzes the primary and secondary relationships of mutations, and calculates the time difference between nodes to obtain mutation priority information.

[0060] All trend segments and signal inflection points are extracted from the recorded electromyography (EMG) and heart rate signals. The magnitude difference before and after each change segment is calculated to determine if it constitutes a mutation event. If the magnitude exceeds a set threshold, it is marked as a mutation node. For example, if the EMG signal changes by 1.5 at 1.8 seconds and the heart rate signal changes by 0.9 at 1.85 seconds, both exceeding the corresponding mutation identification threshold, they are identified as EMG and heart rate mutation nodes, respectively. Timestamps are then extracted from the mutation nodes. Each type of mutation event is sorted by time and the time difference is compared. For example, if the time difference between the two nodes mentioned above is 50 milliseconds, it is identified as a time event. Related events are then classified into primary and secondary signals based on the magnitude of changes at each node. Nodes with larger magnitude changes are considered primary signals, while those with smaller magnitudes are secondary signals. In this example, electromyography (EMG) signals are primary and heart rate signals are secondary, and this is labeled as an "EMG-dominant mutation event." The time difference, primary / secondary order, and signal source number of all event pairs will be included in the mutation priority information. If multiple similar nodes are identified, such as repeated EMG mutations occurring approximately 30 to 60 milliseconds before heart rate mutations during levator flexor training, with significant differences in magnitude, this pattern is classified as a "high-priority EMG-dominant event pattern," and a mutation priority information table for the entire training cycle is generated.

[0061] The synchronization node identification submodule, based on mutation priority information, filters nodes with similar time differences, calculates the number and distribution of synchronization mutation nodes, compares the synchronization changes of various signal types, and analyzes the coupling characteristics of signals in different time periods, using the following formula:

[0062] ;

[0063] Characteristics of synchronous mutation nodes ,in, Representing the The time interval between two signal abrupt changes in a group node. Representing the The number of signal types in the group The time standard deviation of each mutation point represents the time standard deviation of the mutation points. This represents the sum of the absolute values ​​of the time differences between the abrupt changes of each signal. This represents the total number of mutation node groups.

[0064] Synchronous mutation node characteristics refer to the characteristic statistical indicators formed when different physiological signals (such as pelvic floor electromyography signals and heart rate signals) change significantly almost synchronously in time during the same training period. This feature is used to quantitatively describe the synergistic situation of synchronous mutation of various signals during the training movement cycle, reflect the synchronicity and coupling tightness between different signals at key nodes, and is used for further analysis and evaluation of training movement quality, physiological response synchronization, etc. in subsequent stages.

[0065] Mutation node groups with time intervals within ±0.15 seconds were selected. The number of signal types involved in each mutation node group was counted, and the time interval between two mutation signals in each group was recorded. Time standard deviation between mutation points and the sum of the absolute values ​​of the time differences between each mutation signal. To ensure consistency in the dimensions of all participating items, non-time-dimensional data were normalized before being used in subsequent calculations. The number of signal types after normalization were as follows: Group 1 Group 2 Group 3 The normalized standard deviations are as follows: , , The sum of the normalized mutation time differences is: , , When performing multiple calculations using the formula, first calculate the numerator:

[0066] ;

[0067] Substitute the calculation terms into the denominator in sequence:

[0068] Group 1: ;

[0069] Group 2: ;

[0070] Group 3: ;

[0071] Take the average of the denominators: ;

[0072] Substituting into the principal equation, we get: ;

[0073] The results indicate that the calculated characteristics of synchronous mutation nodes... Within the synchronicity evaluation benchmark range After comparison, the result falls near the upper limit of the interval, indicating that the temporal distribution among the mutation nodes has a strong degree of concentration and signal coupling. This value reflects that the synchronous mutation of multiple signals at key nodes during training is relatively clear, the linkage between mutation events between signals tends to be consistent, and the response delay of signal variation is short. Therefore, this result can be used as a quantitative basis for the characteristics of synchronous mutation nodes and as a standard reference for speech guidance optimization and action phase decomposition strategies based on node synchronization in subsequent modules.

[0074] Please see Figure 3 The multi-parameter linkage recognition module includes:

[0075] The signal synchronization extraction submodule analyzes the collected electromyography, heart rate and displacement signals based on the characteristics of synchronous mutation nodes, determines the synchronous change trend of each signal on the time axis, and filters the signal combinations that change synchronously within the same time period to obtain the set of synchronous nodes.

