Orthopedic rehabilitation effect monitoring method and system based on artificial intelligence

Abstract

The invention relates to the technical field of patient monitoring, in particular to an orthopedics rehabilitation effect monitoring method and system based on artificial intelligence, and the method comprises the following steps: obtaining motion time and angle sequences, combining the motion time and angle sequences into a record set, constructing a time chain and associating angle changes, comparing the time difference and angle amplitude to mark rhythms, judging rhythm continuity to form a trend segment, and obtaining a trend segment; according to the method, a traceable action record is constructed through the starting and ending time collected by a sensor and an angle sequence, the completion time of multiple times of training forms a time chain in sequence, the time chain is associated with an angle change section, the fast and slow change of the training rhythm is clearly presented in a continuous structure, and the accuracy of the training rhythm is improved. A recovery efficiency structured input set is combined according to rhythm attributes, performance of different training batches can be compared under the same framework, progress stagnation or efficiency decline is revealed in advance, the rehabilitation process is presented as a trend chain with directivity, and high-income training stage recognition and low-efficiency interval adjustment are facilitated.

Description

Technical Field

[0001] This invention relates to the field of patient monitoring technology, and in particular to an artificial intelligence-based method and system for monitoring orthopedic rehabilitation effects. Background Technology

[0002] The field of patient monitoring technology mainly involves the continuous acquisition and quantitative assessment of patients' physiological and functional states. Core aspects of this technology include measuring motor performance in specific areas, extracting key features of the rehabilitation process, and processing monitoring data in a systematic manner. Methodological aspects include using wearable measuring devices to collect posture parameters, assessing movement range through quantifiable standards, and analyzing functional recovery using temporal changes. Traditional orthopedic rehabilitation effect monitoring methods involve recording the range of motion and muscle strength of joints during the patient's orthopedic functional recovery process. Based on sub-categories such as joint range of motion assessment scales and gait observation scales, rehabilitation status is determined by manually judging movement trajectories and differences at time points. Traditional methods often rely on manual visual recording of movement range and manual comparison of scale levels for monitoring.

[0003] Current technologies for monitoring patient rehabilitation rely on observers recording approximate activity levels and muscle strength grades during the joint function recovery process. They often use visual estimation of movement range and assign grades based on scale items, focusing on a few key moments and lacking continuous tracking of each repetitive movement within a single training session. In this mode, time information is often presented as vague start and end descriptions, making it difficult to create a sequential time series between multiple training sessions. The efficiency of completion at different schedules or stages cannot be directly compared on the same timeline, and assessment results tend to be limited to descriptions of slight improvement or no significant change. Because activity trajectories rely on observer recollection and simple notes, details such as rhythm changes, pauses, accelerations, or decelerations during a single movement are easily overlooked. For example, when a patient completes ten flexion-extension repetitions, the rhythm is stable in the first few repetitions, but slows down significantly in the later ones due to fatigue. Traditional recording often only retains the overall time or a single impression, making it difficult to accurately identify the efficient training range and the efficiency decline range. The monitoring process is highly dependent on experience. Different assessors may have different judgments regarding visual angles, muscle strength perception, and gait interpretation. Even the same patient may receive inconsistent conclusions at different times due to the observer's subjective bias. In long-term follow-up, the recovery curve lacks a quantifiable and reproducible trend structure. When faced with the need to adjust rehabilitation plans, traditional records only provide scattered scale levels and textual notes, making it difficult to make precise judgments on whether the training pace is too fast or too slow, or whether movement control is more coherent. This can easily lead to delayed adjustments in training intensity, failure to identify high-load phases in a timely manner, or prolonged maintenance of an inefficient pace, affecting the overall rehabilitation cycle and the quality of functional reconstruction. Summary of the Invention

[0004] To address the technical problems existing in the prior art, this invention provides an artificial intelligence-based method for monitoring orthopedic rehabilitation effects, comprising the following steps:

[0005] S1: Acquire the start and end times of the movements recorded by the sensors and calculate the completion time. Organize the joint angle sequence in order and combine the completion time with the angle sequence to form a set of original records of rehabilitation movements.

[0006] S2: Based on the original record set of rehabilitation movements, the completion times of the movements from multiple training sessions are arranged in sequence to form a time chain. The time difference between adjacent movements is calculated and associated with the corresponding angle change segments to generate a joint movement time chain structure.

[0007] S3: Based on the joint action time chain structure, compare the time difference and angle change amplitude, mark the rhythm of segments with short time differences and continuous angles with accelerated rhythm, and mark the rhythm of segments with long time differences and gradual angle changes with slowed rhythm, thus forming an action rhythm feature structure.

[0008] S4: Based on the action rhythm feature structure, determine the consistency of adjacent rhythm markers, divide the same marker segments into the same trend continuous segments, set the different marker segments as the starting point of the new segment, and combine them in order to generate a set of restored trend continuous segments;

[0009] S5: Based on the set of continuous segments of the recovery trend, perform attribute determination on the trend segments, classify the segments with accelerated pace as efficiency improvement segments, classify the segments with slowed pace as efficiency deceleration segments, and combine them in the training order to obtain the structured input set of recovery efficiency.

[0010] As a further embodiment of the present invention, the original rehabilitation movement record set includes movement start and end time label data, joint angle change time series data, and training movement identification information. The joint movement time chain structure includes a training sequence completion time chain, information on the time difference between adjacent movement completions, and a mapping relationship of joint angle change segments. The movement rhythm feature structure includes a rhythm acceleration marker sequence, a rhythm deceleration marker sequence, and a rhythm feature pattern index. The recovery trend continuous segment set includes a set of continuous segments with the same trend, records of the starting position of trend changes, and segment order index information. The recovery efficiency structured input set includes a set of efficiency improvement segments, a set of efficiency deceleration segments, and a training order mapping structure.

[0011] As a further aspect of the present invention, the specific steps of S1 are as follows:

[0012] S101: Obtain the start posture signal time information and end posture signal time information of the action, calculate the time between the start and end of the action based on the two time information, verify the action interval boundary by corresponding the time amount and the signal time sequence, and generate the action duration based on the interval boundary.

[0013] S102: Based on the duration of the action, call the joint angle sequence formed by the sensor recording, analyze the correspondence between adjacent angle values ​​and adjacent time points in the angle sequence, and classify the angle changes and the duration of the action into an angle sequence arrangement order, and generate the joint angle change amplitude according to the arrangement order.

[0014] S103: Based on the amplitude of the joint angle change, call the starting posture signal time information and the ending posture signal time information of the movement, sequentially combine the joint angle sequence and the two time information, and integrate the combined angle values ​​and time points to form the structure of the entire movement process, and generate the original record set of rehabilitation movements based on the structure.

[0015] As a further aspect of the present invention, the specific steps of S2 are as follows:

[0016] S201: Based on the original record set of rehabilitation actions, the completion times of the specified rehabilitation actions generated from multiple training sessions are arranged in the training order. The arranged completion times are compared according to the order relationship, and the completion times are chained together in the order of the comparison. An action time chain sequence is generated based on the chained time sequence.

[0017] S202: Call the action time chain sequence quantity to calculate the adjacent completion time in the sequence and analyze the corresponding time difference value with the adjacent time point. The time difference value is classified into continuous segments according to the adjacent relationship and the time difference sequence structure is formed by the difference value change in the segment. The action time difference sequence quantity is generated according to the structure.

[0018] S203: Based on the action time difference sequence, call the joint angle change segment in the original rehabilitation action record set, match the time difference sequence with the angle change segment item by item, and integrate the corresponding sequence value and angle change value in sequence. Only the paired sequence is retained in the integrated content to generate a joint action time chain structure.

[0019] As a further aspect of the present invention, the specific steps of S3 are as follows:

[0020] S301: Based on the joint motion time chain structure, the time difference sequence and the joint angle change amplitude are compared. The time difference value and the angle change amount in the time difference sequence are judged according to the sequence position. The segment with short time difference and continuous angle change is marked as the rhythm acceleration segment based on the judgment result, and the rhythm acceleration segment amount is obtained.

[0021] S302: Based on the amount of the accelerated rhythm segment, the corresponding angle change amount is called to judge the unlabeled segment in the time difference sequence, and the segment with extended time difference and slowed angle change is identified as the slowed rhythm segment based on the judgment result. The slowed segment is then sorted according to the sequence position to generate the amount of the slowed rhythm segment.

[0022] S303: Based on the amount of the slow-down segment, call the amount of the fast-up segment, combine the two types of segments in the original sequence order, and use the combined segment identifier sequence and the joint action time chain structure to perform a position-corresponding rhythm feature arrangement structure to generate an action rhythm feature structure.

[0023] As a further aspect of the present invention, the specific steps of S4 are as follows:

[0024] S401: Based on the action rhythm feature structure, perform consistency judgment on adjacent rhythm marks, compare adjacent rhythm marks one by one according to the sequence position and identify continuous segments with the same mark based on the comparison result, summarize and organize the identified segments according to the continuous relationship, and generate rhythm continuous segment quantity according to the organized continuous segment structure.

