Self-adaptive closed-loop control method and system for hand joint rehabilitation

Abstract

The invention relates to the field of medical rehabilitation training control, in particular to a hand joint rehabilitation self-adaptive closed-loop control method and system. The method comprises the following steps: acquiring equipment and joint parameters, performing movement track configuration and compensation registration, and constructing a fixed configuration structure; on the basis of the generated session configuration and the individualized security threshold, threshold comparison configuration loading is completed; then angle and torque data are collected in real time, motion intention and spasm state judgment is carried out, and a self-adaptive motor control instruction is generated according to the motion intention and spasm state judgment; and finally, performing archiving processing on the training data and generating a progress report and a threshold updating record. Self-adaptive adjustment of the hand joint rehabilitation training process is achieved through closed-loop control, the auxiliary force and the safety boundary can be dynamically adjusted according to the real-time state of a patient, and the training safety and individualized adaptability are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of medical rehabilitation training control, and in particular to an adaptive closed-loop control method and system for hand joint rehabilitation. Background Technology

[0002] In the field of medical rehabilitation training control, existing solutions for hand joint rehabilitation training typically focus on setting motion trajectories based on the geometric parameters of the actuator and the kinematic baseline of the joint. During training, they generate rehabilitation training session configurations and construct real-time data acquisition packages. However, these solutions suffer from limitations such as unclear relationships between motion trajectories, fixed configuration structures, and rehabilitation training session configuration generation; implicit dependencies between the generation of individualized safety threshold version records and the loading of threshold comparison configurations; and inconsistencies between the processing of motion intentions and spasticity state discrimination and the generation of control commands. Existing methods often separate the registration of composite motion trajectory parameters, compensation parameter registration, and fixed configuration field mapping from the loading of threshold comparison configurations. Furthermore, they treat the generation of rehabilitation parameter archive data packages, rehabilitation database index records, and progress report encapsulation as independent post-processing steps. In scenarios where individualized safety threshold version record generation involves constraints, this can easily lead to inconsistencies between the real-time acquisition and threshold comparison configuration structures and the motor output control command sequence structure, and difficulty in tracing the source of fields in the progress report and threshold update record structures. These limitations make it difficult to achieve stable adaptive closed-loop control for hand joint rehabilitation. For the joint processing of actuator geometric parameters and joint kinematic baselines, motion trajectories and fixed configuration structures, real-time acquisition and threshold comparison configuration structures and motor output control command sequence structures, existing technologies generally lack consistent field mapping and constraints between threshold version records and session configurations, discrimination results and control commands, and archived data packets and database index records. This makes it difficult to form a consistent process in hand joint rehabilitation training, including registration of composite motion trajectory parameters and fixed configuration field mapping, construction of real-time acquisition data packets, discrimination of movement intentions and spasticity states, generation of control commands, generation of rehabilitation parameter archived data packets, generation of rehabilitation database index records and encapsulation of progress reports. This results in insufficient continuity between the threshold update record structure and the session link, and affects the management and traceability of rehabilitation training data. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides an adaptive closed-loop control method for hand joint rehabilitation, comprising:

[0004] Obtain the geometric parameters of the actuator and the kinematic baseline of the joint, perform composite motion trajectory parameter registration, compensation parameter registration and fixed configuration field mapping processing, and generate motion trajectory and fixed configuration structure;

[0005] Based on the motion trajectory and fixed configuration structure, perform rehabilitation training session configuration generation, individualized safety threshold version record generation and threshold comparison configuration loading operations to generate real-time acquisition and threshold comparison configuration structure;

[0006] The system acquires the configuration structure for real-time acquisition and threshold comparison, performs real-time acquisition data packet construction, motion intent and spasm state discrimination and processing, and control command generation operations, and generates a motor output control command sequence structure.

[0007] The system acquires the motor output control command sequence structure, generates rehabilitation parameter archive data packages, generates rehabilitation database index records, and encapsulates progress reports, generating progress reports and threshold update record structures.

[0008] Furthermore, the process of registering compensation parameters also includes:

[0009] The compensation parameter registration and processing includes the circular arc trajectory segment parameter field, the Archimedes spiral trajectory segment parameter field, and the splicing point number field from the composite motion trajectory parameter set. It is jointly verified with the geometric parameters of the actuator, completes the selection of the origin and the binding of the direction convention of the compensation coordinate system, and generates the direction compensation parameter field and the distance compensation parameter field. The direction compensation parameter field includes the direction offset angle, the direction offset symbol, the axial mapping relationship identifier, and the compensation effective segment identifier. The distance compensation parameter field includes the zero distance offset, the stroke backlash correction amount, the fixed point tension correction amount, and the compensation effective segment identifier.

[0010] Furthermore, the process of mapping fixed configuration fields also includes:

[0011] The fixed configuration field mapping process includes the coordinates of the end support mounting holes in the actuator geometry parameters and the joint sliding direction convention in the joint kinematic baseline, generating fixed point mapping table fields, strap tension level fields, and training action configuration table fields. The fixed point mapping table fields include the device-side fixed point number, patient-side fixed point number, external force application point number, strap path number, strap contact surface identifier, and fixed point verification time label. The strap tension level field includes the level number, tension amount caliber, locking status mark, and tension verification time label. The training action configuration table fields include the action number field, action type field, segment sequence reference field, splicing point sequence number reference field, intra-segment velocity planning identifier reference field, intra-segment pause time label reference field, direction compensation effective segment identifier reference field, and distance compensation effective segment identifier reference field.

[0012] Furthermore, the process of generating rehabilitation training session configurations also includes:

[0013] The rehabilitation training session configuration generation process includes fields for composite motion trajectory parameter set, compensation parameter set, fixed point mapping table, and training action configuration table in the motion trajectory and fixed configuration structure. It synchronously accesses the patient file data package and threshold update record structure to generate session number field, angle sensor channel number field, torque sensor channel number field, and acquisition scheduling parameter field. The session number field includes device identification segment, patient identification segment, session start time tag segment, and random anti-repetition segment. The angle sensor channel number field includes channel number, sampling pitch diameter, range diameter, zero-point calibration mark, and channel status mark. The torque sensor channel number field includes channel number, sampling pitch diameter, range diameter, zero-point drift calibration mark, and channel status mark. The acquisition scheduling parameter field includes acquisition cycle mark, sliding time window length mark, buffer partition number mark, and missing measurement tolerance mark.

[0014] Furthermore, the process of generating personalized security threshold version records also includes:

[0015] The personalized safety threshold version record generation process includes the patient identity identifier, affected side information, previous assessment record index and training restriction notes from the patient file data package, as well as the threshold version number, threshold time tag and update source identifier from the threshold update record structure. It generates patient ability assessment field, angle safety interval field, torque safety interval field, threshold version number field and threshold time tag field. The patient ability assessment field includes assessment source identifier, assessment time tag, joint active range of motion record index, muscle tone grading record index and pain subjective record index. The angle safety interval field includes lower angle boundary, upper angle boundary, boundary source identifier and boundary constraint mark. The torque safety interval field includes lower torque boundary, upper torque boundary, boundary source identifier and boundary constraint mark.

[0016] Furthermore, the process of constructing real-time data acquisition packets also includes:

[0017] The real-time acquisition data packet construction process includes the angle sensor channel number field, torque sensor channel number field, and acquisition scheduling parameter field in the real-time acquisition and threshold comparison configuration structure. It performs acquisition frame number binding, channel calibration and threshold field loading, and missing segment marking processing to generate a real-time acquisition data packet. The acquisition frame number binding includes aligning the angle sampling sequence field and the torque sampling sequence field at the frame level according to the acquisition frame number sequence field. The channel calibration and threshold field loading includes performing channel zero-point calibration and zero-point drift calibration on the zero-point calibration mark in the angle sensor channel number field and the zero-point drift calibration mark in the torque sensor channel number field, and writing the angle safety interval field and the torque safety interval field into the real-time acquisition data packet. The missing segment marking processing includes performing continuity verification and time interval verification on the acquisition frame number sequence field to generate a missing segment mark.

[0018] Furthermore, the process of discriminating between motor intention and spastic state also includes:

[0019] The motion intent and spasticity state discrimination process includes angle sampling sequence fields and torque sampling sequence fields, as well as angle safety interval fields and torque safety interval fields, from the real-time data acquisition data packet. It performs threshold comparison marker sequence generation, motion state feature extraction, and spasticity trigger fragment annotation to generate motion intent and spasticity state discrimination results. The threshold comparison marker sequence generation includes interval judgment of angle sampling values ​​with the lower and upper boundaries of the angle to generate angle comparison markers, and interval judgment of torque sampling values ​​with the lower and upper boundaries of the torque to generate torque comparison markers. The markers are then merged according to the conflict handling subfield in the threshold comparison rule field to generate the threshold comparison marker sequence field. The motion state feature extraction includes performing window-based angle change trend extraction, torque change trend extraction, and synchronization determination on the angle sampling sequence field and torque sampling sequence field to generate the motion intent label sequence field. The spasticity trigger fragment annotation includes torque abrupt change segments and homing determination results, which are combined with the threshold comparison marker sequence field to generate the spasticity fragment annotation sequence field.

[0020] Furthermore, the process of generating control instructions also includes:

[0021] The control command generation operation includes the motion intention label sequence field, threshold comparison mark sequence field, and spasm fragment annotation sequence field from the motion intention and spasm state discrimination results, as well as the adaptive control parameter table field and acquisition scheduling parameter field from the real-time acquisition and threshold comparison configuration structure. It performs control command mapping, output limiting, and command issuance processing to generate a motor output control command sequence structure. The control command mapping includes mapping the motion intention label sequence field to the motor target output sequence field according to the control command mapping caliber and action segment association caliber in the adaptive control parameter table field. The output limiting includes the threshold comparison mark sequence field and the spasm fragment annotation sequence field. It combines the normal limiting boundary and abnormal limiting boundary in the output limiting parameter field to perform pruning and rate of change constraint on the target output value. The command issuance includes generating a control cycle trigger event according to the acquisition cycle mark in the acquisition scheduling parameter field and writing the target output value into the drive motor control interface.

[0022] Furthermore, the process of packaging and processing progress reports also includes:

[0023] The progress report encapsulation process includes rehabilitation database index records, performs report field extraction and formatting encapsulation, and threshold update record field registration processing to generate a progress report and threshold update record structure. The report field extraction and formatting encapsulation process includes extracting angle range summary fields, velocity change summary fields, torque peak segment summary fields, and control command summary fields from angle archive sequence fields, torque archive sequence fields, velocity archive sequence fields, and command archive sequence fields, and writing them into the report archive object. The threshold update record field registration process includes threshold version number fields and threshold time stamp fields, and combines them with output amplitude limiting parameter fields and spastic fragment annotation sequence fields to generate threshold update record structure fields.

[0024] Furthermore, a hand joint rehabilitation adaptive closed-loop control system includes: a hand joint rehabilitation adaptive closed-loop control system, a rehabilitation training session configuration structure generation module, an individualized safety threshold version recording and real-time acquisition and threshold comparison configuration structure generation module, a real-time acquisition data packet generation module, a movement intention and spasticity state discrimination module, a motor output control command sequence structure generation module, a rehabilitation parameter archiving data packet generation module, and a rehabilitation database index record and progress report and threshold update record structure generation module; used to implement the method described in any of the above.

[0025] The key innovations of this invention include:

[0026] (1) Establish a motion trajectory and fixed configuration structure generation link driven by the geometric parameters of the actuator and the kinematic baseline of the joint. Complete the registration of composite motion trajectory parameters, compensation parameter registration and fixed configuration field mapping within the same link, so that the motion trajectory and fixed configuration structure and the rehabilitation training session configuration generation form a callable input-output relationship in terms of field scope.

[0027] (2) The rehabilitation training session configuration generation, individualized security threshold version record generation and threshold comparison configuration loading operations are connected in the same main link to generate a real-time acquisition and threshold comparison configuration structure, so that the individualized security threshold version record is used as the loading basis for threshold comparison rule field and acquisition scheduling parameter field and participates in the continuous reference of session-level configuration.

[0028] (3) Taking the real-time acquisition and threshold comparison configuration structure as a unified entry point, the real-time acquisition data packet construction, motion intention and spasticity state discrimination processing and control command generation operation closed loop is connected to generate the motor output control command sequence structure. In the subsequent process, the progress report and threshold update record structure are output through rehabilitation parameter archive data packet generation, rehabilitation database index record generation and progress report encapsulation processing, so as to realize the backfilling closed loop of the threshold update record structure for the aforementioned session configuration link.

[0029] The following are its main beneficial effects:

[0030] (1) In response to the problem that the existing scheme separates the motion trajectory setting and fixed configuration field mapping processing, resulting in an implicit dependency between the motion trajectory and fixed configuration structure and the session configuration generation, the composite motion trajectory parameter registration, compensation parameter registration and fixed configuration field mapping processing are uniformly organized into a motion trajectory and fixed configuration structure. This enables the rehabilitation training session configuration generation to directly call the trajectory and fixed configuration fields associated with the training action configuration table fields, reducing session configuration conflicts and duplicate maintenance of configuration items caused by inconsistent field definitions.

