A post-stroke hand spasm real-time sensing module based on multi-modal sensor fusion and intelligent hand exoskeleton rehabilitation system

Abstract

The application discloses a post-stroke hand spasm real-time sensing module based on multi-modal sensor fusion and an intelligent hand exoskeleton rehabilitation system, and relates to the technical field of sensors. The module comprises a flexible carrier unit, a multi-modal sensing unit and a data processing unit. The flexible carrier unit is in a glove-shaped partial fitting structure and adopts a hollow elastic net-shaped design. The multi-modal sensing unit is fixed on the flexible carrier unit and comprises an inertial measurement unit (IMU), a tension sensor unit and a thin film pressure sensor unit. The data processing unit is fixed on the flexible carrier unit and is electrically connected with each subunit of the multi-modal sensing unit. The data processing unit is used for synchronously sampling and fusion processing of motion data, muscle tension data and pressure distribution data collected by the multi-modal sensing unit, and performing spasm state recognition. The data transmission unit is electrically connected with the data processing unit and is used for real-time transmission of spasm sensing data to an external device.

Description

Technical Field

[0001] This application relates to the field of sensor technology, and in particular to a real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion and an intelligent hand exoskeleton rehabilitation system. Background Technology

[0002] The incidence of hand spasticity in stroke patients is approximately 60%-80%, manifesting as involuntary finger tremors, increased muscle tone, weak grip, or abnormal force exertion, severely impacting the rehabilitation process. Intelligent hand exoskeleton rehabilitation systems are core rehabilitation equipment for these patients, and "real-time spasticity sensing" is a crucial prerequisite for achieving closed-loop control of the system. Existing sensing solutions for hand spasticity mainly fall into two categories. The first is a single IMU sensing solution, which collects motion data by placing IMU sensors on the hand. This solution can only identify tremor-related spasticity but cannot detect changes in muscle tone or abnormal grip pressure. It is prone to misinterpreting movements caused by the patient's active exertion as spasticity or missing latent spasticity without obvious tremor but with a sudden increase in muscle tone. The second is a single tension / pressure sensing solution, which only detects muscle tone using a tension sensor or grip force using a pressure sensor. This solution cannot capture dynamic tremor data of the finger joints, and the sensors are mostly rigid structures, resulting in poor wearing comfort and affecting patient training compliance.

[0003] Existing technologies suffer from three main problems. First, they suffer from limited sensing dimensions and high misjudgment rates. Current solutions cover only a single indicator of spasticity, failing to fully reflect the complex state of spasticity and leading to incorrect control responses from the exoskeleton system. Second, they suffer from poor wearing comfort and low compatibility. Existing sensors often use rigid shells or fixed-size carriers, making them unsuitable for patients with different hand shapes, and prolonged wear can easily lead to hand pressure and sweat buildup. Finally, they lack real-time data processing capabilities. Summary of the Invention

[0004] The purpose of this application is to provide a real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion and an intelligent hand exoskeleton rehabilitation system, which can achieve accurate and real-time sensing of post-stroke hand spasticity.

[0005] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion, comprising: The flexible carrier unit has a glove-shaped localized fitting structure and adopts a hollowed-out elastic mesh design to fit the contours of different parts of the human hand; A multimodal sensing unit, fixed on the flexible carrier unit, includes an inertial measurement unit (IMU) subunit, a tension sensor subunit, and a thin-film pressure sensor subunit. The inertial measurement unit is used to collect motion data of the finger joints, the tension sensor subunit remotely transmits muscle tension data of the fingers through a flexible guide wire, and the thin-film pressure sensor subunit is used to collect pressure distribution data of the fingertip gripping. The data processing unit, fixed on the flexible carrier unit, is electrically connected to each sub-unit of the multimodal sensing unit. It is used to synchronously sample and fuse the motion data, muscle tension data, and pressure distribution data collected by the multimodal sensing unit, and to perform spasticity state recognition based on the fused data. The data transmission unit is electrically connected to the data processing unit and is used to transmit the spasm perception data identified by the spasm state to an external device in real time.

[0006] Secondly, this application provides an intelligent hand exoskeleton rehabilitation system, including a real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion, and an exoskeleton glove; the real-time sensing module for post-stroke hand spasticity is physically and functionally integrated with the driving layer of the exoskeleton glove through flexible electronic integration technology; the controller of the exoskeleton glove receives the spasticity sensing data output by the sensing module through the data transmission unit, and executes the corresponding closed-loop control response according to the control instructions output by the data transmission unit.