[0025] S402: Call the rhythm continuous segment quantity to judge the rhythm markers in the action rhythm feature structure that are inconsistent with the continuous segment, take the position of the inconsistent marker as the starting point of the new segment, and connect the starting point with the adjacent rhythm markers to form an independent segment structure. Generate a new rhythm segment quantity based on the connected segment structure.

[0026] S403: Based on the new rhythm segment quantity, call the continuous rhythm segment quantity, combine the two types of segments in the order of the time chain, and form trend arrangement content by corresponding the position of the combined segment sequence in the original action rhythm feature structure. Generate a set of restored trend continuous segments based on the arrangement content.

[0027] As a further aspect of the present invention, the specific steps of S5 are as follows:

[0028] S501: Perform attribute judgment on the trend segment according to the set of continuous segments of the recovery trend, compare the acceleration and deceleration markers in the trend segment with the current identifier of the trend segment according to their sequence positions, and classify the trend segment where the acceleration marker is located into the efficiency improvement category and the trend segment where the deceleration marker is located into the efficiency slowdown category based on the comparison results, and generate the trend attribute classification quantity based on the classification results.

[0029] S502: Based on the trend attribute classification quantity, call the recovery trend continuous segment set, arrange the classified trend segments according to the training order, and compare the time order of the arranged trend segments with the trend attribute classification quantity. Arrange the trend segments of the same category into a combination segment structure according to the continuous sequence, and generate the trend combination segment quantity according to the combination segment structure.

[0030] S503: For the trend combination segment quantity, call the trend attribute classification quantity, integrate the efficiency improvement segment and efficiency slowdown segment in the combination segment according to the original training order, and form the efficiency feature arrangement content by matching the position of the integrated segment sequence with the recovery trend continuous segment set, and generate the recovery efficiency structured input set according to the arrangement content.

[0031] As a further embodiment of the present invention, the sensor is a sensing device that is fixed near the joint to be rehabilitated by the patient and can output posture information reflecting changes in joint angle. The data comes from an inertial measurement unit, an angle encoder or a gyroscope.

[0032] The angle sequence is a sequence of multiple angle values ​​recorded over time by a wearable joint angle sensor during the patient's performance of a designated rehabilitation movement;

[0033] The original record set of rehabilitation movements is a data set obtained by combining the completion time of the specified rehabilitation movements with the corresponding joint angle sequence.

[0034] The time chain is a time sequence formed by arranging the completion times of multiple designated rehabilitation actions in the order in which they occur during training.

[0035] The time difference is a sequence obtained by arranging the differences between the completion times of two adjacent rehabilitation specified actions in the action time chain;

[0036] The angle change segment is a set of angle records within a time range corresponding to a single rehabilitation specified action or a single time difference in the joint angle sequence;

[0037] The joint motion time chain structure is a composite data structure formed by establishing a correlation between the motion time chain, the time difference sequence, and the corresponding joint angle change segment.

[0038] As a further aspect of the present invention, the action rhythm feature structure is a structured sequence obtained by arranging rhythm markers in the rhythm feature pattern in chronological order;

[0039] The increased rhythm is a marker set for joint angle change segments with short time differences and continuous angle changes;

[0040] The slowdown in rhythm is a marker set for joint angle change segments with long time differences and slow angle changes, used to reflect the slowdown in the rhythm of segment action execution;

[0041] The trend continuum is a time segment formed by a continuous arrangement of multiple adjacent rhythm markers with the same marker type in the action rhythm feature structure.

[0042] The recovery trend continuous segment set is a data set obtained by combining multiple trend continuous segments in chronological order.

[0043] The efficiency improvement segment is the trend continuous segment marked with accelerated rhythm, which is concentrated in the trend continuous segment recovery trend;

[0044] The efficiency slowdown segment is a continuous trend segment marked by a slowdown in the rhythm, which is concentrated in the continuous segment of the recovery trend.

[0045] The structured input set for recovery efficiency is a set of structured data formed by combining efficiency improvement segments and efficiency slowdown segments in the training order.

[0046] An artificial intelligence-based orthopedic rehabilitation effect monitoring system includes:

[0047] The motion recording and acquisition module is used to perform S1: acquire the time information of the motion start posture signal and motion end posture signal collected by the wearable joint angle sensor, calculate the completion time of the specified rehabilitation motion, organize the joint angle sequence formed by the sensor in the order of appearance, and combine the completion time information and the angle sequence to generate the original set of rehabilitation motion records.

[0048] The rhythm association construction module is used to execute S2: based on the original record set of rehabilitation movements, the completion time of the specified rehabilitation movements generated by multiple training sessions is arranged in the training order to form a movement time chain. The completion time of adjacent movements in the movement time chain is calculated to obtain a time difference sequence, and the sequence is associated with the corresponding joint angle change segment to generate a joint movement time chain structure.

[0049] The rhythm pattern extraction module is used to execute S3: Based on the joint action time chain structure, the time difference sequence and the joint angle change amplitude are compared. The rhythm of the segment with a short time difference and continuous angle change is accelerated, and the rhythm of the segment with a long time difference and slow angle change is slowed down. The marks are combined in sequence to form a rhythm feature pattern to obtain the action rhythm feature structure.

[0050] The trend segment generation module is used to execute S4: based on the action rhythm feature structure, perform consistency judgment on adjacent rhythm marks, record rhythm segments with the same mark as the same trend continuous segment, take the inconsistent part as the starting point of the new segment, and combine all trend continuous segments in order to generate a set of restored trend continuous segments;

[0051] The efficiency structure classification module is used to execute S5: perform attribute judgment on each trend segment according to the set of continuous recovery trend segments, classify segments with accelerated pace as efficiency improvement segments, classify segments with slowed pace as efficiency deceleration segments, and combine the segments into an overall structure according to the training order to obtain the structured input set of recovery efficiency.

[0052] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0053] In this invention, a traceable action record is constructed by collecting start and end times and angle sequences from sensors. The completion times of multiple training sessions are sequentially formed into a time chain and associated with angle change segments. The changes in training pace are clearly presented in a continuous structure. Based on the rhythm attributes, a structured input set for recovery efficiency is combined. The performance of different training batches can be compared within the same framework. Stagnation of progress or decline in efficiency is revealed in advance. The rehabilitation process presents a directional trend chain, which is more conducive to identifying high-yield training phases and adjusting inefficient intervals. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a schematic diagram of the steps of the present invention;

[0056] Figure 2 This is a detailed schematic diagram of S1 of the present invention;

[0057] Figure 3 This is a detailed schematic diagram of S2 of the present invention;

[0058] Figure 4 This is a detailed schematic diagram of S3 of the present invention;

[0059] Figure 5 This is a detailed schematic diagram of S4 of the present invention;

[0060] Figure 6 This is a detailed schematic diagram of S5 of the present invention;

[0061] Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0062] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0063] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0064] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0065] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0066] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0067] Please see Figure 1 This invention provides an artificial intelligence-based method for monitoring orthopedic rehabilitation effects, comprising the following steps:

[0068] S1: Obtain the time information of the starting posture signal and ending posture signal of the action collected by the wearable joint angle sensor, calculate the completion time of the specified rehabilitation action, organize the joint angle sequence formed by the sensor in the order of appearance, and combine the completion time information and the angle sequence to generate the original record set of rehabilitation actions.

[0069] Wearable joint angle sensors are sensing devices that are fixed near the joints of patients undergoing rehabilitation. They can output posture information that reflects changes in joint angles, with data sourced from inertial measurement units, angle encoders, or gyroscopes.

[0070] The starting posture signal is a posture detection signal generated by the wearable joint angle sensor when it detects that the patient's limb has reached a preset starting angle threshold. It is used to mark the start time of the specified rehabilitation movement.

[0071] The action end posture signal is a posture detection signal generated by the wearable joint angle sensor when it detects that the patient's limb has reached a preset end angle threshold or returned to the initial angle position. It is used to mark the end time of the specified rehabilitation action.

[0072] The completion time of a rehabilitation-specified action is the length of time between the time corresponding to the start posture signal and the time corresponding to the end posture signal of the action, and is used to represent the time taken for one rehabilitation-specified action.

[0073] A joint angle sequence is a sequence of multiple angle values ​​recorded over time by a wearable joint angle sensor during a patient's performance of a designated rehabilitation movement. It is used to represent the trajectory of angle changes of the joint during the movement.

[0074] The original record set of rehabilitation movements is a data set obtained by combining the completion time of the specified rehabilitation movements with the corresponding joint angle sequence, and is used for time chain construction and rhythm feature extraction.

[0075] S2: Based on the original record set of rehabilitation movements, the completion time of the specified rehabilitation movements generated from multiple training sessions is arranged in the training order to form a movement time chain. The completion time of adjacent movements in the movement time chain is calculated to obtain a time difference sequence, and the sequence is associated with the corresponding joint angle change segment to generate a joint movement time chain structure.

