Multi-modal signal fused fine exercise rehabilitation evaluation and regulation method and system

Abstract

The invention belongs to the cross technical field of rehabilitation engineering and neural engineering, and relates to a multi-modal signal fused fine exercise rehabilitation evaluation, regulation and control method and system, and each evaluation comprises the following steps: collecting a pressure distribution signal, a surface electromyogram signal and an electroencephalogram signal of a hand fine exercise; performing preprocessing, feature extraction and normalization processing on the three signals to obtain a pressure feature value, a surface myoelectricity feature value and an electroencephalogram feature value; calculating a conversion coefficient of a current rehabilitation stage based on the number of days that the fine motor function of the patient is damaged, and calculating weights of a pressure characteristic value, a surface myoelectricity characteristic value and an electroencephalogram characteristic value based on the conversion coefficient; performing weighted summation on the pressure characteristic value, the surface myoelectricity characteristic value and the electroencephalogram characteristic value to obtain an evaluation score; rehabilitation training parameters are regulated based on the assessment score and the number of days of impaired patient's fine motor function. According to the method, the limitation of traditional single-mode physiological signal evaluation can be broken through, and the comprehensiveness, the accuracy and the personalized adaptation capability of rehabilitation evaluation can be improved.

Description

Technical Field

[0001] The present invention relates to the interdisciplinary technical field of rehabilitation engineering and neural engineering, and in particular to a fine motor rehabilitation assessment and control method and system using multimodal signal fusion. Background Art

[0002] Stroke is a neurological disorder caused by the sudden rupture or blockage of a cerebral blood vessel, resulting in insufficient blood supply to the brain. As the second-leading cause of death worldwide, my country experiences 3.3 million new stroke cases annually, making it the leading cause of death and disability among adults in the country. Limb spasticity is a common complication after stroke, with an incidence rate as high as 65%. Persistent spasticity can induce pain in the affected limb and gradually develop into muscle atrophy, severely impacting the patient's quality of life. Current stroke rehabilitation follows a closed-loop process of "functional assessment, planning, implementation, and reassessment of effectiveness." Accurate motor function assessment and closed-loop neuromodulation are crucial for ensuring effective rehabilitation. However, current clinical assessments for fine motor skills in the hand primarily rely on rating scales, lacking objective, quantitative metrics. Furthermore, traditional rehabilitation training often employs fixed-parameter electrical stimulation or physical training, making it difficult to dynamically adjust based on the patient's recovery progress, limiting its effectiveness.

[0003] Current methods for assessing hand function in stroke patients primarily include clinical scale assessments based on the Fugl-Meyer Upper Extremity Assessment Scale (FMA-UE), the Nine-Hole Pegboard Test (9HPT), and the Modified Ashworth Spasticity Rating Scale, machine vision-based limb and motor function assessments, and intelligent assessments based on physiological signals. However, these assessment methods still have several technical limitations that need to be addressed when applied to fine motor rehabilitation assessments of the hand:

[0004] (1) Scale assessment: First, scale assessment is time-consuming, the results are inconsistent, and it is highly dependent on the clinical experience of rehabilitation therapists. Second, scale assessment is not sensitive enough to changes in fine motor function of the hands during the short-term rehabilitation phase, making it difficult to accurately reflect the patient's progressive rehabilitation process. Furthermore, there are subjective factors in scale assessment, which affect the accuracy and reliability of the assessment and cannot meet the clinical needs of quantitative assessment of fine motor function.

[0005] (2) Machine vision-based limb and motor function assessment: Patients with hand dysfunction often have hypertonia, which results in small or even difficult fine movements of the hands. Therefore, there are significant limitations in using cameras to capture fine movements. At the same time, occlusion is very likely to occur during the execution of complex fine movements of the hands, affecting the complete capture of the movement trajectory and reducing the accuracy of the assessment. In addition, the overall structure of the assessment system is complex, the equipment cost is high, and its application and promotion are limited.

[0006] (3) Intelligent assessment based on physiological signals: Due to the limited residual muscle activity of the patient's affected limb, the physiological signal acquisition process is easily affected by external interference, resulting in a decrease in signal quality and affecting the stability of the assessment. Currently, most assessment systems rely on only a single sensor or data acquisition system, which is difficult to fully reflect the patient's hand function status. In addition, due to significant physiological differences between individuals, existing machine learning algorithms are difficult to effectively adapt to the characteristics of different patients and lack generalization ability, which seriously affects the accuracy and universality of the assessment. Summary of the Invention

[0007] The purpose of the present invention is to address the problems of existing assessment methods, such as long scale assessment time, poor consistency of assessment results, complex overall structure of the assessment system, high equipment cost, poor assessment accuracy and universality, and propose a fine motor rehabilitation assessment, regulation method and system based on multimodal signal fusion, breaking through the limitations of single-modal physiological signal assessment, constructing a neuromuscular function assessment system based on multimodal signal fusion, and proposing a feature fusion algorithm based on dynamic weight allocation at different rehabilitation stages. By dynamically adjusting the weight coefficient to adapt to the needs of different patients and different rehabilitation stages, the comprehensiveness, accuracy and personalized adaptation ability of rehabilitation assessment are improved.

[0008] In order to achieve the above object, the present invention adopts the following technical solutions:

[0009] In a first aspect, the present invention provides a multimodal signal fusion fine motor rehabilitation assessment and control method, each assessment comprising the following steps:

[0010] S1. Collecting multimodal signals, the multimodal signals including: pressure distribution signals of fine hand movements, surface electromyography signals, and electroencephalogram signals;

[0011] S2. Preprocess, extract features, and normalize the pressure distribution signals, surface electromyography signals, and electroencephalogram (EEG) signals of fine hand movements to obtain pressure eigenvalues, surface electromyography eigenvalues, and EEG eigenvalues;

[0012] S3. Calculate a conversion coefficient for the current rehabilitation stage based on the number of days the patient's fine motor function has been impaired, and calculate weights for the pressure eigenvalue, surface electromyography eigenvalue, and electroencephalogram eigenvalue based on the conversion coefficient;

[0013] S4. performing weighted summation of the pressure characteristic value, the surface electromyography characteristic value, and the electroencephalogram characteristic value to obtain an evaluation score;

[0014] S5. Adjust rehabilitation training parameters based on assessment scores and the number of days the patient's fine motor function is impaired.

[0015] As a possible implementation method, the pressure characteristic value includes the maximum pressure value, the minimum pressure value and the trend quantization value; the surface electromyography characteristic value is the root mean square value of the surface electromyography signal; the EEG characteristic value includes the root mean square value of the EEG signal and event-related desynchronization;

[0016] The evaluation score is calculated using the following method:

[0017] Score

[0018] =100×[α Press (t)(k1y max +k2y max +k3k)+α sEMG (t)RMS sEMG +α EEG (t)(k4RMS EEG +k5ERD)]

[0019] Among them, Score represents the assessment score, t represents the number of days the patient's fine motor function is impaired, α Press (t) represents the weight of the pressure characteristic value when the number of damaged days is t, α sEMG (t) represents the weight of the surface electromyography characteristic value when the number of days of damage is t, α EEG (t) represents the weight of the EEG characteristic value when the number of days of damage is t, y max Indicates the maximum pressure, y min represents the minimum pressure value, k represents the trend quantization value, k1, k2, k3 represent the weights of the maximum pressure value, the minimum pressure value, and the trend quantization value, respectively, and k1+k2+k3=1; RMS sEMG Represents the root mean square value of the surface electromyography signal, RMS EEG represents the RMS value of the EEG signal, ERD represents event-related desynchronization, k4 and k5 represent the RMS value of the EEG signal and the weight of event-related desynchronization, respectively, and k4+k5=1.