[0031] (2) In response to the problem that the individualized security threshold version record and threshold comparison configuration loading are separated in the existing scheme, which makes it difficult for the real-time acquisition and threshold comparison configuration structure to be consistently referenced in different sessions, the rehabilitation training session configuration generation, individualized security threshold version record generation and threshold comparison configuration loading operations are completed in the same main link. This enables the real-time acquisition and threshold comparison configuration structure to stably carry the threshold comparison rule field and acquisition scheduling parameter field under the constraint of the session number field, and enables the threshold update record structure to be continuously included in the reference source of configuration loading, thereby improving the situation of threshold loading caliber drift and session configuration disconnection.

[0032] (3) In response to the problem that the existing schemes have segmented implementation of discrimination and control, and that the archiving and reporting are independent post-processing links, it is difficult to form a traceable closed loop between the motor output control command sequence structure and the progress report and threshold update record structure. By using the real-time acquisition and threshold comparison configuration structure to run through the real-time acquisition data packet construction, motion intention and spasticity state discrimination processing and control command generation operations, a unified output of the motor output control command sequence structure is formed. In the same closed loop, the rehabilitation parameter archiving data packet generation, rehabilitation database index record generation and progress report encapsulation processing are completed, so that the progress report and threshold update record structure can establish an index association with the session-level acquisition, discrimination and control fields and backfill to the configuration link, thereby improving the traceability consistency of the training process record and threshold update record structure when connecting across sessions. Attached Figure Description

[0033] Figure 1 A flowchart illustrating an adaptive closed-loop control method for hand joint rehabilitation provided in this application embodiment;

[0034] Figure 2 This is a structural block diagram of an adaptive closed-loop control system for hand joint rehabilitation provided in an embodiment of this application. Detailed Implementation

[0035] Example 1: Refer to Figure 1 This is a flowchart illustrating an adaptive closed-loop control method for hand joint rehabilitation provided in an embodiment of the present invention. The flowchart may include at least steps S100-S400:

[0036] S100: Obtain the geometric parameters of the actuator and the kinematic baseline of the joint, perform composite motion trajectory parameter registration, compensation parameter registration and fixed configuration field mapping processing, and generate motion trajectory and fixed configuration structure;

[0037] S200, based on motion trajectory and fixed configuration structure, performs rehabilitation training session configuration generation, individualized safety threshold version record generation and threshold comparison configuration loading operations, and generates real-time acquisition and threshold comparison configuration structure;

[0038] S300: Obtain the real-time acquisition and threshold comparison configuration structure, perform real-time acquisition data packet construction, motion intention and spasm state discrimination processing and control command generation operations, and generate motor output control command sequence structure.

[0039] S400: Obtain the motor output control command sequence structure, generate rehabilitation parameter archiving data packets, generate rehabilitation database index records and encapsulate progress reports, and generate progress reports and threshold update record structures.

[0040] Step S100 includes at least steps S110-S130:

[0041] S110. Obtain the geometric parameters of the actuator and the kinematic baseline of the joint, and perform registration of the circular arc trajectory segment parameters and splicing point numbers, the Archimedes spiral trajectory segment parameters, and the concave-convex law constraint terms field to obtain a composite motion trajectory parameter set containing the circular arc trajectory segment parameter field, the Archimedes spiral trajectory segment parameter field, the splicing point number field, and the concave-convex law constraint terms field.

[0042] Specifically, the actuator geometric parameters are taken from the equipment assembly configuration file and factory calibration record. These parameters include at least the drive shaft center position, coordinates of the end support mounting holes, upper and lower limits of the transmission stroke, zero-position mechanical limit position, and unit stroke angle conversion relationship. The joint kinematic baseline is taken from clinical assessment records and pre-training zero-position calibration records. This baseline includes at least an initial angle reference, allowable angle range, joint sliding direction convention, joint rotation direction convention, and baseline time label. When the system triggers the trajectory configuration loading action at the start of the rehabilitation training session, it aligns and verifies the actuator geometric parameters with the joint kinematic baseline using a unified coordinate system. This ensures the coordinate origin, rotation direction, and angle sign are consistent in dimension, and the verification conclusion is written to the trajectory parameter registration log for subsequent traceability. Furthermore, the circular arc trajectory segment parameters characterize the circular arc motion curve of the end support in the main flexion-extension motion plane of the joint. The circular arc trajectory segment parameter fields include the arc radius, arc center coordinates, starting angle, ending angle, intra-segment sampling pitch, intra-segment velocity planning identifier, and intra-segment dwell time label. The system selects the corresponding arc radius and arc center coordinates based on the initial angle reference and allowable angle range recorded in the joint kinematic baseline, and performs boundary truncation on the starting and ending angles according to the upper and lower limits of the transmission stroke recorded in the actuator geometry parameters. It performs minimum step size constraint registration on the sampling pitch within the segment and gear mapping registration on the speed planning identifier within the segment, thus completely writing the arc trajectory segment parameters into the arc trajectory segment parameter field. Furthermore, the Archimedean spiral trajectory segment parameters are used to characterize the composite curve segment under joint rotation and sliding coupling conditions. The Archimedean spiral trajectory segment parameter field includes the initial radius, pitch parameter, angle expansion range, radial increment diameter, sampling pitch within the segment, speed planning identifier within the segment, and dwell time label within the segment. The system calculates the pitch parameter and the radial increment diameter based on the joint rotation, sliding, and rotation combination relationships registered in the concave-convex law constraint field, completing the registration of the Archimedean spiral trajectory segment parameter field. The concave-convex law constraint field describes the constraint relationship and allowable deviation diameter of the synchronous change of joint rotation and sliding, and shares the intra-segment sampling pitch diameter with the circular arc trajectory segment parameter field, maintaining consistent segmented curve sampling. Furthermore, the splicing point number is used to identify the switching position between the circular arc segment and the spiral segment. The splicing point number field includes at least the splicing point sequence number, corresponding segment type identifier, splicing point geometric coordinates, splicing point tangential direction, splicing point velocity continuity marker, and splicing point time label. The system generates the splicing point geometric coordinates based on the arc termination angle and the starting point of the spiral angle expansion interval, and establishes tangential consistency constraints between adjacent segments based on the splicing point tangential direction. Subsequently, the splicing point sequence numbers are registered according to the trajectory execution order.After the system completes the above registration, it encapsulates the arc trajectory segment parameter field, the Archimedes spiral trajectory segment parameter field, the splicing point number field, and the concavity and convexity law constraint field into a composite motion trajectory parameter set. Within this composite motion trajectory parameter set, the parameter version number, generation time tag, and device identification field are recorded synchronously. The composite motion trajectory parameter set serves as the input source for subsequent step S120 within this main step, from which S120 extracts the trajectory segment parameter field and the splicing point number field for further processing. Furthermore, the concavity and convexity law constraint field in this composite motion trajectory parameter set shares the same caliber as the compensation parameter calculation process, satisfying the requirement for segmented trajectory continuity.

[0043] S120. Extract the trajectory segment parameter field and splicing point number field from the composite motion trajectory parameter set, register the direction compensation parameter field and the distance compensation parameter field, and generate a compensation parameter set containing the direction compensation parameter field, the distance compensation parameter field and the compensation parameter version number field.

[0044] Specifically, after receiving the composite motion trajectory parameter set, the system first extracts the sampling pitch within the segment, the velocity planning identifier within the segment, the tangential direction of the splicing point, and the geometric coordinates of the splicing point from the circular arc trajectory segment parameter field, the Archimedes spiral trajectory segment parameter field, and the splicing point number field. This constructs a compensation calculation input package, which is then jointly verified with the actuator's geometric parameters to complete the selection of the origin and binding of the direction convention of the compensation coordinate system. The direction compensation parameter field describes the angular offset and sign convention between the actuator's output direction and the joint's flexion-extension axis direction. This field must include at least the direction offset angle, direction offset sign, axial mapping relationship identifier, and compensation effective segment identifier. The distance compensation parameter field describes the distance offset and stroke correction caliber between the actual contact point of the end support and the theoretical trajectory point. This field must include at least the zero-position distance offset, stroke backlash correction amount, fixed-point tension correction caliber, and compensation effective segment identifier. Furthermore, the registration process for the direction compensation parameter field is triggered by the direction compensation mechanism, which consists of an adjustable direction guide component, an angle locking component, and a direction scale reading component. The direction scale reading component provides the current direction gear reading, the angle locking component provides a gear fixed status mark, and the adjustable direction guide component provides the offset angle adjustment amount. The system compares the direction gear reading with the tangential direction of the splicing point, generates a direction offset angle, and writes it into the direction offset angle field, while also writing it into the axial mapping relationship identifier field. When the angle locking component returns an unlocked status mark, the system writes a compensation anomaly record and marks the direction compensation parameter field as pending review. The pending review status triggers the manual confirmation process together with the fixed point mapping table field in the subsequent S130 configuration encapsulation stage. Furthermore, the registration process for the distance compensation parameter field is triggered by the distance compensation mechanism, which consists of an adjustable slider component, a travel distance measuring component, and a position locking component. The travel distance measuring component outputs the zero-position distance offset reading and the travel hysteresis reading, the position locking component outputs a locking status mark, and the adjustable slider component outputs the travel correction amount. The system calculates the theoretical distance from the theoretical trajectory point to the contact point of the end support component based on the arc radius and pitch parameter diameter in the composite motion trajectory parameter set. It then calculates the difference between the theoretical distance and the output reading of the travel distance measuring component to obtain the zero-position distance offset, which is written into the zero-position distance offset field. Threshold filtering is applied to the travel hysteresis reading to obtain the travel hysteresis correction amount, which is written into the travel hysteresis correction amount field. When the position locking component returns to an unlocked state, the system also writes a compensation anomaly record and marks it as pending review. Furthermore, the compensation parameter version number field is used to bind the effective range of the same batch of direction compensation parameter fields and distance compensation parameter fields. Each time the compensation mechanism parameters change and the locking verification is completed, the system triggers a version increment action, writing the new version number into the compensation parameter version number field and writing the old version number along with the new version number into the version change record.After the system completes the registration of the direction compensation parameter field and the distance compensation parameter field, it encapsulates the compensation parameter set. The compensation parameter set includes at least the direction compensation parameter field, the distance compensation parameter field, and the compensation parameter version number field. It serves as the input source for the subsequent S130 within this main step, allowing S130 to register the fixed point mapping table field, the strap tension level field, and the training action configuration table field for the compensation parameter set and the composite motion trajectory parameter set. At the same time, the compensation parameter version number field is written into the session number field associated record in the rehabilitation training session configuration structure generation stage of the subsequent S210, supporting version traceability across main steps.

[0045] S130. Register the fields of the fixed point mapping table, the strap tension level, and the training action configuration table for the compensation parameter set and the composite motion trajectory parameter set, and generate a motion trajectory and fixed configuration structure containing the composite motion trajectory parameter set field, the compensation parameter set field, the fixed point mapping table field, and the training action configuration table field.

[0046] Specifically, after receiving the compensation parameter set and the composite motion trajectory parameter set, the system first establishes a fixed-point mapping table field. This field describes the correspondence between the adjustable strap's fixed points on the device and on the patient. The device-side fixed points refer to the strap fixing holes or fixing post numbers on the end support, wrist rest, or finger rest. The patient-side fixed points refer to the proximal support position and distal force-bearing position numbers adjacent to the training target joint. The fixed-point mapping table field contains at least the device-side fixed-point number, the patient-side fixed-point number, the external force application point number, the strap path number, the strap contact surface identifier, and the fixed-point calibration time stamp. Based on the agreement between the end support mounting hole coordinates in the actuator's geometric parameters and the joint sliding direction in the joint kinematic baseline, the system generates the strap path number and registers the external force application point number. When the fixed-point calibration time stamp exceeds the validity period allowed by the session configuration, the system triggers a recalibration action, rereads the fixed-point positions, and updates the fixed-point mapping table field. Furthermore, the strap tension level field describes the tension status of the adjustable strap. After the strap is fixed, the system reads the tension level reading of the tensioning mechanism and writes it into the strap tension level field. The strap tension level field includes at least the level number, tension level, locking status mark, and tension verification time label. The tension level shares the same unit as the fixed point tension correction level in the distance compensation parameter field. When registering the strap tension level field, the system simultaneously writes the reference identifier of this shared level and jointly determines the locking status mark with the pending verification status in the compensation parameter set. When a pending verification status or an unlocked status mark appears, the system writes the status into the pre-session check record and marks the training action configuration table field as pending confirmation. The pending confirmation status triggers the session start blocking record when the rehabilitation training session configuration structure is generated in S210. Furthermore, the training action configuration table fields describe the binding relationship between training actions and segmented trajectories of the composite motion trajectory parameter set. The training action configuration table fields must include at least the following fields: action number, action type, segment sequence reference, splicing point sequence number reference, intra-segment velocity planning identifier reference, intra-segment pause time label reference, direction compensation effective segment identifier reference, and distance compensation effective segment identifier reference. The system matches the action type field with the joint rotation direction convention in the joint kinematic baseline, generates the segment sequence reference field, and writes it; it performs consistency checks between the splicing point sequence number reference field and the splicing point number field to complete the registration of the segmented trajectory execution order; and it performs linkage checks between the intra-segment velocity planning identifier reference field and the compensation effective segment identifier to complete the registration of the compensation effective range within the same action. For each modification to the training action configuration table fields, the system records the modifier's identifier, modification timestamp, and modification reason description, and associates them with the compensation parameter version number field to form an auditable version chain.In the engineering implementation scenario, before training begins, the therapist places the patient's forearm on the wrist support, aligns the finger support corresponding to the target joint with the joint axis, and then selects the device-side fixation point number according to the fixation point mapping table field prompts and completes the strap path layout. After the tensioning mechanism is locked, the system reads the gear number and writes it into the strap tension gear field. The system then reads the composite motion trajectory parameter set and combines it with the compensation parameter set to load the training action configuration table field. After completion, the composite motion trajectory parameter set field, compensation parameter set field, fixation point mapping table field, and training action configuration table field are encapsulated to generate a motion trajectory and fixation configuration structure. The motion trajectory and fixation configuration structure serves as the input source for subsequent S210 and is called by S210 when generating the rehabilitation training session configuration structure based on the motion trajectory and fixation configuration structure. Furthermore, the compensation parameter version number field in the motion trajectory and fixation configuration structure, along with the fixation point verification time label, is written into the threshold time label field as reference information during the individualized safety threshold version record generation stage in subsequent S220, realizing the cross-main step configuration association with the threshold version.