[0076] A movement time chain is a time sequence formed by arranging the completion times of multiple specified rehabilitation movements in the order in which they occur during training. It is used to describe the changes in the time taken for multiple movements.

[0077] The time difference sequence is a sequence obtained by arranging the differences between the completion times of two adjacent rehabilitation specified actions in the action time chain, and is used to represent the changes in the time taken for adjacent training actions.

[0078] The joint angle change segment is a set of angle records within a time range corresponding to a single rehabilitation specified action or a single time difference in the joint angle sequence, used to reflect the angle changes of the joint during the action;

[0079] The joint motion time chain structure is a composite data structure formed by establishing a correlation between the motion time chain, the time difference sequence, and the corresponding joint angle change segment, and is used for rhythm feature analysis.

[0080] S3: Based on the joint action time chain structure, compare the time difference sequence with the joint angle change amplitude. For segments with short time differences and continuous angle changes, the marking rhythm is accelerated, while for segments with long time differences and slow angle changes, the marking rhythm is slowed down. The markings are combined in sequence to form a rhythm feature pattern, thus obtaining the action rhythm feature structure.

[0081] The rhythm feature pattern is a recording pattern formed by comparing the time difference sequence with the amplitude of joint angle change and assigning a mark of faster or slower rhythm to the joint angle change segment.

[0082] The action rhythm feature structure is a structured sequence obtained by arranging rhythm markers in the rhythm feature pattern in chronological order, and is used for continuous segment analysis to recover trends;

[0083] The accelerated rhythm marker is an indicator set for joint angle change segments with short time differences and continuous angle changes, used to reflect the accelerated rhythm of the segment's movement execution;

[0084] The rhythm slowdown marker is an indicator set for joint angle change segments with long time differences and slow angle changes, used to reflect the slowdown in the rhythm of segment movement execution.

[0085] S4: Based on the action rhythm feature structure, perform consistency judgment on adjacent rhythm marks, record rhythm segments with the same mark as the same trend continuous segment, take the inconsistent part as the starting point of the new segment, and combine all trend continuous segments in order to generate a set of restored trend continuous segments;

[0086] A trend continuum is a time segment formed by a continuous arrangement of multiple adjacent rhythmic markers with the same marker type in the rhythmic feature structure of an action.

[0087] The recovery trend continuous segment set is a data set obtained by combining multiple trend continuous segments in chronological order. It is used to describe the trend structure of changes in the patient's rehabilitation rhythm as training progresses.

[0088] S5: Perform attribute judgment on each trend segment according to the continuous segment set of recovery trend, classify the segments with accelerated pace as efficiency improvement segments, classify the segments with slowed pace as efficiency deceleration segments, and combine the segments into an overall structure according to the training order to obtain the structured input set of recovery efficiency.

[0089] The efficiency improvement segment is a continuous segment of the recovery trend that corresponds to the accelerated pace, used to indicate that the pace of action in this stage tends to accelerate.

[0090] The efficiency slowdown segment is a continuous segment of the recovery trend that corresponds to the slowdown in the pace of the movement, and is used to indicate that the pace of the movement in this stage tends to slow down.

[0091] The structured input set for recovery efficiency is a set of structured data formed by combining efficiency improvement segments and efficiency slowdown segments in the training order. It is used as input data for an AI-based rehabilitation effect monitoring module.

[0092] The original rehabilitation movement record set includes movement start and end time label data, joint angle change time series data, and training movement identification information. The joint movement time chain structure includes the training sequence completion time chain, the correlation information of the time difference between adjacent movement completion, and the mapping relationship of joint angle change segments. The movement rhythm feature structure includes rhythm acceleration marked sequences, rhythm deceleration marked sequences, and rhythm feature pattern index. The recovery trend continuous segment set includes the set of continuous segments with the same trend, the record of the starting position of trend change, and the segment order index information. The recovery efficiency structured input set includes the set of efficiency improvement segments, the set of efficiency deceleration segments, and the training order mapping structure.

[0093] Please see Figure 2 The specific steps of S1 are as follows:

[0094] S101: Obtain the start posture signal time information and end posture signal time information of the action, calculate the time between the start and end of the action based on the two time information, verify the action interval boundary by corresponding the time amount and the signal time sequence, and generate the action duration based on the interval boundary.

[0095] The motion start attitude signal time information, in the acquisition system, contains multiple raw timestamps and status identifiers. These timestamps are broken down into consecutive records according to the acquisition sequence. By reading the trigger flag corresponding to each time point, the earliest valid start time point is selected as the starting reference. For example, the start time points recorded by the acquisition device may include 12.03s, 12.05s, 12.07s, etc. The time point of 12.03s, which triggers the actual start signal, is taken as the starting benchmark. Similarly, the motion end attitude signal time information is broken down into multiple end time points such as 14.28s, 14.30s, 14.31s, etc. The earliest valid end record, 14.28s, is selected as the interval end. The logical continuity between the end time and the start time is confirmed by comparing their order. If there is an erroneous record where the end time is earlier than the start time, it is discarded and re-selected. Then, the number of data points that should exist in this interval is calculated based on the sampling frequency. For example, if the sampling frequency is 100Hz... When the start and end time interval is 2.25s, there should be approximately 225 sampling points. The actual number of data points acquired is compared with the estimated value. For example, if only 223 data points are actually collected, it is determined that there are two missing segments that need to be filled in during subsequent processing. By checking the time interval where the missing points are located and combining the time difference between the records before and after, the missing points are marked as intermediate state values. Then, by reading the angle, acceleration, and other information of the adjacent valid data points in the interval, the interpolation range is determined. The period with insufficient continuity is corrected to a state consistent with the sampling period. The overall start and end intervals are further verified. By checking whether there are duplicate sampling, erroneous triggers, time jumps, etc. in the interval, it is ensured that the records can maintain a uniform time interval. If a record shows a time jump exceeding 0.02s, it is classified as an anomaly and the interval boundary is re-determined according to the time order. All time records are organized into a single continuous time series according to the collection order, and the difference between the start time and the end time is recorded as the action duration in the time field of the action dataset.

[0096] S102: Based on the duration of the action, call the joint angle sequence recorded by the sensor, analyze the correspondence between adjacent angle values ​​and adjacent time points in the angle sequence, and classify the angle changes and the duration of the action to form the angle sequence arrangement order. Generate the joint angle change amplitude according to the arrangement order.

[0097] The duration of the action is broken down into time series directly corresponding to the sensor's records during actual data acquisition. For example, a duration of 2.25 seconds with a sampling frequency of 100Hz can be broken down into 225 time points. These time points are then sequentially mapped to a joint angle sequence, output by the sensor and containing multiple angle records such as 32 degrees, 35 degrees, 41 degrees, and 47 degrees. These angles are arranged chronologically, and the differences between adjacent angles are read one by one. The magnitude of the angle change between each pair of adjacent points is determined by a simple difference method. For example, the change from 32 degrees to 35 degrees is 3 degrees, from 35 degrees to 41 degrees is 6 degrees, and from 41 degrees to 47 degrees is 6 degrees. Each set of angle changes is paired with its corresponding time record for interval classification. Interval classification can be established based on the type of action. For example, during knee flexion and extension, 0 degrees to 5 degrees can be considered a mild change interval, 5 degrees to 15 degrees a moderate change interval, and more than 15 degrees a large change interval. Then, each... The group of change values ​​are placed into the corresponding intervals and rearranged in chronological order to form the arrangement order of the angle sequence. After the arrangement order is formed, each segment of angle change is read and accumulated segment by segment to form the overall amplitude. For example, the cumulative change corresponding to 3 degrees, 6 degrees, and 6 degrees in the above angle difference sequence is 15 degrees. The cumulative change is used as the basic indicator of the angle change amplitude. Then, a weight coefficient is introduced according to the required action type. For example, the basic weight of the knee bending action is set to 1.0. If the action speed factor needs to be considered, the speed value can be generated by comparing the angle change with the sampling time interval. For example, a time interval of 0.01s and an angle change of 6 degrees can be regarded as a type of change with relatively fast unit time. An additional weight coefficient, such as 0.1, is assigned to this type of change and it is marked independently in the angle change record. Then, the additional value of this type of change is superimposed in the process of cumulative change, thus forming the joint angle change amplitude. After completing all interval classification, change reading, value accumulation, and weight superposition, the amplitude is recorded as the angle change index of the action.

[0098] S103: Based on the amplitude of joint angle change, call the starting posture signal time information and the ending posture signal time information of the movement, sequentially coordinate the joint angle sequence and the two time information, and integrate the coordinated angle values ​​and time points to form the whole process structure of the movement, and generate the original record set of rehabilitation movements based on the structure.

