Rehabilitation platform for expressive aphasia based on joint detection of actions and vocalizations and intelligent glove device

Abstract

This application relates to the field of stroke aphasia rehabilitation technology, and in particular to an expressive aphasia rehabilitation platform based on joint detection of movement and speech, and its intelligent glove device. The platform includes a patient-side device, a wearable glove, a cloud server, and a three-terminal linkage device. The patient-side device runs an interactive training application, displaying scenario demonstration videos via a touchscreen. The wearable glove is placed on the patient's unaffected hand and incorporates a six-axis inertial sensor and a thin-film pressure sensor, which collect movement trajectory and pressure data at sampling frequencies of at least 100Hz and 50Hz, respectively, to assess gesture completion in real time and provide vibration feedback. The cloud server synchronizes data between the patient, family, and doctor ends, using HTTPS protocol for transmission and encryption. The system captures the patient's willingness to speak through a speech activity detection module, dynamically adjusts the training difficulty, and provides multimodal prompts. This application is suitable for expressive aphasia patients with right-sided hemiplegia, activating the brain's language center through gestures to achieve personalized closed-loop rehabilitation training while protecting the patient's confidence in rehabilitation.

Description

Technical Field

[0001] This invention belongs to the field of medical rehabilitation technology, specifically, it relates to an expressive aphasia rehabilitation platform based on joint detection of movement and vocalization and its intelligent glove device. Background Technology

[0002] As is known in existing technologies, various assistive communication tools and rehabilitation systems have been developed for rehabilitation training of patients with post-stroke expressive aphasia. These include static assistive tools such as communication cards and flashcards that primarily use visual cues, as well as mobile software applications such as speech training apps, scenario-based communication apps, and multi-dimensional modular communication apps. In addition, there are professional-grade assistive communication hardware and software systems such as eye-tracking communication devices and speech generation devices.

[0003] It is also known that, to enhance the fun and interactivity of rehabilitation training, existing technologies have developed solutions that combine gesture training with language rehabilitation. Based on embodied cognition theory and mirror neuron system research in cognitive neuroscience, there is a strong positive correlation between limb movements and speech rehabilitation. Physical movements accompanying the hands or limbs can effectively activate the brain's language center, helping patients retrieve words more easily.

[0004] However, existing technological solutions have the following limitations: traditional assistive communication tools are not inclusive enough for hemiplegic patients, most aphasic patients have mild to moderate unilateral motor impairment, and existing equipment cannot fully adapt to the special physiological limitations of this group; daily fixed-time rehabilitation tasks can easily make patients feel bored and stressed, which is not conducive to transferring language ability to daily life; home-based independent rehabilitation lacks immediate multi-sensory feedback, and after being separated from clinical guidance, patients have difficulty confirming the results of their expression, and single visual or auditory stimulation is not enough to maintain the continuous activity of language neural pathways; long-term high-frequency trial and error leads to emotional exhaustion for family caregivers, patients have sound cognition but have difficulty outputting, and repeated guessing errors by family members can easily exhaust the energy of both parties. Summary of the Invention

[0005] The objective of this invention is to provide a rehabilitation training system and method for patients with moderate to severe expressive stroke-induced aphasia. This system and method, while addressing the physiological limitations of right-sided hemiplegia, activates the brain's language center through hand gestures, thereby promoting the patient's language function recovery. This objective is achieved through the rehabilitation training method according to claim 1, the rehabilitation training system according to claim 7, and the three-terminal linkage device according to claim 14. The dependent claims relate to advantageous structural designs of the invention.

[0006] The rehabilitation training method for patients with moderate to severe expressive stroke aphasia according to the present invention includes the following steps: the patient selects a training scenario through a patient-end device; the system displays a scenario demonstration video corresponding to the training scenario to the patient; the patient wears a glove with built-in sensors and performs a target gesture action according to the demonstration video; the six-axis inertial sensor of the glove collects the patient's motion trajectory data, and the thin-film pressure sensor collects the pressure data of the contact surface between the patient's hand and the glove; the system evaluates the gesture completion degree in real time based on the motion trajectory data and pressure data; if the gesture completion degree reaches a preset threshold, the system triggers vibration feedback and plays the speech of the target word; if the gesture completion degree does not reach the preset threshold, the system provides multimodal prompts or reduces the training difficulty; the system captures the patient's vocal intention through a speech activity detection module and records the training results according to the detection results; the system synchronizes the training data to a cloud server.