[0047] Step S200 includes at least steps S210-S230:

[0048] S210. Based on the motion trajectory and fixed configuration structure, perform session number field generation, angle sensor channel number field registration, and torque sensor channel number field registration processing to obtain a rehabilitation training session configuration structure containing session number field, patient file data packet field, threshold update record structure field, angle sensor channel number field, torque sensor channel number field, and acquisition scheduling parameter field.

[0049] Specifically, the motion trajectory and fixed configuration structure originate from the output of preceding step S130. Internally, it includes a composite motion trajectory parameter set field, a compensation parameter set field, a fixed point mapping table field, and a training action configuration table field. When the rehabilitation training start command is triggered, the system reads the motion trajectory and fixed configuration structure and simultaneously accesses the patient file data packet and threshold update record structure as session initialization input. The patient file data packet field describes the structured encapsulation of patient identification, affected side information, previous assessment record index, and training restriction notes. The threshold update record structure field describes the structured encapsulation of the threshold version number, threshold time stamp, and update source identifier generated after the previous training archive. The patient file data packet and the threshold update record structure can originate from the system's local storage partition or cloud synchronization partition. When the cloud synchronization partition is unavailable, the system calls the most recent complete version from the local storage partition and records the synchronization status flag and rollback reason in the session log. These records are bound to the session number field for easy subsequent traceability.

[0050] Furthermore, the session number field is generated by the session management unit, which consists of a session creation module, a session log module, and a session state machine module. The session number field adopts a segmented encoding structure, containing at least a device identifier segment, a patient identifier segment, a session start time tag segment, and a random anti-duplicate segment. The device identifier segment is taken from the device firmware registration information, the patient identifier segment is taken from the patient file data packet field, the session start time tag segment is taken from the system clock, and the random anti-duplicate segment is generated by a pseudo-random sequence generator and written into the anti-duplicate subfield of the session number field. After generating the session number field, the system writes it to the session log module and puts the session state machine module into the "initialization state." Only in the initialization state are the registration actions for the angle sensor channel number field and the torque sensor channel number field allowed. When the session state machine module detects that the motion trajectory and fixed configuration structure are missing training action configuration table fields or that the fixed point mapping table field is in an empty state, the system writes a session anomaly flag and switches the session state machine module to the "blocking state." The blocking state record is bound to the session number field and is referenced by the subsequent progress report field set.

[0051] Furthermore, the registration of the angle sensor channel number field and the torque sensor channel number field is performed by the acquisition channel management module, which includes a channel discovery submodule, a channel binding submodule, and a channel health self-check submodule. The angle sensor channel number field is used to identify the unique number of the angle sensor acquisition link and its caliber specification. The field must include at least the channel number, sampling pitch caliber, range caliber, zero-point calibration mark, and channel status mark. The torque sensor channel number field is used to identify the unique number of the torque sensor acquisition link and its caliber specification. The field must include at least the channel number, sampling pitch caliber, range caliber, zero-point drift calibration mark, and channel status mark. During the channel discovery submodule phase, the system scans available acquisition interfaces and bus addresses to generate a candidate channel list. During the channel binding submodule phase, the candidate channel list is matched with the sensor installation location information recorded in the patient file data packet field to complete the writing of the channel number. The channel health self-check submodule then updates the channel status flag. The update action includes three types of operations: sampling frame increment check, missing fragment ratio statistics, and abnormal mutation fragment detection. When any channel status flag is written with an abnormal status, the system writes the abnormal status into the constraint subfield of the acquisition scheduling parameter field to restrict subsequent acquisition scheduling actions from entering the high-frequency acquisition mode, and synchronously writes the abnormal status into the session log module for auditing.

[0052] Furthermore, the acquisition scheduling parameter field describes the acquisition rhythm, window slicing rules, and caching strategy of angle and torque sensors within the current session period. This field includes at least an acquisition period marker, a sliding time window length marker, a cache partition number marker, and a missing test tolerance caliber marker. When registering the acquisition scheduling parameter field, the system reads the intra-segment sampling pitch caliber and intra-segment dwell time label reference fields from the training action configuration table, writes the acquisition period marker and sliding time window length marker into the acquisition scheduling parameter field, and hash-binds the cache partition number marker with the session number field to form a stable write path. The acquisition scheduling parameter field, along with the angle sensor channel number field and the torque sensor channel number field, are encapsulated into the rehabilitation training session configuration structure. This rehabilitation training session configuration structure is written into the session creation module as the output of this section and is directly referenced in subsequent S220. S220 extracts the patient file data package field and the threshold update record structure field from the rehabilitation training session configuration structure as input to continue the individualized safety threshold version record generation process. At the same time, the session number field in the rehabilitation training session configuration structure is reused in the subsequent rehabilitation parameter archiving data package generation stage in S410, forming an index connection across main steps.

[0053] In the engineering implementation scenario, after the therapist completes patient fixation and equipment posture adjustment in the outpatient rehabilitation training area, they click the "Start Session" button on the system interface to trigger the session management unit to enter the initialization state. The system reads the motion trajectory and fixed configuration structure output by the previous main step, and retrieves the patient file data packet field and threshold update record structure field from the database. The acquisition channel management module performs discovery, binding, and health self-check on the angle sensing channel and torque sensing channel, writes the angle sensing channel number field and torque sensing channel number field, and generates the acquisition scheduling parameter field. After encapsulation, the system outputs the rehabilitation training session configuration structure, and then transmits the rehabilitation training session configuration structure to S220, which extracts the patient file data packet field and threshold update record structure field from it to perform threshold registration.

[0054] S220. Extract patient profile data packet fields and threshold update record structure fields from the rehabilitation training session configuration structure, register patient ability assessment fields, angle safety interval fields, and torque safety interval fields, and generate an individualized safety threshold version record containing threshold version number fields, patient ability assessment fields, angle safety interval fields, torque safety interval fields, and threshold time tag fields.

[0055] Specifically, when the system enters the threshold registration phase, the threshold management module first reads the rehabilitation training session configuration structure output by S210 and extracts the patient file data package field and the threshold update record structure field as input for this section. The threshold management module includes an assessment data access submodule, a threshold caliber generation submodule, a version management submodule, and an audit record submodule. The patient ability assessment field is used to structurally express the current patient's ability baseline at the start of this session. The field includes at least an assessment source identifier, assessment time tag, joint active range of motion record index, muscle tone grading record index, and pain subjective record index. The assessment source identifier is used to distinguish outpatient assessment records, home training records, or remote follow-up records. The assessment time tag establishes a consistency verification relationship with the session start time tag fragment of the session number field to avoid cross-session mismatch. The assessment data access submodule expands the previous assessment record index in the patient file data package field into an assessment record set and uses the threshold version number and threshold time tag in the threshold update record structure field as comparison input to complete the version connection of the same patient in different sessions.

[0056] Furthermore, the angle safety interval field describes the allowable angle range and boundary source for this training, and includes at least the lower angle boundary, upper angle boundary, boundary source identifier, and boundary constraint marker. The torque safety interval field describes the allowable torque range and boundary source for this training, and includes at least the lower torque boundary, upper torque boundary, boundary source identifier, and boundary constraint marker. When generating the angle safety interval field, the threshold caliber generation submodule reads the range of motion corresponding to the joint active range of motion record index in the patient capability assessment field, and combines it with the training restriction notes in the patient file data package field to write the lower angle boundary and upper angle boundary into the angle safety interval field. When the training restriction notes include postoperative restrictions or fixation brace restrictions, the boundary constraint marker is written as "strong constraint," and the constraint source is registered in the audit record submodule. When generating the torque safety interval field, the threshold caliber generation submodule reads the grading information corresponding to the muscle tone grading record index and the pain subjective record index in the patient ability assessment field, and compares it with the previous training archive summary in the threshold update record structure field. Then, it writes the lower and upper boundaries of the torque into the torque safety interval field. When the comparison finds an abnormal marker in the previous session, the boundary constraint marker is written as "convergence constraint," and a convergence basis explanation is registered in the audit record submodule. The aforementioned angle safety interval field and torque safety interval field constitute the minimum set of parameters for individualized safety thresholds, facilitating subsequent threshold comparison rule loading. Furthermore, the system can optionally load extended threshold fields, such as speed limit caliber or dwell time upper limit caliber. Extended threshold fields do not affect the main input-output relationship of the closed-loop link of this invention.

[0057] Furthermore, the version management submodule is responsible for generating and writing the threshold version number field and the threshold timestamp field. The threshold version number field identifies the threshold evolution sequence for the same patient between adjacent sessions. This field contains at least a major version number fragment, a minor version number fragment, and a version change reason fragment. The version change reason fragment is taken from the update source identifier in the threshold update record structure field and mapped to the assessment source identifier in the current patient capability assessment field. The threshold timestamp field records the time caliber and effective window marker for this threshold registration. This field contains at least a registration timestamp, an effective start timestamp, and an expiration timestamp. The registration timestamp is checked for consistency with the session start timestamp fragment in the session number field. The effective start timestamp is written from the same source as the registration timestamp. The expiration timestamp is written according to the session end event. When the system exits abnormally before the session end event arrives, the expiration timestamp is written as the abnormal exit timestamp, and the abnormal exit marker is added to the audit record submodule. After the system completes the registration of patient capability assessment fields, angle safety interval fields, and torque safety interval fields, it synchronously writes the threshold version number field and threshold time tag field, encapsulating and generating an individualized safety threshold version record. This individualized safety threshold version record is used as the output product of this subsection in the invention chain and is directly input into subsequent S230. S230 loads the threshold field, acquisition scheduling parameter field, and threshold comparison rule field into the individualized safety threshold version record, completing the generation of the real-time acquisition and threshold comparison configuration structure. At the same time, the threshold version number field is referenced in the subsequent progress report encapsulation stage of S430 and enters the threshold update record structure field, forming a closed-loop backfill across the main steps.

[0058] In the engineering implementation scenario, after the therapist completes the session initialization, they select the assessment source identifier for this training and enter the assessment record index for the day on the system interface. The system reads the patient file data packet fields to expand the assessment record set, and identifies the previous threshold version number and abnormal marker by combining the threshold update record structure fields. Then, the angle safety interval field and torque safety interval field are generated in the threshold caliber generation submodule. The version management submodule writes the threshold version number field and the threshold time tag field. The system outputs the individualized safety threshold version record and transmits it to S230 as input.

[0059] S230. Load threshold field, acquisition scheduling parameter field, and threshold comparison rule field into the individualized security threshold version record to generate a real-time acquisition and threshold comparison configuration structure containing individualized security threshold version record field, acquisition scheduling parameter field, threshold comparison rule field, and adaptive control parameter table field.

[0060] Specifically, when the system enters the configuration loading phase, the control configuration loading module receives the individualized safety threshold version record output by S220 and simultaneously reads the rehabilitation training session configuration structure output by S210. It extracts the acquisition scheduling parameter field, angle sensor channel number field, and torque sensor channel number field as parallel inputs, and then enters the threshold field loading process. Threshold field loading refers to the process of unpacking the angle safety interval field and torque safety interval field from the individualized safety threshold version record and writing them into the runtime configuration cache. The runtime configuration cache consists of a session-level cache partition and a channel-level cache partition. The session-level cache partition is indexed by the session number field, and the channel-level cache partition is indexed by the angle sensor channel number field and the torque sensor channel number field. When writing the angle safety interval field and torque safety interval field, the control configuration loading module simultaneously writes the threshold version number field and the threshold timestamp field, forming an auditable runtime snapshot. It also registers a "threshold loading complete" status flag in the session log module as one of the conditions for subsequent acquisition initiation.