[0099] The amplitude of joint angle change is typically used in motion sequence integration along with the start and end posture signal time information for time alignment. The start and end times are broken down into the complete time range of the actual motion execution. For example, if the start time is 12.03s and the end time is 14.28s, the total motion duration is 2.25s, corresponding to a time interval of 0.01s at a sampling frequency of 100Hz. Each angle point in the joint angle sequence is sequentially mapped onto a continuous time axis; for example, point 1 is mapped to 12.03s, point 2 to 12.04s, point 3 to 12.05s, and so on until near the end time. By pairing the angle sequence and time sequence point by point, it is confirmed whether there is any offset in the time position of each angle point. The difference between the time stamps in the original records and the theoretical mapped time is used to determine the error. If the difference exceeds a specified range, such as more than 0.02 seconds, the point is marked as an anomaly and re-inserted into a more reasonable time position. The method of distributing the offset time in half between two points is used to simply correct some discontinuous time periods. That is, the offset time is split into two parts and the time records of two nearby points are added to each part, so that the time axis is closer to the linear sampling mode. Then, the corrected time series is integrated with the corresponding angle series. The integration method is to establish a one-to-one pair structure of time and angle point by point. All records are stored as a continuous time series set to form the structure of the entire movement process. Then, this structure and the angle change amplitude are recorded together to obtain the original record set of rehabilitation movements, which is used in subsequent movement analysis or data comparison processing steps.

[0100] Please see Figure 3 The specific steps of S2 are as follows:

[0101] S201: Based on the original record set of rehabilitation movements, the completion time of the specified rehabilitation movements generated from multiple training sessions is arranged in the training order. The arranged completion times are compared according to the order relationship, and the completion times are linked in a chain according to the order of the comparison. The movement time chain sequence is generated based on the linked time sequence.

[0102] The original record set of rehabilitation movements includes a complete time record of each training movement, the start and end times of the movement, the joint movement structure, and the corresponding sequence number information. The completion time of each training movement is broken down into independent time values, such as the completion time of the first training session being 12.30s, the completion time of the second training session being 12.10s, and the completion time of the third training session being 12.45s, etc. When arranging these completion times according to the training sequence, the training number must first be read and combined with the time tag to establish a sequential linked list. For example, when the training numbers are arranged from 1 to 3, the corresponding completion time can be organized as 12. 30s, 12.10s, 12.45s. Then, the order of the completed times is compared. This comparison is achieved by determining the order of two adjacent completed times. Direct comparison can be used, such as determining that 12.30s is greater than 12.10s, and arranging them in a sequential order. Each comparison result is written into a comparison record for subsequent chaining. When chaining the obtained sequential order into a chain structure, one time is used as the starting point, and the next time point is connected sequentially. For example, starting with 12.10s... Using the earliest time point as the head of the chain, 12.30s and 12.45s are connected to form a continuous time chain. A continuity check is then performed on the chained results, comparing the time difference between adjacent times to determine if there are any abnormal jumps. If the difference between the completion time and the previous time point exceeds a set interval (e.g., more than 5s), the record is marked as needing inspection. The interval division is determined based on the type of training action. For example, a normal training interval between 0s and 5s is considered a reasonable range, while intervals exceeding 5s are considered deviations. All records are then linked together to form an action time sequence, which is then used as the action sequence. In the process of creating a time chain sequence, it is important to pay attention to the parsing method of each time value. The completion time value is generated by the acquisition system based on the action end trigger signal, and may be recorded in milliseconds, such as 12.304s. Before processing, all times with different precisions need to be unified to seconds for comparison. For example, 12.304s can be kept as 12.30s or 12.304s with three decimal places as a unified format. Then, the sequence number is compared to ensure that the completion time corresponds to which training session in the original action record set. The entire sequence chain after the connection is completed is used as the action time chain sequence.

[0103] S202: Call the action time chain sequence quantity to calculate the adjacent completion time in the sequence and analyze the corresponding time difference value with the adjacent time point. The time difference value is classified into continuous segments according to the adjacent relationship and the time difference sequence structure is formed by the difference value change in the segment. The action time difference sequence quantity is generated according to the structure.

[0104] When analyzing continuously trained actions, the action time chain sequence is broken down into a time chain consisting of multiple completion times. For example, if three action completion times are recorded in a training cycle as 12.10s, 12.30s, and 12.45s, adjacent times in the time chain, such as 12.10s and 12.30s, are grouped together for calculation. The calculation is based on the difference between the two, i.e., the difference in time interval 22.20s, which is accomplished using a simple subtraction operation. The difference is represented as a time difference and linked to the correspondence between the two. This second group of time differences... The difference between values ​​such as 12.30s and 12.45s forms another record and is combined into a set of adjacent time differences. Each time difference is analyzed for continuity according to the time chain sequence. The analysis method is to determine whether these differences form a complete segment in the continuous structure. For example, if the difference shows a gradually increasing or decreasing trend, it can be considered a continuous segment. When classifying time differences into continuous segments, a range of difference variation needs to be set. For example, 0s to 3s can be considered a low difference range, 3s to 8s can be considered a medium difference range, and 8s... The above can be considered as a relatively high difference range. Based on the training data, 22.20s can be regarded as a relatively high segment that significantly exceeds the training difference range. It is marked as an independent segment. Then, the adjacent differences are sorted according to the segment number to form a time difference sequence structure. The time difference sequence structure consists of the segment order, the difference size, and the corresponding training number index. For example, the time difference structure obtained from a certain training may include a difference of 2.20s for segment 1, a difference of 1.50s for segment 2, a difference of 3.80s for segment 3, etc. The resulting structure is combined in sequence to form the action time difference sequence quantity. This sequence quantity serves as the input content for the subsequent action chain structure. If a benchmark value is needed during the segment division process, the average difference of the most recent N training sessions can be used as the difference benchmark. For example, if the average time difference of the most recent 10 training sessions is 3.2s, then 3s can be set as the intermediate benchmark for segment division. Setting the medium segment as 3s to 8s is a reasonable range. The entire difference sequence is obtained by going through steps such as continuity judgment, difference segment division, benchmark value reference, and segment sorting to obtain the complete action time difference sequence quantity.

[0105] S203: Based on the action time difference sequence, call the joint angle change segment in the original rehabilitation action record set, match the time difference sequence with the angle change segment item by item, and integrate the corresponding sequence value with the angle change value in sequence. Only the paired sequence is retained in the integrated content to generate the joint action time chain structure.

[0106] The time difference sequence of movements consists of multiple differences, such as 2.20s, 1.50s, and 3.80s obtained during a training process. When establishing the corresponding structure of joint movements, this sequence needs to be synchronized with the joint angle change segments in the original rehabilitation movement record set. The angle change segments are broken down into the angle changes corresponding to each training session. For example, the first training session corresponds to a mild angle change segment with a change of 10 degrees, the second training session corresponds to a moderate angle change segment with a change of 27 degrees, and the third training session corresponds to a high angle change segment with a change of 35 degrees. When corresponding the time difference sequence to the angle change segments item by item, the training order should be used as the basis. For example, a time difference of 2.20s corresponds to a 10-degree angle change in the first training session, a time difference of 1.50s corresponds to a 27-degree angle change in the second training session, and a time difference of 3.80s corresponds to a 35-degree angle change in the third training session. During the correspondence process, it is necessary to ensure that the length of the time difference sequence corresponds to the angle change segment. If the quantity is consistent and a missing training data item is found, the missing training number needs to be retrieved from the original record set and the record needs to be filled in according to the number order. The time difference value and angle change value after pairing are organized in a sequence manner. Data that is not paired is removed, and only the sequence content with training mapping relationship is retained. The integrated content can adopt a structured recording method, with each pairing content as an entry and the time difference and angle change amount as two fields in the entry. By integrating them one by one, a joint action time chain structure is formed. If a weight coefficient is required during the integration process, different weight values ​​can be set according to the joint type. For example, the weight of the knee joint can be set to 1.0, and the weight of the shoulder joint can be set to 0.8. The weight adjustment amount is obtained by multiplying the weight by the angle change value and written into the integrated entry. For example, an angle change of 27 degrees in the shoulder joint scenario will be multiplied by 0.8 to get 21.6 degrees as the weighted record. Then, all the integrated entries are recorded in the training order to form a joint action time chain structure.

[0107] Please see Figure 4 The specific steps of S3 are as follows:

[0108] S301: Based on the joint motion time chain structure, compare the time difference sequence with the joint angle change amplitude, make corresponding judgments on the time difference value and angle change amount in the time difference sequence according to the sequence position, and mark the segment with short time difference and continuous angle change as the rhythm acceleration segment based on the judgment result, and obtain the rhythm acceleration segment amount.