[0076] The system retrieves a list of all previously identified synchronous mutation nodes. This list contains time stamps and amplitude changes in electromyography (EMG), heart rate, and displacement signals during training at different time points. The time axis of each signal type is standardized, aligning all signals within the same training cycle. The corresponding time precision is uniformly set to 0.01 seconds, and the nodes are arranged according to their time stamps. For each node, it checks whether the other two signal types also mutate within a window of ±0.05 seconds at that time point. If all three signal types exhibit amplitude changes exceeding their respective mutation identification thresholds within this window, the node is identified as a preliminary synchronization node. For example, if at 7.32 seconds, the EMG signal increases by more than 1.0, the heart rate signal decreases by 0.7, and the displacement signal shifts upward by 0.9, with time intervals all within 0.04 seconds, then this node is considered a preliminary synchronization node. The time point is marked as a synchronous mutation node. Then, all candidate synchronous nodes are further compared to determine whether the change direction of the three types of signals of each node is consistent. That is, whether the signals all show an upward, downward or unidirectional trend at that moment. For example, if electromyography and displacement increase at the same time, and heart rate decreases slightly but does not exceed the direction consistency threshold, it is marked as a "partially consistent" node. Only when the three types of signals meet the consistency judgment conditions in both the direction of change and the rate of change level are they included in the final set of synchronous nodes. The direction consistency judgment is based on the direction sign judgment, and the rate level difference does not exceed the upper limit of the set level difference as a reference. For example, the rate level is set to three levels: slow, medium speed, and fast. If two of the three types of signals are medium speed and one is fast, it meets the condition. Otherwise, the node is removed. Finally, the nodes that meet all the conditions are output as the set of synchronous nodes.

[0077] The node matching and discrimination submodule is based on the set of synchronous nodes. It compares the direction of change and fluctuation of each signal within the recording period, judges the consistency of the performance of different signals on the same node, analyzes the correspondence between time series and trend, and obtains the node coupling distribution group.

[0078] Synchronization nodes are extracted one by one, and the direction and amplitude of change of electromyography (EMG), heart rate, and displacement signals at that node are read during the recording period. Continuous change data segments of each signal within 0.2 seconds before and after the node are extracted, and the amplitude difference between the starting and ending points is compared to determine the trend. If the three signals all change in the same node segment in either an increasing or decreasing direction, it is marked as a consistent direction event. If one signal's direction differs from the other two, it is marked as a difference signal. The amplitude difference between the difference signal and the other two signals is calculated. If the difference is within a set matching error threshold (e.g., setting the tolerance for fluctuation difference between EMG and heart rate to ±0.4, and the measured difference is 0.35), it is considered a fluctuation match. The above process filters out event nodes where all signals exhibit consistent behavior at the same node. Simultaneously, it combines data such as directional consistency level, amplitude difference, and fluctuation correspondence into a comprehensive node feature group. Further analysis is performed on the start-up time difference of each signal's change trend at the node. If the start-up time is within 0.1 seconds, it is marked as high timing correspondence; if it exceeds 0.2 seconds, it is judged as low correspondence. This rule is applied to all synchronization nodes, forming three categories of indicators for each node: directional consistency, amplitude correspondence, and timing response difference. By setting comprehensive indicator thresholds, for example, if at least two of the three indicators of a node meet a set reference standard, the node can be judged as an effective coupled node. All nodes that pass this judgment are then grouped into the node coupling distribution group.

[0079] The linkage feature generation submodule is based on the node coupling distribution group, judges the combination of changes of each signal, analyzes the synchronicity and linkage features of the signal under key nodes, filters the related change types, and obtains the three-parameter linkage feature group.