[0020] As a possible implementation, S3 includes the following sub-steps:

[0021] S30. Determine the number of days t in which the patient's fine motor function is impaired;

[0022] S31. Calculate the conversion coefficient of the current recovery stage based on the number of days of impairment t, where the conversion coefficient includes a β coefficient and a γ coefficient;

[0023] S32. Determine the initial weights of the pressure characteristic value, surface electromyography characteristic value and electroencephalogram characteristic value based on the number of days of injury t, and calculate the weights of the pressure characteristic value, surface electromyography characteristic value and electroencephalogram characteristic value based on the β coefficient and the γ coefficient.

[0024] As a possible implementation method, the β coefficient is calculated using the following method:

[0025]

[0026] The γ coefficient is calculated by the following method:

[0027]

[0028] Where, t represents the number of days when the patient's fine motor function is impaired, 0 < t ≤ 14 indicates that the current rehabilitation stage is the acute phase, 14 < t ≤ 180 indicates that the current rehabilitation stage is the recovery phase, and t > 180 indicates that the current rehabilitation stage is the chronic phase.

[0029] As a possible implementation method, the following method is used to determine the initial weights:

[0030] If it is the first evaluation, then according to the number of impaired days, the initial weight ratios of the pressure characteristic value, surface electromyogram characteristic value, and electroencephalogram characteristic value are set as follows: Acute phase: The initial weight ratio is 1:1:4 to 1:1:6;

[0031] Recovery phase: The initial weight ratios of the pressure characteristic value, surface electromyogram characteristic value, and electroencephalogram characteristic value are set as 1:2:2 to 1:2:3;

[0032] Chronic phase: The initial weight ratios of the pressure characteristic value, surface electromyogram characteristic value, and electroencephalogram characteristic value are set as 3:1:1 to 4:1:1;

[0033] If it is not the first evaluation, then the weighted weights of the pressure characteristic value, surface electromyogram characteristic value, and electroencephalogram characteristic value calculated during the previous evaluation are used as the initial weights for this evaluation.

[0034] As a possible implementation method, the following method is used to calculate the weight of the electroencephalogram characteristic value:

[0035] α EEG (t) = β(t) × a1 + (1 - β(t)) × (1 - γ(t)) × a2 + γ(t) × a3

[0036] The following method is used to calculate the weight of the surface electromyogram characteristic value:

[0037] α sEMG (t) = β(t) × b1 + (1 - β(t)) × (1 - γ(t)) × b2 + γ(t) × b3

[0038] The following method is used to calculate the weight of the pressure characteristic value:

[0039] α press (t) = β(t) × c1 + (1 - β(t)) × (1 - γ(t)) × c2 + γ(t) × c3

[0040] And it satisfies:

[0041] α EEG (t)+α sEMG (t)+α press (t)=1

[0042] Among them, α EEG (t) represents the weight of the EEG eigenvalue, α sEMG (t) represents the weight of the surface electromyography eigenvalue, α Press (t) represents the weight of the stress characteristic value, a1, a2, and a3 represent the initial weights of the EEG characteristic values ​​in the acute, recovery, and chronic stages, respectively; b1, b2, and b3 represent the initial weights of the surface electromyography characteristic values ​​in the acute, recovery, and chronic stages, respectively; c1, c2, and c3 represent the initial weights of the stress characteristic values ​​in the acute, recovery, and chronic stages, respectively.

[0043] As a possible implementation, S5 is specifically as follows:

[0044] S50. The assessment scores are divided into low-level scores, medium-level scores, and high-level scores; wherein the low-level scores are assessment scores of 0 to 40 points, the medium-level scores are assessment scores of 41 to 70 points, and the high-level scores are assessment scores of 71 to 100 points;

[0045] S51. Adjust rehabilitation training parameters based on the assessment scores and the number of days the patient's fine motor function is impaired. Specifically:

[0046] If the patient's current rehabilitation stage is acute and the assessment score is low, set the current amplitude to 12-14 mA, the frequency to 15-20 Hz, the pulse width to 180-220 μs, and the number of training times to 10-12 times per set.

[0047] If the patient's current rehabilitation stage is acute and the assessment score is medium, set the current amplitude to 14-16 mA, the frequency to 20 Hz, the pulse width to 200 μs, and the number of training times to 12-14 times per group.

[0048] If the patient's current rehabilitation stage is acute and the assessment score is high, set the current amplitude to 15-17 mA, the frequency to 20 Hz, the pulse width to 200 μs, and the number of training times to 12-15 times per group.

[0049] If the patient's current rehabilitation stage is the recovery stage and the assessment score is low, set the current amplitude to 15-17 mA, the frequency to 25 Hz, the pulse width to 220 μs, and the number of training times to 15-18 times per set.

[0050] If the patient's current rehabilitation stage is the recovery stage and the assessment score is a medium-level score, set the current amplitude to 18-20 mA, the frequency to 30 Hz, the pulse width to 230 μs, and the number of training times to 18-20 times per set;

[0051] If the patient's current rehabilitation stage is the recovery stage and the assessment score is a high-level score, set the current amplitude to 20-22 mA, the frequency to 30-35 Hz, the pulse width to 240-250 μs, and the number of training times to 20-22 times per set;

[0052] If the patient's current rehabilitation stage is the chronic stage and the assessment score is low, set the current amplitude to 18-20 mA, the frequency to 30 Hz, the pulse width to 240 μs, and the number of training times to 18-20 times per group.

[0053] If the patient's current rehabilitation stage is the chronic stage and the assessment score is a medium-level score, the current amplitude is set to 22-24 mA, the frequency to 35 Hz, the pulse width to 250 μs, and the number of training times to 22 times per group;

[0054] When the patient's current rehabilitation stage is the chronic stage and the assessment score is a high-level score, the current amplitude is set to 24-26 mA, the frequency is 40 Hz, the pulse width is 260-280 μs, and the number of training times is 25 times / group.

[0055] In a second aspect, the present invention provides a multimodal signal fusion fine motor rehabilitation assessment and control system, comprising:

[0056] A multimodal signal acquisition module is used to collect multimodal signals, including pressure distribution signals of fine hand movements, surface electromyography signals, and electroencephalogram signals;

[0057] The data processing module is used to pre-process, extract features and normalize the collected pressure distribution signals, surface electromyography signals and electroencephalogram signals to obtain pressure features, surface electromyography features and electroencephalogram features respectively;

[0058] Rehabilitation assessment module, used to automatically evaluate rehabilitation progress based on feature-level fusion algorithm;

[0059] The feedback and adaptive control module is used to dynamically control the weights of pressure distribution signals, surface electromyography signals, and electroencephalogram signals based on rehabilitation progress.