[0109] Before performing judgments, the joint motion time chain structure needs to be broken down into directly comparable corresponding items, such as time difference sequences, angle change amplitudes, etc. For example, the time difference sequence might contain data such as 2.20s, 1.50s, and 3.80s, while the angle change amplitude sequence might contain data such as 10 degrees, 27 degrees, and 35 degrees. These sequences are then organized into pairs according to their indices, such as (2.20s, 10 degrees), (1.50s, 27 degrees), and (3.80s, 35 degrees). A comparison step is then performed on each pair of data. The comparison method is based on judging whether the time difference is within a shorter interval and whether the angle change is within a continuous interval. The comparison is further differentiated by setting the interval range of the time difference. The time difference ranges are categorized into short, medium, and long ranges, which can be set based on the characteristics of rehabilitation training movements. For example, 0s to 2s can be classified as a short range, 2s to 6s as a medium range, and 6s to 12s as a long range. A time difference of 1.50s is classified as a short range, 2.20s as a medium range, and 3.80s as a slightly above-medium range. The angle change is further divided according to its continuity. The determination of angle continuity can be based on a threshold for continuous change ranges. For example, an angle change of 0 to 15 degrees is a slightly continuous range, 15 to 30 degrees is a moderately continuous range, and more than 30 degrees is a large-amplitude continuous range. Using this standard, 10 degrees is classified as a slightly continuous range, and 27 degrees as a moderately continuous range. Continuing the segmentation, 35 degrees is categorized as a large continuous segment. Then, each pair of structures is evaluated, with two parts: first, whether the time difference belongs to a shorter segment; second, whether the degree of angle continuity belongs to a continuously increasing segment. If both conditions are met, it is labeled as an accelerating segment. When detailing the evaluation process, the data processing method needs to be explained. For example, taking 1.50s and 27 degrees as an example, first, 1.50s is determined to be a shorter segment, then 27 degrees is determined to be a moderately continuous segment, and because both conditions are met simultaneously, this group is labeled as an accelerating segment. To avoid inconsistent interval standards, a benchmark system needs to be established for dividing time difference intervals and angle change intervals. Based on the time difference, the benchmark value can be the average time difference of the last five training sessions. For example, if the average time difference of the last five training sessions is 2.8s, then the shorter segment can be set to a range of about 40% below the average, i.e., 0s to 1.7s, the medium segment can be set to 1.7s to 5s, and the segment longer than 5s can be classified as the longer segment. The benchmark value for the angle change can be the average value of the typical angle change of the movement, such as 20 degrees, as the basis. The light segment can be set to 0 degrees to 15 degrees, the medium segment to 15 degrees to 30 degrees, and the heavy segment to 30 degrees or more. Then, the above segments are compared sequentially through training records and the rhythm change trend is marked. By sorting out all segments that meet the conditions, the amount of rhythm acceleration segment is obtained.

[0110] S302: Based on the amount of the accelerated rhythm segment, the corresponding angle change amount is called to judge the unlabeled segment in the time difference sequence, and the segment with extended time difference and slowed angle change is identified as the slowed rhythm segment based on the judgment result. The slowed segment is then sorted according to the sequence position to generate the amount of the slowed rhythm segment.

[0111] When further processing time difference sequences, the acceleration segment needs to re-evaluate the angle change values ​​for unlabeled segments. The unlabeled segments in the time difference sequence are broken down item by item and evaluated using the angle change values. For example, if a time difference sequence contains 2.20s, 1.50s, and 3.80s, but the acceleration segment only labels 1.50s, then the unlabeled segments are 2.20s and 3.80s. 2.20s is paired with its corresponding angle of 10 degrees to form a pair to be evaluated, and 3.80s is paired with its corresponding angle of 35 degrees to form another pair. This process is then repeated. The process first requires judging the time difference interval. By setting thresholds, the time difference is categorized by magnitude. For example, 0s to 2s is considered short, 2s to 6s is medium, and over 6s is long. Therefore, 2.20s and 3.80s both fall within the medium interval. Next, the continuity of the angle change is judged. By comparing the angle change intervals, 10 degrees corresponds to a slightly continuous segment, and 35 degrees corresponds to a heavily continuous segment. A joint judgment is made based on the time difference and angle change results. The judgment logic can be expressed as: when the time difference belongs to the medium to long interval and the angle change is in the slightly or moderate range... The sections can be categorized as rhythm slowdown sections. The 2.20s and 10-degree sections are classified as rhythm slowdown sections. The 3.80s and 35-degree sections, due to their higher angle changes, require further comparison to determine if the intensity of the angle change exceeds the range used for slowdown judgment. An upper limit for the angle change used for slowdown judgment can be set, for example, 30 degrees. Changes exceeding 30 degrees need to be adjusted using a weighting coefficient. For example, combining an angle change of 35 degrees with a weighting coefficient of 0.7 yields approximately 24.5 degrees as a reference value. This value falls into the moderate change section and meets the conditions for a slowdown section; therefore, it is designated as a rhythm slowdown section. To ensure consistency in the judgment criteria, an angle change benchmark value can be set, such as 20 degrees, as the benchmark for the moderate segment. An angle below 20 degrees is considered mild, 20 to 30 degrees is considered moderate, and above 30 degrees is considered high. The high-amplitude portion is processed through a weighted conversion method to make it closer to the judgment criteria for the rhythm slowdown segment. Then, all the marked slowdown segments are arranged according to their sequence positions. For example, if the original sequence order is the first item 2.20s, the second item 1.50s, and the third item 3.80s, then the slowdown segments correspond to the first and third items. The amount of rhythm slowdown segment is then determined by arranging them in this order.

[0112] S303: Based on the amount of the slow-down segment, call the amount of the fast-up segment, combine the two types of segments in the original sequence order, and use the combined segment identifier sequence and the joint action time chain structure to perform position-corresponding rhythm feature arrangement structure to generate the action rhythm feature structure.

[0113] When integrating the tempo-slowing segments, they need to be combined with the tempo-accelerating segments in sequence combination. When combining the two types of segments according to the original time chain order, the sequence index information in the joint action time chain structure must first be read. The sequence number, corresponding time difference, and angle change segment of each action are arranged sequentially. Then, the tempo-accelerating and tempo-slowing segments are labeled with their respective sequence numbers. For example, the accelerating segment is the 2nd item, and the slowing segment is the 1st and 3rd item. The label sets of the two types of segments are arranged in sequence to form a segment identifier sequence, such as slowing down, accelerating, slowing down. The segment identifier sequence is then matched item by item with the joint action time chain structure. A one-to-one correspondence structure is established by comparing the sequence item number with the segment identifier number sequentially. For example, the 1st item in the time chain is labeled as slowing down, the 2nd item as accelerating, and the 3rd item as releasing. Next, the angle changes within the sequence and the rhythm identifiers are written together into the rhythm feature arrangement structure. To handle different intensities of angle changes, rhythm feature coefficients can be introduced. For example, the feature coefficient for the accelerated rhythm section can be set to 1.2, and the feature coefficient for the decelerated rhythm section can be set to 0.8. The rhythm feature value is obtained by multiplying the coefficient by the angle change. For example, the second angle change of 27 degrees multiplied by the coefficient 1.2 gives 32.4 degrees, the first angle change of 10 degrees multiplied by the coefficient 0.8 gives 8 degrees, and the third angle change of 35 degrees multiplied by the coefficient 0.8 gives 28 degrees. The rhythm feature value is written into the corresponding field of the arrangement structure. During the integration process, it is necessary to ensure that the identifier order of all sections is completely consistent with the original time chain order, without inserting or deleting any items, thus forming the action rhythm feature structure.

[0114] Please see Figure 5 The specific steps of S4 are as follows:

[0115] S401: Based on the action rhythm feature structure, perform consistency judgment on adjacent rhythm markers, compare adjacent rhythm markers one by one according to the sequence position, and identify continuous segments with the same markers based on the comparison results. Summarize and organize the identified segments according to the continuous relationship, and generate the number of continuous rhythm segments based on the organized continuous segment structure.

[0116] Before performing consistency judgment on adjacent rhythm markers, the rhythm marker sequence needs to be broken down into comparable data units. For example, rhythm markers may contain "speed up, speed up, slow down, slow down, speed up, slow down," etc. These markers are organized by index into the following order: 1st item speed up, 2nd item speed up, 3rd item slow down, 4th item slow down, 5th item speed up, 6th item slow down. Then, the adjacent marker comparison operation within the sequence is performed. The comparison method is to compare the content of each pair of adjacent items one by one. For example, if the 1st and 2nd items are both speed up, they are judged to be consistent; if the 2nd and 3rd items are both speed up, they are judged to be consistent. A slowdown is considered inconsistent, and this judgment can be made through textual comparison. Setting a consistency judgment value of 1 and an inconsistency judgment value of 0 facilitates subsequent generalization. For example, consistent pairs are recorded as 1, and inconsistent pairs as 0. A value of 1 is considered an extension marker of a continuous segment. The judgment results of all adjacent pairs in the rhythm markers are then organized into a simple consistency vector, such as 1, 0, 1, 0, 0, etc. When establishing the consistency judgment logic, a classification reference standard for the rhythm markers needs to be defined. For example, a rhythm acceleration segment can be identified by a rhythm feature coefficient greater than 1.0, and a rhythm slowdown segment... This can be confirmed by a rhythm feature coefficient less than 1.0. The rhythm feature coefficient is obtained by establishing a weighted relationship between the angle change and the rhythm category. For example, if the angle change in a training session is 27 degrees and the coefficient for the accelerated rhythm segment is set to 1.2, then the rhythm feature value can be obtained by multiplying 27 degrees by the coefficient 1.2 to get 32.4 degrees. Based on the condition that the feature value is greater than the angle baseline value of 20 degrees, this segment is identified as an accelerated rhythm segment. These rhythm markers and rhythm feature values ​​are written together into the sequence structure as the basis for consistency judgment. After continuously comparing all adjacent markers, a grouped continuous segment structure is formed, and segments with the same marker are linked together. The continuation paragraphs are organized according to the index continuity. For example, if the first and second items are both accelerations, they are summarized as a continuous acceleration paragraph. The third and fourth items are summarized as a continuous deceleration paragraph. The fifth item is independently classified into a single acceleration paragraph, and the sixth item is classified into a single deceleration paragraph. All continuous paragraphs are numbered, and the start and end positions of the continuous paragraphs, the marked content within the paragraphs, and the paragraph length are used as the summarization results. The paragraph length is calculated based on the number of rhythm items within the paragraph. For example, the length of the first continuous acceleration paragraph is 2, the length of the continuous deceleration paragraph is 2, and the length of the subsequent acceleration and deceleration paragraphs is 1. Finally, the summarized continuous paragraph structure is used as the number of rhythm continuous paragraphs.

