A lower limb rehabilitation training evaluation device based on a plantar multi-point pressure array

Abstract

The application discloses a lower limb rehabilitation training evaluation device based on a plantar multi-point pressure array, which comprises an ankle joint fixing support, which is adapted to the plantar contour of a human body, is worn on the lower limbs of a tested person and moves synchronously with the tested person performing a toe lifting action; a plantar pressure acquisition sensor, which is integrated on the inner plantar surface of the ankle joint fixing support, detects the pressure distribution of each region of the plantar surface of the tested person in the toe lifting process in real time and outputs corresponding sensing signals; and a processor, which is in communication connection with the plantar pressure acquisition sensor, is used for receiving and processing the sensing signals, predicting the toe lifting action angle based on a preset model and quantitatively evaluating the lower limb function recovery state of the tested person in combination with the angle change trend and outputting a rehabilitation evaluation result. The training device based on the combination of the plantar multi-point pressure array and the adjustable inclined platform calibration is constructed, the angle of the plantar surface during the lower limb movement is predicted and the quantitative evaluation of the rehabilitation progress is realized.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, and more specifically to a lower limb rehabilitation training and assessment device based on a multi-point pressure array on the sole of the foot. Background Technology

[0002] Existing lower limb rehabilitation training assessment devices mostly rely on visual observation, optical motion capture systems, or inertial sensor measurements. However, inertial sensors have problems such as large drift errors and weak anti-interference ability in low-speed and small-angle movements. Optical systems are expensive and complex to set up, making them unsuitable for long-term or home use.

[0003] In addition, although some foot-sensing solutions can collect force signals, they lack a motion angle calibration mechanism, which makes it impossible to effectively train and validate the angle prediction model.

[0004] Plantar pressure signals are highly correlated with movement angles, but existing plantar pressure systems are mostly used for gait detection or sports biomechanics research, and the sensor deployment and algorithms are not suitable for rehabilitation training scenarios.

[0005] For example, in the prior art:

[0006] Pressure distribution systems often employ dense matrix arrays, resulting in high equipment costs; rehabilitation training platforms primarily focus on mechanical measurements and cannot predict angles; data processing algorithms are simple and lack individualized calibration and cross-cycle generalization capabilities.

[0007] Therefore, providing a plantar pressure device that can accurately predict movement angles in rehabilitation training scenarios is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0008] In view of the above problems, the present invention is proposed to provide a lower limb rehabilitation training assessment device based on a multi-point pressure array of the foot to overcome or at least partially solve the above problems. By constructing a training device based on a multi-point pressure array of the foot combined with an adjustable tilt platform calibration, the device predicts the angle of the foot during lower limb movement and realizes a quantitative assessment of rehabilitation progress.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] This invention provides a lower limb rehabilitation training and assessment device based on a multi-point pressure array on the sole of the foot, comprising:

[0011] Ankle fixation brace, adapted to the contour of the human foot, is worn on the lower limb of the test subject and moves synchronously with the test subject's heel raise movement;

[0012] A plantar pressure sensor is integrated into the medial plantar surface of the ankle joint fixation brace to detect the pressure distribution in different areas of the sole of the foot in real time during the subject's heel raise and output the corresponding sensing signal.

[0013] The processor is communicatively connected to the plantar pressure acquisition sensor, and is used to receive and process the sensing signals, predict the heel raise angle based on a preset model, and quantitatively assess the lower limb function recovery status of the subject by combining the angle change trend, and output the rehabilitation assessment results.

[0014] Furthermore, it also includes:

[0015] The ruler limiting structure is fixedly installed on the heel position on the outside of the ankle joint fixation brace. Its upper edge is flush with the sole plane of the subject's foot when the subject is at rest, and serves as an angle trigger reference during the heel raise movement.

[0016] An adjustable tilting platform, which is on the same horizontal plane as the ankle joint fixation brace, includes an angle-adjustable tilting surface and an angle meter; the angle meter is used to display the angle between the tilting surface and the horizontal plane in real time; when the upper edge of the ruler limiting structure contacts the tilting surface, it indicates that the subject's heel raise has reached the target angle set by the tilting surface.

[0017] Furthermore, the ankle joint fixation brace also includes straps and / or buckle structures on both sides.

[0018] Furthermore, the plantar pressure acquisition sensor includes:

[0019] The flexible pressure sensor array in the heel area consists of five flexible pressure sensor units arranged in a "five-point array" layout. One sensor unit is located in the center, and the other four are respectively arranged above, below, left and right of the center, covering the key pressure areas of the heel.

[0020] The flexible pressure sensor array in the forefoot area consists of three flexible pressure sensor units arranged in a "one-line three-point" configuration along the lateral direction, corresponding to the main force points in the metatarsal head area of ​​the forefoot.

[0021] Each flexible pressure sensor unit is electrically connected to the data port via flexible leads and communicates with the processor.

[0022] Furthermore, the angle gauge is a magnetic angle gauge, fixed to the inclined surface.

