Rehabilitation training process risk prediction system and method

Abstract

The invention discloses a rehabilitation training process risk prediction system and method, and relates to the technical field of rehabilitation risk prediction, and the method comprises the steps: monitoring an illumination parameter of a rehabilitation training region, patient kinematics data and pressure center trajectory data, and generating an environment interference early warning signal when the illumination parameter deviates from a preset comfort interval; based on the environment interference early warning signal, motion sequence stage analysis of the kinematics data of the patient is started, an expected motion type of the patient is identified, a starting time point of the expected motion type of the patient is determined, and a pre-coping strategy analysis instruction is generated; based on a pre-coping strategy analysis instruction, pressure center trajectory data within a time window before a starting time point of a patient's expected action type is backtracked and analyzed to generate a strategy validity index. According to the method, the pressure center track data is backtracked and analyzed in the time window before the expected action starting time point of the patient, so that the quantitative evaluation of the pre-posture regulation and control capability of the patient under the environment interference is realized.

Description

Technical Field

[0001] This invention relates to the field of rehabilitation risk prediction technology, and in particular to a risk prediction system and method for the rehabilitation training process. Background Technology

[0002] In rehabilitation training, dynamic assessment and intervention of fall risk are crucial for ensuring patient safety and improving rehabilitation efficiency. Effective risk prediction not only requires accurately capturing the patient's current motor state but also the ability to proactively perceive potential disturbances, thereby establishing a basis for intervention in the early stages of risk development. An ideal prediction mechanism should integrate multi-dimensional physiological and environmental information, identify abnormal triggers before movement execution, and dynamically adjust the risk level based on individual behavioral patterns to support precise and timely clinical responses. This is of great significance for building a closed-loop, adaptive, intelligent rehabilitation support system and is the core foundation for achieving personalized rehabilitation intervention.

[0003] Current rehabilitation risk prediction technologies have significant limitations in modeling coupled environmental variables, especially under common environmental disturbances such as changes in light intensity. They lack a mechanism for linking patient movement intentions with the effectiveness of coping strategies. Existing technologies typically rely on single kinematic data for risk assessment, failing to trigger targeted movement intention identification and strategy retrospection at the initial stage of environmental disturbances. Therefore, current rehabilitation training risk prediction technologies still suffer from response lag and insufficient strategy adaptation in environment-behavior interaction modeling. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a risk prediction method for the rehabilitation training process to address the problems of response lag and insufficient strategy adaptation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for predicting risks in the rehabilitation training process, which includes monitoring the illumination parameters of the rehabilitation training area, patient kinematic data, and stress center trajectory data, and generating an environmental interference warning signal when the illumination parameters deviate from the preset comfort range. Based on environmental interference early warning signals, the system initiates phased analysis of the patient's kinematic data to identify the patient's expected movement type and determine the starting time point of the patient's expected movement type, and generates pre-coping strategy analysis instructions. Based on the pre-coping strategy analysis instructions, the stress center trajectory data within the time window before the start time of the patient's expected action type is retrospectively analyzed to generate a strategy effectiveness index. An interactive evaluation of environmental interference early warning signals and strategy effectiveness indices is conducted to dynamically generate real-time fall risk levels.

[0007] In a preferred embodiment of the risk prediction method for the rehabilitation training process described in this invention, the generation of environmental interference early warning signals is specifically as follows: Ambient light sensors, inertial measurement sensor nodes, and plantar pressure distribution instruments are deployed in the rehabilitation training area, and a synchronous timing reference is established. Based on the synchronous timing reference, the illumination parameters collected by the ambient light sensor, the patient kinematic data collected by the inertial measurement sensor node, and the pressure center trajectory data collected by the plantar pressure distribution instrument are continuously read and stored in the circular data buffer. Retrieve illumination parameters from the circular data buffer and query the database of illumination comfort zones that match the current rehabilitation training task; The acquired illumination parameters are compared with a lighting comfort zone database. When the illumination parameters deviate from the preset comfort zone in the database, an environmental interference warning signal is generated.

[0008] As a preferred embodiment of the risk prediction method for the rehabilitation training process described in this invention, the stepwise analysis of the motion sequence of the patient's kinematic data involves performing motion signal denoising and reconstruction processing on the patient's kinematic data, using an adaptive window sliding method to segment the motion flow, and extracting multi-scale motion feature vectors.