[0007] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion and an intelligent hand exoskeleton rehabilitation system. By fusing three types of sensor data—inertial measurement unit (IMU), tension sensor, and thin-film pressure sensor—it simultaneously captures three features of spasticity: finger joint tremors, overall muscle tone changes, and fingertip gripping pressure distribution. This covers the multi-dimensional manifestations of post-stroke hand spasticity, effectively reducing the false positive and false negative rates. The flexible carrier design with a hollowed-out elastic mesh distinguishes between a high-precision fitting area and a dynamic compliance area, ensuring sensor measurement accuracy while adapting to different hand shapes, improving wearing comfort, and compatibility with most existing exoskeleton rehabilitation devices. By integrating a hardware acceleration module into the data processing unit, it achieves synchronous sampling, end-to-end fusion, and real-time recognition, enabling spasticity state recognition within milliseconds, ensuring timely emergency intervention by the exoskeleton system and preventing joint damage to the patient. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a schematic diagram of the structure of a real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion, provided as an embodiment of this application. Detailed Implementation

[0010] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0011] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0012] Example 1 like Figure 1 As shown in the schematic diagrams of the back of hand A and the palm of hand B, this embodiment provides a real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion, including: The flexible carrier unit has a glove-shaped localized fitting structure and adopts a hollowed-out elastic mesh design to fit the contours of different parts of the human hand; A multimodal sensing unit, fixed on the flexible carrier unit, includes an inertial measurement unit (IMU) subunit, a tension sensor subunit, and a thin-film pressure sensor subunit. The inertial measurement unit is used to collect motion data of the finger joints, the tension sensor subunit remotely transmits muscle tension data of the fingers through a flexible guide wire, and the thin-film pressure sensor subunit is used to collect pressure distribution data of the fingertip gripping. The data processing unit, fixed on the flexible carrier unit, is electrically connected to each sub-unit of the multimodal sensing unit. It is used to synchronously sample and fuse the motion data, muscle tension data, and pressure distribution data collected by the multimodal sensing unit, and to perform spasticity state recognition based on the fused data. The data transmission unit, electrically connected to the data processing unit, is used to transmit the identified spasticity perception data to an external device in real time. The various units of this real-time post-stroke hand spasticity sensing module are connected via wires or wireless communication links, and the entire module is packaged as a flexible module adapted for hand wear, which can be integrated into corresponding positions within a smart hand exoskeleton rehabilitation glove. "Corresponding positions" specifically refers to the combination of anatomical functional and mechanical structural correspondences. Although both the post-stroke hand spasticity sensing module and the exoskeleton glove are glove-shaped, they belong to a "sensing layer" and a "drive / support layer," respectively. The specific integration scheme is as follows: The specific definition of the corresponding position: 1) Inertial Measurement Unit (IMU): Corresponds to the rotation center position of the interphalangeal joints of the exoskeleton glove (i.e., the dorsal side of the proximal, middle, and distal phalanges), used to synchronously collect joint motion data.

[0013] 2) Tension sensor subunit: corresponding to the anchor point of the drive tendon rope of the exoskeleton glove (i.e., the base of the fingers on the palm side), used to monitor the drive tension.

[0014] 3) Thin-film pressure sensor subunit: corresponding to the fingertip contact surface of the exoskeleton glove (i.e., the fingertip on the palm side), used to monitor the gripping contact force.

[0015] In this embodiment, the structure and connection relationship of each unit are as follows: The flexible carrier unit is made of a composite material of medical-grade silicone and elastic fabric, with an overall glove-like partial bonding structure. The thickness is 0.5-1mm, and it has the characteristics of tensile strength ≥20%, sweat resistance, and good biocompatibility. The carrier surface has a pre-reserved sensor mounting groove (depth 0.3mm, diameter matching the corresponding sensor) and a wire storage channel (width 0.2mm) to ensure that the sensor fits tightly to the skin of the hand without affecting finger movement.

[0016] Regarding the morphological definition and functional layout of the flexible carrier unit, the real-time hand spasticity sensing module after stroke adopts a design that combines anatomical structure adaptation with functional zoning: Glove-like overall configuration: The overall outline of the flexible carrier unit is highly consistent with the physiological and anatomical structure of the human hand, possessing a complete five-finger structure. This configuration ensures that the module can cover all core rehabilitation sensing areas of the stroke patient's hand, providing a stable spatial reference for the multimodal sensing unit.

[0017] Partial Fit Function Layout: To balance sensing accuracy and wearing comfort, the flexible carrier unit, instead of a fully enclosed solid structure, utilizes a perforated elastic mesh structure to achieve differentiated fit. Specifically, it is divided into the following areas: 1) High-precision bonding area: At key sampling points where the sensor is positioned (such as the dorsal side of each finger bone, fingertip, and finger root areas), the carrier adopts a solid silicone pressure structure. Utilizing the elastic recoil force of medical-grade silicone, the sensor is pressed tightly against the skin surface, ensuring the stability and consistency of signal acquisition.