[0061] Furthermore, the loading of acquisition scheduling parameter fields refers to the process of writing the acquisition period marker, sliding time window length marker, cache partition number marker, and missing test tolerance caliber marker into the acquisition scheduling executor. The acquisition scheduling executor consists of a timed trigger submodule, a window slicing submodule, and a cache writing submodule. The timed trigger submodule generates acquisition trigger events according to the acquisition period marker. The window slicing submodule performs segmented aggregation on the angle sampling sequence and torque sampling sequence according to the sliding time window length marker. The cache writing submodule writes the sampling frames into the session-level cache partition according to the cache partition number marker. During runtime, the acquisition scheduling executor checks the missing test tolerance caliber marker. When the proportion of missing test segments exceeds the missing test tolerance caliber marker, the acquisition scheduling executor writes a missing test alarm marker and binds the alarm marker to the session number field. The alarm marker is then referenced by the subsequent S310 acquisition frame number and the missing test segment marker processing, forming a consistent abnormal link across sections.

[0062] Furthermore, the threshold comparison rule field loading refers to the process of writing the threshold judgment criteria into the comparison rule engine, which consists of a threshold mapping submodule, a boundary judgment submodule, and a tag output submodule. The threshold comparison rule field includes at least an angle comparison rule subfield, a torque comparison rule subfield, and a conflict handling subfield. The angle comparison rule subfield specifies the comparison method between the angle sampling sequence and the angle safety interval field; the torque comparison rule subfield specifies the comparison method between the torque sampling sequence and the torque safety interval field; and the conflict handling subfield specifies the tag merging criteria when the angle comparison result and the torque comparison result are inconsistent. During the loading phase, the comparison rule engine writes the lower and upper boundaries of the angle safety interval field and the lower and upper boundaries of the torque safety interval field into the threshold mapping submodule, and registers boundary constraint tags in the boundary judgment submodule. When the boundary constraint tag is "strong constraint" or "convergence constraint," the conflict handling subfield writes the corresponding merging priority to avoid rule conflicts causing unstable tag output. After the threshold comparison rule field is loaded, the mark output submodule generates a rule ready status mark and writes the rule ready status mark to the session log module. The rule ready status mark and the "threshold loading completed" status mark together constitute the data acquisition start gating condition. After the gating condition is met, the data acquisition scheduler enters the running state.

[0063] Furthermore, the adaptive control parameter table field describes the set of operating parameters required for control command mapping and output limiting. The field includes at least the control command mapping caliber, the output limiting parameter caliber, and the action segment association caliber. The action segment association caliber maps the action number field and segment sequence reference field in the training action configuration table field to the scope of the control command mapping caliber. When generating the adaptive control parameter table field, the control configuration loading module reads the training action configuration table field from the motion trajectory and fixed configuration structure, extracts the action number field, the intra-segment velocity planning identifier reference field, the direction compensation effective segment identifier reference field, and the distance compensation effective segment identifier reference field, and associates the extraction results with the threshold version number field to form a session-wide parameter consistency boundary. When the training action configuration table field lacks the action number field or the segment sequence reference field, the control configuration loading module writes a loading failure flag and blocks the acquisition start gating condition, while simultaneously recording the failure reason in the session log module. After the system completes the loading of threshold fields, acquisition scheduling parameter fields, and threshold comparison rule fields, it encapsulates the individualized safety threshold version record field, acquisition scheduling parameter field, threshold comparison rule field, and adaptive control parameter table field to generate a real-time acquisition and threshold comparison configuration structure. This real-time acquisition and threshold comparison configuration structure, as the output of this section, is directly input into subsequent S310. S310 obtains the real-time acquisition and threshold comparison configuration structure and performs acquisition frame number binding, channel calibration and threshold field loading, and missing segment marking processing. Simultaneously, its internal adaptive control parameter table field is reused in subsequent S330 control command mapping and output limiting processing. Furthermore, the threshold version number field and threshold time stamp field are written back to the threshold update record structure field in subsequent S430 threshold update record field registration processing and are re-accessed during the session initialization phase of S210, forming a closed-loop transmission relationship.

[0064] In the engineering implementation scenario, after the threshold registration is completed, the system automatically triggers the control configuration loading module to run. The loading module unpacks the angle safety interval field and torque safety interval field from the individualized safety threshold version record and writes them into the runtime configuration cache. At the same time, it loads the acquisition scheduling parameter field from the rehabilitation training session configuration structure and starts the gating check of the acquisition scheduling executor. The comparison rule engine loads the threshold comparison rule field and generates a rule ready status flag. Subsequently, the loading module generates the adaptive control parameter table field and completes the encapsulation, outputs the real-time acquisition and threshold comparison configuration structure, and transmits it to S310 as the real-time acquisition entry configuration.

[0065] Step S300 includes at least steps S310-S330:

[0066] S310. Obtain the real-time acquisition and threshold comparison configuration structure, perform acquisition frame number binding, channel calibration and threshold field loading, and missing segment marking processing to obtain a real-time acquisition data packet containing acquisition frame number sequence field, acquisition time tag sequence field, angle sampling sequence field, torque sampling sequence field, angle safety interval field, and torque safety interval field.

[0067] Specifically, after the rehabilitation training session enters the running state, the system receives the real-time acquisition and threshold comparison configuration structure output by the preceding S230, and reads the angle sensor channel number field, the torque sensor channel number field, and the acquisition scheduling parameter field from the rehabilitation training session configuration structure output by S210, as parallel input sources for the real-time acquisition entry. The real-time acquisition and threshold comparison configuration structure includes at least an individualized safety threshold version record field, an acquisition scheduling parameter field, a threshold comparison rule field, and an adaptive control parameter table field. The adaptive control parameter table field only participates in writing runtime associated information in this main step and does not participate in the minimum set of threshold field loading. The system establishes a session-level cache partition in the runtime configuration cache and writes the session number field as the index primary key into the cache metadata. Subsequently, it unpacks the angle safety interval field and the torque safety interval field from the individualized safety threshold version record field and writes them into the session-level cache partition, forming a threshold field loading snapshot. The threshold field loading snapshot is bound to the threshold version number field and the threshold time stamp field, and synchronously written to the session log module as the caliber source for subsequent threshold comparison marker sequence generation. Subsequently, the system triggers the acquisition action by the acquisition scheduling executor. The acquisition scheduling executor generates acquisition trigger events according to the acquisition cycle. When each acquisition trigger event arrives, the channel-side acquisition link reads the original angle sample from the angle sensing channel identified by the angle sensing channel number field and writes it into the angle sampling buffer. It also reads the original torque sample from the torque sensing channel identified by the torque sensing channel number field and writes it into the torque sampling buffer. At the same time, the system generates an acquisition time tag and writes it into the acquisition time tag sequence field. The acquisition time tag is checked for consistency between the system clock and the session start time tag segment in the session number field to prevent cross-session time segment mixing. The acquisition frame number binding process is used to bind samples generated by the angle sampling buffer and the torque sampling buffer under the same acquisition trigger event into the same acquisition frame. The system generates an incrementing acquisition frame number in each acquisition trigger event and writes it into the acquisition frame number sequence field. At the same time, the acquisition frame number is written into the frame header metadata of the corresponding sample in the angle sampling sequence field and the torque sampling sequence field, thereby forming a frame-level alignment. When the acquisition scheduler detects jitter or repeated triggering of the acquisition trigger event, the system writes the repeated acquisition frame number into the abnormal frame flag subfield and writes the abnormal frame flag subfield into the session log module for subsequent archiving and traceability.

[0068] Furthermore, the channel calibration and threshold field loading process operates under a dual-trigger condition during the real-time acquisition phase: an initialization trigger upon entering the session running state and a periodic trigger when the sliding time window length marker reaches its boundary. During the initialization trigger phase, the system reads the zero-point calibration marker from the angle sensing channel number field and the zero-point drift calibration marker from the torque sensing channel number field, performs channel zero-point calibration and zero-point drift calibration, and writes the calibration bias into the channel calibration record subfield. The channel calibration record subfield is associated with the acquisition frame number sequence field, allowing for retrospective analysis of the effective calibration caliber in any subsequent acquisition frame. During the periodic trigger phase, the system statistically analyzes the baseline drift trend of the angle sampling sequence field and the torque sampling sequence field within the sliding time window and writes the drift estimate into the drift subfield of the channel calibration record subfield. This drift subfield does not change the original stored values ​​of the angle sampling sequence field and the torque sampling sequence field, but is referenced as an auxiliary record in the session log module when the real-time acquisition data packet is output in this main step, maintaining the originality of the sampling sequence fields. The threshold field loading is executed synchronously during the periodic triggering phase. The system reads the angle safety interval field and the torque safety interval field from the runtime configuration cache, and writes the angle safety interval field and the torque safety interval field into the corresponding field positions of the real-time acquisition data packet. The angle safety interval field contains at least the lower boundary and the upper boundary of the angle, and the torque safety interval field contains at least the lower boundary and the upper boundary of the torque. The above four types of boundaries constitute the minimum threshold set of the real-time acquisition data packet in this main step, which can be directly referenced by S320 to complete the generation of the threshold comparison mark sequence.

[0069] Furthermore, the missing segment marking process is executed after the acquisition frame number binding process. The system performs a continuity check on the acquisition frame number sequence field and a time interval check on the acquisition time tag sequence field. When the continuity check finds a jump in adjacent acquisition frame numbers, or the time interval check finds that the interval between adjacent acquisition time tags exceeds the missing tolerance caliber mark, the system writes a missing segment mark in the session-level cache partition and binds the missing segment mark to the missing start acquisition frame number and the missing end acquisition frame number. At the same time, the system writes a missing association mark in the frame header metadata of the angle sampling sequence field and the torque sampling sequence field, indicating that the frame is at the leading or trailing edge of the missing segment. The missing segment mark, as a process record of this main step, is not included in the minimum field set of the real-time acquisition data packet. However, when the system outputs the real-time acquisition data packet, it writes the missing segment mark to the session log module and binds it to the session number field. This binding relationship can be read during subsequent S410 merging and archiving to complete the audit link. After completing the above-mentioned acquisition frame number binding, channel calibration and threshold field loading, and missing segment marking, the system encapsulates the acquisition frame number sequence field, acquisition time tag sequence field, angle sampling sequence field, torque sampling sequence field, angle safety interval field, and torque safety interval field to generate a real-time acquisition data packet. The real-time acquisition data packet is used as the input source of S320. S320 extracts the angle sampling sequence field and torque sampling sequence field from the real-time acquisition data packet to continue the threshold comparison mark sequence generation and motion state feature extraction processing. At the same time, the angle safety interval field and torque safety interval field in the real-time acquisition data packet participate in the threshold comparison mark sequence generation in S320 to form a comparison input with the threshold comparison rule field caliber consistent with that in S230.

[0070] S320. Extract angle sampling sequence field and torque sampling sequence field from real-time data acquisition data, perform threshold comparison marker sequence generation, motion state feature extraction, and spasm trigger fragment annotation processing to generate motion intention and spasm state discrimination results containing motion intention label sequence field, spasm fragment annotation sequence field, and threshold comparison marker sequence field.

[0071] Specifically, each time the sliding time window length mark triggers the window slice boundary, the system receives the real-time acquisition data packet output by the preceding S310, and extracts the angle sampling sequence field and torque sampling sequence field from the real-time acquisition data packet as discrimination input. At the same time, it reads the angle safety interval field and torque safety interval field in the real-time acquisition data packet as threshold boundary input. In parallel, the system reads the threshold comparison rule field in the real-time acquisition and threshold comparison configuration structure as the comparison caliber input. The threshold comparison rule field includes an angle comparison rule subfield, a torque comparison rule subfield, and a conflict handling subfield. The conflict handling subfield is used to define merging rules when angle and torque out-of-bounds marks appear simultaneously or alternately. The threshold comparison mark sequence generation process first performs boundary determination on the angle sampling sequence field. The system compares the angle sampling value of each acquisition frame with the lower and upper boundaries of the angle and outputs an angle comparison mark according to the angle comparison rule sub-field. Similarly, the system compares the torque sampling value of each acquisition frame with the lower and upper boundaries of the torque and outputs a torque comparison mark according to the torque comparison rule sub-field. The system then merges the angle comparison mark and torque comparison mark according to the conflict handling sub-field to generate a threshold comparison mark sequence field. The threshold comparison mark sequence field contains at least a frame-level mark value, a mark source identifier, and a merging rule identifier. The mark source identifier indicates that the mark comes from an angle out-of-bounds error, a torque out-of-bounds error, or a merging conflict. The merging rule identifier indicates the conflict handling sub-field caliber used. The threshold comparison mark sequence field, as one of the minimum set fields output in this section, participates in the amplitude limiting caliber selection in the subsequent output amplitude limiting processing in S330, and is also associated with the audit record in the instruction archiving sequence field during the archiving stage in S410.