[0023] Furthermore, the processor includes:

[0024] The data acquisition module is used to simultaneously acquire resistance change signals from eight channels of the flexible pressure sensor array in the heel area and the flexible pressure sensor array in the forefoot area.

[0025] The data processing module is used to filter the acquired signals, perform initial resistance calibration and normalization, and extract the resistance change rate, time gradient and dynamic features.

[0026] Angle prediction module is used to receive the feature vector output by the data processing module, perform nonlinear fitting using a multi-layer residual random forest regression model, correct the prediction residual layer by layer, and output the predicted angle value of the lower limb heel raise movement.

[0027] The rehabilitation assessment module is used to calculate corresponding indicators based on the predicted angle values ​​and their changing trends, and to divide the rehabilitation stages according to preset thresholds to generate quantitative assessment results.

[0028] Furthermore, the multilayer residual random forest regression model includes cascaded multilayer random forest regressors; each layer takes the original feature vector as input, and its training objective is the residual between the prediction result of the previous layer and the true angle, which is used by the next layer to learn and fit the prediction error of the previous layer.

[0029] Furthermore, the random forest regressor consists of multiple decision regression trees. Each decision regression tree uses a bootstrap sampling method to randomly select a subset of all input features as candidate split features during the node splitting process of the tree, thereby enhancing the diversity of the model. The output of the leaf node of each tree is the mean of the target values ​​corresponding to all training samples falling into that leaf node. The final output of this layer of random forest regressor is the arithmetic mean of the prediction results of all decision trees in this layer.

[0030] Furthermore, the rehabilitation assessment module,

[0031] This is used to receive the sequence of predicted angle values ​​continuously output by the subject when performing calf raises within a training cycle, and to calculate the following indicators based on the sequence and its dynamic characteristics: the maximum calf raise angle in a single movement, the rate of angle change, and the movement smoothness index across movements.

[0032] It is also used to divide the rehabilitation process into pre-rehabilitation, mid-rehabilitation and post-rehabilitation based on preset multi-level thresholds, and generate quantitative rehabilitation assessment results for the training cycle by combining the statistical characteristics and stability of the indicators throughout the whole cycle.

[0033] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a lower limb rehabilitation training and assessment device based on a multi-point pressure array on the sole of the foot, which has the following beneficial effects:

[0034] This invention provides an ankle joint fixation brace that simultaneously performs four functions: support, positioning, sensing, and limiting triggering. It features a compact structure and is easy to wear. Employing a modular sensor layout with a sparse arrangement of a five-point array on the heel and a three-point array on the forefoot, it balances coverage of key stress areas with cost control, avoiding the high complexity of dense arrays.

[0035] This invention utilizes a multi-layer residual random forest model combined with multi-point pressure characteristics of the foot to achieve real-time prediction of the angle of heel raises and objective quantitative assessment of the rehabilitation stage. It significantly reduces hardware costs without requiring expensive equipment, while meeting the accuracy requirements of clinical rehabilitation assessment. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0037] Figure 1 This is a front view of the structure of the lower limb rehabilitation training and assessment device based on a multi-point pressure array on the sole of the foot provided in this embodiment of the invention;

[0038] Figure 2 This is a top view of the lower limb rehabilitation training and assessment device based on a multi-point pressure array on the sole of the foot provided in this embodiment of the invention;

[0039] Figure 3 This is a rear view of the structure of the lower limb rehabilitation training and assessment device based on a multi-point pressure array on the sole of the foot provided in an embodiment of the present invention;

[0040] Figure 4 The curve showing the heel-lift angle prediction result of the angle prediction module provided in this embodiment of the invention;

[0041] Figure 5 These are the prediction curves of heel lift angle under different cycles provided in the embodiments of the present invention;

[0042] In the figure, 1-ankle joint fixation brace, 2-foot pressure acquisition sensor, 3-ruler limiting structure, 4-adjustable tilt platform, 21-heel area flexible pressure sensor array, 22-forefoot area flexible pressure sensor array, 41-tilted surface, 42-angle meter. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] This invention discloses a lower limb rehabilitation training and assessment device based on a multi-point pressure array on the sole of the foot, referring to... Figure 1 As shown, it includes:

[0045] Ankle joint fixation brace 1, adapted to the contour of the human foot, is worn on the lower limb of the test subject and moves synchronously with the test subject when performing heel raises;

[0046] The plantar pressure sensor 2 is integrated into the inner plantar surface of the ankle joint fixation brace 1. It can detect the pressure distribution in different areas of the sole of the foot in real time during the subject's heel raise and output the corresponding sensing signal.

[0047] The ruler limiting structure 3 is fixedly installed on the heel position on the outside of the ankle joint fixation brace 1, with its upper edge flush with the sole plane of the subject's foot in a static state, and serves as an angle trigger reference during the heel raise movement.

[0048] The adjustable tilting platform 4 is on the same horizontal plane as the ankle joint fixation brace 1, and includes an angle-adjustable tilting surface 41 and an angle meter 42; the angle meter 42 is used to display the angle between the tilting surface and the horizontal plane in real time.