[0060] As a possible implementation, the multimodal signal acquisition module includes a pressure distribution signal acquisition unit, a surface electromyography signal acquisition unit, and an electroencephalogram signal acquisition unit;

[0061] The pressure distribution signal acquisition unit includes an array-type flexible pressure sensor subunit, a pressure distribution signal processing subunit and a host computer. The array-type flexible pressure sensor subunit includes a hand-grip sensor array and a sleeve sensor array. The hand-grip sensor array includes high-density independent sensitive points, the center spacing between adjacent sensitive points is 0.5 to 2 mm, and the scanning frequency is 50 to 500 Hz; the sleeve sensor array includes high-density independent sensitive points, the center spacing between adjacent sensitive points is 0.5 to 1 mm, and the scanning frequency is 50 to 500 Hz; the array-type flexible pressure sensor subunit and the pressure distribution signal processing subunit are connected through a flexible circuit board. The pressure distribution signal processing subunit adopts a sliding differential algorithm to eliminate the static error introduced by the deformation of the sensor base, and transmits the pressure data to the host computer after filtering. The host computer is used to calculate the pressure value of each sensitive point, map the pressure value of each sensitive point to the standard anatomical coordinate system, and generate a pressure thermodynamic map in real time to display the pressure distribution of each area of ​​the hand;

[0062] The surface electromyography signal acquisition unit uses an 8-channel electromyography acquisition bracelet to collect surface electromyography signals;

[0063] The EEG signal acquisition unit uses saline electrodes to collect EEG signals.

[0064] As a possible implementation method, the handheld sensor array includes 4000 independent sensitive points, and the arrangement of the 4000 independent sensitive points is 80 rows * 50 columns; the sleeve sensor array includes 1000 independent sensitive points, and the arrangement of the 1000 independent sensitive points is 40 rows * 25 columns.

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] 1. The multimodal signal fusion fine motor rehabilitation assessment and regulation method proposed in the present invention breaks through the limitations of traditional single-modal physiological signal assessment, constructs a neuromuscular function assessment system based on multimodal signal fusion, and proposes a feature fusion algorithm based on dynamic weight allocation at different rehabilitation stages. By dynamically adjusting the weight coefficient to adapt to the needs of different patients and different rehabilitation stages, the comprehensiveness, accuracy and personalized adaptation capability of rehabilitation assessment are improved.

[0067] 2. The multimodal signal fusion fine motor rehabilitation assessment, control method and system proposed in the present invention adopts a high-density distributed flexible pressure sensing matrix, which can accurately collect and quantify the pressure distribution and dynamic changes of patients in fine hand movements such as grasping and pointing in real time, overcoming the limitations of traditional single-point measurement and improving assessment accuracy.

[0068] 3. The multimodal signal fusion-based fine motor rehabilitation assessment and control method proposed in this paper constructs a 3×3 joint control matrix of assessment scores and rehabilitation stages, ensuring that electrical stimulation parameters are dynamically matched to the patient's motor state. This method overcomes the limitations of traditional fixed-parameter electrical stimulation protocols, significantly improving the accuracy, personalization, and effectiveness of rehabilitation training, enhancing the autonomous activation of the patient's neuromuscular system, promoting neuroplasticity, and supporting long-term recovery.

[0069] 4. The multimodal signal fusion fine motor rehabilitation assessment and regulation method proposed in the present invention determines whether it is the first assessment based on the patient's actual situation. During the first assessment, the initial weight is dynamically assigned based on the patient's rehabilitation stage, and different modal signals dominate in different rehabilitation stages. When it is not the first assessment, the weight of the previous assessment is inherited. The inheritance setting can ensure the continuity of the evaluation indicators between multiple training sessions and fully retain the dynamic evolution trend, which can ensure the accuracy and continuity of the assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] The accompanying drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0071] Figure 1 A flow chart of a multimodal signal fusion fine motor rehabilitation assessment and control method provided in an embodiment of the present invention;

[0072] Figure 2 A schematic diagram of the structure of a fine motor rehabilitation assessment and control system using multimodal signal fusion provided by an embodiment of the present invention;

[0073] Figure 3 Schematic diagram of the structure of the pressure distribution signal acquisition unit in an embodiment of the present invention.

[0074] Reference numerals

[0075] 1-Multimodal signal acquisition module, 10-Pressure distribution signal acquisition unit, 100-Array flexible pressure sensor subunit, 1000-Hand grip sensor array, 1001-Cuff sensor array, 101-Pressure distribution signal processing subunit, 102-Upper computer, 11-Surface electromyography signal acquisition unit, 12-EEG signal acquisition unit, 2-Data processing module, 3-Rehabilitation assessment module, 4-Feedback and adaptive control module. DETAILED DESCRIPTION

[0076] To facilitate a clear description of the technical solutions of the embodiments of the present invention, the words "first" and "second" are used in the embodiments of the present invention to distinguish between identical or similar items with substantially the same functions and effects. For example, the first threshold and the second threshold are merely used to distinguish between different thresholds and do not limit their order. Those skilled in the art will understand that the words "first" and "second" do not limit the quantity or execution order, and the words "first" and "second" do not necessarily mean different.

[0077] It should be noted that, in the present invention, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the present invention should not be construed as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0078] In the present invention, "at least one" refers to one or more, and "more" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. The following at least one item (item) or similar expressions thereof refer to any combination of these items, including any combination of single items (items) or plural items (items). For example, at least one item (item) of a, b or c can represent: a, b, c, the combination of a and b, the combination of a and c, the combination of b and c, or the combination of a, b and c, where a, b, c can be single or multiple.

[0079] The embodiments of the present invention aim to provide a fine motor rehabilitation assessment and regulation method based on multimodal signal fusion. By means of pressure distribution signals, surface electromyography signals and electroencephalogram signals of fine hand movements, key physiological parameters are acquired in real time. Combined with the duration of the patient's fine motor function impairment, the assessment score is calculated, and the rehabilitation training parameters are dynamically adjusted to construct a full-process personalized rehabilitation program covering "signal acquisition-multimodal signal fusion-quantized output-closed-loop control", thereby improving the accuracy, effectiveness, personalization and adaptability of rehabilitation training.

[0080] In the first aspect, the present invention provides a multimodal signal fusion fine motor rehabilitation assessment and control method, see Figure 1 Each assessment includes the following steps:

[0081] S1. Collect multimodal signals, including pressure distribution signals of fine hand movements, surface electromyography signals, and electroencephalogram (EEG) signals.