[0080] The system analyzes the combination of changes in electromyography (EMG), heart rate, and displacement signals within each node. For each node, the system extracts the change path type of its signal combination within the corresponding time period. For example, the combination of increased EMG, decreased heart rate, and increased displacement is categorized as "contraction activation - autonomic regulation - movement propulsion." Each combination is recorded as an independent linkage mode number, and the frequency and time distribution of each combination within the complete training cycle are statistically analyzed to determine whether they are correlated. If a certain signal combination appears in the initial, middle, and final stages of training, and the interval between each appearance does not exceed a set time limit, such as 10 seconds, the combination is marked as a periodic linkage mode. Then, by calculating the consistency ratio of the three types of signal change directions, the amplitude synchronization rate, and the initial response difference, it is determined whether the combination meets the linkage judgment conditions. If at least two of the three items meet the set thresholds, such as a consistency ratio higher than 70% and an amplitude synchronization rate higher than 65%, it is considered to have correlated linkage characteristics. All node combinations that meet the above criteria are identified as three-parameter linkage events. Finally, a three-parameter linkage feature group is constructed and output to support the recognition of action segmentation structure and the assignment of voice guidance strategies in the subsequent training process.

[0081] Please see Figure 4 The three-stage motion decomposition module includes:

[0082] The stage identification submodule is based on the three-parameter linkage feature group to analyze the coordinated performance of each signal, determine the rising, stable and falling trends of the displacement signal curve in different time periods, optimize the start and end nodes of stage division, screen the effective intervals, and obtain the stage boundary time series.

[0083] Iterate through all linked feature nodes, extracting the trend sequences of electromyography (EMG), heart rate, and displacement signals within a 2-second range before and after each node. Establish fluctuation curve groups for the three types of signals according to the sampling time axis. Prioritize judging the numerical fluctuation of the displacement signal during this period, marking all upward trend segments as candidates for the rising phase, stable fluctuation segments as candidates for the maintenance phase, and continuously decreasing segments as candidates for the falling phase. The judgment criterion is whether the direction of the difference between consecutive sampling points remains consistent. Further screening of the trend nature is combined with the magnitude of the rate difference between each segment. For example, in a user's contraction action, if the displacement signal continuously rises between 2.1 and 2.9 seconds with an amplitude change exceeding 1.0, it is marked as a candidate segment for the rising phase. Immediately following, from 2.9 to 3.3 seconds, the signal fluctuation amplitude is less than 0.1, and the segment is judged as a candidate segment for the falling phase. The segment is defined as a stable segment. Then, the displacement signal continuously decreases from 3.3 seconds to 4.0 seconds, which is determined to be a declining segment. After identifying the trend of each segment, the consistency of the three types of signals in this segment is further compared. If the electromyographic signal rises synchronously and the accompanying change in the heart rate signal is within the threshold, the segment is confirmed as a valid segment. The start and end times of all valid segment nodes are extracted for segment boundary calculation. Then, all action cycles are numbered according to the start and end times. Each action cycle is divided into three segments. Then, segments whose amplitude fluctuation does not reach the set recognition benchmark value are screened out from all segments. For example, segments with fluctuation amplitude less than 0.3 or segment duration less than 0.4 seconds will not be included in the segment sequence. Finally, all segment numbers that meet the amplitude threshold and duration requirements are confirmed, and the start and end times are combined to form a segment boundary time series for output.

[0084] The signal comparison submodule compares the displacement signal and electromyographic signal trajectory at each stage based on the stage boundary time series, analyzes the fluctuation amplitude and consistency of the two signal trajectories at each stage, and uses the following formula:

[0085] ;

[0086] Calculate the fluctuation differences between trajectories at each stage By aggregating fluctuation differences, the amplitude of stage signal differences is obtained, where... Indicates the first Phase 1 A displacement signal trajectory Indicates the first The level of equilibrium of all displacement signal trajectories within a stage. Indicates the first Phase 1 The amplitude of the electromyographic signal, Indicates the first The number of samples for the phase displacement signal trajectory or electromyographic signal amplitude;

[0087] The difference in oscillation between trajectories refers to the degree of dispersion between the change amplitude of the user's displacement signal and the overall average performance during each stage of the action, which is used to quantitatively analyze the action. Combined with the response intensity of electromyographic signals within the same stage, it comprehensively reflects the synchronous oscillation and coordinated performance of different signals during the action.