[0018] 2) Dynamic compliance zone: In non-sensing areas (such as gaps between fingers, non-stress areas of the palm, etc.), the carrier is designed with a large-aperture mesh or retains a specific physical relaxation. This design avoids the restriction of joint movement caused by tight wrapping of the entire glove surface.

[0019] Therefore, the effects that flexible carrier units can achieve include: 1) Prevent displacement: By applying tight pressure to key nodes, the sensor is ensured not to undergo anatomical displacement when the finger is severely bent or spasmed, thus guaranteeing the authenticity and reliability of the sensed data.

[0020] 2) Improve compliance: The hollowed-out mesh design solves the heat dissipation problem for stroke patients during rehabilitation training that lasts for several hours, effectively avoiding skin allergies or discomfort caused by sweat accumulation.

[0021] 3) Ensuring freedom of movement: The combination of localized relaxation and highly elastic materials minimizes the mechanical resistance of the sensing module to the patient's active / passive rehabilitation movements, achieving the optimal solution between sensing accuracy and wearing compliance.

[0022] Regarding the fit to the skin of the hand, the sensor installation and tactile interaction mechanism are as follows: The real-time hand spasticity sensing module after stroke achieves mechanical coupling between the sensors in the multimodal sensing unit and the hand through a partitioned installation structure, ensuring the stability of signal acquisition and the safety of wearing it. The graded tactile structure is designed as follows: 1) Embedded window type (for pressure sensing): A mounting position (through hole or groove) adapted to the size of the sensor is provided on the inner side (skin-contact side) of the flexible carrier unit. The sensor is embedded in it, and its sensing surface is flush with or slightly protruding (0.1-0.2mm) from the inner surface of the carrier.

[0023] 2) Elastic adaptive compression: Utilizing the elastic tension of the flexible carrier unit (medical-grade silicone), a centripetal micro-pressure is generated when wearing the glove, causing the sensor sensing surface to naturally press against the corresponding anatomical part of the hand, effectively filtering artifact noise caused by glove slippage.

[0024] 3) Closed sandwich type (for motion sensing): For inertial measurement IMU sub-units, the carrier adopts a closed sandwich pocket structure, which completely wraps the sensor between flexible materials, which can accurately capture skeletal movements and avoid contact between electronic components and skin.

[0025] Biocompatibility and protective measures are as follows: 1) Anti-allergy isolation layer: Optionally, a 0.1mm thick biocompatible film (such as medical-grade TPU or a special coating) can be applied between the sensor sensing surface and the skin. This film blocks direct contact between the sensor encapsulation material and the skin without hindering pressure transmission, thus solving the problem of allergies or skin pressure injuries that may occur with long-term wear by stroke patients.

[0026] 2) Physical protection: The embedded design wraps the sensor around the edge of the carrier, avoiding the stinging sensation on the hand caused by the hard object at the edge of the sensor, and ensuring the flexibility of the finger when bending it significantly.

[0027] The connection is as follows: each sub-unit of the multimodal sensing unit is fixed to the corresponding mounting slot of the carrier with medical-grade double-sided adhesive; the data processing unit, data transmission unit, and power supply unit are detachably fixed to the back of the carrier via a snap-fit ​​structure, and the tightness of the fixation can be adjusted by Velcro.

[0028] In this embodiment, the multimodal sensing unit is specifically as follows: 1) Inertial Measurement Unit (IMU): Employs a miniature six-axis IMU sensor (MPU6050 optional), measuring 3mm×3mm×1mm, with a sampling rate of 200Hz, capable of acquiring angular velocity (measurement range ±2000° / s) and acceleration (measurement range ±16g) data of finger joints; Each IMU subunit is electrically connected to the mixed signal interface of the data processing unit via a wire, which is stored within the wire channel of the carrier to avoid damage from pulling.

[0029] The specific allocation of the mixed-signal interface is as follows: Digital Interface (I2C / SPI): The inertial measurement unit (IMU) uses a digital interface. Since the MPU6050 integrates an A / D converter, the data processing unit directly reads the processed attitude data via the digital bus.

[0030] Analog Interface (ADC): The tension sensor subunit and the thin-film pressure sensor subunit output continuous voltage signals, which are connected to the 12-bit ADC channel of the data processing unit for digital-to-analog conversion.