[0082] As an example, a rehabilitation torsion bar was used as a carrier, and an array of flexible pressure sensors with 80 rows and 50 columns was integrated onto the torsion bar. The sensor thickness can be adjusted between 0.1mm and 0.3mm, forming 4000 independent pressure-sensing points. The center spacing between adjacent sensitive points is 2mm, and the scanning frequency is 50Hz. The pressure sensors convert the pressure signals into electrical signals, achieving highly sensitive pressure detection. This design can adapt to different patient hand sizes and accurately measure the pressure distribution of each finger during the grasping process of stroke patients at different rehabilitation stages. It also dynamically records the pressure changes over time, ensuring the accuracy and universality of rehabilitation training. Based on the array of flexible pressure sensors and medical elastic fabric, a 40-row by 25-column array of pressure sensors (1000 sensitive points with 1mm spacing between each sensitive point) was designed to cover the thumb tip. This formed a thumb cuff-type pressure sensing unit, which can collect real-time dynamic pressure changes during grasping and finger pointing, and evaluate the dynamic mechanical characteristics of fine hand movements through time series analysis. A multi-channel portable myoelectric wristband collects surface electromyographic (sEMG) signals from key muscle groups associated with rehabilitation movements to assess muscle activity and motor execution. A saline electrode cap collects EEG signals from movement-related brain regions to decode movement intention and assess the brain's ability to control hand movements.

[0083] The present invention accurately collects the pressure distribution signals of fine hand movements, and simultaneously integrates surface electromyography signals and electroencephalogram signals, which can extract the coordinated control characteristics of nerves and muscles, realize the synchronous collection, feature extraction and dynamic evaluation of multimodal motion information, and help improve the effect of fine hand movement rehabilitation training.

[0084] S2. Preprocess, extract features, and normalize the pressure distribution signals, surface electromyography signals, and electroencephalogram (EEG) signals of fine hand movements to obtain pressure eigenvalues, surface electromyography eigenvalues, and EEG eigenvalues;

[0085] For example, to ensure the accuracy of rehabilitation assessment and regulation, it is necessary to perform baseline calibration, filtering and other operations on the collected pressure distribution signals, surface electromyography signals and electroencephalogram signals to improve signal quality.

[0086] As an example, the pressure characteristic value includes the maximum pressure value, the minimum pressure value and the trend quantization value; the surface electromyography characteristic value is the root mean square value of the surface electromyography signal; and the EEG characteristic value includes the root mean square value of the EEG signal and event-related desynchronization.

[0087] In the specific implementation, the pressure distribution signal is collected cyclically at a frequency of 50 Hz, and a 100 ms sliding window is used to perform median filtering with a window length of 5 points, as shown in the following formula (1), to remove the noise in the signal and maintain the original characteristics of the signal as much as possible during filtering.

[0088] y(n)=median{x(n-2),x(n-1),x(n),x(n+1),x(n+2)} (1)

[0089] Where n represents the number of data points in the window, x(n) represents the original pressure signal value collected at the nth time point, and y(n) represents the denoised pressure signal value after median filtering at the nth time point.

[0090] Extract the maximum pressure y max , minimum value y min And the trend quantization value k is used as the pressure characteristic value, as shown in the following formulas (2) and (3):

[0091] y max =max(y(n)),y min = min (y(n)) (2)

[0092]

[0093] Among them, t i is the time point of the sliding window.

[0094] A 4th-order Butterworth filter was used to perform band-pass filtering from 20 to 500 Hz, and the root mean square value (RMS) of the surface electromyography signal was calculated as the surface electromyography characteristic value according to the following formula (4):

[0095]

[0096] Among them, x i is the surface electromyography signal sampling point, and N is the window size.

[0097] When collecting EEG signals, a bandpass filter of 0.1 Hz to 35 Hz and a notch filter of 50 Hz are performed to extract the signals of the α band (8 to 13 Hz) and β band (13 to 30 Hz) related to movement. The root mean square value (RMS) and event-related desynchronization (ERD) of the EEG signals are extracted as EEG feature values, as shown in the following equations (5) and (6):

[0098]

[0099] Among them, x i is the EEG signal sampling point, and N is the window size.

[0100]

[0101] Among them, P bascline is the average power (resting state) 1s before the action, P task is the average power during the 2-s action period (task state).

[0102] To ensure that signals of different modalities have the same scale and facilitate subsequent multimodal signal fusion, a maximum threshold is set based on the pressure sensor range and the reference value of healthy people, and the original value of the pressure distribution signal is mapped to the interval [0,1]. For surface electromyographic signals, a threshold is set according to the maximum expected electromyographic activity intensity, and normalization is performed to avoid individual differences or noise interference. For electroencephalographic signals, normalization is performed based on resting-state baseline data, and relative changes are used to eliminate individual baseline differences.

[0103] S3. Calculate a conversion coefficient for the current rehabilitation stage based on the number of days the patient's fine motor function has been impaired, and calculate weights for the pressure eigenvalue, surface electromyography eigenvalue, and electroencephalogram eigenvalue based on the conversion coefficient;

[0104] As a possible implementation, S3 includes the following sub-steps:

[0105] S30. Determine the number of days t in which the patient's fine motor function is impaired;

[0106] S31. Calculate the conversion coefficient of the current recovery stage based on the number of days of impairment t, where the conversion coefficient includes a β coefficient and a γ coefficient;

[0107] In specific implementation, for each time window t, a multi-dimensional feature vector is constructed, as shown in formula (7):

[0108] F(t)=[y max ,y max ,k,RMS sEMG ,RMS EEG ,ERD] (7)

[0109] Based on this feature vector, a feature-level fusion algorithm based on dynamic weight allocation is constructed. The algorithm is based on the theory of the time window for neurological rehabilitation, combines the complementarity between multi-modal signals (pressure distribution signal, surface electromyogram signal, electroencephalogram signal), designs a dynamic weight allocation mechanism, and adjusts the weights of each signal at different recovery stages to optimize the accuracy of data fusion during the patient's rehabilitation process. The theory of the time window for neurological rehabilitation divides the rehabilitation stage into five stages: acute stage (0 - 14 days), sub-acute stage (15 - 60 days), recovery stage (60 days - 180 days), chronic stage (180 days - 365 days), and long-term chronic stage (more than 1 year). The present invention constructs a β coefficient based on the above theory of the time window for neurological rehabilitation to characterize the degree of conversion of the patient from the acute stage to the recovery stage, and constructs a γ coefficient to characterize the degree of conversion of the patient from the chronic stage to the recovery stage.

[0110] As a possible implementation method, the following method is used to calculate the β coefficient:

[0111]

[0112] The following method is used to calculate the γ coefficient:

[0113]

[0114] Where, t represents the number of days when the patient's fine motor function is impaired. 0 < t ≤ 14 indicates that the current rehabilitation stage is the acute stage, 14 < t ≤ 180 indicates that the current rehabilitation stage is the recovery stage, and t > 180 indicates that the current rehabilitation stage is the chronic stage.

[0115] S32. Determine the initial weights of the pressure feature value, surface electromyogram feature value, and electroencephalogram feature value based on the number of impaired days t, and calculate the weights of the pressure feature value, surface electromyogram feature value, and electroencephalogram feature value based on the β coefficient and γ coefficient.

[0116] As a possible implementation method, the following method is used to determine the initial weights:

[0117] If it is the first evaluation, then according to the number of impaired days, the initial weight ratio of the pressure feature value, surface electromyogram feature value, and electroencephalogram feature value is set as: acute stage: the initial weight ratio is 1:1:4 to 1:1:6; among them, the weight of the electroencephalogram signal is the largest, highlighting its important role in decoding the movement intention in the early rehabilitation.