[0072] Furthermore, the motion state feature extraction process is performed after the threshold comparison marker sequence is generated. The system performs synchronous feature extraction on the angle sampling sequence field and the torque sampling sequence field within the same sliding time window. In this invention, motion state features refer to the set of state descriptions composed of the change trends, abrupt changes, and stable segments of the angle sampling sequence and the torque sampling sequence within the window. The feature extraction process includes at least three types of operations: First, angle change trend extraction within the window: the system performs differential operations on adjacent acquisition frames for the angle sampling sequence field within the window and generates an angle change sequence, and then performs stable segment identification on the angle change sequence and writes it into the stable segment marker subfield; Second, torque change trend extraction within the window: the system performs differential operations on adjacent acquisition frames for the torque sampling sequence field within the window and generates a torque change sequence, and then performs abrupt segment identification on the torque change sequence and writes it into the abrupt segment marker subfield; Third, synchronization determination of the angle change sequence and the torque change sequence: the system performs same-direction determination on the angle change value and the torque change value in the same frame and generates a same-direction marker subfield. The aforementioned stable segment marker subfield, abrupt segment marker subfield, and unidirectional marker subfield together constitute the minimum set output for motion state feature extraction. The system maps this to the discriminative input for the motion intent label sequence field. In this invention, the motion intent label refers to the tagged record of the active motion, passive stretching, or maintenance state of each acquisition frame or frame segment. The system generates frame-level labels for the motion intent label sequence field based on the stable segment marker subfield and the unidirectional marker subfield, and writes the threshold comparison marker sequence field associated with the label generation into the label metadata, thereby aligning the motion intent label sequence field with the threshold comparison marker sequence field at the frame level.

[0073] Furthermore, the spasticity trigger segment annotation process is performed after motion state feature extraction. In this invention, a spasticity trigger segment refers to a set of frames where torque abrupt changes and angle changes are asynchronous within the sliding time window, and the threshold comparison marker sequence field shows continuous or high-frequency alternating boundary crossings. The system first reads the abrupt change segment marker subfield and the same-direction marker subfield, filtering out candidate segments with torque abrupt changes and same-direction determinations of different directions. Subsequently, it reads the threshold comparison marker sequence field, and counts the continuous length and alternating boundary crossing frequency of boundary crossing markers within the candidate segments. When the continuous length or alternating frequency meets the trigger caliber defined by the threshold comparison rule field, the system writes the candidate segment into the spasticity segment annotation sequence field. The spasticity segment annotation sequence field includes at least the segment start acquisition frame number, the segment end acquisition frame number, the trigger type identifier, and the trigger caliber identifier. The trigger type identifier is used to distinguish between continuous boundary crossing triggers and alternating boundary crossing triggers, and the trigger caliber identifier is used to record the threshold comparison rule field version association information referenced in the trigger determination. When writing the spasticity fragment annotation sequence field, the system simultaneously writes to the session log module and binds the fragment to the session number field. This allows the system to reference the spasticity fragment annotation sequence field to form a threshold update source identifier when registering the threshold update record field in S430. After completing the generation of the threshold comparison marker sequence, extraction of motion state features, and annotation of the spasticity trigger fragment, the system encapsulates the motion intention label sequence field, the spasticity fragment annotation sequence field, and the threshold comparison marker sequence field to generate a motion intention and spasticity state discrimination result. This result is then used as the input source for S330, which further performs control command mapping, output limiting, and command issuance processing on the motion intention and spasticity state discrimination result and the real-time acquisition and threshold comparison configuration structure. Simultaneously, the motion intention and spasticity state discrimination result is archived into the associated record of the command archive sequence field during the rehabilitation parameter archive data package generation stage in S410, supporting cross-main step traceability of the discrimination basis.

[0074] S330: The system performs control command mapping, output limiting, and command issuance processing on the judgment results of movement intention and spasticity state and the configuration structure of real-time acquisition and threshold comparison, generating a motor output control command sequence structure containing command time tag sequence field, motor target output sequence field, and output limiting parameter field.

[0075] Specifically, after outputting the discrimination result in each sliding time window, the system receives the motion intention and spasm state discrimination result output by the preceding S320, and reads the real-time acquisition and threshold comparison configuration structure output by S230 in parallel. The adaptive control parameter table field in the real-time acquisition and threshold comparison configuration structure serves as the core input for control command mapping, while the threshold comparison rule field and acquisition scheduling parameter field serve as the input for output limiting and command issuance rhythm. In this invention, control command mapping refers to the process of mapping the motion intention label sequence field to the motor target output sequence field. The motor target output sequence field describes the target output of the drive motor in the next control cycle, and the field contains at least the target output value, the target output direction identifier, and the action number field association marker. The system reads the control command mapping caliber and action segment association caliber in the adaptive control parameter table field, associates the motion intention label sequence field with the action number field of the training action configuration table field, selects the corresponding control command mapping caliber entry according to the action segment association caliber, and then maps the motion intention label of each frame segment to the target output value and the target output direction identifier, and writes it into the motor target output sequence field. The instruction time tag sequence field is generated synchronously during the mapping phase. The system checks the consistency between the session start time tag fragment in the session number field and the current system clock time tag, and then writes it into the instruction time tag sequence field. The system also establishes a one-to-one correspondence between the instruction time tag and the motor target output sequence field entries, so that when the S410 merges and archives, the instruction archive sequence field can be aligned with the angle archive sequence field and the torque archive sequence field according to the time tag.

[0076] Furthermore, the output limiting process is executed after the control command mapping. In this invention, output limiting refers to the process of performing boundary clipping and rate of change constraint on the target output value after the target output sequence field of the motor is generated, in combination with the threshold comparison mark sequence field and the spasm fragment annotation sequence field. The system reads the threshold comparison mark sequence field and the spasm fragment annotation sequence field in the motion intention and spasm state discrimination result, and reads the output limiting parameter caliber in the real-time acquisition and threshold comparison configuration structure. The output limiting parameter caliber includes at least two types of boundary sets: normal limiting boundaries and abnormal limiting boundaries. The normal limiting boundaries are associated with the threshold version number field, and the abnormal limiting boundaries are associated with the trigger caliber identifier. When the threshold comparison marker sequence field shows that the current frame segment has an out-of-bounds marker, or the spasm fragment annotation sequence field covers the current frame segment, the system selects an abnormal amplitude limiting boundary and performs clipping on the target output value. Simultaneously, it applies a rate of change constraint to the target output value of adjacent control cycles, writing the clipping boundary, rate of change constraint caliber, and trigger source identifier into the output amplitude limiting parameter field. When the threshold comparison marker sequence field shows that the current frame segment does not have an out-of-bounds marker and the spasm fragment annotation sequence field does not cover the current frame segment, the system selects a regular amplitude limiting boundary and performs clipping, writing the clipping boundary caliber into the output amplitude limiting parameter field. The output amplitude limiting parameter field, as one of the minimum set fields output in this section, can be referenced in the subsequent threshold update record field registration stage of S430, forming a supplementary basis for the threshold update source identifier.

[0077] Furthermore, the instruction issuance processing is executed after output limiting. The system generates control cycle trigger events according to the acquisition cycle flag and window slicing rules in the acquisition scheduling parameter field. When each control cycle trigger event arrives, the system retrieves the target output value and target output direction identifier corresponding to the instruction time tag from the motor target output sequence field, and writes them, along with the output limiting parameter field, to the control interface of the drive motor. The system simultaneously writes the instruction issuance record subfield, which includes at least the issuance time tag, issuance channel identifier, and issuance status flag. In abnormal states, the issuance status flag is written with a failure reason code and bound to the session number field. Subsequently, when the S410 merges and archives, this binding relationship is written to the associated metadata of the instruction archiving sequence field. The instruction issuance processing is also linked with the session log module. When the session state machine module is in a blocked or paused state, the system writes the control cycle trigger event to the blocking record and stops writing the target output value to the control interface, thereby maintaining consistency between the session state and the instruction issuance. After completing the control command mapping, output limiting, and command issuance, the system encapsulates the command time stamp sequence field, the motor target output sequence field, and the output limiting parameter field to generate a motor output control command sequence structure. This structure is then output to S410 for subsequent processing, including torque and angle record merging, time stamp alignment, and speed field generation. Simultaneously, the command time stamp sequence field within the motor output control command sequence structure is aligned with the acquired time stamp sequence field in S410, ensuring time consistency during session-level archiving.

[0078] In the engineering implementation scenario, after the therapist immobilizes the patient and starts the session, the acquisition scheduling actuator continuously outputs samples according to the acquisition cycle, marking the driving angle sensing channel and torque sensing channel. The system generates real-time acquisition data packets in S310 and triggers discrimination processing in S320 at the sliding time window boundary. When the patient shows an active flexion and extension intention, the motion intention label sequence field displays active motion labels in multiple consecutive frame segments. The system maps this label to the motor target output sequence field in S330 and issues it according to the conventional amplitude limiting boundary. When a torque mutation occurs and the same direction is determined to be different in the frame segment, the system writes the spastic segment label sequence field in S320, and selects the abnormal amplitude limiting boundary in S330 to perform pruning on the target output value and issue it. At the same time, the trigger caliber identifier is written into the output amplitude limiting parameter field and written into the session log module. The subsequent archiving and reporting stages can read this record to complete the generation and backfilling of the threshold update record structure.

[0079] In summary, the technical effects of this step are as follows: Within the same session running link, the frame number of the acquisition is bound to the threshold field, and the snapshot is loaded with the threshold comparison marker sequence to establish a frame-level alignment relationship. The spasm fragment annotation sequence field is directly used to write the output limiting parameter field, so that the control command mapping and limiting clipping run under the same threshold version number field caliber, reducing the caliber deviation between the discrimination link and the command link.

[0080] Step S400 includes at least steps S410-S430:

[0081] S410. Obtain the motor output control instruction sequence structure, merge torque and angle record items, align time tags, and generate speed fields to obtain a rehabilitation parameter archive data package containing session number field, angle archive sequence field, torque archive sequence field, speed archive sequence field, and instruction archive sequence field.

[0082] The motor output control command sequence structure originates from the output of the preceding S330 and includes at least a command time tag sequence field, a motor target output sequence field, and an output limiting parameter field. It is continuously written to the session-level cache partition during session execution. Specifically, the archiving and merging unit triggers an archiving construction action when it detects that the session state machine module corresponding to the rehabilitation training session configuration structure has entered an "end state" or "pause state." Simultaneously, during session execution, the archiving and merging unit triggers incremental archiving actions based on the sliding time window length given by the acquisition scheduling parameter field, thereby forming multiple batches of archived segments in long-term training scenarios. Upon triggering, the archiving and merging unit first reads the cached fragments of the real-time acquisition data packets within the current session from the session-level cache partition. These cached fragments include the acquisition timestamp sequence field, the angle sampling sequence field, and the torque sampling sequence field. Simultaneously, the archiving and merging unit reads the instruction timestamp sequence field and the motor target output sequence field corresponding to the motor output control instruction sequence structure from the instruction cache partition to form an archiving input set. This archiving input set forms a minimum set parameter in this step. The minimum set parameter includes at least the session number field, the acquisition timestamp sequence field, the instruction timestamp sequence field, the angle sampling sequence field, and the torque sampling sequence field. The output limiting parameter field is used as an optional extended field for writing audit records.

[0083] Furthermore, the torque and angle record merging process refers to the process of pairing the angle sampling sequence field and the torque sampling sequence field under the same session number field according to the acquisition frame number sequence field, and writing the pairing result into the archive sequence. The archive merging unit first performs frame header metadata verification on the angle sampling sequence field and the torque sampling sequence field. The frame header metadata includes at least the acquisition frame number, channel status flag, and missing measurement association flag. When the channel status flag has an abnormal state or the missing measurement association flag indicates the leading or trailing edge of the missing measurement segment, the archive merging unit writes the abnormal flag into the archive metadata field and retains the original sampled value without replacement, thereby maintaining the traceability of the audit link. Subsequently, the archive merging unit writes the paired angle sampled values ​​into the angle archive sequence field in chronological order, and writes the paired torque sampled values ​​into the torque archive sequence field in the same order. It also writes a consistent acquisition time tag within both types of archive sequence fields. The acquisition time tag comes from the acquisition time tag sequence field, thus completing the record merging. Understandably, the angle archive sequence field and the torque archive sequence field use the same sampling pitch in this step. The sampling pitch comes from the acquisition period marker in the acquisition scheduling parameter field. When the acquisition period marker is adjusted during session operation, the archive merging unit divides the archive segments according to the adjustment boundary and records the acquisition frame number corresponding to the division boundary in the archive metadata field, so as to identify the segment range of different sampling pitches in the subsequent report field extraction stage.