[0117] S402: Call the rhythm continuous segment quantity to judge the rhythm markers in the action rhythm feature structure that are inconsistent with the continuous segment, take the position of the inconsistent marker as the starting point of the new segment, and connect the starting point with the adjacent rhythm markers to form an independent segment structure. Generate a new rhythm segment quantity based on the connected segment structure.

[0118] When further distinguishing the rhythmic feature structure of an action, the rhythm continuous segment quantity needs to judge the rhythmic markers not included in the continuous segment. Each marker in the action rhythmic feature structure is compared with the rhythm continuous segment quantity. If a marker does not appear in the corresponding index of the continuous segment, it is considered an inconsistent marker. This index is used as the starting point of a new segment to establish an independent segment structure. For example, if the rhythmic marker sequence is speed up, speed up, slow down, slow down, speed up, slow down, and the continuous segment recorded in the continuous segment quantity is (1-2, speed up) and (3-4, slow down), then the 5th (speed up) and 6th (slow down) are the segments not included. During the judgment process, a rhythmic marker classification benchmark value needs to be set first. For example, a speeding up marker can be judged if the rhythmic feature value is greater than the angle benchmark value of 20 degrees, and a slowing down marker can be judged if the feature value is less than 20 degrees. The angle benchmark value of 20 degrees is derived from the average of the angle values ​​of the last ten training iterations. This benchmark is used as a judgment condition when the rhythmic labels cannot be directly read. When judging the 5th and 6th... When performing independent judgment, it is necessary to read the angle change of the corresponding action. For example, the angle change of the 5th item is 27 degrees, which is greater than the benchmark range, so it is marked as accelerated. The angle change of the 6th item is 35 degrees, which is also greater than the benchmark, but the slowdown coefficient can be set to 0.8 by adjusting the rhythm coefficient. Multiplying 35 degrees by 0.8 gives 28 degrees, so that it falls back into the rhythm slowdown range, forming a rhythm slowdown mark. Then, the mark is used as the segment classification. The starting point of the new segment is connected with the adjacent rhythm items in sequence. By constructing a continuous sequence structure, an independent segment structure is formed. For example, the accelerated 5th item can become the new segment A, and the slowed 6th item can become the new segment B. The two new segments are organized into new segment quantities according to the sequence number. If a step that requires the use of a threshold occurs in the whole process, a rhythm difference threshold needs to be set to distinguish between rapid changes and slow changes. For example, the threshold is set to consider rhythm feature value differences exceeding 10 degrees as different rhythms. The difference condition is used as a supplementary item for inconsistency judgment. All inconsistent segments are organized into new rhythm segment quantities.

[0119] S403: Based on the new rhythm segment quantity, call the rhythm continuous segment quantity, combine the two types of segments in the order of the time chain, and form trend arrangement content by corresponding the position of the combined segment sequence in the original action rhythm feature structure. Generate the recovery trend continuous segment set based on the arrangement content.

[0120] During the trend structure consolidation phase, the new rhythm segment quantity needs to be combined with the continuous rhythm segment quantity. Before combining the two types of segments in time chain order, the original index information of the time chain sequence needs to be read first. Each item in the sequence is bound to the rhythm mark and related angle change quantity. For example, if the time chain is T1, T2, T3, T4, T5, T6, a three-element field containing rhythm category, angle quantity, and rhythm feature value is created for each item. The continuous rhythm segment quantity and the new rhythm segment quantity are mapped to the time chain index respectively. For example, the continuous segment is recorded as (1-2 acceleration segment) and (3-4 deceleration segment), and the new segment is recorded as (5 acceleration) and (6 deceleration). Then, they are combined in index order to form a sequence of acceleration, acceleration, deceleration, deceleration, acceleration, and deceleration. This combined sequence is written into the segment sequence content, and its position is matched with the original action rhythm feature structure item by item. The index consistency judgment ensures that the combined segment sequence and the time order of the original data are consistent. To ensure consistency, when establishing trend arrangement content, the rhythm change trend needs to be presented through sorting. For example, this can be achieved through rhythm marking encoding, encoding acceleration as 1 and deceleration as 0. The encoding sequence is then arranged in chronological order as 1, 1, 0, 0, 1, 0. The encoding sequence is written into the trend data structure, and then the trend data structure is integrated with the segment combination list so that each data item contains rhythm attributes, angle change amount, and encoding value. When dealing with the importance of different rhythm types, trend weight coefficients can be set. For example, the weight of the acceleration segment is set to 1.1, and the weight of the deceleration segment is set to 0.9. The trend amplitude value is constructed by multiplying the weight by the angle change amount. For example, the first item with an angle of 10 degrees corresponds to the acceleration segment, so the trend amplitude value is 11 degrees. The sixth item with an angle of 35 degrees corresponds to the deceleration segment, so the trend amplitude value is 31.5 degrees. All trend amplitude values ​​are written into the arrangement content, and the trend information of each segment is organized in sequence to form a set of continuous segments for restoring the trend.

[0121] Please see Figure 6 The specific steps of S5 are as follows:

[0122] S501: Perform attribute judgment on the trend segment based on the set of continuous segments of the recovery trend, compare the acceleration and deceleration markers in the trend segment with the current identifier of the trend segment according to their sequence positions, and classify the trend segment where the acceleration marker is located into the efficiency improvement category and the trend segment where the deceleration marker is located into the efficiency slowdown category based on the comparison results, and generate the trend attribute classification quantity based on the classification results.

[0123] Before performing trend segment attribute judgment, the rhythm marker of each trend segment in the continuous segment set needs to be decomposed into directly comparable field content. For example, the continuous segment set may contain six trend segments with a rhythm marker sequence of acceleration, acceleration, deceleration, deceleration, acceleration, deceleration, corresponding to T1 to T6 respectively according to the training order. After reading these trend segments one by one, they are expanded according to the sequence position as (T1, acceleration), (T2, acceleration), (T3, deceleration), (T4, deceleration), (T5, acceleration), (T6, deceleration). When performing attribute judgment, the rhythm marker of each trend segment needs to be... The comparison is performed with the current label field of the trend segment. This comparison can be done using text comparison. For example, if both T1 and T2 show acceleration, the comparison result is consistent. If T3 shows deceleration and the current label also shows deceleration, the result is also consistent. The consistent result is recorded as 1 (Boolean value), and the inconsistent result as 0. The comparison result is written into the label judgment vector (e.g., 1, 1, 1, 1, 1) for subsequent processing. Before establishing the judgment rules, a classification benchmark value for acceleration and deceleration needs to be set. The benchmark value is determined by the rhythm feature coefficient and the angle change. For example, for an acceleration segment, a rhythm feature coefficient can be set. A coefficient greater than 1.0 is used as the criterion for acceleration, while a coefficient less than 1.0 can be set as the criterion for deceleration. The rhythm characteristic coefficient is derived from the angle change multiplied by the coefficient of different rhythm types. For example, if the angle change in T1 is 20 degrees and the acceleration coefficient is 1.2, the characteristic value is 24 degrees; if the angle change in T3 is 35 degrees and the deceleration coefficient is 0.8, the characteristic value is 28 degrees and it falls into the deceleration category. These characteristic values ​​are written into the trend segment attribute field for auxiliary reference during judgment. In attribute judgment, if the trend segment rhythm is marked as acceleration and the comparison result is consistent, then the trend segment is classified as efficiency improvement. If the trend segment's rhythm is marked as slowing down, it is classified into the efficiency slowdown category. During the classification process, if the rhythm feature value is close to the boundary, interval division judgment is required. For example, the rhythm feature interval can be set as a mild interval (0 degrees to 20 degrees), a moderate interval (20 degrees to 35 degrees), and a high interval (above 35 degrees). The validity of the corresponding rhythm mark is confirmed by judging which interval the feature value belongs to. For example, the T3 feature value of 28 degrees is in the moderate interval and the corresponding slowdown mark can remain unchanged. All trend segments are classified and organized according to the efficiency improvement category or the efficiency slowdown category, and trend attribute classification quantity is generated.