[0009] As a preferred embodiment of the risk prediction method for the rehabilitation training process described in this invention, the step of identifying the patient's expected movement type and determining the starting time point of the patient's expected movement type is as follows: Retrieve all movement templates from the standard rehabilitation movement library; A first-level coarse matching of hierarchical dynamic time warping matching is performed on the multi-scale motion feature vectors and the retrieved action templates to filter out a subset of candidate action templates; A second-level fine-matching process using hierarchical dynamic time warping matching is performed on a subset of candidate action templates to obtain precise matching results. Identify the patient's expected action type based on precise matching results; Based on the dynamic time warping alignment path in the second-layer fine matching process, the starting point of the adaptive sliding window is determined as the starting time point of the patient's expected action type.

[0010] In a preferred embodiment of the risk prediction method for the rehabilitation training process described in this invention, the generation of the strategy effectiveness index is specifically as follows: Extract the start time point of the patient's expected action type from the pre-coping strategy analysis instructions; Determine the start and end times of the retrospective time window based on the start time of the patient's expected action type; Based on the determined start and end times of the backtracking time window, extract the pressure center trajectory data for the corresponding time period from the circular data buffer; Multi-scale entropy analysis was performed on the intercepted pressure center trajectory data to obtain the trajectory complexity index. At the same time, dynamic time warping matching was performed on the intercepted pressure center trajectory data to obtain the trajectory similarity index. By fusing the trajectory complexity index and the trajectory similarity index, a strategy effectiveness index is generated.

[0011] As a preferred embodiment of the risk prediction method for the rehabilitation training process described in this invention, the following steps are taken: Multi-scale entropy analysis is performed on the intercepted pressure center trajectory data to obtain a trajectory complexity index; simultaneously, dynamic time warping matching is performed on the intercepted pressure center trajectory data to obtain a trajectory similarity index, as detailed below. The captured pressure center trajectory data is subjected to multi-scale coarse-graining processing to generate a multi-scale time series. Perform sample entropy analysis on multi-scale time series to obtain sample entropy values, and aggregate the sample entropy values ​​to generate a trajectory complexity index; The captured pressure center trajectory data is dynamically time-warped and matched with the ideal pressure center trajectory template in the standard rehabilitation movement library. The optimal alignment path is found through the dynamic time warping algorithm, and the corresponding cumulative distance is obtained from the optimal alignment path. The cumulative distance is converted into a trajectory similarity index.

[0012] As a preferred embodiment of the risk prediction method for the rehabilitation training process described in this invention, the optimal alignment path refers to the curved path that achieves optimal nonlinear alignment between the extracted pressure center trajectory data and the ideal pressure center trajectory template in the standard rehabilitation movement library using a dynamic time warping algorithm.

[0013] As a preferred embodiment of the risk prediction method for the rehabilitation training process described in this invention, the dynamic generation of real-time fall risk levels is specifically as follows: Receive environmental interference early warning signals and strategy effectiveness index, and read the deviation parameters contained in the environmental interference early warning signals; Input the deviation degree parameter and the strategy effectiveness index into the fuzzification transformation rule to convert the deviation degree parameter and the strategy effectiveness index into fuzzy linguistic variables of high, medium and low environmental interference and high, medium and low strategy effectiveness; The transformed fuzzy linguistic variables of environmental disturbances and policy effectiveness are input into a predefined fuzzy rule base for fuzzy inference to generate fuzzy output quantities. The fuzzy output is defuzzified to obtain the risk value; Based on the preset risk level mapping table, the risk value is mapped to a specific real-time fall risk level.

[0014] In a preferred embodiment of the risk prediction method for the rehabilitation training process described in this invention, the risk value is obtained as follows: The weighted average method is used to process the fuzzy output, and the high, medium and low risks in the fuzzy output are assigned preset weights respectively; The fuzzy values ​​for high, medium, and low risk are weighted and summed with their corresponding preset weights to generate a single risk value.