[0084] Furthermore, time stamp alignment processing refers to establishing a mapping relationship between the acquisition time stamp sequence field and the instruction time stamp sequence field, and writing the mapped instruction records into the instruction archive sequence field. The archive merging unit performs deduplication and monotonicity checks on the instruction time stamp sequence field. When duplicate instruction time stamps or monotonicity conflicts exist, the archive merging unit writes the conflict record into the archived copy of the instruction issuance record subfield and binds the conflict record to the session number field for storage. Subsequently, the archive merging unit searches for instruction time stamps that are time-adjacent to each acquisition time stamp according to the time axis of the acquisition time stamp sequence field, and extracts the corresponding motor target output sequence field entries and writes them into the instruction archive sequence field if the adjacent window caliber is satisfied. The adjacent window caliber is taken from the sliding time window length marker or the control cycle trigger event caliber of the acquisition scheduling parameter field. The archive merging unit writes the adjacent window caliber into the metadata subfield of the instruction archive sequence field, so that the alignment caliber can be reused in the subsequent database writing stage. The output limiting parameter field can be optionally written during the alignment stage. The archiving and merging unit adds a limiting caliber summary to each entry in the instruction archiving sequence field. The limiting caliber summary comes from the clipping boundary record and trigger source identifier of the output limiting parameter field, and establishes an association record with the trigger caliber identifier of the same threshold comparison mark sequence field.

[0085] Furthermore, the velocity field generation process refers to the process of generating a velocity archive sequence field by differentially analyzing adjacent angle archive records and combining the acquisition time stamps after the angle archive sequence field is generated. In this invention, the velocity field refers to a derived record set composed of angle changes and time intervals. The archiving and merging unit, without changing the original value of the angle archive sequence field, reads the angle value and time stamp difference between two adjacent angle archive records, calculates the angle change and time interval, and writes the calculation results into the velocity archive sequence field. When an abnormal interval occurs due to missing measurement segments, the archiving and merging unit marks the velocity record as missing measurement associated, and this missing measurement associated status comes from the missing measurement associated marker. Understandably, the velocity archive sequence field is a core parameter set in this step, as it, along with the angle archive sequence field and the torque archive sequence field, constitutes the basic input for subsequent progress report field extraction. Simultaneously, the velocity archive sequence field generation process is constrained by the channel calibration record subfield. When the channel calibration record subfield indicates a change in calibration bias, the archiving and merging unit resets the velocity calculation window at the change boundary and writes the change boundary metadata to avoid cross-caliber segment miscalculation.

[0086] In the engineering implementation scenario, after the therapist immobilizes the patient in the outpatient training area, the system initiates a session. During the session's running state, the system continuously generates a motor output control command sequence structure and writes it to the command cache partition. Simultaneously, the angle and torque sensing channels write the sequence to the sampling cache partition. When the therapist triggers an "End Session" operation on the interface or the session state machine module receives an automatic end event, the archiving and merging unit triggers this step. It pairs the angle sampling sequence field and torque sampling sequence field of this session and writes them to the angle archiving sequence field and torque archiving sequence field, maps the motor target output sequence field to the command archiving sequence field according to the time label, and generates a speed archiving sequence field. After completing the above processing, the archiving and merging unit encapsulates and generates a rehabilitation parameter archive data package, writing the session number field to the primary key position of the data package. Simultaneously, it generates a parameter index field within the data package. The parameter index field is used to identify the partition number, segment start and end time labels, and segment verification flags of each archive sequence field within this data package. The rehabilitation parameter archive data package, as the output of this section, is called in subsequent step S420. Step S420 extracts the session number field and parameter index field from the rehabilitation parameter archive data package to perform database write processing and index update processing.

[0087] S420. Extract the session number field and parameter index field from the rehabilitation parameter archive data packet, perform database write processing and index update processing, and generate a rehabilitation database index record containing the index primary key field and storage location field.

[0088] The rehabilitation parameter archive data packet originates from the output of preceding S410. The session number field serves as a unique session-level identifier, and the parameter index field serves as the archive fragment directory and partition mapping information. Specifically, after receiving the rehabilitation parameter archive data packet, the database write module first establishes a write transaction context in the local persistent storage partition and then establishes a synchronization transaction context when the cloud synchronization partition becomes available. Both types of transaction contexts are bound to the session number field and the transaction identifier is written in the session log module. During the write phase, the database write module writes the angle archive sequence field, the torque archive sequence field, the speed archive sequence field, and the instruction archive sequence field to the corresponding data partition according to the partition number given by the parameter index field. When the parameter index field indicates the existence of multiple archive fragments, the database write module appends the data according to the fragment start and end time tags and writes a fragment verification mark for each fragment. The fragment verification mark consists of a verification digest generated from the fragment data content and is used for subsequent readback verification. For any abnormal states that occur during the writing process, the database writing module records the exception reason code, the failed segment number, and the retry count within the writing transaction context, and binds this record to the session number field. When the cloud synchronization partition is unavailable, the database writing module marks the cloud synchronization transaction context as suspended, and replays the suspended transaction when the subsequent network status recovery event is triggered. During the replay process, the replay timestamp and the number of replays are also recorded.

[0089] Furthermore, the index update process refers to the process of generating or updating a searchable index after the data partition is written. Before the write transaction is committed, the index update module extracts fragment directory information from the parameter index field to generate an index primary key field and a storage location field. In this invention, the index primary key field points to the index entry point of the rehabilitation training session. The index primary key field includes at least a session number field fragment, an archive batch marker fragment, and an index version fragment. The archive batch marker fragment comes from the fragment sequence number information in the parameter index field, and the index version fragment is incremented by the index update module each time the index changes. In this invention, the storage location field points to the location information of the archived data within the storage system. The storage location field includes at least a local storage partition path identifier, a cloud synchronization partition path identifier, and a set of data partition numbers. The cloud synchronization partition path identifier is written with a placeholder status mark when cloud synchronization is unavailable, and is updated to the actual path identifier by the synchronization receipt event after synchronization is completed. Understandably, when generating the storage location field, the index update module synchronously writes the fragment verification tag set and the channel status tag digest, so that consistency verification can be completed in the subsequent reading stage. When the channel status tag digest indicates that there is an abnormal channel status in this session, the index update module writes the abnormal digest into the index metadata subfield, which is used as the filtering input in the subsequent progress report field extraction stage.

[0090] In the engineering implementation scenario, the system immediately writes the rehabilitation parameter archive data packet to the local persistent storage partition after training, and synchronously writes the same archive data packet to the cloud synchronization partition when the network is available. Before committing the local write transaction, the index update module generates an index primary key field and records the local storage partition path identifier, while marking the cloud synchronization partition path identifier as pending, updating it to the actual path identifier after the synchronization receipt event arrives. After completing the database write and index update processes, the system outputs a rehabilitation database index record. This rehabilitation database index record includes an index primary key field and a storage location field, and is input as the output of this section in subsequent S430. S430 extracts and formats the report field of the rehabilitation database index record, and simultaneously registers the threshold update record field to generate a progress report and threshold update record structure.

[0091] S430. Extract report fields, format and encapsulate, and register threshold update record fields for the rehabilitation database index records to generate a progress report and threshold update record structure containing a progress report field set and threshold update record structure fields.

[0092] The rehabilitation database index record is derived from the output of preceding step S420 and includes an index primary key field and a storage location field. Specifically, after receiving the rehabilitation database index record, the report generation module retrieves the corresponding session metadata by index primary key field and locates and reads the angle archive sequence field, torque archive sequence field, speed archive sequence field, and command archive sequence field by storage location field to form the report extraction input set. The minimum set parameter of the report extraction input set in this step includes at least the session number field, angle archive sequence field, torque archive sequence field, speed archive sequence field, and command archive sequence field. The index metadata subfield and channel status marker summary are optional extended parameters for report annotation writing. The report field extraction process refers to the process of extracting the statistical fields, event fields, and version fields required to form the progress report field set from the report extraction input set. During the extraction phase, the report generation module first parses the session number field to obtain training date stamps and device identification fragments, and writes them into the session information subfield of the progress report field set. Subsequently, it performs angle range fragment recognition on the angle archive sequence field to generate an angle range summary field, and performs speed change fragment recognition on the speed archive sequence field to generate a speed change summary field. Simultaneously, it performs torque peak segment recognition on the torque archive sequence field to generate a torque peak segment summary field. For the command archive sequence field, the report generation module extracts target output change fragments and output limiting parameter summaries to generate a control command summary field, and aligns the control command summary field with the torque peak segment summary field according to time stamps, thereby forming cross-domain aligned record pairs in the report field set. Understandably, the aforementioned angle range summary field, speed change summary field, torque peak segment summary field, and control command summary field constitute the core field set of the progress report field set. This core field set does not rely on additional sensors or external evaluation tools; it can be generated solely based on the archived products within the closed-loop link of this system.

[0093] Furthermore, the formatted encapsulation process refers to the process of writing the progress report field set into the report carrier and generating a searchable and printable structured output. During the encapsulation phase, the report generation module provides a report template unit and a rendering unit. The report template unit defines the field layout, field naming conventions, and unit conventions. The rendering unit is responsible for writing the progress report field set into the template and generating a report archive object. In this invention, the report archive object is also written into the storage system, and the report object path identifier is written into the index metadata subfield of the rehabilitation database index record, forming a bidirectional reference relationship between the report and the original archived data. For abnormal states during the report encapsulation process, the report generation module records the template version number, rendering error code, and failed field name, and binds this record to the index primary key field and writes it into the session log module for subsequent regeneration. When the cloud synchronization partition is in a suspended state, the report archive object is preferentially written to the local persistent storage partition, and synchronous writing is performed after the synchronization receipt event arrives.

[0094] Furthermore, the threshold update record field registration process refers to the process of synchronously generating the threshold update record structure field and completing the version chain registration during the report generation stage. In this invention, the threshold update record structure field refers to a set of fields that structurally encapsulate the evolution information of individualized security threshold version records between adjacent sessions, including at least the threshold version number, threshold timestamp, update source identifier, and update summary field. During the registration stage, the threshold update record generation unit first reads the threshold version number and threshold timestamp corresponding to the current session from the session metadata, and generates an update summary field by combining the output amplitude limit parameter summary, torque peak segment summary field, and angle range summary field within the instruction archive sequence field. Simultaneously, the threshold update record generation unit reads the associated records of the spasm fragment annotation sequence field and the threshold comparison mark sequence field bound to the session number field from the session log module, generating an update source identifier. The update source identifier indicates that the threshold update is triggered by a combination of spasm fragment annotations or threshold out-of-bounds markers. Subsequently, the threshold update record generation unit writes the threshold update record structure field into the threshold version chain storage area and binds the version chain to the index primary key field of the receipt, forming an auditable version evolution trajectory. After completing the registration of the threshold update record field, the report generation module encapsulates the progress report field set with the threshold update record structure field to generate a progress report and a threshold update record structure. The progress report and threshold update record structure, as the output of this section, are input to the threshold update record structure field of S210 in the subsequent closed-loop link, for access in the next rehabilitation training session configuration structure generation stage, thereby forming a cross-main step connection with the individualized safety threshold version record generation link of S200. At the same time, the progress report field set can be referenced by the patient file data packet field in the clinical follow-up scenario, forming a homologous index relationship between training records and evaluation records.

[0095] In summary, the technical effects of this step are as follows: In S410, a mapping relationship is established between the sampling archive and the instruction archive based on time tags, and a speed archive sequence field is generated. In S420, the archived fragments are written into the storage location field bound to the primary key field of the index. Then, in S430, the output amplitude limiting parameter summary and spasm fragment annotation associated records are registered with the threshold version chain. After the session ends, the system forms a traceable archive-index-report chain and provides the same source input to the threshold update record structure field of the next session.