[0124] S502: Based on the trend attribute classification, call the recovery trend continuous segment set, arrange the classified trend segments according to the training order, and compare the time order of the arranged trend segments with the trend attribute classification. Organize the trend segments of the same category into a combination segment structure according to the continuous sequence, and generate the trend combination segment quantity according to the combination segment structure.

[0125] After further retrieving the continuous trend segment set, the trend attribute classification value needs to be rearranged according to the training order. The training number, rhythm marker, angle change, and trend attribute classification result of each trend segment are organized into a sequence structure. For example, the trend attribute classification value may include (T1, increase), (T2, increase), (T3, slowing down), (T4, slowing down), (T5, increase), and (T6, slowing down). After arranging these segments according to the training order, a trend attribute sequence is established as increase, increase, slowing down, slowing down, increase, slowing down. Each of the rearranged trend segments is then compared with the trend attribute classification value. The comparison method is to check whether the category of the trend segment in the classification value is consistent with the attribute in the sequence. For example, if T1 is increase in the classification value and its position in the sequence corresponds to increase, then they are consistent. If an inconsistency occurs, the feature fields of the trend segment need to be read and it needs to be re-evaluated whether the category shift is caused by the feature value boundary. Correction is performed by comparing the trend feature value with the baseline value range. For example, if the rhythm feature value is 19.8 degrees and the baseline division boundary is 20 degrees, then... Values ​​at critical positions can be determined by setting a classification threshold offset, such as ±1 degree. Trend segments within the offset range are grouped into the same category as their neighbors according to the nearest category merging principle to maintain sequence integrity. Trend segments of the same category are then arranged into combined segments in consecutive order. For example, if T1 and T2 are both increasing, they can form combined segment A; if T3 and T4 are both decreasing, they can form combined segment B; if T5 is increasing, it forms combined segment C; and if T6 is decreasing, it forms combined segment D. During the arrangement process, the start and end points of the combined segments need to be determined. The start point is the consecutive category. The first position of a segment is the position of the last item before the category change occurs. For example, combination segment A is (1 to 2), combination segment B is (3 to 4), combination segment C is (5 to 5), and combination segment D is (6 to 6). When forming the combination segment structure, characteristic statistics can be set for each segment, such as the average angle or rhythm characteristic of the segment, for subsequent correction operations. For example, the angle change in combination segment B is 35 degrees and 32 degrees, and the average value of 33.5 degrees can be calculated as its characteristic record. These combination segments are arranged in order to form the trend combination segment quantity.

[0126] S503: For trend combination segment quantity, call trend attribute classification quantity, integrate the efficiency improvement segment and efficiency slowdown segment in the combination segment according to the original training order, and form efficiency feature arrangement content by position correspondence between the integrated segment sequence and the recovery trend continuous segment set, and generate recovery efficiency structured input set according to the arrangement content;

[0127] When integrating trend combination segments, it is necessary to call the trend attribute classification quantity and organize the efficiency improvement segments and efficiency deceleration segments into a continuous segment sequence according to the training order. The segment identifier, start and end positions, intra-segment angle feature values, and intra-segment rhythm category of the combination segment are written into the combination structure. For example, in the previous content, combination segments A, B, C, and D are improvement, deceleration, improvement, and deceleration respectively. These segments are arranged in the training order as improvement → deceleration → improvement → deceleration. Then, an integration operation is performed on each combination segment, mapping the segment sequence in the trend combination segment quantity to the corresponding position in the recovery trend continuous segment set. For example, positions T1 and T2 are mapped to segment A, positions T3 and T4 to segment B, position T5 to segment C, and position T6 to segment D. These segment identifiers are then written into the efficiency feature arrangement content. When integrating, separate settings are needed for efficiency improvement segments and efficiency deceleration segments. Efficiency coefficients are defined, for example, the efficiency improvement coefficient is set to 1.1 and the efficiency slowdown coefficient is set to 0.9. The efficiency coefficients are then used to generate efficiency metrics by performing simple calculations with the angle change. For example, the T1 angle of 20 degrees multiplied by 1.1 results in 22 degrees, which is recorded as the efficiency metric. The T3 angle of 35 degrees multiplied by 0.9 results in 31.5 degrees, which is recorded as the slowdown efficiency metric. Each efficiency metric is written into the corresponding sequence field. If a feature value is located at the interval threshold, such as around 20 degrees, the interval boundary correction method is used, that is, a ±1 degree floating interval is set to determine whether the category change needs to be adjusted. The adjusted category is compared with the original category, and only the category is written into the structure. After processing all combination segments, the sequence is bound item by item to fields such as training number, rhythm mark, trend category, and efficiency metric, forming a structured input set for restored efficiency.

[0128] Please see Figure 7 An artificial intelligence-based orthopedic rehabilitation effect monitoring system includes:

[0129] The motion recording and acquisition module is used to perform S1: acquire the time information of the motion start posture signal and motion end posture signal collected by the wearable joint angle sensor, calculate the completion time of the specified rehabilitation motion, organize the joint angle sequence formed by the sensor in the order of appearance, and combine the completion time information and the angle sequence to generate the original set of rehabilitation motion records.

[0130] The rhythm association construction module is used to execute S2: based on the original record set of rehabilitation movements, the completion time of the specified rehabilitation movements generated by multiple training sessions is arranged in the training order to form a movement time chain. The completion time of adjacent movements in the movement time chain is calculated to obtain a time difference sequence, and the sequence is associated with the corresponding joint angle change segment to generate a joint movement time chain structure.

[0131] The rhythm pattern extraction module is used to execute S3: Based on the joint action time chain structure, the time difference sequence and the joint angle change amplitude are compared. For segments with short time differences and continuous angle changes, the rhythm is accelerated; for segments with long time differences and slow angle changes, the rhythm is slowed. The marks are combined in sequence to form a rhythm feature pattern, thus obtaining the action rhythm feature structure.

[0132] The trend segment generation module is used to execute S4: based on the action rhythm feature structure, it performs consistency judgment on adjacent rhythm marks, records rhythm segments with the same mark as the same trend continuous segment, takes the inconsistent parts as the starting point of the new segment, and combines all trend continuous segments in sequence to generate a set of restored trend continuous segments;

[0133] The efficiency structure classification module is used to execute S5: based on the set of continuous segments of the recovery trend, it performs attribute judgment on each trend segment, classifies segments marked with accelerated pace as efficiency improvement segments, classifies segments marked with slowed pace as efficiency deceleration segments, and combines the segments into an overall structure according to the training order to obtain the structured input set of recovery efficiency.

[0134] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for monitoring the effectiveness of orthopedic rehabilitation based on artificial intelligence, characterized in that, Includes the following steps: S1: Acquire the start and end times of the movements recorded by the sensors and calculate the completion time. Organize the joint angle sequence in order and combine the completion time with the angle sequence to form a set of original records of rehabilitation movements. S2: Based on the original record set of rehabilitation movements, the completion times of the movements from multiple training sessions are arranged in sequence to form a time chain. The time difference between adjacent movements is calculated and associated with the corresponding angle change segments to generate a joint movement time chain structure. S3: Based on the joint action time chain structure, compare the time difference and angle change amplitude, mark the rhythm of segments with short time differences and continuous angles with accelerated rhythm, and mark the rhythm of segments with long time differences and gradual angle changes with slowed rhythm, thus forming an action rhythm feature structure. S4: Based on the action rhythm feature structure, determine the consistency of adjacent rhythm markers, divide the same marker segments into the same trend continuous segments, set the different marker segments as the starting point of the new segment, and combine them in order to generate a set of restored trend continuous segments; S5: Based on the set of continuous segments of the recovery trend, perform attribute determination on the trend segments, classify the segments with accelerated pace as efficiency improvement segments, classify the segments with slowed pace as efficiency deceleration segments, and combine them in the training order to obtain the structured input set of recovery efficiency.

2. The method for monitoring orthopedic rehabilitation effects based on artificial intelligence according to claim 1, characterized in that, The original rehabilitation movement record set includes movement start and end time label data, joint angle change time series data, and training movement identification information. The joint movement time chain structure includes a training sequence completion time chain, information on the time difference between adjacent movement completions, and a mapping relationship of joint angle change segments. The movement rhythm feature structure includes a rhythm acceleration marker sequence, a rhythm deceleration marker sequence, and a rhythm feature pattern index. The recovery trend continuous segment set includes a set of continuous segments with the same trend, records of the starting position of trend changes, and segment order index information. The recovery efficiency structured input set includes a set of efficiency improvement segments, a set of efficiency deceleration segments, and a training order mapping structure.