[0015] Secondly, the present invention provides a risk prediction system for the rehabilitation training process, comprising, The monitoring module is used to monitor the illumination parameters, patient kinematic data, and center of pressure trajectory data in the rehabilitation training area. When the illumination parameters deviate from the preset comfort range, an environmental interference warning signal is generated. The instruction module is used to initiate phased analysis of the patient's kinematic data based on environmental interference warning signals, identify the patient's expected action type and determine the start time point of the patient's expected action type, and generate pre-response strategy analysis instructions. The analysis module is used to retrospectively analyze the stress center trajectory data within a time window before the start time of the patient's expected action type, based on the pre-response strategy analysis instructions, and generate a strategy effectiveness index. The assessment module is used to interactively evaluate environmental interference early warning signals and strategy effectiveness indices, and dynamically generate real-time fall risk levels.

[0016] The beneficial effects of this invention are as follows: By retrospectively analyzing the center of pressure trajectory data within a time window prior to the patient's expected action initiation time, and combining multi-scale entropy analysis and dynamic time warping matching to extract trajectory complexity and trajectory similarity indices respectively, a quantitative assessment of the patient's pre-positional control ability under environmental interference is achieved. This method overcomes the limitations of traditional risk assessment methods that rely solely on data from the action execution phase, enabling the strategy effectiveness index to truly reflect an individual's stability readiness state before action initiation. This provides a forward-looking basis with physiological significance and temporal logic for the dynamic generation of subsequent fall risk levels. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the 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.

[0018] Figure 1 A flowchart for risk prediction methods in the rehabilitation training process; Figure 2A schematic diagram of a risk prediction system for the rehabilitation training process; Figure 3 A flowchart for generating environmental interference early warning signals; Figure 4 A flowchart for generating a strategy effectiveness index. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0022] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a method for predicting risks in the rehabilitation training process, including the following steps: S1. Monitor the illumination parameters, patient kinematic data, and center of pressure trajectory data of the rehabilitation training area. When the illumination parameters deviate from the preset comfort range, generate an environmental interference warning signal.

[0023] Ambient light sensors, inertial measurement sensor nodes, and plantar pressure distribution instruments are deployed in the rehabilitation training area, and a synchronous time-series benchmark is established. The synchronous time-series benchmark is a reference standard for time alignment of the data collected by the ambient light sensors, inertial measurement sensor nodes, and plantar pressure distribution instruments through a unified timestamp, ensuring the consistency of illumination parameters, patient kinematic data, and pressure center trajectory data on the time axis. Based on the synchronous timing reference, the illumination parameters collected by the ambient light sensor, the patient kinematic data collected by the inertial measurement sensor node, and the pressure center trajectory data collected by the plantar pressure distribution instrument are continuously read and stored in the circular data buffer. Illumination parameters are obtained from the circular data buffer, and the illumination comfort zone database that matches the current rehabilitation training task is queried. The illumination comfort zone database is an existing database containing preset comfort zones, which are used to assess whether the illumination intensity collected by the ambient light sensor is suitable for the patient's visual comfort and the safety of rehabilitation training.

[0024] The acquired illumination parameters are compared with a lighting comfort zone database. When the illumination parameters deviate from the preset comfort zone in the database, an environmental interference warning signal is generated. The preset comfort zone is set based on lighting requirements (i.e., the specified light intensity to ensure visual comfort and safety, suitable for work or rehabilitation environments, avoiding misjudgment due to excessive darkness or glare due to excessive brightness). For example, the value is set between 100 and 500 lux. 100 to 500 lux supports visual comfort and rehabilitation safety, below 100 lux may cause misjudgment of actions, and above 500 lux may cause glare.

[0025] S2. Based on the environmental interference early warning signal, initiate the phased analysis of the patient's kinematic data to identify the patient's expected action type and determine the starting time point of the patient's expected action type, and generate pre-response strategy analysis instructions.