[0088] The sampled raw displacement velocity and electromyographic amplitude signals were grouped according to stage numbers. Within each stage, the normalized signal sequence of each group was processed segmentally. Then, the dispersion between the normalized displacement signal sample and its equilibrium level within the stage was analyzed. The amplitude of change was measured using the trajectory squared difference method. At the same time, the absolute response level of the normalized electromyographic signal amplitude within the stage was superimposed. Based on this, a formula was used for calculation. Taking stage 2 as an example, its normalized sampled data is as follows:

[0089] Displacement signal sequence: ;

[0090] Electromyographic signal sequence: ;

[0091] First calculate :

[0092] ;

[0093] Next, calculate the sum of the squared differences in the numerator:

[0094] ;

[0095] ;

[0096] Calculate the absolute values ​​of the electromyographic response intensity in the denominator and add 1:

[0097] ;

[0098] Substitute into the formula to calculate:

[0099] ;

[0100] The result indicates that the calculated difference in trajectory fluctuations during stage 2 is... Compare this value with the volatility difference grading interval. If the set normalized volatility difference grading interval is:

[0101] Low volatility segment ( );

[0102] Mid-range fluctuation ( );

[0103] High volatility segment ( If the value is within the medium fluctuation range, it indicates that the displacement trajectory changes to a certain extent compared to the equilibrium level during this stage, but does not exhibit overly violent or extreme fluctuations. Combined with the continuous response of the electromyographic signal amplitude, it can be deduced that there are certain fluctuations in the movement rhythm during this stage, and the signal coordination is relatively obvious, but it is still maintained within the controllable range of the structure. Therefore, this value result can be used as one of the basic characteristic indicators for determining the stage type, to support the construction and classification basis of the stage signal difference amplitude.

[0104] The morphological induction submodule determines the key stage number based on the difference amplitude of the stage signal, analyzes the change process of the displacement signal and the electromyographic signal response of the stage, summarizes the difference characteristics of the three stages, and obtains the morphological characteristics of the action segment.

[0105] The start and end time ranges and corresponding displacement and electromyographic (EMG) signal changes for each stage are extracted from the identified stage sequence. First, the total change value, direction, and duration of the displacement signal within each stage are calculated. Then, the synchronization of EMG signal changes within that stage is statistically analyzed. For example, in the rising stage, if the total displacement signal change exceeds 1.5, the average rate is greater than 0.3 per second, and the duration is 1.2 seconds, this stage is recorded as a high-intensity change stage. Next, the fluctuation amplitude and initial response time of the EMG signal within the same time period are extracted to determine the response delay between it and the displacement signal. If the delay is less than 0.2 seconds and the EMG fluctuation amplitude exceeds 0.8, this stage is determined to be an enhanced EMG-displacement coupling segment. Finally, the heart rate signal is compared within this segment. If the direction of heart rate fluctuation is consistent with the electromyography trend and the amplitude change exceeds 0.5, the stage is marked as a three-parameter linkage salient segment. The feature summarization operation of each stage is completed in sequence, and all stages are archived according to three states: rising phase, maintenance phase, and falling phase. In each state, the corresponding displacement response amplitude range, average electromyography change value, and heart rate fluctuation amplitude are recorded to form a movement stage morphology database. The morphological features of the above three stages are arranged and combined to determine which combination of morphology appears most frequently in the training cycle, has the highest similarity of change path, and has a high degree of overlap of signal fluctuation features. The repetitive structure is confirmed as the core morphological feature of the movement. After marking and numbering it, the complete movement segment morphological feature output is generated.

[0106] Please see Figure 5The time-series voice guidance module includes:

[0107] The voice node alignment submodule is based on the action segmentation morphological features and the corresponding time point of the voice prompt. It compares the sequence of each stage signal node and the voice prompt, and selects signal nodes with similar time to pair with the voice prompt to obtain the node alignment time interval group.

[0108] The system calls upon the start and end times of each action segment and the corresponding list of signal nodes output by the three-segment action decomposition module. Then, it reads the preset sequence of prompt voice playback times from the voice prompt module. It compares the trigger time of each voice prompt with the time of each signal node in the action segment, recording the time interval between each voice prompt and all signal nodes within a 0.5-second range before and after it. If the absolute time difference between a signal node's time and the voice prompt within a certain segment is less than a set alignment tolerance threshold (e.g., 0.3 seconds), that signal node becomes a candidate pairing. All candidate pairings are sorted by time interval, and the signal node with the shortest time interval is extracted and paired with the voice prompt. For example, in the maintenance phase, the voice prompt might be at 4.2 seconds. The playback proceeds, and signal nodes at 4.05 seconds, 4.25 seconds, and 4.45 seconds are present in this stage. The time differences are calculated to be 0.15 seconds, 0.05 seconds, and 0.25 seconds, respectively. Since 0.05 seconds is the smallest, the signal node at 4.25 seconds is selected as the pairing node with the voice prompt. This process is repeated for each stage, generating a one-to-one pairing result between the signal nodes and voice prompts for each stage. The time interval information for all pairings is recorded as a node alignment time interval group. Each record in this group contains the stage number, signal node time, voice prompt time, and their time difference value. For example, the output content may indicate: Stage 2, signal node 4.25 seconds, voice prompt 4.2 seconds, interval 0.05 seconds. This data is used for subsequent analysis of the pairing matching degree.