[0007] According to the present invention, the glove worn by the patient is placed on the unaffected hand. This is because the brain injury in patients with expressive stroke-induced aphasia is located close to the motor cortex, and the adjacent motor cortex is often damaged simultaneously, resulting in approximately 75% of patients experiencing right-sided hemiplegia or weakness. Therefore, placing the glove on the patient's unaffected hand, i.e., the left hand, ensures that the patient can perform gesture training movements normally. In this way, the problem of insufficient tolerance for hemiplegia in traditional assistive communication tools in the prior art is solved, allowing patients to complete rehabilitation training tasks within the limited range of their limb motor abilities.

[0008] If the contact force detected by the glove's pressure sensor is greater than a first preset threshold and the duration is greater than a first preset duration, the grasping gesture is determined to be complete; if the contact force detected by the pressure sensor is within a second preset threshold and the IMU sensor detects the finger pinching action, the pinching gesture is determined to be complete. However, different gesture types can also be designed according to different training scenarios to adapt to different rehabilitation training goals.

[0009] The gloves are preferably made of breathable fabric and have a built-in flexible circuit board. This flexible circuit board integrates a six-axis inertial sensor, a thin-film pressure sensor, and a miniature vibration motor. The six-axis inertial sensor is used to collect triaxial acceleration and triaxial angular velocity data. The thin-film pressure sensor is arrayed on the fingertips and palm of the glove. The miniature vibration motor is located on the back of the hand to provide tactile feedback to the patient. The first and second blanks are preferably made of unvulcanized rubber.

[0010] Furthermore, according to the present invention, the system can dynamically adjust the training difficulty coefficient based on the patient's training performance. Specifically, when the patient fails to complete the same target gesture three times consecutively, the system automatically lowers the completion threshold of that gesture or switches the current training vocabulary to the next vocabulary; when the patient successfully completes the target gesture three times consecutively, the system automatically raises the completion threshold of that gesture. This method enables personalized rehabilitation training with adaptive difficulty. According to existing technologies, rehabilitation training applications typically use training tasks of fixed difficulty, preventing patients from adjusting the training intensity according to their own rehabilitation progress. According to the present invention, the system can also record the patient's training behavior data, including but not limited to single training session dwell time, gesture timeout rate, vocabulary completion rate, and vocalization detection pass rate, and generate a patient ability profile based on the behavioral data. As an optional feature, the system automatically updates the training baseline difficulty weekly based on the patient's ability profile to ensure that the training tasks are neither too easy, causing the patient to lose interest, nor too difficult, causing the patient to feel frustrated.

[0011] A speech activity detection module is installed in the audio processing unit of the patient's device to analyze in real time whether the patient is emitting sound signals. The speech activity detection module uses an energy threshold detection algorithm; when the energy value of the input audio signal exceeds a preset energy threshold, it is determined that speech has been detected; when the energy value of the input audio signal is lower than the preset energy threshold, it is determined that no speech has been emitted. As an optional solution, the speech activity detection module can use an endpoint detection algorithm based on Mel-frequency cepstral coefficients to improve detection accuracy in complex noisy home environments. In this invention, the speech activity detection module is only used to assess the patient's vocal activity level and does not judge the accuracy of the patient's pronunciation. This is because patients with moderate to severe expressive aphasia usually have low pronunciation clarity, and demanding high pronunciation accuracy is not conducive to protecting the patient's confidence in rehabilitation.