[0026] Receive environmental interference warning signals to trigger phased analysis of the patient's kinematic data motion sequence; The patient's kinematic data underwent motion signal denoising and reconstruction. An adaptive window sliding method was used to segment the motion flow, and multi-scale motion feature vectors were extracted, as detailed below: The process of performing motion signal denoising and reconstruction on patient kinematic data involves decomposing the patient kinematic data into signal components of different frequencies, removing high-frequency noise signal components from the different frequency signal components, and then combining the remaining signal components to form denoised patient kinematic data. The adaptive window sliding method for segmenting motion flow adjusts the window size based on acceleration changes in the denoised patient kinematic data, thus separating independent motion phases from the continuously denoised patient kinematic data. Extracting multi-scale motion feature vectors involves processing denoised patient kinematic data using wavelet transform. At a coarse-scale level, the time series of the denoised patient kinematic data is analyzed to identify overall motion trends and record the energy distribution values. Specifically, at the coarse-scale level, low-frequency components of the denoised patient kinematic data are extracted using wavelet transform as a coarse-scale time series. Repetitive or continuous fluctuation patterns are observed in the coarse-scale time series to identify overall motion trends (such as periodic fluctuations in gait cycles). The coarse-scale time series is divided into fixed time windows (e.g., 1 second), and the squares of the data points within each time window are summed. Record energy distribution values ​​to reflect movement intensity; analyze the time series of denoised patient kinematic data at a fine-scale level to identify local movement changes and record the peak positions of the denoised patient kinematic data. Specifically, compare data point values ​​in the fine-scale time series of denoised patient kinematic data through a sliding window, record the timestamp or index corresponding to the maximum or minimum value of the denoised patient kinematic data, mark the peak position (such as the acceleration peak in the arm raising movement), and record the peak position to reflect the instantaneous characteristics of the movement; concatenate the energy distribution values ​​and peak positions of the denoised patient kinematic data to form a multi-scale motion feature vector containing energy distribution values ​​and peak positions.

[0027] The multi-scale motion feature vectors are hierarchically and dynamically time-warped and matched with motion templates in a standard rehabilitation motion library to identify the patient's expected motion type and determine the starting point of the adaptive sliding window as the starting time point of the patient's expected motion type, as detailed below: Retrieve all movement templates from the standard rehabilitation movement library; the standard rehabilitation movement library is an existing database that stores various movement templates (such as gait and arm raising movements) based on clinical rehabilitation training data or standard movement samples, which are used to match with the denoised patient kinematic data to identify the movement type. A first-level coarse matching of hierarchical dynamic time warping is performed on the multi-scale motion feature vectors and the retrieved action templates to filter out a subset of candidate action templates. The first-level coarse matching of hierarchical dynamic time warping is achieved by comparing the overall features of the multi-scale motion feature vectors with the action templates in the standard rehabilitation action library using the dynamic time warping algorithm, and initially filtering out the subset of action templates with the highest similarity to the denoised patient kinematic data as the candidate action template subset. The dynamic time warping algorithm is an existing algorithm that finds the matching path with the minimum cumulative distance by non-linearly aligning two time series (such as multi-scale motion feature vectors and action templates) on the time axis, and is used to compare action similarity. The second layer of fine matching of hierarchical dynamic time warping matching is performed on the subset of candidate action templates to obtain the precise matching result. The second layer of fine matching of hierarchical dynamic time warping matching is to compare the local features of the subset of candidate action templates with the multi-scale motion feature vectors in detail through the dynamic time warping algorithm, and obtain the action template that is most similar to the denoised patient kinematic data as the precise matching result. Identify the patient's expected action type based on precise matching results; Based on the dynamic time warping alignment path in the second-layer fine matching process, the starting point of the adaptive sliding window is determined as the starting time point of the patient's expected action type. The dynamic time warping alignment path is a curved path generated by the dynamic time warping algorithm, which performs optimal nonlinear alignment of the multi-scale motion feature vector and the action template on the time axis to mark the starting time point of the patient's expected action type.

[0028] The identified patient expected action types and their start times are encapsulated to generate pre-response strategy analysis instructions.

[0029] S3. Based on the pre-coping strategy analysis instructions, retrospectively analyze the stress center trajectory data within the time window before the start time of the patient's expected action type, and generate a strategy effectiveness index.