[0109] The response difference calculation submodule, based on node alignment time interval groups, compares the time interval between each pair of signal nodes and the voice prompt, and calculates the deviation between the synchronous changes of the signal and the voice using the following formula:

[0110] ;

[0111] Obtain the average response difference magnitude ,in, This indicates the total number of pairs of signal nodes and voice prompts. Indicates the first Timing of signal nodes in the group, Indicates the first The timing of voice prompts within a group;

[0112] The average response difference amplitude refers to the overall dispersion of the temporal response differences between all signal nodes and their corresponding voice prompts. It reflects whether the responses of each group of signal nodes and voice prompts are synchronized and whether the deviation is large. The smaller the value, the closer the responses of the signal and the voice prompts are and the higher the synchronization; the larger the value, the more obvious the temporal difference between the two and the lower the synchronization. This indicator is used to measure the closeness of cooperation between action execution and voice guidance and is a core quantitative basis for evaluating the effectiveness of voice guidance.

[0113] The temporal response differences between each pair of signal nodes and the voice prompts were analyzed. The original acquisition time of each group of signal nodes and the trigger time of the corresponding voice prompts were obtained. Five pairs of paired node data were formed in the current training period. The original data are as follows:

[0114] The first group of signal nodes has a duration of 3.21 seconds, and the corresponding voice prompt is 3.20 seconds.

[0115] The second set consisted of 5.45 seconds and 5.41 seconds;

[0116] The third set consisted of 7.62 seconds and 7.59 seconds;

[0117] The fourth set consisted of 10.88 seconds and 10.86 seconds;

[0118] The fifth set consisted of 13.74 seconds and 13.69 seconds;

[0119] After normalizing all time data, the normalized signal time series are obtained as follows:

[0120] [0.276, 0.451, 0.63, 0.902, 1];

[0121] The normalized speech time series is:

[0122] [0.275, 0.445, 0.627, 0.9, 0.997];

[0123] Calculate the squared difference of the response time difference for each pair using the formula, and obtain:

[0124] Group 1 is ;

[0125] Group 2 is ;

[0126] Group 3 is ;

[0127] Group 4 is ;

[0128] Group 5 is ;

[0129] Sum the above results as follows: Substituting into the formula, we get:

[0130] ;

[0131] The results indicate that the current five groups of signal nodes and voice prompts exhibit a high degree of synchronization response at a unified time scale, and the calculated average response difference amplitude is [data missing]. The value is far below the upper limit of the temporal synchronization benchmark interval of 0.01, which belongs to the fully synchronized response interval. This means that by quantifying the degree of response deviation between the signal and the voice prompt, a criterion is provided for subsequent pairing relationship screening and optimization. Therefore, this result not only reflects the degree of temporal coupling between nodes, but can also be used as a pairing judgment condition in the guided matching construction submodule. Based on this value, the difference between each group and the mean deviation can be further compared, thereby deriving and forming the signal-voice one-to-one matching relationship supported by the average response difference amplitude.

[0132] The guided matching construction submodule optimizes the pairing results between signal nodes and voice prompts at each stage based on the average response difference amplitude, selects the node combination with the best pairing effect, establishes a one-to-one correspondence between signal nodes and voice prompts, and obtains the voice guidance temporal features.