[0012] The system comprises patient-side devices, family-side devices, and doctor-side devices, all synchronized via a cloud server. The patient-side device is a tablet running an interactive hand training application called "Echo Hand"; the family-side device is a smartphone running a family-side application; and the doctor-side device is a personal computer running a doctor-side assessment program. Training data collected by the patient-side devices is encrypted and uploaded to the cloud server, which then distributes the data to the family-side and doctor-side devices. This again creates a collaborative advantage, as family members can monitor the patient's rehabilitation progress in real time, doctors can remotely adjust the rehabilitation plan, and patients can receive professional guidance in their home environment.

[0013] Preferably, after the patient completes the day's training tasks, the system automatically plays a pre-recorded encouraging message from a family member. This enhances the emotional connection between the patient and their family, improving patient adherence to home-based rehabilitation. Additionally, the family-side application automatically generates a daily rehabilitation dashboard, including indicators such as training duration, number of words completed, and vocal activity level. Finally, as an optional feature, the doctor-side assessment program may automatically generate a weekly assessment report, including trends in the patient's ability profile and suggested adjustments to training parameters.

[0014] According to the present invention, the rehabilitation training system for moderate to severe expressive stroke aphasia can also be implemented using the following architecture as an optional solution: the patient-side device includes a touch screen, an audio playback unit, a microphone acquisition unit, a glove connection unit, and a main control unit; the main control unit is connected to the touch screen, audio playback unit, microphone acquisition unit, and glove connection unit respectively; the glove connection unit establishes a connection with the sensor module built into the glove via a Bluetooth wireless communication protocol. The main control unit is used to execute rehabilitation training control logic, including scene rendering, gesture assessment, speech activity detection, and data synchronization. In this case, only one patient-side device is needed to complete the main interactive functions of rehabilitation training, rather than relying on multiple independent devices. However, instead of concentrating all functions on a single device, it is more precisely a modular design, with the glove connected to the patient-side device as an independent peripheral via wireless communication.

[0015] The preferred approach is to achieve data interaction between the three devices via a cloud server, specifically using HTTPS protocol for data transmission and AES-256 encryption. Training data can then be stored in a cloud database and organized according to time series, facilitating long-term rehabilitation effect assessment by physicians. In this way, patients can access the latest individual ability profile and training baseline settings from the cloud each time they train, achieving truly personalized closed-loop rehabilitation.

[0016] In the context of this invention, the rehabilitation training system is generally preferred for patients with moderate to severe expressive aphasia. These patients have impaired language expression ability but have largely retained their cognitive function, and are able to understand training instructions and cooperate in performing gestures.

[0017] The method for rehabilitation training according to the present invention generally preferably includes the final step: the system classifies the training results as "both gesture and vocalization were successful", "gesture only was successful" or "gesture failed" based on the patient's vocalization detection results, and adjusts the prompting strategy for the next round of training based on the classification results.

[0018] According to the present invention, a rehabilitation training system for moderate to severe expressive stroke aphasia includes: a patient-side device equipped with a touch screen, an audio playback unit, a microphone acquisition unit, and a main control unit; a wearable glove equipped with a six-axis inertial sensor, a thin-film pressure sensor, and a vibration feedback unit; a cloud server for storing training data and achieving data synchronization across the three terminals; a family-side device running a family-side application; and a doctor-side device running a doctor-side assessment program. This approach addresses the problems of lack of professional guidance, multi-sensory feedback, and family support in existing home rehabilitation methods. Furthermore, an adaptive algorithm module can be included to dynamically adjust the training difficulty based on the patient's training performance.

[0019] In addition, a module for providing multimodal cues is preferably provided, including a visual cues module, an auditory cues module, and a tactile cues module. The visual cues module is used to display scenario demonstration videos and gesture animations, the auditory cues module is used to play target vocabulary speech and training instructions, and the tactile cues module is used to provide vibration feedback.

[0020] Generally, it is preferred that the system configuration according to the present invention is for implementing the rehabilitation training method according to the present invention.

[0021] According to the present invention, the wearable glove is configured to be worn on the unaffected hand of the patient, and the sampling frequencies of the six-axis inertial sensor and the thin-film pressure sensor are not less than 100Hz and 50Hz, respectively, to ensure the real-time performance and accuracy of the hand gestures.