[0030] Extract the start time point of the patient's expected action type from the pre-coping strategy analysis instructions; Determine the start and end times of the retrospective time window based on the start time of the patient's expected action type; Based on the determined start and end times of the backtracking time window, extract the pressure center trajectory data for the corresponding time period from the circular data buffer; Multi-scale entropy analysis was performed on the extracted pressure center trajectory data to obtain the trajectory complexity index. Simultaneously, dynamic time warping matching was performed on the extracted pressure center trajectory data to obtain the trajectory similarity index. Details are as follows: The extracted pressure center trajectory data undergoes multi-scale coarse-grained processing to generate a multi-scale time series. Specifically, the extracted pressure center trajectory data is divided into fixed-length time windows (e.g., 5 seconds) according to time sequence. Within each time window, the mean of all data points in the extracted pressure center trajectory data is calculated to generate a sequence point representing the time window. The sequence points of all time windows are connected sequentially to form a coarse-scale time series. The multi-scale coarse-grained processing process is repeated on the extracted pressure center trajectory data, using larger time windows (e.g., 10 seconds, 20 seconds, etc.) to generate even coarser-scale time series, thus forming a set of coarse-scale time series containing different scales, which serves as the multi-scale time series. Sample entropy analysis is performed on multi-scale time series to obtain sample entropy values, and these values ​​are aggregated to generate a trajectory complexity index. Specifically, fixed-length subsequences are extracted from the multi-scale time series. The similarity between subsequences is compared using a preset similarity threshold (e.g., Euclidean distance less than 0.01), and the number of similar subsequence pairs that meet the threshold is counted. Based on the proportion of similar subsequence pairs, sample entropy values ​​are generated, reflecting the complexity of the multi-scale time series. The sample entropy values ​​from different scales are then weighted and averaged to generate the trajectory complexity index. The preset similarity threshold (e.g., Euclidean distance less than 0.01) is set based on the signal amplitude and noise level of the time series. A Euclidean distance less than 0.01 effectively distinguishes similar subsequences without being overly sensitive. A value that is too small (e.g., 0.001) will ignore some similar sequences, while a value that is too large (e.g., 0.1) will include too many noisy sequences. The extracted pressure center trajectory data is dynamically time-warped and matched with the ideal pressure center trajectory template in the standard rehabilitation movement library. The optimal alignment path is found through the dynamic time warping algorithm, and the corresponding cumulative distance is obtained from the optimal alignment path. The optimal alignment path is the curved path with the minimum cumulative distance that achieves the best non-linear alignment between the extracted pressure center trajectory data and the ideal pressure center trajectory template in the standard rehabilitation movement library on the time axis through the dynamic time warping algorithm. Dynamic time warping matching treats the captured pressure center trajectory data as a pressure center trajectory time series and the ideal pressure center trajectory template in the standard rehabilitation exercise library as an ideal pressure center trajectory time series. It calculates the Euclidean distance between each trajectory point in the pressure center trajectory time series and each template point in the ideal pressure center trajectory time series, and fills each Euclidean distance value into the row of the corresponding trajectory point in the pressure center trajectory time series and the column of the corresponding template point in the ideal pressure center trajectory time series in the distance matrix to construct the distance matrix. Each distance element in the distance matrix represents the local distance between the trajectory point in the pressure center trajectory time series and the template point in the ideal pressure center trajectory time series.

[0031] The Euclidean distance formula is as follows: ; In the formula, The Euclidean distance is the distance between the trajectory points in the time series of the pressure center trajectory and the template points in the time series of the ideal pressure center trajectory. For the trajectory points in the time series of the pressure center trajectory Axis coordinates For the trajectory points in the time series of the pressure center trajectory Axis coordinates Template points in the time series of the ideal pressure center trajectory Axis coordinates Template points in the time series of the ideal pressure center trajectory Axis coordinates; The optimal alignment path is found using the dynamic time warping algorithm by traversing each element of the distance matrix in priority order. The priority order starts with the first element of the first row of the distance matrix, processing all elements in the first row from left to right, then processing all elements in the second row, and so on, until all elements in the last row are processed. For each element position, the path from the left element position, the next element position, or the lower left element position is examined, and the shortest path is selected to continue to the next element position, thus constructing the path graph of the entire distance matrix. Starting from the upper right element position of the distance matrix, the path is gradually returned to the lower left element position along the shortest path graph, and the returned path is recorded as the optimal alignment path, representing the best nonlinear alignment path between the pressure center trajectory time series and the ideal pressure center trajectory time series. Obtaining the corresponding cumulative distance from the optimal alignment path involves extracting the sum of distance values ​​from the matrix elements along the optimal alignment path from the distance matrix, which is then used as the cumulative distance. The cumulative distance is converted into a trajectory similarity index. The conversion process involves taking the reciprocal of the cumulative distance and normalizing it to generate a trajectory similarity index, which reflects the similarity between the time series of the pressure center trajectory and the time series of the ideal pressure center trajectory. By fusing the trajectory complexity index and the trajectory similarity index, a strategy effectiveness index is generated.