[0133] Extract the time interval value of each pairing group from the node alignment time interval group. Calculate the average response difference value of all pairs in each of the three action stages. For example, if there are three pairs in the rising stage with time differences of 0.12 seconds, 0.08 seconds, and 0.15 seconds, respectively, calculate their average value of 0.116 seconds. Use this value as the baseline for the response difference amplitude in the rising stage, and then compare it with the set reference response difference threshold. If the average response difference is lower than the threshold value, such as 0.2 seconds, the current pairing in this stage is valid; otherwise, nodes need to be reselected and pairing combinations re-established. Reorder all available pairing nodes in each stage, try replacing one or more signal nodes to find new pairing combinations, and recalculate the average response difference of the replaced combinations. If a new combination response is found... The difference is smaller; for example, after replacement, the time difference of the three groups becomes 0.09 seconds, 0.07 seconds, and 0.10 seconds, and the average value drops to 0.086 seconds. This combination is taken as the optimal pairing. Subsequently, the combination with the smallest response difference is selected in each stage. In each pairing, a one-to-one correspondence is formed between the final signal node and the voice prompt. This relationship group consists of the stage number, the signal node time, and the voice prompt time, constructing a complete voice guidance timing feature data structure. The final output result identifies the voice guidance trigger point of each stage and the key response point of the signal change to form a pair. For example, in stage three, the pairing result is a signal node of 6.35 seconds and a voice prompt of 6.4 seconds, with a recorded response time difference of 0.05 seconds. This structure supports the rhythmic release of prompt instructions in the subsequent voice guidance module.

[0134] Please see Figure 6 The deviation assessment module includes:

[0135] The training period extraction submodule is based on the timing features of voice guidance. It matches the instruction time points marked by the voice prompts with the start and end nodes of the user's actions one by one. By judging the degree of overlap of nodes and the degree of temporal closeness, it extracts the start and end times of the corresponding actions in each stage to obtain the stage execution time period sequence.

[0136] The system reads the instruction time points corresponding to the voice prompts at each stage and retrieves the previously identified action segment signal nodes. It compares the start time of the voice prompt for each stage with the time of the start signal node for that stage, calculating the time difference. If the time difference is within a tolerance threshold (e.g., no more than 0.25 seconds), it is marked as "closely aligned"; if the time difference is between 0.25 and 0.5 seconds, it is marked as "delayed aligned"; and if it exceeds 0.5 seconds, it is marked as "not aligned". Then, it sequentially determines the time difference between each voice prompt time and its subsequent signal termination node, recording the time interval between all voice prompt time points and signal node times. For example, in a contraction maintenance training exercise, the voice prompt at the start of the maintenance phase is identified at 3.5 seconds, while the corresponding electromyographic signal node is at 3.7 seconds, with a time difference of 0.2 seconds. This pairing is marked as closely aligned. Subsequently, all voice prompts and signal nodes within the movement phase are paired, and the start and end times of each phase are sequentially formed. For example, the rising phase is from 2.2 seconds to 3.1 seconds, the maintenance phase is from 3.5 seconds to 4.4 seconds, and the falling phase is from 4.6 seconds to 5.3 seconds. The start and end times of each phase after the voice signal and movement node are aligned are uniformly organized to form a complete sequence of phase execution time periods.

[0137] The phase difference calculation submodule compares the phase execution time sequence with the preset action sequence arrangement, focuses on the distribution and differences of the start and end nodes of each phase action in the time sequence, and determines the time sequence error type of each phase by sorting out the offset performance of the nodes before and after, and obtains the time matching offset set.

[0138] The system invokes a built-in training standard movement template. This template records the standard rhythm arrangement and preset start and end times for each training movement. For example, the template sets the ascent phase to be completed between 2.0 and 3.0 seconds, the sustain phase to be between 3.0 and 4.0 seconds, and the descent phase to be between 4.0 and 5.0 seconds. Then, it compares the user's actual start and end times for each phase with the template times, calculating the start and end offset values ​​sequentially, and classifying the offset direction as either ahead or behind. For example, if the user's ascent phase starts at 2.2 seconds and ends at 3.1 seconds, the start offset... The starting offset is +0.2 seconds, and the ending offset is +0.1 seconds. This stage is marked as "lagging at both start and end". For example, the maintenance stage starts at 3.5 seconds and ends at 4.4 seconds, lagging by 0.5 seconds and 0.4 seconds respectively. According to the offset interval classification standard, 0.0 to ±0.2 seconds is considered a slight offset, ±0.2 to ±0.5 seconds is a moderate offset, and more than ±0.5 seconds is a severe offset. In the above example, the maintenance stage is judged to be a moderately lagging type. The start and end offset values, offset types and corresponding rhythm numbers of each stage are organized into structured data and named the time-matching offset set.