[0022] The beneficial effects of this invention are as follows: By collecting patients' hand gesture and pressure data through a glove-type wearable device, it can adapt to the physiological limitations of patients with expressive aphasia accompanied by right-sided hemiplegia; by stimulating the activity of the patient's brain's language center through multimodal cues (visual, auditory, and tactile), it achieves synergistic promotion of hand gestures and language rehabilitation based on embodied cognition theory and mirror neuron systems; by enabling real-time connection between patients' home rehabilitation and family companionship and doctor guidance through a three-terminal linkage architecture; by achieving personalized closed-loop rehabilitation training through a difficulty-adaptive algorithm; and by evaluating only vocal activity without demanding perfect pronunciation through a speech activity detection module, effectively protecting the patient's confidence in rehabilitation. Attached Figure Description

[0023] Figure 1 The overall architecture of a rehabilitation training system according to an embodiment of the present invention is shown, which includes a patient-side device, wearable gloves, a cloud server, a family-side device, and a doctor-side device; Figure 2 A schematic diagram of a wearable glove according to an embodiment of the present invention is shown, which is provided with a six-axis inertial sensor, a thin-film pressure sensor and a vibration feedback unit; Figure 3 The module composition of a patient terminal device according to an embodiment of the present invention is shown, which includes a touch screen, an audio playback unit, a microphone acquisition unit, and a main control unit; Figure 4 The present invention illustrates a three-terminal linkage data interaction process according to an embodiment of the present invention, wherein the cloud server establishes data connections with the patient-end device, the family-end device, and the doctor-end device respectively; Figure 5 A flowchart of a rehabilitation training method according to an embodiment of the present invention is shown, which includes steps of scene selection, situation demonstration, gesture execution, gesture assessment, feedback triggering, and data synchronization. Detailed Implementation

[0024] Figure 1 A rehabilitation training system 100 according to a first embodiment of the present invention is shown, which includes a patient-side device 10, a wearable glove 20, a cloud server 30, a family-side device 40, and a doctor-side device 50. Since the devices synchronize data through the cloud server 30, the patient-side device 10, the family-side device 40, and the doctor-side device 50 can be independently placed in different geographical locations. Therefore, in... Figure 1 The view of each device is considered as its position view within system 100.

[0025] The patient-side device 10 according to the present invention preferably includes a touch screen 11, an audio playback unit 12, a microphone acquisition unit 13, and a main control unit 14, wherein one or more modules can be used for scene rendering, gesture assessment, voice activity detection, and data transmission for rehabilitation training. In particular, the patient-side device 10 is preferably a tablet computer running an interactive hand training application called "Echo Hand" (see...). Figure 1 ).

[0026] The touchscreen display 11 of the patient-side device 10 preferably has a display size of 10 to 15 inches, particularly 10.1 to 13.3 inches, and more preferably 11.6 to 12.3 inches, with a resolution of, for example, 1920×1080 to 2560×1600 pixels, particularly 1920×1200 pixels. The audio playback unit 12 preferably has an output power of 2 to 5 watts, particularly 3 to 4 watts, and a frequency response range of, for example, 100Hz to 20kHz. The microphone acquisition unit 13 preferably has a sensitivity of -30dB to -50dB, particularly -38dB, and a signal-to-noise ratio of not less than 60dB.

[0027] Figure 2A wearable glove 20 according to an embodiment of the present invention is shown, which is worn on the unaffected hand (i.e., the left hand) of a patient. The wearable glove 20 is preferably made of a breathable fabric material and has a built-in flexible circuit board 21, on which a six-axis inertial sensor 22, a thin-film pressure sensor 23 and a micro vibration motor 24 are integrated.

[0028] A six-axis inertial sensor 22 is used to acquire triaxial acceleration and triaxial angular velocity data. The sampling frequency is preferably not less than 100 Hz, particularly 120 Hz to 200 Hz, and more preferably 150 Hz. The acceleration range is, for example, ±2g to ±16g, particularly ±8g, and the angular velocity range is, for example, ±250° / s to ±2000° / s, particularly ±1000° / s. Thin-film pressure sensors 23 are arranged in an array on the finger and palm portions of the glove. The sampling frequency is preferably not less than 50 Hz, particularly 50 Hz to 100 Hz, and more preferably 80 Hz. The pressure detection range is, for example, 0 to 100 kPa, particularly 0 to 50 kPa, and more preferably 0 to 20 kPa.