[0032] S4. Conduct interactive evaluation of environmental interference early warning signals and strategy effectiveness index to dynamically generate real-time fall risk levels.

[0033] Receive environmental interference early warning signals and strategy effectiveness index, and read the deviation parameters contained in the environmental interference early warning signals; The deviation degree parameter and the strategy effectiveness index are input into the fuzzification transformation rule to convert them into fuzzy linguistic variables of high, medium, and low environmental interference and high, medium, and low strategy effectiveness, as follows: The process involves acquiring deviation parameters and a strategy effectiveness index. Based on a preset range in the fuzzification transformation rules, the deviation parameters are compared with the preset deviation range to determine the corresponding fuzzy linguistic variables (high, medium, or low). Simultaneously, the strategy effectiveness index is compared with the preset effectiveness range to determine the corresponding fuzzy linguistic variables (high, medium, or low), thus generating environmental interference fuzzy linguistic variables and strategy effectiveness fuzzy linguistic variables. The preset deviation range is set based on the lighting requirements of the rehabilitation environment, which are the light intensity range specified for the rehabilitation training environment (e.g., 0-50 lux is low, 50-150 lux is medium, and above 150 lux is high), ensuring patient visual comfort and movement safety, and avoiding misjudgment due to excessive darkness or misjudgment due to excessive brightness. Glare emission, for example, with a value range of 0-50 lux (low), 50-150 lux (medium), and above 150 lux (high), reflects the sensitivity of light to the patient's vision and movement. The advantage is that it accurately distinguishes the degree of interference. Too narrow a range (such as 0-30 lux) may ignore moderate interference, while too wide a range (such as 0-100 lux) may reduce the warning sensitivity. The preset effectiveness range is set based on the standard rehabilitation movement library, for example, with a value range of 0-0.4 (low), 0.4-0.7 (medium), and 0.7-1.0 (high). These values ​​effectively distinguish the patient's movement response ability. The advantage is that it accurately reflects the stability of movement. Too narrow a range (such as 0-0.3) may misjudge normal movements, while too wide a range (such as 0-0.6) may ignore subtle deviations. The transformed fuzzy linguistic variables of environmental disturbance and strategy effectiveness are input into a predefined fuzzy rule base for fuzzy inference to generate fuzzy output. The fuzzy rule base contains a set of predefined conditional rules for evaluating the risk level corresponding to different combinations of fuzzy linguistic variables. The predefined fuzzy rule base is based on fuzzy logic control theory and rehabilitation training risk assessment experience. It is an existing database that stores the set of conditional rules corresponding to the risk level of combinations of fuzzy linguistic variables of environmental disturbance and fuzzy linguistic variables of strategy effectiveness. Fuzzy reasoning refers to: acquiring fuzzy linguistic variables for environmental interference (e.g., high) and fuzzy linguistic variables for strategy effectiveness (e.g., low); retrieving conditional rules from a predefined fuzzy rule base that match the combination of environmental interference and strategy effectiveness (e.g., "if the environmental interference fuzzy linguistic variable is high and the strategy effectiveness fuzzy linguistic variable is low, then the risk is high"); determining the corresponding fuzzy risk level value (e.g., high risk) based on the matching conditional rules; and aggregating the fuzzy risk level values ​​of all matching conditional rules to generate a fuzzy output quantity, representing the comprehensive risk assessment of the combination of environmental interference and strategy effectiveness fuzzy linguistic variables. The fuzzy output is defuzzified to obtain the risk value. The defuzzification process is as follows: obtain the fuzzy output; process the fuzzy output using a weighted average method, assigning preset weights (e.g., high risk is 0.8, medium risk is 0.5, and low risk is 0.2) to the high, medium, and low risk values ​​respectively; and sum the fuzzy values ​​of high, medium, and low risk with their corresponding preset weights to generate a single risk value. According to the preset risk level mapping table, the risk value is mapped to the specific real-time fall risk level. The preset risk level mapping table is set based on rehabilitation training risk assessment experience and clinical fall risk standards, and includes risk value ranges (e.g. 0-0.3 is low, 0.3-0.6 is medium and 0.6-1.0 is high) and the corresponding real-time fall risk levels (low, medium and high).