3. The method for monitoring orthopedic rehabilitation effects based on artificial intelligence according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Obtain the start posture signal time information and end posture signal time information of the action, calculate the time between the start and end of the action based on the two time information, verify the action interval boundary by corresponding the time amount and the signal time sequence, and generate the action duration based on the interval boundary. S102: Based on the duration of the action, call the joint angle sequence formed by the sensor recording, analyze the correspondence between adjacent angle values ​​and adjacent time points in the angle sequence, and classify the angle changes and the duration of the action into an angle sequence arrangement order, and generate the joint angle change amplitude according to the arrangement order. S103: Based on the amplitude of the joint angle change, call the starting posture signal time information and the ending posture signal time information of the movement, sequentially combine the joint angle sequence and the two time information, and integrate the combined angle values ​​and time points to form the structure of the entire movement process, and generate the original record set of rehabilitation movements based on the structure.

4. The method for monitoring orthopedic rehabilitation effects based on artificial intelligence according to claim 3, characterized in that, The specific steps of S2 are as follows: S201: Based on the original record set of rehabilitation actions, the completion times of the specified rehabilitation actions generated from multiple training sessions are arranged in the training order. The arranged completion times are compared according to the order relationship, and the completion times are chained together in the order of the comparison. An action time chain sequence is generated based on the chained time sequence. S202: Call the action time chain sequence quantity to calculate the adjacent completion time in the sequence and analyze the corresponding time difference value with the adjacent time point. The time difference value is classified into continuous segments according to the adjacent relationship and the time difference sequence structure is formed by the difference value change in the segment. The action time difference sequence quantity is generated according to the structure. S203: Based on the action time difference sequence, call the joint angle change segment in the original rehabilitation action record set, match the time difference sequence with the angle change segment item by item, and integrate the corresponding sequence value and angle change value in sequence. Only the paired sequence is retained in the integrated content to generate a joint action time chain structure.

5. The method for monitoring orthopedic rehabilitation effects based on artificial intelligence according to claim 4, characterized in that, The specific steps for S3 are as follows: S301: Based on the joint motion time chain structure, the time difference sequence and the joint angle change amplitude are compared. The time difference value and the angle change amount in the time difference sequence are judged according to the sequence position. The segment with short time difference and continuous angle change is marked as the rhythm acceleration segment based on the judgment result, and the rhythm acceleration segment amount is obtained. S302: Based on the amount of the accelerated rhythm segment, the corresponding angle change amount is called to judge the unlabeled segment in the time difference sequence, and the segment with extended time difference and slowed angle change is identified as the slowed rhythm segment based on the judgment result. The slowed segment is then sorted according to the sequence position to generate the amount of the slowed rhythm segment. S303: Based on the amount of the slow-down segment, call the amount of the fast-up segment, combine the two types of segments in the original sequence order, and use the combined segment identifier sequence and the joint action time chain structure to perform a position-corresponding rhythm feature arrangement structure to generate an action rhythm feature structure.

6. The method for monitoring orthopedic rehabilitation effects based on artificial intelligence according to claim 5, characterized in that, The specific steps of S4 are as follows: S401: Based on the action rhythm feature structure, perform consistency judgment on adjacent rhythm marks, compare adjacent rhythm marks one by one according to the sequence position and identify continuous segments with the same mark based on the comparison result, summarize and organize the identified segments according to the continuous relationship, and generate rhythm continuous segment quantity according to the organized continuous segment structure. S402: Call the rhythm continuous segment quantity to judge the rhythm markers in the action rhythm feature structure that are inconsistent with the continuous segment, take the position of the inconsistent marker as the starting point of the new segment, and connect the starting point with the adjacent rhythm markers to form an independent segment structure. Generate a new rhythm segment quantity based on the connected segment structure. S403: Based on the new rhythm segment quantity, call the continuous rhythm segment quantity, combine the two types of segments in the order of the time chain, and form trend arrangement content by corresponding the position of the combined segment sequence in the original action rhythm feature structure. Generate a set of restored trend continuous segments based on the arrangement content.

7. The method for monitoring orthopedic rehabilitation effects based on artificial intelligence according to claim 6, characterized in that, The specific steps of S5 are as follows: S501: Perform attribute judgment on the trend segment according to the set of continuous segments of the recovery trend, compare the acceleration and deceleration markers in the trend segment with the current identifier of the trend segment according to their sequence positions, and classify the trend segment where the acceleration marker is located into the efficiency improvement category and the trend segment where the deceleration marker is located into the efficiency slowdown category based on the comparison results, and generate the trend attribute classification quantity based on the classification results. S502: Based on the trend attribute classification quantity, call the recovery trend continuous segment set, arrange the classified trend segments according to the training order, and compare the time order of the arranged trend segments with the trend attribute classification quantity. Arrange the trend segments of the same category into a combination segment structure according to the continuous sequence, and generate the trend combination segment quantity according to the combination segment structure. S503: For the trend combination segment quantity, call the trend attribute classification quantity, integrate the efficiency improvement segment and efficiency slowdown segment in the combination segment according to the original training order, and form the efficiency feature arrangement content by matching the position of the integrated segment sequence with the recovery trend continuous segment set, and generate the recovery efficiency structured input set according to the arrangement content.

8. The method for monitoring orthopedic rehabilitation effects based on artificial intelligence according to claim 1, characterized in that, The sensor is a sensing device that is fixed near the joint to be rehabilitated by the patient. It can output posture information that reflects changes in joint angle. The data comes from an inertial measurement unit, an angle encoder, or a gyroscope. The angle sequence is a sequence of multiple angle values ​​recorded over time by a wearable joint angle sensor during the patient's performance of a designated rehabilitation movement; The original record set of rehabilitation movements is a data set obtained by combining the completion time of the specified rehabilitation movements with the corresponding joint angle sequence. The time chain is a time sequence formed by arranging the completion times of multiple designated rehabilitation actions in the order in which they occur during training. The time difference is a sequence obtained by arranging the differences between the completion times of two adjacent rehabilitation specified actions in the action time chain; The angle change segment is a set of angle records within a time range corresponding to a single rehabilitation specified action or a single time difference in the joint angle sequence; The joint motion time chain structure is a composite data structure formed by establishing a correlation between the motion time chain, the time difference sequence, and the corresponding joint angle change segment.

9. The method for monitoring orthopedic rehabilitation effects based on artificial intelligence according to claim 1, characterized in that, The action rhythm feature structure is a structured sequence obtained by arranging rhythm markers in the rhythm feature pattern in chronological order; The increased rhythm is a marker set for joint angle change segments with short time differences and continuous angle changes; The slowdown in rhythm is a marker set for joint angle change segments with long time differences and slow angle changes, used to reflect the slowdown in the rhythm of segment action execution. The trend continuum is a time segment formed by a continuous arrangement of multiple adjacent rhythm markers with the same marker type in the action rhythm feature structure. The recovery trend continuous segment set is a data set obtained by combining multiple trend continuous segments in chronological order. The efficiency improvement segment is the trend continuous segment marked with accelerated rhythm, which is concentrated in the trend continuous segment recovery trend; The efficiency slowdown segment is a continuous trend segment marked by a slowdown in the rhythm, which is concentrated in the continuous segment of the recovery trend. The structured input set for restoring efficiency is a set of structured data formed by combining efficiency-enhancing segments and efficiency-decelerating segments in the training order.

10. An artificial intelligence-based orthopedic rehabilitation effect monitoring system, characterized in that, The system is used to implement the artificial intelligence-based orthopedic rehabilitation effect monitoring method according to any one of claims 1-9, the system comprising: The motion recording and acquisition module is used to perform S1: acquire the time information of the motion start posture signal and motion end posture signal collected by the wearable joint angle sensor, calculate the completion time of the specified rehabilitation motion, organize the joint angle sequence formed by the sensor in the order of appearance, and combine the completion time information and the angle sequence to generate the original set of rehabilitation motion records. The rhythm association construction module is used to execute S2: based on the original record set of rehabilitation movements, the completion time of the specified rehabilitation movements generated by multiple training sessions is arranged in the training order to form a movement time chain. The completion time of adjacent movements in the movement time chain is calculated to obtain a time difference sequence, and the sequence is associated with the corresponding joint angle change segment to generate a joint movement time chain structure. The rhythm pattern extraction module is used to execute S3: Based on the joint action time chain structure, the time difference sequence and the joint angle change amplitude are compared. The rhythm of the segment with a short time difference and continuous angle change is accelerated, and the rhythm of the segment with a long time difference and slow angle change is slowed down. The marks are combined in sequence to form a rhythm feature pattern to obtain the action rhythm feature structure. The trend segment generation module is used to execute S4: based on the action rhythm feature structure, perform consistency judgment on adjacent rhythm marks, record rhythm segments with the same mark as the same trend continuous segment, take the inconsistent part as the starting point of the new segment, and combine all trend continuous segments in order to generate a set of restored trend continuous segments; The efficiency structure classification module is used to execute S5: perform attribute judgment on each trend segment according to the set of continuous recovery trend segments, classify segments with accelerated pace as efficiency improvement segments, classify segments with slowed pace as efficiency deceleration segments, and combine the segments into an overall structure according to the training order to obtain the structured input set of recovery efficiency.

Images