[0029] A miniature vibration motor 24 is disposed on the back of the hand of the glove for providing tactile feedback to the patient, with a vibration frequency of, for example, 100Hz to 300Hz, particularly 150Hz to 250Hz, more preferably 200Hz, and a vibration amplitude of, for example, 0.5mm to 2mm, particularly 0.8mm to 1.5mm.

[0030] The wearable glove 20 establishes a connection with the patient terminal device 10 via a Bluetooth wireless communication protocol, with a communication distance of, for example, 3 to 10 meters, particularly 5 to 8 meters, and a Bluetooth version of, for example, 4.0 to 5.0, particularly BLE 5.0.

[0031] Figure 3 A cloud server 30 and its data interaction architecture according to an embodiment of the present invention are illustrated. The cloud server 30 includes a data storage module 31, a data encryption module 32, and a synchronization and distribution module 33. The data storage module 31 is used to store the patient's training behavior data, including but not limited to single training session duration, gesture timeout rate, vocabulary completion rate, and vocalization detection pass rate. The data storage is organized in a time series manner, preferably indexed by training date and timestamp.

[0032] The data encryption module 32 uses the AES-256 algorithm for data encryption, and preferably transmits data via the HTTPS protocol. The synchronization distribution module 33 distributes the encrypted training data collected by the patient-side device 10 to the family-side device 40 and the doctor-side device 50, with a synchronization delay preferably not exceeding 5 seconds, and particularly 1 to 3 seconds.

[0033] The family-side device 40 is preferably a smartphone running a family-side application. The family-side application automatically generates a daily rehabilitation dashboard, including indicators such as training duration, number of words completed, and vocal activity level. The doctor-side device 50 is preferably a personal computer running a doctor-side assessment program. The doctor-side assessment program automatically generates a weekly assessment report, including trends in the patient's ability profile and suggested adjustments to training parameters.

[0034] Methods for assessing gesture completion include Figure 4 As shown. Gesture completion rate. Based on the completion rate of the motion trajectory and pressure completion rate The weighted calculation is as follows:

[0035] in, and As the weighting coefficient, preferred Up to 0.8, especially 0.7, Up to 0.4, especially 0.3. Motion trajectory completion rate. Pressure completion rate is calculated by comparing the similarity between the patient's actual movement trajectory and the target movement trajectory. The similarity is calculated by comparing the patient's actual pressure distribution with the target pressure distribution.

[0036] If the pressure sensor on the glove detects the contact force Greater than the first preset threshold And duration Greater than the first preset duration If so, the grasping gesture is considered complete. Preferably, For example, 3 kPa to 8 kPa, especially 5 kPa, For example, 0.5 seconds to 2 seconds, especially 1 second. If the pressure sensor detects the contact force... Second preset threshold range If the IMU sensor detects a pinching motion, the pinching gesture is considered complete. Preferably, For example, 0.5 kPa, For example, 3 kPa.

[0037] The system dynamically adjusts the training difficulty coefficient based on the patient's training performance. When a patient fails to complete the same target gesture three times consecutively, the system automatically lowers the completion threshold for that gesture or switches the current training vocabulary to the next vocabulary. The reduction in the completion threshold is preferably 5% to 15%, particularly 10%. When a patient successfully completes the target gesture three times consecutively, the system automatically raises the completion threshold for that gesture, preferably by 3% to 10%, particularly 5%.

[0038] Figure 5 The workflow of a voice activity detection module according to an embodiment of the present invention is illustrated. The voice activity detection module is disposed in the audio processing unit of the patient-side device 10 and is used to analyze in real time whether the patient is emitting a sound signal. The voice activity detection module employs an energy threshold detection algorithm, which determines the energy value of the input audio signal. Exceeding the preset energy threshold When the energy value of the input audio signal is [value missing], it is determined that sound has been detected; Below the preset energy threshold When this occurs, it is determined that no sound has been emitted. Preferably, For example, -40dB to -60dB, especially -50dB.