[0031] 2) Tension sensor subunit: A miniature tension / compression sensor (range 0-50N, accuracy ±0.1N) is used. It is connected to the fixed anchor point at the base of the finger and the back of the hand through a flexible guide wire (made of ultra-high molecular weight polyethylene, diameter 0.1mm). One end of the guide wire is mechanically connected to the force-bearing end of the sensor, and the other end is fixed to the carrier at the base of the finger through a miniature buckle. The signal output end of the sensor is electrically connected to the analog signal input interface of the data processing unit through a wire.

[0032] To accurately capture changes in finger muscle tension, a design scheme of discrete anchoring and remote sensing was adopted, considering the distribution location of the tension sensors and the connection method of the guide wires. The core components of the multimodal sensing unit are distributed as follows: The tension sensor subunit adopts a remote sensing layout. It is installed on the wrist side of the hand (avoiding the finger root area), a position that not only reduces the thickness of the hand covering but also utilizes the relative stability of the wrist to improve the accuracy of the sampling reference. A flexible guide wire of a set diameter is connected to a mechanical anchoring end at the proximal phalanx of the finger, forming a long-range tension transmission link across the joint. When any finger undergoes spastic flexion, the relative displacement between the finger root and the wrist is converted into axial tension on the guide wire, which is measured by the tension sensor subunit located on the wrist. The mechanical anchoring end (fixed point A) is fixed to a flexible carrier at the proximal phalanx (finger root) using a micro-clip. The flexible carrier is a reinforced structure achieved through an integrated molding process, ensuring non-destructive deformation under stress.

[0033] Specifically, the tension sensor subunit is installed on the palm side of the wrist, utilizing the distal artificial tendon conduction mechanism: 1) Mechanical principle: This guide wire constitutes a complete artificial tendon system. Although the tension sensor is installed in the stable wrist area (fixed end), it is connected to the proximal phalanx of the finger (force-bearing end) through an ultra-high molecular weight polyethylene guide wire.

[0034] 2) Sensing Mechanism: When a patient experiences hand spasm and involuntary finger flexion, a minute relative displacement occurs in the anatomical distance between the finger base and the wrist. This displacement is instantly converted into axial tension on the guide wire. The sensor monitors the relative tension value across the joint in real time, rather than the movement of the wrist muscles themselves. This design achieves a transformation from "local direct measurement" to "remote kinetic link measurement".

[0035] Furthermore, moving the sensor components to the wrist was based on a combination of engineering and medical considerations: 1) Avoiding mechanical interference: The anatomical space on the surface of the fingers is extremely small and the skin is extremely thin. If a sensor with a certain volume and rigidity is placed directly on the knuckle, it will severely hinder the already stiff finger movement of the patient, causing serious mechanical interference and making it impossible to truly reproduce the motor characteristics during spasticity.

[0036] 2) Improve signal-to-noise ratio (reduce anatomical displacement): The muscles and bones at the wrist are relatively stable, making it the optimal mounting reference point. Fixing the sensor here avoids sensor position shift (anatomical displacement) caused by severe finger spasms, thus ensuring the purity and consistency of the acquired signal.

[0037] 3) Lightweight design: The heavier sensing elements are placed far away from the fingers, which greatly reduces the load on the fingers, ensures the flexibility of the rehabilitation gloves when performing rehabilitation training, and improves the patient's wearing compliance.

[0038] This embodiment constructs a multimodal sensing network that combines near and far sensing. The proximal end (finger region) consists of an IMU subunit responsible for capturing joint dynamics and a thin-film pressure sensor that senses the gripping pressure of the fingertips, focusing on microscopic and local details. The distal end (wrist region) uses a tension sensor to capture the abnormal evolution trend of the overall muscle tension of the fingers from a global perspective. This near-far combined design enables the system to process both "high-frequency tremor (local feature)" and "sudden increase in muscle tension (overall feature)" simultaneously, which is superior to the solution of simply stacking sensors on the fingers in terms of spatial layout and sensing depth.

[0039] The force transmission path and material of the tension transmission link are as follows: 1) Guide line design: Made of 0.1mm diameter ultra-high molecular weight polyethylene (UHMWPE) material, which has high strength, low elongation and excellent self-lubricating properties.

[0040] 2) Connection Logic: One end of the guide wire is connected to the force-receiving end of the tension sensor, and the other end is securely fixed by a miniature clip at the base of the finger. The wire crosses the finger joint, forming a complete tension transmission link.

[0041] The working principle and monitoring logic of the multimodal sensing unit are as follows: 1) When a stroke patient experiences hand spasm that causes involuntary finger bending or a sudden increase in muscle tone, the relative displacement of the sensor at the base of the finger and the wrist causes tension in the guide wire.

[0042] 2) The tension is applied directly to the force-bearing end of the sensor through the wire.