[0118] Recovery stage: Set the initial weight ratio of the pressure feature value, surface electromyogram feature value, and electroencephalogram feature value to 1:2:2 to 1:2:3; in this stage, the electroencephalogram signal still dominates, and at the same time, the weight of the electromyogram feature is significantly increased, reflecting the trend of enhanced coordination between the central movement intention and muscle execution.

[0119] Chronic stage: The initial weight ratio of pressure characteristic value, surface electromyography characteristic value and EEG characteristic value is set to 3:1:1 to 4:1:1; in this stage, the weight of pressure signal gradually becomes dominant to highlight the core position of terminal execution ability and fine motor recovery.

[0120] As an example, the initial weight distribution of pressure eigenvalues, surface electromyography eigenvalues, and EEG eigenvalues ​​is shown in Table 1:

[0121] Table 1 Initial weight distribution of pressure eigenvalues, surface electromyography eigenvalues ​​and EEG eigenvalues

[0122]

[0123] If this is not the first evaluation, the weighted weights of the pressure characteristic values, surface electromyography characteristic values, and EEG characteristic values ​​calculated during the previous evaluation are used as the initial weights for this evaluation. The inheritance setting can ensure the continuity of evaluation indicators between multiple training sessions and fully retain the dynamic evolution trend.

[0124] As a possible implementation method, the following method is used to calculate the weight of the EEG eigenvalue:

[0125] α EEG (t)=β(t)×a1+(1―β(t))×(1―γ(t))×a2+γ(t)×a3 (10)

[0126] The weight of the surface electromyography eigenvalue is calculated using the following method:

[0127] α sEMG (t)=β(t)×b1+(1―β(t))×(1―γ(t))×b2+γ(t)×b3 (11)

[0128] The weight of the pressure eigenvalue is calculated using the following method:

[0129] α press (t)=β(t)×c1+(1―β(t))×(1―γ(t))×c2+γ(t)×c3 (12)

[0130] And satisfy:

[0131] α EEG (t)+α sEMG (t)+α press (t)=1 (13)

[0132] Among them, α EEG (t) represents the weight of the EEG eigenvalue, α sEMG (t) represents the weight of the surface electromyography eigenvalue, α Press(t) represents the weight of the stress characteristic value, a1, a2, and a3 represent the initial weights of the EEG characteristic values ​​in the acute, recovery, and chronic stages, respectively; b1, b2, and b3 represent the initial weights of the surface electromyography characteristic values ​​in the acute, recovery, and chronic stages, respectively; c1, c2, and c3 represent the initial weights of the stress characteristic values ​​in the acute, recovery, and chronic stages, respectively.

[0133] S4. performing weighted summation of the pressure characteristic value, the surface electromyography characteristic value, and the electroencephalogram characteristic value to obtain an evaluation score;

[0134] As a possible implementation method, the following method is used to calculate the evaluation score:

[0135] Score = 100 × [α Press (t)(k1y max +k2y min +k3k)+α sEMG (t)RMS sEMG +α EEG (t)(k4RMS EEG +k5ERD)] (14)

[0137] Among them, Score represents the assessment score, t represents the number of days the patient's fine motor function is impaired, α Press (t)

[0138] represents the weight of the stress characteristic value when the number of damaged days is t, α sEMG (t) represents the weight of the surface electromyography characteristic value when the number of days of damage is t, α EEG (t) represents the weight of the EEG characteristic value when the number of days of damage is t, y max Indicates the maximum pressure, y min represents the minimum pressure value, k represents the trend quantization value, k1, k2, k3 represent the weights of the maximum pressure value, the minimum pressure value, and the trend quantization value, respectively, and k1+k2+k3=1; RMS sEMG Represents the root mean square value of the surface electromyography signal, RMS EEG represents the root mean square value of the EEG signal, ERD represents event-related desynchronization, k4 and k5 represent the root mean square value of the EEG signal and the weight of event-related desynchronization, and k4

[0139] +k5=1.

[0140] S5. Adjust rehabilitation training parameters based on assessment scores and the number of days the patient's fine motor function is impaired.

[0141] As a possible implementation, S5 is specifically as follows:

[0142] S50. The assessment scores are divided into low-level scores, medium-level scores, and high-level scores; wherein the low-level scores are assessment scores of 0 to 40 points, the medium-level scores are assessment scores of 41 to 70 points, and the high-level scores are assessment scores of 71 to 100 points;

[0143] The present invention divides the assessment scores into three levels: low-level scores (0-40 points), medium-level scores (41-70 points) and high-level scores (71-100 points); at the same time, the rehabilitation stages corresponding to the number of days of functional impairment are divided into: acute stage (≤14 days), recovery stage (15-180 days) and chronic stage (>180 days), forming a 3×3 joint regulation matrix.

[0144] S51. Adjust rehabilitation training parameters based on the assessment scores and the number of days the patient's fine motor function is impaired. Specifically:

[0145] If the patient's current rehabilitation stage is acute and the assessment score is low, the current amplitude is set to 12-14 mA, the frequency to 15-20 Hz, the pulse width to 180-220 μs, and the number of training times to 10-12 times per set. The goal is to activate movement intention and central nervous system response.

[0146] If the patient's current rehabilitation stage is acute and the assessment score is moderate, the current amplitude is set to 14-16 mA, the frequency to 20 Hz, the pulse width to 200 μs, and the number of training times to 12-14 times per set. The goal is to establish a preliminary neuromuscular pathway.

[0147] If the patient's current rehabilitation stage is acute and the assessment score is high, the current amplitude is set to 15-17 mA, the frequency to 20 Hz, the pulse width to 200 μs, and the number of training times to 12-15 times per set, with the goal of stabilizing central control performance.

[0148] If the patient's current rehabilitation stage is recovery and the assessment score is low, the current amplitude is set to 15-17 mA, the frequency is 25 Hz, the pulse width is 220 μs, and the number of training times is 15-18 times per set. The goal is to induce muscle contraction and establish coordinated control.

[0149] If the patient's current rehabilitation stage is the recovery stage and the assessment score is medium, set the current amplitude to 18-20 mA, the frequency to 30 Hz, the pulse width to 230 μs, and the number of training times to 18-20 times per set to strengthen muscle strength and endurance training.

[0150] If the patient's current rehabilitation stage is recovery and the assessment score is high, the current amplitude is set to 20-22 mA, the frequency to 30-35 Hz, the pulse width to 240-250 μs, and the number of training times to 20-22 times per set. The goal is to improve fine manipulation ability and neuromuscular coordination.

[0151] If the patient's current rehabilitation stage is the chronic stage and the assessment score is low, set the current amplitude to 18-20 mA, the frequency to 30 Hz, the pulse width to 240 μs, and the number of training times to 18-20 times per set to strengthen the residual motor ability.

[0152] If the patient's current rehabilitation stage is the chronic stage and the assessment score is medium, set the current amplitude to 22-24 mA, frequency to 35 Hz, pulse width to 250 μs, and training times to 22 times / set to enhance sustained force and fine control.

[0153] When the patient's current rehabilitation stage is the chronic stage and the assessment score is a high-level score, the current amplitude is set to 24-26 mA, the frequency is 40 Hz, the pulse width is 260-280 μs, and the number of training times is 25 times / group to achieve comprehensive reconstruction of fine motor function.