[0039] As an optional embodiment, the voice activity detection module can employ an endpoint detection algorithm based on Mel-frequency cepstral coefficients (MFCC) to improve detection accuracy in complex noisy home environments. MFCC feature extraction preferably uses 13 to 40 Mel filters, particularly 20 to 26 Mel filters, with a frame length of, for example, 20 ms to 40 ms, particularly 25 ms, and a frame shift of, for example, 10 ms to 20 ms, particularly 10 ms.

[0040] The system categorizes training results into "both gesture and vocalization successful," "gesture only successful," or "gesture failed" based on the patient's vocalization test results, and adjusts the prompting strategy for the next round of training according to the categorization results. The categorization rules are shown in the table below:

[0041] in, The threshold for gesture passing is preferably 70% to 90%, and especially 80%.

[0042] The multimodal cueing module includes a visual cueing module, an auditory cueing module, and a tactile cueing module. The visual cueing module displays contextual demonstration videos and gesture animations, with a video frame rate of, for example, 24fps to 60fps, particularly 30fps, and a resolution of, for example, 720p to 1080p. The auditory cueing module plays target vocabulary speech and training instructions, with a speech sampling rate of, for example, 16kHz to 48kHz, particularly 44.1kHz, and a bitrate of, for example, 64kbps to 256kbps, particularly 128kbps. The tactile cueing module provides vibration feedback, with vibration modes including, but not limited to, single vibrations, continuous vibrations, and rhythmic vibrations.

[0043] Although in the illustrated embodiment the patient device 10 and wearable glove 20 are connected via Bluetooth, according to the present invention, the connection can also be established via other wireless communication protocols, such as Wi-Fi, ZigBee or Bluetooth Low Energy (BLE).

[0044] After the patient completes the day's training tasks, the system automatically plays a pre-recorded encouraging message from a family member. The duration of the encouraging message is preferably 3 to 15 seconds, especially 5 to 10 seconds, and the audio format is, for example, MP3 or WAV, with a sampling rate of, for example, 16kHz or 44.1kHz.

[0045] The system records the patient's training behavior data and generates a patient competency profile based on this data. The competency profile includes, but is not limited to, the distribution of gesture completion, trends in vocal activity, changes in training duration, and difficulty adaptation level. The system automatically updates the baseline training difficulty weekly based on the patient's competency profile to ensure that the training tasks are neither too easy (leading to loss of interest) nor too difficult (causing frustration). The preferred update range for baseline difficulty is 5% to 20%, particularly 10%.

[0046] In the tire factory, the strip 13 with integrated electronic components 3 is combined with the rubber used to manufacture the tire and vulcanized together, so that the electronic components are finally integrated into the tire.

[0047] Although the system uses a tablet computer as the patient terminal device 10 in the above embodiments, according to the present invention, a smartphone can also be used as the patient terminal device 10, or an all-in-one device with a touch screen can be used.

[0048] Furthermore, according to the present invention, the wearable glove 20 can be configured with a replaceable sensor module to adapt to different training scenarios and rehabilitation goals. For example, for patients in the early stages of rehabilitation, a pressure sensor with a large detection range can be selected; for patients in the middle stages of rehabilitation, an inertial sensor with a high sampling rate can be selected. The sensor module is preferably replaced by a plug-in structure, fixed by a snap-fit ​​or magnetic attraction.

[0049] Furthermore, according to the present invention, a statistical analysis module for evaluating rehabilitation effectiveness can be provided. This module generates a rehabilitation effectiveness evaluation report based on the patient's historical training data, including trends in gesture completion improvement, vocal activity changes, and a comprehensive rehabilitation index. The formula for calculating the comprehensive rehabilitation index is:

[0050] in, For the comprehensive rehabilitation index, This represents the change in the completion rate of the motion trajectory. This represents the change in pressure completion rate. This represents the change in vocal activity. , and As the weighting coefficient, preferred , , .