[0043] 3) The data processing unit monitors the absolute magnitude and rate of change of tension values ​​to achieve real-time and accurate capture of abnormal muscle tone states.

[0044] In this embodiment, the thin-film pressure sensor subunit adopts a capacitive thin-film pressure sensor (range 0-10N, accuracy ±0.05N), with dimensions of 5mm×5mm×0.2mm and a surface covered with anti-slip texture; the sensor's signal output terminal is electrically connected to the digital signal input interface of the data processing unit 3 through a flexible FPC cable (thickness 0.1mm). The FPC cable is embedded in the conductor channel of the carrier to ensure signal stability when bent.

[0045] To accurately capture the pressure distribution and abnormal muscle tone of a stroke patient's fingers during the grasping process, the thin-film pressure sensor subunit adopts a "sandwich" mechanical sensing structure, the principle and structure of which are as follows: Working Principle (Pierre Resistance Effect): The main body of the thin-film pressure sensor consists of two conductive thin films and a pressure-sensitive material in between. Based on the piezoresistive effect of the conductive material, when an external force is applied to the sensor, the contact area between the upper and lower electrodes and the deformation of the pressure-sensitive material produce a dynamic response, resulting in a continuous change in resistance. The data processing unit collects this resistance change and converts it into a pressure signal.

[0046] Core innovative structure (sandwich layout): In order to accurately collect the "active grip force" of the fingers rather than simply the skin contact pressure, the following structure is introduced in this embodiment: 1) Rigid support pad: A rigid pad (support sheet) with a hardness higher than silicone is tightly placed on the back of the thin-film pressure sensor (away from the fingertip). When a gripping action occurs due to spasm, the stiff contraction of the fingertip muscles squeezes the pressure onto the pad, forming a sandwich structure of "fingertip-sensor-rigid support", ensuring that the sensor can capture the true gripping pressure distribution.

[0047] 2) Micro-bump sensitive design: The sensing surface of the thin film pressure sensor is designed with a micron-scale array of micro-bumps, which significantly increases the force concentration and thus greatly improves the sensing sensitivity under slight pressure, ensuring accurate capture of slight gripping pressure.

[0048] The thin-film pressure sensor subunit is embedded in the fingertip groove of the flexible carrier unit, with the sensing surface facing the palm side. It is softly encapsulated with medical-grade silicone, which not only ensures the comfort of flexible contact with the skin, but also improves the biocompatibility for long-term wear.

[0049] It uses a flexible FPC cable embedded in the carrier conductor channel, which effectively avoids the risk of pulling caused by frequent bending of the fingers and ensures the continuity of electrical signal transmission.

[0050] The built-in temperature compensation algorithm eliminates the influence of ambient temperature on the drift of the pressure-sensitive material; at the same time, the sealed packaging process provides waterproof and sweatproof capabilities to adapt to long-term wear scenarios in stroke rehabilitation training.

[0051] The core chip of the data processing unit adopts a microcontroller with ARM Cortex-M4 architecture (model STM32L431 optional), which integrates a 12-bit ADC module (sampling rate 1MHz) and a Kalman filter algorithm hardware acceleration module; the unit shell is made of ABS material, with dimensions of 20mm×30mm×5mm, and heat dissipation holes (hole diameter 1mm, spacing 2mm) on the surface.

[0052] The data processing unit is electrically connected to each sub-unit of the multimodal sensing unit, the data transmission unit, and the power supply unit via wires; the ADC module is used to convert the analog signals of the tension sensor and the thin-film pressure sensor sub-unit into digital signals, and the Kalman filter algorithm module is used to perform fusion processing on the multimodal data.

[0053] Specifically, the data processing unit includes: The feature extraction subunit is used to extract tremor dimension features, tension dimension features, and pressure dimension features from the fused data; The threshold judgment subunit is used to compare the tremor dimension feature, tension dimension feature and pressure dimension feature with their corresponding judgment thresholds to obtain the tremor judgment result, tension judgment result and pressure judgment result; The alarm subunit is used to output a graded spasm alarm when at least one of the tremor judgment result, tension judgment result, and pressure judgment result is abnormal, and to send control commands to external devices through the data transmission unit; the graded spasm alarm is used to distinguish different degrees of spasm.

[0054] The data transmission unit uses a Bluetooth 5.0 low-power module (model nRF52832 optional), with a transmission distance of ≥10m, power consumption ≤10mA, and supports encrypted data transmission (AES-128 encryption).

[0055] The data transmission unit is electrically connected to the data processing unit 3 via the SPI interface. It can transmit the processed spasticity perception data to the controller of the intelligent hand exoskeleton rehabilitation system in real time, and at the same time receive parameter configuration instructions (such as sensor sampling rate adjustment instructions) issued by the controller.