[0154] Next, this method was used to evaluate the effect of a 58-year-old male patient named Wang, who had been in the chronic stage of rehabilitation one year after a stroke (t = 365 days). He also had problems with poor coordination and strength in the thumb-index finger opposition movements of the affected limb. His rehabilitation goal was to improve his fine motor skills, especially the rehabilitation of finger opposition function. He had completed one week of rehabilitation training based on this evaluation system, and was undergoing phased evaluation and adjustment of rehabilitation training parameters.

[0155] Patients wore a thumb-cuff-type flexible pressure sensor (40 rows and 25 columns, totaling 1,000 sensitive points) and completed thumb-index finger opposition rehabilitation training driven by functional electrical stimulation. The thumb-cuff-type flexible pressure sensor collected the pressure between the thumb and index finger during opposition. An 8-channel portable myoelectric wristband, worn on the extensor and flexor muscles of the forearm, collected myoelectric signals. Saline electrodes collected EEG signals from the motor area. Patients visualized the movement by viewing a virtual animation of the opposition, decoding the movement intention to trigger the functional electrical stimulator.

[0156] The stimulation electrodes were placed on the key muscle groups (flexor pollicis longus, flexor digitorum profundus, etc.) that complete the thumb and index finger opposition, and the stimulation parameters were set as default values ​​according to the previous rehabilitation training (e.g., current intensity 17 mA, pulse width 200 μs).

[0157] During the rehabilitation training process, patients completed 10 thumb and index finger exercises in each block under the guidance of functional electrical stimulation, and simultaneously collected pressure signals, surface electromyography signals and electroencephalogram signals and performed normalization processing.

[0158] The maximum pressure recorded when completing the finger-to-finger movement was 1.3N, the minimum was 0.35N, and the trend quantization k was 1.1N / s. After normalization based on empirical values, the values ​​were 0.65, 0.35, and 0.55, respectively. The root mean square value of the patient's electromyographic signal was 0.9mV, which was normalized to 0.3. The patient's EEG signal was synchronously collected, and the root mean square value of the EEG signal collected was calculated to be 6.5uV. The event-related desynchronization in the α band was 35%, which was normalized to 0.65 and 0.35, respectively. Combined with the patient's current recovery stage and formulas (8), (9), (10), (11), and (12), the weights of the EEG characteristic value, surface EMG characteristic value, and pressure characteristic value were calculated to be 0.433, 0.2835, and 0.2835, respectively. The final evaluation score was 46 points calculated according to formula (14).

[0159] Based on the control method proposed in this embodiment, this score is a medium-level score, and the patient is in the chronic stage. According to this score and the current rehabilitation stage, the current amplitude is set to 22-24mA, the frequency is 35Hz, the pulse width is 250μs, and the number of training times is 22 times / group to enhance sustained force and fine control. During the experiment, Wang's current amplitude was adjusted from 19mA to 23mA, and the pulse width was increased from 240us to 250us. The training intensity was increased, and 3 groups of thumb and index finger fine motor training were added to each block. Combined with brain-computer interface technology, visual stimulation was introduced to enhance the accuracy of motor intention decoding. After two weeks and ten sessions of new rehabilitation training, patient Wang's evaluation system score increased from 46 points to 52 points, indicating that the patient's thumb and index finger ability was significantly improved, and the personalized rehabilitation training program was effective.

[0160] In the second aspect, the embodiment of the present invention provides a fine motor rehabilitation assessment and control system based on multimodal signal fusion, see Figure 2 ,include:

[0161] Multimodal signal acquisition module 1, used to acquire multimodal signals, including pressure distribution signals of fine hand movements, surface electromyography signals, and electroencephalogram signals;

[0162] See also Figure 2 , as a possible implementation, the multimodal signal acquisition module 1 includes a pressure distribution signal acquisition unit 10, a surface electromyography signal acquisition unit 11 and an electroencephalogram signal acquisition unit 12;

[0163] See also Figure 3The pressure distribution signal acquisition unit 10 includes an array-type flexible pressure sensor subunit 100, a pressure distribution signal processing subunit 101 and a host computer 102. The array-type flexible pressure sensor subunit 100 includes a hand-held sensor array 1000 and a sleeve sensor array 1001. The hand-held sensor array 1000 includes high-density independent sensitive points, the center spacing between adjacent sensitive points is 0.5 to 2 mm, and the scanning frequency is 50 to 500 Hz; the sleeve sensor array includes high-density independent sensitive points, the center spacing between adjacent sensitive points is 0.5 to 1 mm, with a scanning frequency of 50 to 500 Hz; the array-type flexible pressure sensor subunit 100 is connected to the pressure distribution signal processing subunit 101 via a flexible circuit board. The pressure distribution signal processing subunit 101 uses a sliding differential algorithm to eliminate static errors introduced by sensor substrate deformation, filters the pressure data, and transmits it to the host computer 102; the host computer 102 is used to calculate the pressure value of each sensitive point, map the pressure value of each sensitive point to a standard anatomical coordinate system, and generate a pressure heat map in real time to display the pressure distribution of each area of ​​the hand;

[0164] As a possible implementation method, the handheld sensor array includes 4000 independent sensitive points, and the arrangement of the 4000 independent sensitive points is 80 rows * 50 columns; the sleeve sensor array includes 1000 independent sensitive points, and the arrangement of the 1000 independent sensitive points is 40 rows * 25 columns.

[0165] See also Figure 3 As an example, the grip sensor array 1000 uses a rehabilitation torsion bar as a carrier, with an array of 80 rows and 50 columns of flexible pressure sensors applied to its surface. The sensor thickness can be selected between 0.1mm and 0.3mm, forming 4,000 independent pressure-sensitive points. The center spacing between adjacent sensitive points is 2mm, and the scanning frequency is 50Hz. The pressure sensors convert the pressure signal into an electrical signal, achieving highly sensitive pressure detection. This design can adapt to the hand sizes of different patients and can accurately measure the pressure distribution of each finger during the grasping process of stroke patients at different stages of rehabilitation. It can also dynamically record the pressure changes over time, ensuring the accuracy and universality of rehabilitation training. The sleeve sensor array 1001 combines an array of flexible pressure sensors with medical elastic fabric to form an ergonomic sleeve that covers the distal phalanx area. The cuff sensor array 1001 integrates 40 rows and 25 columns of miniature flexible pressure sensing units, with a spacing of 1 mm between adjacent sensitive points and a total of 1,000 independent sensitive points. It can accurately monitor the pressure changes on the fingertips during fine hand movements such as finger pointing, extract mechanical characteristics, and quantify rehabilitation progress.

[0166] As an example, the array-type flexible pressure sensor subunit 100 is connected to the pressure distribution signal processing subunit 101 through a flexible circuit board to realize array scanning, and a sliding differential algorithm is used to eliminate the static error introduced by the deformation of the sensor base. After filtering, the pressure data is transmitted to the host computer 102 by wired or wireless means. The host computer 102 is responsible for real-time data processing, including calculating the pressure value of each sensitive point, extracting key mechanical features such as maximum value, minimum value and trend quantification. At the same time, based on the biomechanical model of the hand and fingers, the pressure data is mapped to the standard anatomical coordinate system to achieve precise positioning. In addition, the host computer can generate a pressure thermogram in real time, intuitively display the pressure distribution of each area, and allow users to view the specific pressure value of each sensitive point, providing an efficient and intuitive rehabilitation assessment tool.