[0034] This embodiment also provides a risk prediction system for the rehabilitation training process, including: The monitoring module is used to monitor the illumination parameters, patient kinematic data, and center of pressure trajectory data in the rehabilitation training area. When the illumination parameters deviate from the preset comfort range, an environmental interference warning signal is generated. The instruction module is used to initiate phased analysis of the patient's kinematic data based on environmental interference warning signals, identify the patient's expected action type and determine the start time point of the patient's expected action type, and generate pre-response strategy analysis instructions. The analysis module is used to retrospectively analyze the stress center trajectory data within a time window before the start time of the patient's expected action type, based on the pre-response strategy analysis instructions, and generate a strategy effectiveness index. The assessment module is used to interactively evaluate environmental interference early warning signals and strategy effectiveness indices, and dynamically generate real-time fall risk levels.

[0035] This embodiment also provides a computer device applicable to the risk prediction method for rehabilitation training process, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the risk prediction method for rehabilitation training process proposed in the above embodiment.

[0036] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0037] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the risk prediction method for the rehabilitation training process as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0038] In summary, this invention achieves a quantitative assessment of a patient's pre-positional control ability under environmental disturbances by retrospectively analyzing the center of pressure trajectory data within a time window prior to the patient's expected action initiation time, and by combining multi-scale entropy analysis and dynamic time warping matching to extract trajectory complexity and trajectory similarity indices, respectively. This method overcomes the limitations of traditional risk assessment methods that rely solely on data from the action execution phase, enabling the strategy effectiveness index to truly reflect an individual's stability readiness state before action initiation. This provides a forward-looking basis with physiological significance and temporal logic for the dynamic generation of subsequent fall risk levels.

[0039] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A rehabilitation training process risk prediction method, characterized by: The method comprises the steps of: monitoring the light parameter, the patient kinematics data and the center of pressure trajectory data in the rehabilitation training area, and generating an environmental disturbance warning signal when the light parameter deviates from the preset comfortable interval; based on the environmental disturbance warning signal, starting the motion sequence stage analysis of the patient kinematics data, identifying the patient's expected action type and determining the starting time point of the patient's expected action type, and generating a pre-responding strategy analysis instruction; based on the pre-responding strategy analysis instruction, backtracking the center of pressure trajectory data within the time window before the starting time point of the patient's expected action type, and generating a strategy effectiveness index; interactively evaluating the environmental disturbance warning signal and the strategy effectiveness index, and dynamically generating a real-time fall risk level.

2. The rehabilitation training process risk prediction method of claim 1, wherein: The method for generating an environmental disturbance warning signal comprises the following steps: deploying an environmental light sensor, an inertial measurement sensor node and a plantar pressure distribution instrument in the rehabilitation training area, and establishing a synchronous timing reference; based on the synchronous timing reference, continuously reading the light parameter collected by the environmental light sensor, the patient kinematics data collected by the inertial measurement sensor node, and the center of pressure trajectory data collected by the plantar pressure distribution instrument, and storing them in a cyclic data buffer; obtaining the light parameter from the cyclic data buffer and querying the light comfort interval database matched with the current rehabilitation training task; comparing the obtained light parameter with the light comfort interval database, and generating an environmental disturbance warning signal when the light parameter deviates from the preset comfortable interval of the light comfort interval database.

3. The rehabilitation training process risk prediction method of claim 1, wherein: The motion sequence stage analysis of the patient kinematics data is to perform motion signal denoising reconstruction processing on the patient kinematics data, use an adaptive window sliding method to segment the motion stream, and extract a multi-scale motion feature vector.

4. The rehabilitation training process risk prediction method of claim 1, wherein: The method for identifying the patient's expected action type and determining the starting time point of the patient's expected action type comprises the following steps: retrieve all action templates from a standard rehabilitation action library; performing a first layer coarse matching of hierarchical dynamic time warping matching on the multi-scale motion feature vector and the retrieved action templates to filter out a candidate action template subset; performing a second layer fine matching of hierarchical dynamic time warping matching on the candidate action template subset to obtain an accurate matching result; identifying the patient's expected action type according to the accurate matching result; determining the starting point of the adaptive sliding window as the starting time point of the patient's expected action type based on the dynamic time warping alignment path in the second layer fine matching process.