Claims

1. A rehabilitation training method for patients with moderate to severe expressive stroke aphasia, characterized in that, The method includes the following steps: selecting a training scenario via a patient-side device; showing the patient a scenario demonstration video corresponding to the training scenario; the patient wearing a glove with built-in sensors and performing a target gesture action according to the demonstration video; collecting the patient's motion trajectory data through a six-axis inertial sensor on the glove, and collecting pressure data of the contact surface between the patient's hand and the glove through a thin-film pressure sensor; evaluating the gesture completion rate in real time based on the motion trajectory data and pressure data; if the gesture completion rate reaches a preset threshold, triggering vibration feedback and playing the speech of the target word; if the gesture completion rate does not reach the preset threshold, providing multimodal prompts or reducing the training difficulty; capturing the patient's vocal intention through a speech activity detection module and recording the training results based on the detection results; and synchronizing the training data to a cloud server, wherein the glove is placed on the patient's unaffected hand.

2. The rehabilitation training method according to claim 1, characterized in that, The gesture completion degree is calculated by weighting the motion trajectory completion degree and the pressure completion degree, wherein the weight of the motion trajectory completion degree is 0.6 to 0.8 and the weight of the pressure completion degree is 0.2 to 0.

4.

3. The rehabilitation training method according to claim 1 or 2, characterized in that, When the contact force detected by the pressure sensor is greater than the first preset threshold and the duration is greater than the first preset duration, it is determined that the grasping gesture is completed; when the contact force detected by the pressure sensor is within the second preset threshold range and the inertial sensor detects the pinching action of the fingers, it is determined that the pinching gesture is completed.

4. The rehabilitation training method according to claim 1 or 2, characterized in that, When a patient fails to complete the same target gesture three times in a row, the system automatically lowers the completion threshold of that gesture or switches the current training vocabulary to the next vocabulary; when a patient successfully completes the target gesture three times in a row, the system automatically raises the completion threshold of that gesture.

5. The rehabilitation training method according to claim 1 or 2, characterized in that, The training difficulty coefficient is dynamically adjusted based on the patient's training performance, and the patient's training behavior data is recorded, including single training session dwell time, gesture timeout rate, vocabulary completion rate, and vocalization detection pass rate. The patient's ability profile is then generated based on the behavioral data.

6. The rehabilitation training method according to claim 1 or 2, characterized in that, Based on the patient's vocalization test results, the training results are categorized as "both gesture and vocalization were successful", "gesture only was successful", or "gesture failed", and the prompting strategy for the next round of training is adjusted according to the categorization results.

7. The rehabilitation training method according to any one of claims 1 to 6, characterized in that, After the patient completes the day's training tasks, the system automatically plays a pre-recorded encouraging message from the family member.

8. A rehabilitation training system for implementing the rehabilitation training method according to any one of claims 1 to 7, characterized in that, The system includes: a patient-side device equipped with a touch screen, audio playback unit, microphone acquisition unit, and main control unit; a wearable glove equipped with a six-axis inertial sensor, a thin-film pressure sensor, and a vibration feedback unit; a cloud server for storing training data and synchronizing data across the three devices; a family-side device running a family application; and a doctor-side device running a doctor assessment program.

9. The rehabilitation training system according to claim 8, characterized in that, The wearable glove is made of breathable fabric and has a built-in flexible circuit board. The flexible circuit board integrates a six-axis inertial sensor, a thin-film pressure sensor, and a micro vibration motor. The sampling frequency of the six-axis inertial sensor is not less than 100Hz, and the sampling frequency of the thin-film pressure sensor is not less than 50Hz.

10. The rehabilitation training system according to claim 8 or 9, characterized in that, The system is equipped with an adaptive algorithm module to dynamically adjust the training difficulty based on the patient's training performance; the cloud server uses the AES-256 algorithm for data encryption and transmits data via the HTTPS protocol; the patient-side device and the wearable glove establish a connection via the Bluetooth wireless communication protocol.

Images