[0167] The surface electromyography signal acquisition unit 11 uses an 8-channel electromyography acquisition bracelet to collect surface electromyography signals;

[0168] Even in the case of severe motor impairment, patients may still have low-amplitude spontaneous electromyographic activity and generate corresponding electromyographic responses during passive movement induced by functional electrical stimulation (FES). This embodiment uses a highly sensitive portable electromyographic bracelet to collect surface electromyographic signals. An 8-channel electromyographic acquisition bracelet is used, based on metal electrodes, and integrates an active signal acquisition unit with high input impedance, which can effectively reduce external noise interference and improve the signal-to-noise ratio and stability of surface electromyographic signals. The Bluetooth module is used to achieve wireless transmission of surface electromyographic signals to ensure comfort and convenience during evaluation. According to specific fine motor rehabilitation tasks and different rehabilitation stages, the myoelectric bracelet is worn on the key motor muscle groups of the corresponding movements to achieve surface electromyographic signal acquisition. By accurately capturing muscle activity patterns, the bracelet can support dynamic evaluation of fine motor rehabilitation of the hands of stroke patients and provide reliable data support for the formulation of personalized rehabilitation intervention strategies.

[0169] The EEG signal acquisition unit 12 uses saline electrodes to acquire EEG signals.

[0170] As an example, saline electrodes are used to collect EEG signals, and KCl or NaCl solutions are used as conductive media to effectively reduce the contact impedance between the electrodes and the skin. This overcomes the tedious process of applying conductive paste to traditional wet electrodes and requiring patients to clean them, and enables convenient collection of high-quality EEG signals. During the collection process, a 64-lead standard EEG cap is used, and the international 10-20 system standard electrode placement method is followed to ensure that the electrodes remain moist but do not cause short circuits between electrodes. When the patient performs fine motor tasks of the hand (grasping, pointing fingers, etc.), EEG signals related to movement intentions are collected and analyzed in real time, and movement intention features are extracted to provide reliable neurofeedback data for brain-controlled functional electrical stimulation.

[0171] Data processing module 2 is used to preprocess, extract features and normalize the collected pressure distribution signals, surface electromyography signals and electroencephalogram signals to obtain pressure features, surface electromyography features and electroencephalogram features respectively;

[0172] Rehabilitation Assessment Module 3, used to automatically assess rehabilitation progress based on a feature-level fusion algorithm;

[0173] The feedback and adaptive control module 4 is used to dynamically control the weights of the pressure distribution signal, surface electromyography signal and electroencephalogram signal based on the rehabilitation progress.

[0174] Although the present invention is described herein in conjunction with various embodiments, in the process of implementing the claimed invention, those skilled in the art can understand and implement other variations of the disclosed embodiments by viewing the drawings, the disclosure, and the accompanying drawings. In the specification, the word "comprising" does not exclude other components or steps, and "one" or "an" does not exclude multiple situations. A single processor or other unit can implement several functions listed in the specification. Certain measures are recorded in different embodiments, but this does not mean that these measures cannot be combined to produce good results.

[0175] Although the present invention has been described with reference to specific features and embodiments thereof, it will be apparent that various modifications and combinations thereof may be made without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely illustrative of the present invention and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the invention. It will be apparent that various modifications and variations of the present invention may be made by those skilled in the art without departing from the spirit and scope of the invention. Thus, the present invention is intended to include such modifications and variations as fall within the scope of the invention and its equivalents.

Claims

1. A multimodal signal fusion fine motor rehabilitation assessment and control method, characterized in that: Each evaluation includes the following steps: S1. Collect multimodal signals, which include: pressure distribution signals of fine hand movements, surface electromyography signals, and electroencephalogram signals; S2. Perform preprocessing, feature extraction, and normalization on the pressure distribution signals, surface electromyography signals, and electroencephalogram signals of the fine hand movements respectively, to obtain pressure feature values, surface electromyography feature values, and electroencephalogram feature values; S3. Calculate the conversion coefficient of the current rehabilitation stage based on the number of days of impaired fine motor function of the patient, and calculate the weights of the pressure feature values, surface electromyography feature values, and electroencephalogram feature values based on the conversion coefficient; S4. Perform weighted summation on the pressure feature values, surface electromyography feature values, and electroencephalogram feature values to obtain an evaluation score; S5. Regulate the rehabilitation training parameters based on the evaluation score and the number of days of impaired fine motor function of the patient.

2. The multimodal signal fusion fine motor rehabilitation assessment and control method according to claim 1, characterized in that: The pressure feature values include the maximum pressure, minimum pressure, and trend quantization value; the surface electromyography feature value is the root mean square value of the surface electromyography signal; the electroencephalogram feature values include the root mean square value of the electroencephalogram signal and event-related desynchronization; The evaluation score is calculated by the following method: Score =100×[a Press (t)(k1y max +k2y min +k3k)+a sEMG (t)RMS sEMG +α EEG (t)(k4RMS EEG +k5ERD)] Among them, Score represents the assessment score, t represents the number of days the patient's fine motor function is impaired, α Press (t) represents the weight of the pressure characteristic value when the number of damaged days is t, α sEMG (t) represents the weight of the surface electromyography characteristic value when the number of days of damage is t, α EEG (t) represents the weight of the EEG characteristic value when the number of days of damage is t, y max Indicates the maximum pressure, y min represents the minimum pressure value, k represents the trend quantization value, k1, k2, k3 represent the weights of the maximum pressure value, the minimum pressure value, and the trend quantization value, respectively, and k1+k2+k3=1; RMS sEMG Represents the root mean square value of the surface electromyographic signal, RMS EEG represents the RMS value of the EEG signal, ERD represents event-related desynchronization, k4 and k5 represent the RMS value of the EEG signal and the weight of event-related desynchronization, respectively, and k4+k5=1.

3. The multimodal signal fusion fine motor rehabilitation assessment and control method according to claim 1, characterized in that: The S3 includes the following sub-steps: S30. Determine the number of days t of impaired fine motor function of the patient; S31. Calculate the conversion coefficient of the current rehabilitation stage based on the number of days of impairment t, and the conversion coefficient includes the β coefficient and γ coefficient; S32. Determine the initial weights of the pressure feature values, surface electromyography feature values, and electroencephalogram feature values based on the number of days of impairment t, and calculate the weights of the pressure feature values, surface electromyography feature values, and electroencephalogram feature values based on the β coefficient and γ coefficient.

4. The multimodal signal fusion fine motor rehabilitation assessment and control method according to claim 3, characterized in that: The β coefficient is calculated by the following method: The γ coefficient is calculated by the following method: where, t represents the number of days of impaired fine motor function of the patient, 0 < t ≤ 14 indicates that the current rehabilitation stage is the acute phase, 14 < t ≤ 180 indicates that the current rehabilitation stage is the recovery phase, and t > 180 indicates that the current rehabilitation stage is the chronic phase.