5. The rehabilitation training process risk prediction method of claim 1, wherein: The method for generating a strategy effectiveness index comprises the following steps: extracting the starting time point of the patient's expected action type from the pre-responding strategy analysis instruction; determining the starting time and the ending time of the backtracking time window according to the starting time point of the patient's expected action type; cutting out the center of pressure trajectory data of the corresponding time period from the cyclic data buffer according to the determined starting time and ending time of the backtracking time window; performing multi-scale entropy analysis on the cut-out center of pressure trajectory data to obtain a trajectory complexity index, and performing dynamic time warping matching on the cut-out center of pressure trajectory data to obtain a trajectory similarity index; fusing the trajectory complexity index and the trajectory similarity index to generate a strategy effectiveness index.

6. The rehabilitation training process risk prediction method of claim 5, wherein: The multiscale entropy analysis is performed on the intercepted pressure center trajectory data to obtain a trajectory complexity index, and meanwhile, dynamic time warping matching is performed on the intercepted pressure center trajectory data to obtain a trajectory similarity index, specifically as follows, The multiscale time series is generated by performing multiscale coarse-graining processing on the intercepted pressure center trajectory data; The sample entropy analysis is performed on the multiscale time series to obtain a sample entropy value, and the sample entropy values are aggregated to generate a trajectory complexity index; The dynamic time warping matching is performed on the intercepted pressure center trajectory data and the ideal pressure center trajectory template in the standard rehabilitation action library, and the optimal alignment path is found by the dynamic time warping algorithm, and the corresponding cumulative distance is obtained from the optimal alignment path; The cumulative distance is converted into a trajectory similarity index.

7. The rehabilitation training process risk prediction method of claim 6, wherein: The optimal alignment path refers to the curved path that is best nonlinearly aligned on the time axis between the intercepted pressure center trajectory data and the ideal pressure center trajectory template in the standard rehabilitation action library by the dynamic time warping algorithm.

8. The rehabilitation training process risk prediction method of claim 1, wherein: The real-time fall risk level is dynamically generated, specifically as follows, The environmental interference early warning signal and the strategy effectiveness index are received, and the deviation degree parameter contained in the environmental interference early warning signal is read; The deviation degree parameter and the strategy effectiveness index are input into the fuzzification conversion rule to convert the deviation degree parameter and the strategy effectiveness index into fuzzy language variables of high, medium and low environmental interference and high, medium and low strategy effectiveness; The converted fuzzy language variables of the environmental interference and the strategy effectiveness are input into the pre-defined fuzzy rule base for fuzzy reasoning to generate a fuzzy output quantity; The fuzzy output quantity is de-fuzzified to obtain a risk value; According to a pre-set risk level mapping table, the risk value is mapped to a specific real-time fall risk level.

9. The rehabilitation training process risk prediction method of claim 8, wherein: The risk value is obtained, specifically as follows, The fuzzy output quantity is processed using the weighted average method, and the risk high, medium and low in the fuzzy output quantity are respectively assigned a pre-set weight; The fuzzy values of risk high, medium and low and the corresponding pre-set weights are weighted and summed to generate a single risk value.

10. A rehabilitation training process risk prediction system based on the rehabilitation training process risk prediction method according to any one of claims 1-9, characterized in that: It comprises, A monitoring module for monitoring the illumination parameter, the patient kinematics data and the pressure center trajectory data of the rehabilitation training area, and generating an environmental interference early warning signal when the illumination parameter deviates from a pre-set comfortable interval; An instruction module for starting the motion sequence phased analysis of the patient kinematics data based on the environmental interference early warning signal, identifying the patient's expected action type and determining the starting time point of the patient's expected action type, and generating a pre-responding strategy analysis instruction; An analysis module for backtracking the pressure center trajectory data within a time window before the starting time point of the patient's expected action type based on the pre-responding strategy analysis instruction, and generating a strategy effectiveness index; An evaluation module for interactive evaluation of the environmental interference early warning signal and the strategy effectiveness index, and dynamically generating a real-time fall risk level.

Images