[0096] Example 2: Figure 2 A structural block diagram of an adaptive closed-loop control system for hand joint rehabilitation according to an embodiment of the present invention is shown. Figure 2 As shown, the structure may include:

[0097] The motion trajectory and fixed configuration structure generation module 01 is used to acquire the geometric parameters of the actuator and the kinematic baseline of the joint, register the parameter fields of the circular arc trajectory segment, the parameter field of the Archimedes' spiral trajectory segment, the splicing point number field, and the concavity and convexity law constraint term field to generate a composite motion trajectory parameter set, and register the direction compensation parameter field, the distance compensation parameter field, and the compensation parameter version number field to generate a compensation parameter set. Then, it registers the fixed point mapping table field, the strap tension level field, and the training action configuration table field to generate the motion trajectory and fixed configuration structure, and provides the motion trajectory and fixed configuration structure to the rehabilitation training session configuration structure generation module; specifically, the motion trajectory and... When the training start trigger or a change in the training action configuration table field triggers, the fixed configuration structure generation module receives the actuator geometric parameters and joint kinematic baseline as input objects. The actuator geometric parameters include geometric constraints related to the actuator's installation position, and the joint kinematic baseline includes baseline constraints consistent with joint kinematics. After receiving the input, the motion trajectory and fixed configuration structure generation module performs field registration processing on the arc trajectory segment parameter field, Archimedean spiral trajectory segment parameter field, splicing point number field, and concavity-convexity law constraint field. The field registration processing includes field naming consistency verification, field value range verification, and splicing. The uniqueness of the contact point number field is verified, and after successful verification, a composite motion trajectory parameter set is encapsulated. This composite motion trajectory parameter set is retained as the upstream input object for subsequent compensation registration. Subsequently, the motion trajectory and fixed configuration structure generation module extracts the trajectory segment parameter field and the splicing point number field from the composite motion trajectory parameter set, performs direction compensation parameter field registration and distance compensation parameter field registration, and writes the compensation parameter version number field. The compensation parameter version number field is associated with the current registration time tag, thereby generating the compensation parameter set. After the compensation parameter set is generated, the motion trajectory and fixed configuration structure generation module processes the compensation parameter set and the composite motion trajectory parameter set. The motion trajectory parameter set is used to register fields in the fixed point mapping table, the strap tension setting, and the training action configuration table. The fixed point mapping table fields maintain the correspondence between the splicing point number field and the fixed point position. The strap tension setting field maintains the consistency between the fixed state caliber and the training action configuration table fields. The training action configuration table fields maintain the reference relationship between the trajectory segment sequence and the compensation parameter version number field. When a field range check or uniqueness check fails, the motion trajectory and fixed configuration structure generation module writes a null value as a placeholder in the current output and retains the original input record to prevent incomplete fields from being propagated to downstream modules. The motion trajectory and fixed configuration structure is provided as an output to the rehabilitation training session configuration structure generation module and is also called as an input object for the rehabilitation training session configuration structure generation module.

[0098] The rehabilitation training session configuration structure generation module 02 is used to receive the motion trajectory and fixed configuration structure, generate a session number field, register the angle sensor channel number field and the torque sensor channel number field, generate acquisition scheduling parameter fields and encapsulate them to obtain the rehabilitation training session configuration structure, and provide the rehabilitation training session configuration structure to the individualized safety threshold version record and real-time acquisition and threshold comparison configuration structure generation module; specifically, after receiving the motion trajectory and fixed configuration structure output from the motion trajectory and fixed configuration structure generation module, the rehabilitation training session configuration structure generation module associates the motion trajectory and fixed configuration structure with the patient file data packet field and threshold update record structure field of the current session and writes them into the session context, and performs session number field generation processing, which includes writing the anti-repetition rule for the session number field and associating the session number field with the patient file data packet field and the threshold update record structure field of the current session. The training action configuration table fields are bound and registered. Subsequently, the rehabilitation training session configuration structure generation module performs registration processing on the angle sensor channel number field and the torque sensor channel number field. The registration processing includes writing the channel number caliber, writing the channel status flag, and writing the channel binding record. The channel binding record is used to maintain a consistent mapping between the channel number and the session number field. After completing the channel field registration, the rehabilitation training session configuration structure generation module generates the acquisition scheduling parameter field. The acquisition scheduling parameter field includes a combination of writing the acquisition cycle flag, the sliding time window length flag, and the missing measurement tolerance caliber flag, and establishes an association record with the angle sensor channel number field and the torque sensor channel number field. When the channel status flag has an abnormal state, the rehabilitation training session configuration structure generation module writes the limiting caliber into the acquisition scheduling parameter field and binds the limiting caliber to the session number field for archiving. The rehabilitation training session configuration structure is packaged as an output product and provided to the individualized safety threshold version record and real-time acquisition and threshold comparison configuration structure generation module, so that it can extract the patient file data package field and the threshold update record structure field from the rehabilitation training session configuration structure for subsequent registration and loading processing.

[0099] The individualized safety threshold version record and real-time acquisition and threshold comparison configuration structure generation module 03 is used to extract patient file data packet fields and threshold update record structure fields from the rehabilitation training session configuration structure, register angle safety interval fields, torque safety interval fields, threshold version number fields, and threshold time tag fields to generate individualized safety threshold version records, and load threshold comparison rule fields and acquisition scheduling parameter fields to generate real-time acquisition and threshold comparison configuration structures. This real-time acquisition and threshold comparison configuration structure is then provided to the real-time acquisition data packet generation module, the movement intention and spasticity state discrimination module, and the motor output control command sequence structure generation module. Specifically, the individualized safety threshold version record and real-time acquisition... After receiving the rehabilitation training session configuration structure, the threshold comparison configuration structure generation module extracts the patient file data packet field and the threshold update record structure field from the rehabilitation training session configuration structure as input objects. The patient file data packet field contains patient identification information and previous assessment index information, and the threshold update record structure field contains previous registration records of the threshold version number field and the threshold time tag field. The individualized safety threshold version record and real-time acquisition and threshold comparison configuration structure generation module performs consistency verification with the previous registration records in the threshold update record structure field based on the assessment index caliber in the patient file data packet field, and then registers the angle safety interval field and the torque safety interval field. The angle safety interval field contains registration records for the lower and upper boundaries of the angle, and the torque safety interval field contains registration records for the lower and upper boundaries of the torque. The angle safety interval field and the torque safety interval field are bound to the current session number field. Subsequently, the individualized safety threshold version record and the real-time acquisition and threshold comparison configuration structure generation module register the threshold version number field and the threshold time tag field. The threshold version number field is written according to the incremental caliber of previous registration records, and the threshold time tag field is bound to the current session time caliber, thereby generating an individualized safety threshold version record. This individualized safety threshold version record is retained as an upstream input object for subsequent loading processing. The individualized safety threshold... After the version record is generated, the individualized security threshold version record and the real-time acquisition and threshold comparison configuration structure generation module load the threshold comparison rule field and the acquisition scheduling parameter field. The threshold comparison rule field includes the loading records of angle comparison rule subfield, torque comparison rule subfield and conflict handling subfield. The acquisition scheduling parameter field is copied and loaded from the rehabilitation training session configuration structure and maintains the same association with the session number field, thereby generating the real-time acquisition and threshold comparison configuration structure. When the consistency check fails, the individualized security threshold version record and the real-time acquisition and threshold comparison configuration structure generation module maintains the previous registration record and writes the current check mark to prevent inconsistent thresholds from entering the downstream comparison link.The real-time acquisition and threshold comparison configuration structure is provided as an output product to the real-time acquisition data packet generation module, the motion intention and spasm state discrimination module, and the motor output control command sequence structure generation module, so that downstream modules can call the threshold comparison rule field, angle safety interval field, torque safety interval field, and acquisition scheduling parameter field to carry out acquisition, discrimination, and control processing.

[0100] The real-time data packet generation module 04 is used to receive the real-time acquisition and threshold comparison configuration structure and call the angle sensing channel number field, the torque sensing channel number field, and the acquisition scheduling parameter field. It performs acquisition frame number sequence field binding, channel calibration, and missing segment marking to generate a real-time data packet containing the acquisition frame number sequence field, acquisition time label sequence field, angle sampling sequence field, torque sampling sequence field, angle safety interval field, and torque safety interval field. The module then provides the real-time data packet to the motion intention and spasticity state discrimination module. Specifically, after receiving the real-time acquisition and threshold comparison configuration structure, the real-time data packet generation module reads the acquisition cycle marker and sliding time window length marker in the acquisition scheduling parameter field as trigger inputs, and calls the angle sensing channel number field and the torque sensing channel number field from the rehabilitation training session configuration structure as acquisition channel inputs. When the acquisition cycle marker is triggered, the real-time acquisition data packet generation module... The real-time data packet generation module accesses the sampling records of the angle sampling sequence field and torque sampling sequence field according to the channel number caliber, and synchronously generates the acquisition time tag sequence field. The real-time data packet generation module performs acquisition frame number sequence field binding on the angle sampling records and torque sampling records formed within the same trigger period. The acquisition frame number sequence field is used to maintain the correspondence between the angle sampling records and torque sampling records within the same acquisition frame, and writes the acquisition frame number into the frame header metadata of the angle sampling sequence field and torque sampling sequence field. The real-time data packet generation module performs channel calibration at the start of the sampling link. Channel calibration includes writing zero-position calibration records and zero-point drift calibration records, and triggers the writing of calibration check records according to the sliding time window length during the operation of the sampling link. When the acquisition time tag sequence field is interrupted or the acquisition frame number sequence field jumps, the real-time data packet generation module performs missing segment marking and binds the missing segment marking to the session number field. When encapsulating the real-time acquisition data packet, the real-time acquisition data packet generation module copies and writes the angle safety interval field and the torque safety interval field from the real-time acquisition and threshold comparison configuration structure, so that the real-time acquisition data packet simultaneously contains the sampling sequence field and the threshold boundary field, and provides the real-time acquisition data packet to the motion intention and spasm state discrimination module for it to extract the angle sampling sequence field and the torque sampling sequence field to perform threshold comparison mark sequence generation processing.

[0101] The motion intention and spasm state discrimination module 05 is used to extract angle sampling sequence fields and torque sampling sequence fields from the real-time acquisition data packet, and perform threshold comparison mark sequence generation, motion state feature extraction, and spasm trigger fragment labeling on the angle safety interval field and the torque safety interval field according to the threshold comparison rule field, generating a motion intention and spasm state discrimination result containing motion intention label sequence fields, spasm fragment label sequence fields, and threshold comparison mark sequence fields, and providing the motion intention and spasm state discrimination result to the motor output control command sequence structure generation module; specifically, after receiving the real-time acquisition data packet output from the real-time acquisition data packet generation module, the motion intention and spasm state discrimination module extracts the angle sampling sequence field, torque sampling sequence field, angle safety interval field, and torque safety interval field as discrimination input, and reads the threshold comparison rule field from the real-time acquisition and threshold comparison configuration structure as rule input; the motion intention The image and spasm state discrimination module performs window slicing on the angle sampling sequence field and torque sampling sequence field according to the sliding time window length. Within each window, it performs interval judgment on the angle sampling record and the torque sampling record with the same angle safety interval field, and merges the judgment records according to the conflict handling sub-field within the threshold comparison rule field, thereby generating the threshold comparison mark sequence field. The motion state feature extraction performs change sequence generation, stable segment identification, and abrupt segment identification on the angle sampling sequence field and torque sampling sequence field within the same window, and maps the extraction results into the motion intention label sequence field. The motion intention label sequence field maintains a frame-level alignment relationship with the threshold comparison mark sequence field. The spasm trigger segment annotation generates the segment start and end acquisition frame number record under the joint constraints of the abrupt segment identification result and the threshold comparison mark sequence field, and writes it into the spasm segment annotation sequence field. The spasm segment annotation sequence field maintains a reference relationship with the acquisition frame number sequence field. The results of the motion intention and spastic state discrimination are packaged as output products and provided to the motor output control command sequence structure generation module, so that it can read the motion intention label sequence field, the spastic fragment label sequence field and the threshold comparison mark sequence field to carry out control command mapping and output limiting processing.

[0102] The motor output control command sequence structure generation module 06 is used to receive the motion intention and spasticity state discrimination result and read the adaptive control parameter table field in the real-time acquisition and threshold comparison configuration structure, perform control command mapping, output limiting, and command issuance, generate a motor output control command sequence structure containing a command time tag sequence field, a motor target output sequence field, and an output limiting parameter field, and provide the motor output control command sequence structure to the rehabilitation parameter archiving data package generation module; specifically, after receiving the motion intention and spasticity state discrimination result output by the motion intention and spasticity state discrimination module, the motor output control command sequence structure generation module reads the adaptive control parameter table field in the real-time acquisition and threshold comparison configuration structure as a mapping input, and uses the motion intention tag sequence field as the command generation input, and uses the threshold comparison tag field as the command generation input, and performs control command mapping, output limiting, and command issuance, generating a motor output control command sequence structure containing a command time tag sequence field, a motor target output sequence field, and an output limiting parameter field, and provides the motor output control command sequence structure to the rehabilitation parameter archiving data package generation module; specifically, after receiving the motion intention and spasticity state discrimination result from the motion intention and spasticity state discrimination module, the motor output control command sequence structure generation module reads the adaptive control parameter table field in the real-time acquisition and threshold comparison configuration structure as a mapping input, and uses the motion intention tag sequence ... The sequence field and the spastic fragment annotation sequence field are used as amplitude limiting trigger inputs. The control command mapping maps the motion intention label sequence field to generate the motor target output sequence field according to the mapping caliber in the adaptive control parameter table fields, and simultaneously generates the command time label sequence field. The command time label sequence field is bound to the session number field time caliber in the record. When the threshold comparison mark sequence field or the spastic fragment annotation sequence field has a trigger record, the output amplitude limiting is written into the output amplitude limiting parameter field according to the amplitude limiting caliber in the adaptive control parameter table fields, and a correspondence is established between the output amplitude limiting parameter field and the motor target output sequence field entry. The command issuance is triggered according to the control cycle caliber of the acquisition scheduling parameter field, and the motor target output sequence field entry and the output amplitude limiting parameter field entry are written into the issuance record, and the issuance record is bound to the command time label sequence field for archiving. The motor output control command sequence structure is encapsulated as the output product and provided to the rehabilitation parameter archiving data package generation module, so that it can read the command time label sequence field, the motor target output sequence field, and the output amplitude limiting parameter field, align and archive the execution time label of the acquisition side archived sequence.