[0056] The power supply unit uses a micro lithium polymer battery (capacity 100mAh, voltage 3.7V), paired with a charging management chip (model TP4056 optional), supports USB-C interface charging, charging time ≤1.5h, and single-charge battery life ≥8h.

[0057] The power supply unit is electrically connected to the power interface of the data processing unit via wires to provide a stable power supply to the entire sensing module; the battery is equipped with overcharge, over-discharge, and short-circuit protection circuits to ensure safe use.

[0058] In this embodiment, the flexible carrier unit provides a mounting base for each sensor and electronic component that fits the hand, while ensuring wearing comfort and freedom of finger movement, avoiding sensing errors caused by excessively hard materials or poor fit; the IMU subunit collects dynamic motion data of the finger joints in real time, and identifies involuntary tremors of the hand in stroke patients by changes in angular velocity and acceleration (e.g., tremor frequency > 5Hz is judged as a precursor to spasticity); the tension sensor subunit indirectly detects changes in finger muscle tension through a guide wire (the tension of the guide wire is equal to the magnitude of muscle tension), and when the tension suddenly increases (e.g., an increase of > 5N within 1 second), it is judged as an abnormal increase in muscle tension, which may be accompanied by spasticity; the thin-film pressure sensor subunit captures pressure distribution data when the fingers grasp, and if the pressure value of a certain finger is consistently 0 or much lower than that of other fingers (e.g., the difference > 3N), it can help determine whether the finger is unable to exert force normally due to spasticity.

[0059] The spasm recognition function is assigned to the data processing unit. This unit integrates an embedded microcontroller and firmware algorithms, serving as the core brain of the sensing system. Its specific responsibilities include: real-time data synchronization from multiple sensors, Kalman filter fusion, feature parameter calculation, and logical determination of the spasm state. This design deeply integrates perception and decision-making, ensuring high real-time performance of data processing and low latency in closed-loop feedback.

[0060] Hand spasm symptom identification process (cascaded processing) The real-time perception module for post-stroke hand spasticity adopts a recognition process of "multimodal fusion perception - feature cascade extraction - multidimensional matrix decision", the specific steps of which are as follows: Step 1: Multi-channel synchronous sampling (data acquisition).

[0061] The data processing unit simultaneously acquires raw signals from three types of sensors at a high sampling rate (e.g., above 100Hz): Inertial Measurement Unit (IMU) Subunit: Acquires triaxial angular velocities of finger joints. With acceleration .

[0062] Tension sensor subunit: Collects the tension T of the guide wire across the joint.

[0063] Thin-film pressure sensor subunit: Collects pressure distribution data P at various points on the fingertip.

[0064] Step 2: Kalman filter fusion (preprocessing and data purification).

[0065] A cascaded data processing architecture is adopted to perform filtering and fusion simultaneously.

[0066] Dynamic prediction: IMU data is used as the predicted value of hand movement state.

[0067] Observation correction: Tension T and pressure P are used as observed variables. The predicted values ​​are updated and corrected through Kalman recursion to suppress the integral drift error of the IMU and to dynamically filter out the interference of ambient temperature through the covariance matrix.

[0068] Step 3: Feature parameters.

[0069] Extracting 3D feature vectors from the fused data: Motion characteristics: Extract the main frequency f, amplitude A, and duration t from the IMU signal.

[0070] Mechanical characteristics: Extracting muscle tone value T and rate of change from tension signals .

[0071] Grasp characteristics: Calculating the variance of pressure distribution from the pressure signal And the total pressure of each indicator, Ptotal.

[0072] Step 4: Multidimensional Symptom Recognition Matrix (Decision Logic) The system performs closed-loop identification based on the criteria in Table 1 and comprehensively assesses the spasticity level: Table 1 Judgment Indicators

[0073] Step 5: Tiered Alarm and Closed-Loop Response (Response Mechanism) Once the above logical triggering conditions are met (such as an anomaly in any dimension or a composite anomaly across multiple dimensions), the data processing unit immediately executes the following response: Recording and Early Warning: Based on the number and severity of triggering conditions, they are classified into three levels: Level 1 (reminder), Level 2 (intervention), and Level 3 (emergency alarm).

[0074] Closed-loop control: Emergency execution commands (such as emergency stop of active drive, switch to passive slow stretching mode) are sent to the exoskeleton controller via Bluetooth 5.0 protocol to ensure that the overall delay from spasm recognition to system response is controlled within 100ms, effectively avoiding secondary injuries caused by spasms.