[0139] The deviation degree summary submodule is based on the time matching offset set, judges the relationship between rhythm features in the action structure parameters, analyzes the phenomenon of advance, lag and rhythm asynchrony of the actual action in the difference stage, and integrates the offset performance of all stages to obtain the execution deviation degree index.

[0140] Extract the start and end offset values ​​and their rhythmic feature numbers for each stage. Compare the difference between the start and end rhythm numbers with the standard rhythm interval to determine if there is an abnormality in rhythmic continuity for that stage. If the difference between the start and end rhythms exceeds the standard rhythm length tolerance (e.g., 0.3 seconds), it is marked as a rhythmic discontinuity segment. Then, statistically analyze the offset direction of each stage. If two consecutive stages have a delayed start or end, the combination is marked as an "overall delayed segment"; otherwise, it is marked as an "advanced segment" or "rhythmic inconsistency segment." For example, if the user's actual behavior is a 0.2-second delay in the first stage and a 0.2-second advance in the second stage... If the first stage is delayed by 0.1 seconds and the third stage is delayed by 0.6 seconds, it is determined that there are rhythm breaks and termination delays between stages in this training cycle. Then, the deviation types of each stage are assigned scores according to mild, moderate and severe, with mild being 1 point, moderate being 2 points and severe being 3 points. The deviation scores of each stage are then weighted and calculated. If the total score after weighting exceeds the preset upper limit value, such as 6 points, the training is marked as "significant deviation", otherwise it is marked as "slight deviation" or "good rhythm". The output execution deviation index records the deviation score of each stage, the rhythm misalignment distribution and the comprehensive score of the training cycle for subsequent quality analysis and feedback control.

[0141] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A pelvic floor muscle training system based on a mobile terminal, characterized in that, The system includes: The signal collaborative acquisition module analyzes pelvic floor electromyography and heart rate signals based on the training process, determines the synchronous fluctuations of the two on the same time axis, compares the order of mutation occurrence, identifies synchronous mutation nodes and counts their number, and obtains the characteristics of synchronous mutation nodes. Based on the characteristics of the synchronous mutation nodes, the multi-parameter linkage identification module filters signal synchronization nodes, compares the acquisition time of electromyography, photoelectric heart rate and triaxial displacement signals, analyzes the correspondence, judges the node linkage performance, and obtains a three-parameter linkage feature group. The three-segment motion decomposition module, based on the three-parameter linkage feature group, determines the rising, maintaining, and falling phases of the displacement signal. Through continuous analysis, it compares the signal changes in the different phases, summarizes the differences in each phase, and obtains the motion segmentation morphological characteristics. Based on the action segmentation morphological features, the timing-based voice guidance module compares signal nodes with voice prompt triggering points, analyzes key nodes with voice command response timing, filters nodes with the best response differences, establishes corresponding relationships, and obtains voice guidance timing features.

2. The pelvic floor muscle training system based on a mobile terminal according to claim 1, characterized in that, The synchronous mutation node features include node occurrence time, synchronous distribution type, and associated signal number; the three-parameter linkage feature group includes multi-signal coupling type, event discrimination category, and linkage amplitude identifier; the action segmentation morphology features include action structure parameters, segmentation morphology category, and action cycle marker; and the voice guidance timing features include trigger signal type, segmented voice nodes, and guidance timing annotation.

3. The pelvic floor muscle training system based on a mobile terminal according to claim 1, characterized in that, The signal collaborative acquisition module includes: The temporal fluctuation analysis submodule analyzes the collected pelvic floor electromyography and heart rate signals based on the training process, identifies the continuous change segments of each signal during the training process, calculates the rate of change of each signal segment, judges the trend change of the signal on the time axis, optimizes the location of signal inflection points, and obtains the temporal change trend quantity. The mutation relationship judgment submodule, based on the time-series change trend, filters key nodes with obvious signal jumps, determines the order of occurrence of various signal mutations, analyzes the primary and secondary relationships of mutation occurrences, and calculates the time difference between nodes to obtain mutation priority information. Based on the mutation priority information, the synchronization node identification submodule filters nodes with time differences, calculates the number and distribution of synchronous mutation nodes, compares the synchronization changes of each signal type, analyzes the coupling characteristics of signals in each time period, and obtains the characteristics of synchronous mutation nodes.