[0103] The rehabilitation parameter archive data packet generation module 07 is used to receive the motor output control command sequence structure and read the real-time acquisition data packet, perform torque and angle record item merging, time tag alignment, and speed field generation, and generate a rehabilitation parameter archive data packet containing a session number field, angle archive sequence field, torque archive sequence field, speed archive sequence field, and command archive sequence field. A parameter index field is written into the rehabilitation parameter archive data packet, and the rehabilitation parameter archive data packet is provided to the rehabilitation database index record and progress report and threshold update record structure generation module. Specifically, after receiving the motor output control command sequence structure output by the motor output control command sequence structure generation module, the rehabilitation parameter archive data packet generation module reads the real-time acquisition data packet output by the real-time acquisition data packet generation module as sampling input. The system includes acquisition time stamp sequence fields, angle sampling sequence fields, and torque sampling sequence fields. The rehabilitation parameter archiving data package generation module performs record item pairing for the angle sampling sequence fields and torque sampling sequence fields according to the acquisition frame number sequence field and writes them into the angle archiving sequence fields and torque archiving sequence fields. During the pairing process, it retains the missing segment marker association records and binds them to the session number field for archiving. The time stamp alignment establishes a proximity mapping relationship between the acquisition time stamp sequence fields and the instruction time stamp sequence fields, writes the motor target output sequence field entries into the instruction archiving sequence field, and associates the output limiting parameter field entries with the instruction archiving sequence field entries. The speed field generation reads the time interval between the angle change of adjacent records in the angle archiving sequence field and the acquisition time stamp, generates a speed archiving sequence field, and writes a missing segment association status marker within the missing segment marker association range. When encapsulating the rehabilitation parameter archiving data package, the rehabilitation parameter archiving data package generation module writes a parameter index field, which includes the archived segment directory scope and the segment start and end time stamp scope. It then provides the rehabilitation parameter archiving data package to the rehabilitation database index record and progress report and threshold update record structure generation module, allowing it to extract the session number field and parameter index field for database writing and index update processing.

[0104] The rehabilitation database index record and progress report and threshold update record structure generation module 08 is used to extract the session number field and parameter index field from the rehabilitation parameter archive data packet and perform database write processing and index update processing to generate a rehabilitation database index record containing an index primary key field and a storage location field. Based on the rehabilitation database index record, it extracts the progress report field set and registers the threshold update record structure field to generate the progress report and threshold update record structure. The threshold update record structure field from the progress report and threshold update record structure is then backfilled into the individualized security threshold version record and real-time acquisition and threshold comparison configuration structure generation module. Specifically, after receiving the rehabilitation parameter archive data packet output by the rehabilitation parameter archive data packet generation module, the rehabilitation database index record and progress report and threshold update record structure generation module extracts the session number field and parameter index field as write input, and performs database write processing according to the archive fragment directory scope given by the parameter index field, including the angle archive sequence field, torque archive sequence field, and velocity. The archive sequence field and instruction archive sequence field are written to the corresponding storage location and written to the fragment verification record; the index update process generates the index primary key field and storage location field after the database write process is completed. The index primary key field establishes a primary key mapping relationship with the session number field, and the storage location field establishes a positioning mapping relationship with the fragment directory caliber in the parameter index field, thereby generating the rehabilitation database index record; the progress report field set extraction, after receiving the rehabilitation database index record, locates the session metadata according to the index primary key field, reads the angle archive sequence field, torque archive sequence field, speed archive sequence field and instruction archive sequence field according to the storage location field, generates the progress report field set and performs formatting encapsulation to write it into the progress report and threshold update record structure; the threshold update record structure field is registered during the progress report field set extraction and encapsulation, synchronously reads the current session threshold version number field and threshold time tag field, and writes the update summary field in combination with the associated record of the output limiting parameter field, forming the threshold update record structure field and writing it into the progress report and threshold update record structure. The threshold update record structure field is backfilled to the personalized security threshold version record and real-time acquisition and threshold comparison configuration structure generation module, and is called by subsequent sessions as the input source of the threshold update record structure field, thereby forming a continuous field flow relationship with the patient file data packet field and the rehabilitation training session configuration structure.

Claims

1. A method for adaptive closed-loop control in hand joint rehabilitation, characterized in that, include: Obtain the geometric parameters of the actuator and the kinematic baseline of the joint, perform composite motion trajectory parameter registration, compensation parameter registration and fixed configuration field mapping processing, and generate motion trajectory and fixed configuration structure; Based on the motion trajectory and fixed configuration structure, perform rehabilitation training session configuration generation, individualized safety threshold version record generation and threshold comparison configuration loading operations to generate real-time acquisition and threshold comparison configuration structure; The system acquires the configuration structure for real-time acquisition and threshold comparison, performs real-time acquisition data packet construction, motion intent and spasm state discrimination and processing, and control command generation operations, and generates a motor output control command sequence structure. The system acquires the motor output control command sequence structure, generates rehabilitation parameter archive data packages, generates rehabilitation database index records, and encapsulates progress reports, generating progress reports and threshold update record structures.

2. The method according to claim 1, characterized in that, The process of registering compensation parameters also includes: The compensation parameter registration and processing includes the circular arc trajectory segment parameter field, the Archimedes spiral trajectory segment parameter field, and the splicing point number field from the composite motion trajectory parameter set. It is jointly verified with the geometric parameters of the actuator, completes the selection of the origin and the binding of the direction convention of the compensation coordinate system, and generates the direction compensation parameter field and the distance compensation parameter field. The direction compensation parameter field includes the direction offset angle, the direction offset symbol, the axial mapping relationship identifier, and the compensation effective segment identifier. The distance compensation parameter field includes the zero distance offset, the stroke backlash correction amount, the fixed point tension correction amount, and the compensation effective segment identifier.

3. The method according to claim 1, characterized in that, The process of mapping fixed configuration fields also includes: The fixed configuration field mapping process includes the coordinates of the end support mounting holes in the actuator geometry parameters and the joint sliding direction convention in the joint kinematic baseline, generating fixed point mapping table fields, strap tension level fields, and training action configuration table fields. The fixed point mapping table fields include the device-side fixed point number, patient-side fixed point number, external force application point number, strap path number, strap contact surface identifier, and fixed point verification time label. The strap tension level field includes the level number, tension amount caliber, locking status mark, and tension verification time label. The training action configuration table fields include the action number field, action type field, segment sequence reference field, splicing point sequence number reference field, intra-segment velocity planning identifier reference field, intra-segment pause time label reference field, direction compensation effective segment identifier reference field, and distance compensation effective segment identifier reference field.

4. The method according to claim 1, characterized in that, The process of generating rehabilitation training session configurations also includes: The rehabilitation training session configuration generation process includes fields for composite motion trajectory parameter set, compensation parameter set, fixed point mapping table, and training action configuration table in the motion trajectory and fixed configuration structure. It synchronously accesses the patient file data package and threshold update record structure to generate session number field, angle sensor channel number field, torque sensor channel number field, and acquisition scheduling parameter field. The session number field includes device identification segment, patient identification segment, session start time tag segment, and random anti-repetition segment. The angle sensor channel number field includes channel number, sampling pitch diameter, range diameter, zero-point calibration mark, and channel status mark. The torque sensor channel number field includes channel number, sampling pitch diameter, range diameter, zero-point drift calibration mark, and channel status mark. The acquisition scheduling parameter field includes acquisition cycle mark, sliding time window length mark, buffer partition number mark, and missing measurement tolerance mark.

5. The method according to claim 1, characterized in that, The process of generating a personalized security threshold version record also includes: The personalized safety threshold version record generation process includes the patient identity identifier, affected side information, previous assessment record index and training restriction notes from the patient file data package, as well as the threshold version number, threshold time tag and update source identifier from the threshold update record structure. It generates patient ability assessment field, angle safety interval field, torque safety interval field, threshold version number field and threshold time tag field. The patient ability assessment field includes assessment source identifier, assessment time tag, joint active range of motion record index, muscle tone grading record index and pain subjective record index. The angle safety interval field includes lower angle boundary, upper angle boundary, boundary source identifier and boundary constraint mark. The torque safety interval field includes lower torque boundary, upper torque boundary, boundary source identifier and boundary constraint mark.

6. The method according to claim 1, characterized in that, The process of constructing real-time data acquisition packets also includes: The real-time acquisition data packet construction process includes the angle sensor channel number field, torque sensor channel number field, and acquisition scheduling parameter field in the real-time acquisition and threshold comparison configuration structure. It performs acquisition frame number binding, channel calibration and threshold field loading, and missing segment marking processing to generate a real-time acquisition data packet. The acquisition frame number binding includes aligning the angle sampling sequence field and the torque sampling sequence field at the frame level according to the acquisition frame number sequence field. The channel calibration and threshold field loading includes performing channel zero-point calibration and zero-point drift calibration on the zero-point calibration mark in the angle sensor channel number field and the zero-point drift calibration mark in the torque sensor channel number field, and writing the angle safety interval field and the torque safety interval field into the real-time acquisition data packet. The missing segment marking processing includes performing continuity verification and time interval verification on the acquisition frame number sequence field to generate a missing segment mark.

7. The method according to claim 1, characterized in that, The process of discriminating between motor intention and spasticity also includes: The motion intent and spasticity state discrimination process includes angle sampling sequence fields and torque sampling sequence fields, as well as angle safety interval fields and torque safety interval fields, from the real-time data acquisition data packet. It performs threshold comparison marker sequence generation, motion state feature extraction, and spasticity trigger fragment annotation to generate motion intent and spasticity state discrimination results. The threshold comparison marker sequence generation includes interval judgment of angle sampling values ​​with the lower and upper boundaries of the angle to generate angle comparison markers, and interval judgment of torque sampling values ​​with the lower and upper boundaries of the torque to generate torque comparison markers. The markers are then merged according to the conflict handling subfield in the threshold comparison rule field to generate the threshold comparison marker sequence field. The motion state feature extraction includes performing window-based angle change trend extraction, torque change trend extraction, and synchronization determination on the angle sampling sequence field and torque sampling sequence field to generate the motion intent label sequence field. The spasticity trigger fragment annotation includes torque abrupt change segments and homing determination results, which are combined with the threshold comparison marker sequence field to generate the spasticity fragment annotation sequence field.

8. The method according to claim 1, characterized in that, The process of generating control instructions also includes: The control command generation operation includes the motion intention label sequence field, threshold comparison mark sequence field, and spasm fragment annotation sequence field from the motion intention and spasm state discrimination results, as well as the adaptive control parameter table field and acquisition scheduling parameter field from the real-time acquisition and threshold comparison configuration structure. It performs control command mapping, output limiting, and command issuance processing to generate a motor output control command sequence structure. The control command mapping includes mapping the motion intention label sequence field to the motor target output sequence field according to the control command mapping caliber and action segment association caliber in the adaptive control parameter table field. The output limiting includes the threshold comparison mark sequence field and the spasm fragment annotation sequence field. It combines the normal limiting boundary and abnormal limiting boundary in the output limiting parameter field to perform pruning and rate of change constraint on the target output value. The command issuance includes generating a control cycle trigger event according to the acquisition cycle mark in the acquisition scheduling parameter field and writing the target output value into the drive motor control interface.

9. The method according to claim 1, characterized in that, The process of packaging and processing progress reports also includes: The progress report encapsulation process includes rehabilitation database index records, performs report field extraction and formatting encapsulation, and threshold update record field registration processing to generate a progress report and threshold update record structure. The report field extraction and formatting encapsulation process includes extracting angle range summary fields, velocity change summary fields, torque peak segment summary fields, and control command summary fields from angle archive sequence fields, torque archive sequence fields, velocity archive sequence fields, and command archive sequence fields, and writing them into the report archive object. The threshold update record field registration process includes threshold version number fields and threshold time stamp fields, and combines them with output amplitude limiting parameter fields and spastic fragment annotation sequence fields to generate threshold update record structure fields.

10. An adaptive closed-loop control system for hand joint rehabilitation, characterized in that, include: An adaptive closed-loop control system for hand joint rehabilitation, comprising a rehabilitation training session configuration structure generation module, an individualized safety threshold version recording and real-time acquisition and threshold comparison configuration structure generation module, a real-time acquisition data packet generation module, a movement intention and spasticity state discrimination module, a motor output control command sequence structure generation module, a rehabilitation parameter archive data packet generation module, and a rehabilitation database index record, progress report, and threshold update record structure generation module; used to implement the method described in any one of claims 1-9 above.

Images