[0075] Data processing unit: Receives raw data from each sensor, fuses motion data from the IMU, muscle tone data from the tension sensor, and pressure data from the thin-film pressure sensor using a Kalman filter algorithm, and outputs three-dimensional spasticity state data of "tremor frequency-muscle tone value-pressure distribution", filtering out noise interference from individual sensors (such as IMU drift error and ambient temperature interference from the pressure sensor). Data transmission unit: Enables real-time data interaction between the sensing module and the exoskeleton controller, ensuring that spasticity data is fed back to the closed-loop control system in a timely manner, and at the same time receives parameter adjustment instructions from the controller; Power supply unit: Provides long-term and safe power supply for the entire module, meeting the battery life requirements of patients for daily rehabilitation training (usually 4-6 times, 30 minutes each time).

[0076] Example 2 This embodiment provides an intelligent hand exoskeleton rehabilitation system, including a real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion, and an exoskeleton glove; the real-time sensing module for post-stroke hand spasticity is physically and functionally integrated with the driving layer of the exoskeleton glove through flexible electronic integration technology; the controller of the exoskeleton glove receives spasticity sensing data output by the sensing module through the data transmission unit, and executes corresponding closed-loop control responses according to the control commands output by the data transmission unit.

[0077] Specifically, the integration method is as follows: The real-time hand spasticity sensing module after stroke, serving as an inner lining layer (sensing layer), is physically and functionally integrated with the exoskeleton glove (driving layer) through the following three methods: 1) Laminated integration: The flexible carrier unit (medical-grade silicone / fabric) of the real-time sensing module for post-stroke hand spasticity is hot-pressed or ultrasonically welded to the inner lining of the exoskeleton glove to form a "sandwich" structure, ensuring that the sensor position is fixed and does not shift.

[0078] 2) Flexible overlapping integration: At the data processing unit and power supply unit locations, Velcro (hook and loop fasteners) or snap-on structures are used to detachably fix the modules to the back of the exoskeleton glove, facilitating maintenance and charging.

[0079] 3) Concealed integration of wires: The wires of the sensing module are housed in the wire channel of the carrier. This channel is aligned with the internal wiring groove of the exoskeleton glove, achieving concealed integration of the cables and avoiding interference.

[0080] In summary, this application has the following technical effects: High accuracy in spasticity detection: The multimodal sensing unit uses IMU, tension sensor, and thin-film pressure sensor to collect data collaboratively (avoiding blind spots of a single sensor), and the data processing unit uses Kalman filtering algorithm (filtering noise and improving data reliability). Compared with existing single IMU sensing solutions (which only detect tremors through IMU), this module improves the accuracy of spasticity recognition to over 95% (the accuracy of a single IMU solution is about 70%) through multimodal data fusion. It can effectively distinguish between "normal movement tremor" and "spastic tremor" and avoid misjudgment.

[0081] Comprehensive Perception Dimensions: The multimodal sensing unit features a three-dimensional sub-unit design (targeting three types of spasticity-related indicators: tremor, muscle tone, and pressure), and the installation position of each sub-unit is precisely matched to the physiological structure of the hand (e.g., the IMU covers each phalanx, and the tension sensor corresponds to the force point at the base of the finger). Existing solutions only focus on a single dimension of tremor or muscle tone, while this module simultaneously covers three-dimensional data of "tremor-muscle tone-grasping pressure," which can fully present the complex state of hand spasticity in stroke patients (e.g., some patients experience spasticity accompanied by increased muscle tone but no obvious tremor, while some patients experience tremor but have normal muscle tone).

[0082] Good wearing comfort and compatibility: The flexible carrier unit uses medical-grade composite materials (soft and fits well) and a detachable fixing structure (clasp + Velcro), as well as miniaturized design of each sensor (size ≤5mm×5mm); the flexible carrier and miniature sensor design make the module weight ≤30g (the weight of existing similar sensor modules is usually ≥50g), with an elongation rate ≥20%, which can be adapted to patients with different hand shapes (palm width 7-10cm) without affecting rehabilitation movements such as finger bending and extension.

[0083] High real-time data performance: The high sampling rate (200Hz) of the IMU subunit, the hardware acceleration algorithm (Kalman filter hardware module) of the data processing unit, and the low-power Bluetooth 5.0 (transmission delay ≤50ms) of the data transmission unit can complete status recognition and feedback to the exoskeleton system within 100ms of the occurrence of the patient's spasm, which buys time for emergency response (such as pausing training and reverse relief) and avoids secondary injury.