5. The multimodal signal fusion fine motor rehabilitation assessment and control method according to claim 4, characterized in that: The initial weights are determined by the following method: If it is the first evaluation, then according to the number of days of impairment, the initial weight ratio of the pressure feature value, surface electromyography feature value, and electroencephalogram feature value is set as: acute phase: the initial weight ratio is 1:1:4 to 1:1:6; recovery phase: the initial weight ratio of the pressure feature value, surface electromyography feature value, and electroencephalogram feature value is set as 1:2:2 to 1:2:3; chronic phase: the initial weight ratio of the pressure feature value, surface electromyography feature value, and electroencephalogram feature value is set as 3:1:1 to 4:1:1; If it is not the first evaluation, then the weighted weights of the pressure feature value, surface electromyography feature value, and electroencephalogram feature value calculated during the previous evaluation are used as the initial weights for this evaluation.

6. The multimodal signal fusion fine motor rehabilitation assessment and control method according to claim 4, characterized in that: The weight of the electroencephalogram feature value is calculated by the following method: a EEG (t)=β(t)×a1+(1―β(t))×(1―γ(t))×a2+γ(t)×a3 The weight of the surface electromyography feature value is calculated by the following method: a sEMG (t)=β(t)×b1+(1―β(t))×(1―γ(t))×b2+γ(t)×b3 The weight of the pressure feature value is calculated by the following method: a press (t)=β(t)×c1+(1―β(t))×(1―γ(t))×c2+γ(t)×c3 And it satisfies: a EEG (t)+a sEMG (t)+a press (t)=1 Among them, α EEG (t) represents the weight of the EEG eigenvalue, α sEMG (t) represents the weight of the surface electromyography eigenvalue, α Press (t) represents the weight of the stress characteristic value, a1, a2, and a3 represent the initial weights of the EEG characteristic values ​​in the acute, recovery, and chronic stages, respectively; b1, b2, and b3 represent the initial weights of the surface electromyography characteristic values ​​in the acute, recovery, and chronic stages, respectively; c1, c2, and c3 represent the initial weights of the stress characteristic values ​​in the acute, recovery, and chronic stages, respectively.

7. The multimodal signal fusion fine motor rehabilitation assessment and control method according to claim 4, characterized in that: The S5 is specifically: S50. Divide the evaluation score into low-level scores, medium-level scores, and high-level scores; where, the low-level score is the evaluation score of 0 - 40 points, the medium-level score is the evaluation score of 41 - 70 points, and the high-level score is the evaluation score of 71 - 100 points; S51. Adjust rehabilitation training parameters based on the assessment scores and the number of days the patient's fine motor function is impaired. Specifically: If the patient's current rehabilitation stage is acute and the assessment score is low, set the current amplitude to 12-14 mA, the frequency to 15-20 Hz, the pulse width to 180-220 μs, and the number of training times to 10-12 times per set. If the patient's current rehabilitation stage is acute and the assessment score is medium, set the current amplitude to 14-16 mA, the frequency to 20 Hz, the pulse width to 200 μs, and the number of training times to 12-14 times per group. If the patient's current rehabilitation stage is acute and the assessment score is high, set the current amplitude to 15-17 mA, the frequency to 20 Hz, the pulse width to 200 μs, and the number of training times to 12-15 times per group. If the patient's current rehabilitation stage is the recovery stage and the assessment score is low, set the current amplitude to 15-17 mA, the frequency to 25 Hz, the pulse width to 220 μs, and the number of training times to 15-18 times per set. If the patient's current rehabilitation stage is the recovery stage and the assessment score is a medium-level score, set the current amplitude to 18-20 mA, the frequency to 30 Hz, the pulse width to 230 μs, and the number of training times to 18-20 times per set; If the patient's current rehabilitation stage is the recovery stage and the assessment score is a high-level score, set the current amplitude to 20-22 mA, the frequency to 30-35 Hz, the pulse width to 240-250 μs, and the number of training times to 20-22 times per set; If the patient's current rehabilitation stage is the chronic stage and the assessment score is low, set the current amplitude to 18-20 mA, the frequency to 30 Hz, the pulse width to 240 μs, and the number of training times to 18-20 times per group. If the patient's current rehabilitation stage is the chronic stage and the assessment score is a medium-level score, the current amplitude is set to 22-24 mA, the frequency to 35 Hz, the pulse width to 250 μs, and the number of training times to 22 times per group; When the patient's current rehabilitation stage is the chronic stage and the assessment score is a high-level score, the current amplitude is set to 24-26 mA, the frequency is 40 Hz, the pulse width is 260-280 μs, and the number of training times is 25 times / group.

8. A multimodal signal fusion fine motor rehabilitation assessment and control system, characterized by: include: A multimodal signal acquisition module, configured to acquire multimodal signals, including pressure distribution signals of fine hand movements, surface electromyography signals, and electroencephalogram signals; The data processing module is used to pre-process, extract features and normalize the collected pressure distribution signals, surface electromyography signals and electroencephalogram signals to obtain pressure features, surface electromyography features and electroencephalogram features respectively; Rehabilitation assessment module, used to automatically evaluate rehabilitation progress based on feature-level fusion algorithm; The feedback and adaptive control module is used to dynamically control the weights of pressure distribution signals, surface electromyography signals, and electroencephalogram signals based on rehabilitation progress.

9. The multimodal signal fusion fine motor rehabilitation assessment and control system according to claim 8, characterized in that: The multimodal signal acquisition module includes a pressure distribution signal acquisition unit, a surface electromyography signal acquisition unit and an electroencephalogram signal acquisition unit; The pressure distribution signal acquisition unit includes an array-type flexible pressure sensor subunit, a pressure distribution signal processing subunit and a host computer. The array-type flexible pressure sensor subunit includes a hand-grip sensor array and a sleeve sensor array. The hand-grip sensor array includes high-density independent sensitive points, the center spacing between adjacent sensitive points is 0.5 to 2 mm, and the scanning frequency is 50 to 500 Hz; the sleeve sensor array includes high-density independent sensitive points, the center spacing between adjacent sensitive points is 0.5 to 1 mm, and the scanning frequency is 50 to 500 Hz; the array-type flexible pressure sensor subunit is connected to the pressure distribution signal processing subunit through a flexible circuit board. The pressure distribution signal processing subunit adopts a sliding differential algorithm to eliminate the static error introduced by the deformation of the sensor base, and transmits the pressure data to the host computer after filtering. The host computer is used to calculate the pressure value of each sensitive point, map the pressure value of each sensitive point to the standard anatomical coordinate system, and generate a pressure thermodynamic map in real time to display the pressure distribution of each area of ​​the hand; The surface electromyography signal acquisition unit uses an 8-channel electromyography acquisition bracelet to collect surface electromyography signals; The electroencephalogram signal acquisition unit uses saline electrodes to acquire electroencephalogram signals.

10. The multimodal signal fusion fine motor rehabilitation assessment and control system according to claim 9, characterized in that: The hand-grip sensor array includes 4000 independent sensitive points, and the arrangement of the 4000 independent sensitive points is 80 rows * 50 columns; the cuff sensor array includes 1000 independent sensitive points, and the arrangement of the 1000 independent sensitive points is 40 rows * 25 columns.

Images