4. The pelvic floor muscle training system based on a mobile terminal according to claim 1, characterized in that, The multi-parameter linkage recognition module includes: Based on the characteristics of the synchronous mutation nodes, the signal synchronization extraction submodule analyzes the collected electromyographic signals, heart rate signals, and displacement signals, determines the synchronous change trend of each signal on the time axis, and filters the signal combinations that change synchronously within the same time period to obtain the set of synchronous nodes. The node matching and discrimination submodule, based on the set of synchronous nodes, compares the direction of change and fluctuation of each signal within the recording period, judges the consistency of the performance of the different signals on the same node, analyzes the correspondence between time series and trend, and obtains the node coupling distribution group. The linkage feature generation submodule, based on the node coupling distribution group, judges the combination of changes in each signal, analyzes the synchronicity and linkage features of the signals at key nodes, filters out the related change types, and obtains a three-parameter linkage feature group.

5. The pelvic floor muscle training system based on a mobile terminal according to claim 1, characterized in that, The three-segment action decomposition module includes: The stage identification submodule analyzes the coordinated performance of each signal based on the three-parameter linkage feature group, determines the rising, stable and falling trends of the displacement signal curve in different time periods, optimizes the start and end nodes of stage division, filters effective intervals, and obtains the stage boundary time series. The signal comparison submodule compares the displacement signal and electromyographic signal trajectory at each stage based on the stage boundary time series, analyzes the fluctuation amplitude and consistency of the two types of signal trajectories at each stage, calculates the fluctuation difference between the trajectories at each stage, and obtains the stage signal difference amplitude by collecting the fluctuation differences. Based on the difference amplitude of the stage signals, the morphological induction submodule determines the key stage number, analyzes the change process of the displacement signal and the electromyographic signal response of the stage, summarizes the difference characteristics of the three stages, and obtains the morphological characteristics of the action segmentation.

6. The pelvic floor muscle training system based on a mobile terminal according to claim 1, characterized in that, The time-series voice guidance module includes: The voice node alignment submodule, based on the action segmentation morphological features and the corresponding time point of the voice prompt, compares the sequence of each stage signal node and the voice prompt in terms of timing, and selects signal nodes with similar times to pair with the voice prompt to obtain a node alignment time interval group. The response difference calculation submodule compares the time interval between each pair of signal nodes and the voice prompt based on the node alignment time interval group, calculates the deviation of the synchronous change of signal and voice, and obtains the average response difference amplitude. Based on the average response difference amplitude, the guidance matching construction submodule optimizes the pairing results of signal nodes and voice prompts at each stage, selects the node combination with the best pairing effect, establishes a one-to-one correspondence between signal nodes and voice prompts, and obtains the voice guidance timing features.

7. The pelvic floor muscle training system based on a mobile terminal according to claim 1, characterized in that, The system also includes: The execution deviation assessment module calculates the difference between the voice guidance timing characteristics and the start and end times of each stage of the user's actual actions, analyzes the execution time, weights the stage differences, judges the consistency with the target, and obtains the execution deviation index. The execution deviation indicators include the time sequence deviation level, action phase evaluation, and execution consistency grading.

8. The pelvic floor muscle training system based on a mobile terminal according to claim 7, characterized in that, The execution deviation evaluation module includes: The training period extraction submodule, based on the voice guidance timing features, matches the instruction time points marked by the voice prompts with the start and end nodes of the user's actions one by one. By judging the degree of overlap of nodes and the degree of temporal closeness, it extracts the start and end times of the corresponding actions in each stage to obtain the stage execution time period sequence. The stage difference calculation submodule compares the stage execution time period sequence with the preset action timing arrangement, focuses on the distribution and differences of the start and end nodes of each stage action in the time sequence, and determines the timing error type of each stage by sorting out the offset performance of the nodes before and after, and obtains the time matching offset set. The deviation degree summary submodule, based on the time matching offset set, determines the relationship between rhythmic features in the action structure parameters, analyzes the phenomenon of advance, lag and rhythm asynchrony in the actual action at the difference stage, and integrates the offset performance of all stages to obtain the execution deviation degree index.

Images