[0084] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0085] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion, characterized in that, include: The flexible carrier unit has a glove-shaped localized fitting structure and adopts a hollowed-out elastic mesh design to fit the contours of different parts of the human hand; A multimodal sensing unit, fixed on the flexible carrier unit, includes an inertial measurement unit (IMU) subunit, a tension sensor subunit, and a thin-film pressure sensor subunit. The inertial measurement unit is used to collect motion data of the finger joints, the tension sensor subunit remotely transmits muscle tension data of the fingers through a flexible guide wire, and the thin-film pressure sensor subunit is used to collect pressure distribution data of the fingertip gripping. The data processing unit, fixed on the flexible carrier unit, is electrically connected to each sub-unit of the multimodal sensing unit. It is used to synchronously sample and fuse the motion data, muscle tension data, and pressure distribution data collected by the multimodal sensing unit, and to perform spasticity state recognition based on the fused data. The data transmission unit is electrically connected to the data processing unit and is used to transmit the spasm perception data identified by the spasm state to an external device in real time.

2. The real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion according to claim 1, characterized in that, The hollowed-out elastic mesh design of the flexible carrier unit is specifically as follows: A solid silicone pressure structure is used as a high-precision bonding area in the sensor placement area; Large-aperture meshes or physical relaxation are used as dynamic compliance zones in the gaps between fingers and the non-stressed areas of the palm.

3. The real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion according to claim 1, characterized in that, The tension sensor subunit adopts a remote sensing layout; the tension sensor subunit is installed on the wrist position on the palm side, and is connected to the mechanical anchoring end of the proximal phalanx of the finger through a flexible guide line of a set diameter, forming a remote tension transmission link across the joint; when any finger spasms and flexes, the relative displacement between the base of the finger and the wrist is converted into the axial tension of the guide line, which is measured by the tension sensor subunit located on the wrist.

4. A real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion as described in claim 3, characterized in that, The inertial measurement unit (IMU) subunit is arranged on the dorsal side of each phalanx of the hand, and the thin-film pressure sensor subunit is arranged on the fingertip of each finger. The inertial measurement unit and the thin-film pressure sensor subunit form a proximal sensing network for capturing finger joint dynamics and local pressure details. The tension sensor subunit, together with the inertial measurement unit and the thin-film pressure sensor subunit, forms a distal sensing network for capturing abnormal evolution trends of overall finger muscle tension through the flexible guide line.

5. A real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion according to claim 1, characterized in that, The thin-film pressure sensor subunit adopts a sandwich-type mechanical sensing structure. A rigid support pad is provided on the back of the sensor, and a micro-protrusion array is provided on the sensing surface. When a spasmodic grip occurs, the contraction pressure of the fingertip muscles is squeezed towards the rigid support pad, forming a sandwich structure of fingertip-sensor-rigid support.

6. A real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion according to claim 1, characterized in that, The data processing unit integrates a Kalman filter algorithm hardware acceleration module and uses a cascaded data processing architecture to perform multimodal data fusion.

7. A real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion according to claim 1, characterized in that, The data processing unit specifically includes: The feature extraction subunit is used to extract tremor dimension features, tension dimension features, and pressure dimension features from the fused data; The threshold judgment subunit is used to compare the tremor dimension feature, tension dimension feature and pressure dimension feature with their corresponding judgment thresholds to obtain the tremor judgment result, tension judgment result and pressure judgment result; The alarm subunit is used to output a graded spasm alarm when at least one of the tremor judgment result, tension judgment result, and pressure judgment result is abnormal, and to send control commands to external devices through the data transmission unit; the graded spasm alarm is used to distinguish different degrees of spasm.

8. A real-time sensing module for post-stroke hand spasticity based on multimodal sensor fusion according to claim 7, characterized in that, The threshold values ​​for the tremor dimension include the dominant frequency threshold and the duration threshold; the threshold values ​​for the tension dimension feature are the rate of change of muscle tone in the tension dimension; and the threshold values ​​for the pressure dimension feature include the pressure distribution variance threshold and the total pressure change threshold.

9. An intelligent hand exoskeleton rehabilitation system, characterized in that, The invention includes a real-time sensing module for post-stroke hand spasticity based on multimodal sensing fusion as described in any one of claims 1-8, and an exoskeleton glove; the real-time sensing module for post-stroke hand spasticity is physically and functionally integrated with the driving layer of the exoskeleton glove through flexible electronic integration technology; the controller of the exoskeleton glove receives spasticity sensing data output by the sensing module through the data transmission unit, and executes corresponding closed-loop control responses according to the control commands output by the data transmission unit.

10. The intelligent hand exoskeleton rehabilitation system according to claim 9, characterized in that, The flexible electronic integration technology is lamination integration technology, flexible overlap integration technology, or wire concealment integration technology.

Images