[0049] The processor is communicatively connected to the plantar pressure acquisition sensor 2, and is used to receive and process the sensing signals, predict the heel raise angle based on a preset model, and quantitatively assess the lower limb function recovery status of the subject by combining the angle change trend, and output the rehabilitation assessment results.

[0050] The lower limb rehabilitation training assessment device constructed in this embodiment uses an ankle joint fixation brace 1 as the main body to provide mechanical support and positioning for the subject's lower limbs and serves as a fixed carrier for the plantar pressure acquisition sensor 2. An adjustable tilting platform 4 provides an angle calibration benchmark; when the subject raises their heels, the ruler fixed to the heel end of the brace touches the adjusted tilting platform, indicating that the target angle has been reached. This device acquires the resistance change signal from the pressure sensor and performs filtering, normalization, and initial resistance calibration; it utilizes a multi-layer machine learning model to achieve angle prediction and recovery status assessment, and achieves cross-time generalization through multi-cycle data training. This device can accurately predict lower limb movement angles, providing objective quantitative indicators for the rehabilitation process of patients with mild foot drop, and can be widely applied in the fields of rehabilitation training monitoring, gait assessment, and lower limb function testing.

[0051] The structure and composition of this device are described in detail below.

[0052] 1) Ankle joint fixation brace.

[0053] The ankle joint fixation brace 1 in this embodiment is used to provide mechanical support and restraint positioning for the ankle joint and sole during lower limb rehabilitation training. This brace not only stabilizes the ankle joint posture and restricts non-target movements, but also serves as a fixed carrier for the flexible pressure sensor array, providing stable support for the accurate acquisition of plantar pressure signals. In this embodiment, the ankle joint fixation brace 1 is made of various materials, such as polyurethane and ethylene. Materials such as vinyl acetate copolymers primarily provide support and fixation, adapt to the shape of the foot, and effectively transmit foot pressure to the flexible pressure sensor array 21 in the heel area and the flexible pressure sensor array 22 in the forefoot area.

[0054] In this embodiment, the ankle joint fixation brace 1 is fixedly connected to the subject's lower leg and sole via straps and buckles, ensuring the foot remains in contact with the bottom of the brace. This guarantees that changes in plantar pressure during heel raises or tilting movements are transmitted completely and stably to the flexible pressure sensor array. This structure effectively reduces sensor displacement and strain coupling errors during testing, ensuring the repeatability and reliability of plantar pressure signal acquisition. Furthermore, the geometric relationship between the sole portion of the ankle joint fixation brace 1 and the adjustable tilt platform 4 allows the device to achieve angle calibration without external goniometer assistance through a "brace ruler-tilt platform contact" method, enabling real-time determination of heel raise angles and calibration of training progress. Through this design, the ankle joint fixation brace 1 not only achieves a stable connection between the flexible sensor array and the sole but also combines posture restriction and angle calibration functions, making it one of the key structures for the device to predict movement angles and perform rehabilitation assessments.

[0055] 2) Foot pressure acquisition sensor.

[0056] In this embodiment, the plantar pressure sensor 2 is integrated into the medial plantar surface of the ankle joint fixation brace 1, as shown in the reference. Figure 2 As shown, it includes: a flexible pressure sensor array 21 in the heel area and a flexible pressure sensor array 22 in the forefoot area.

[0057] The flexible pressure sensor array 21 in the heel area consists of five independent flexible pressure sensor units. Each sensor unit is arranged in a shape similar to a "five-point" die, including a sensor unit located in the center and four sensor units distributed around the center unit on the top, bottom, left, and right.

[0058] In this embodiment, the horizontal and vertical spacing between the center points of each flexible pressure sensor unit is 3.5 cm, which is used to form a local pressure distribution detection network covering the key force area of ​​the heel.

[0059] The five sensor units are independent of each other and are fixed to the inner surface of the ankle joint fixation brace 1 at the corresponding position of the heel with tape in order to cover the key stress points in the heel area and detect the changes in heel pressure of the subject during heel raise, forward lean or backward leaning movements.

[0060] The flexible pressure sensor array 22 in the forefoot area consists of three independent flexible pressure sensor units, which are arranged laterally along the foot in a similar "I" shape.

[0061] In this embodiment, the distance between the center points of adjacent sensors is 3.5cm to cover the main stress points in the metatarsal head area of ​​the forefoot.

[0062] The aforementioned sensor arrays are all used to convert plantar pressure into a resistance signal related to the pressure magnitude. To achieve signal acquisition, each sensor unit of the flexible pressure sensor array 21 in the heel area and the flexible pressure sensor array 22 in the forefoot area is electrically connected via electrodes led out from flexible copper foil. The copper foil runs along the inner wall of the ankle joint fixation brace 1 to the wire outlet at the upper end of the brace, and finally connects to the external data port.

[0063] The copper foil leads in this embodiment have the characteristics of good flexibility, small bending radius and high fatigue resistance, which can avoid signal drift or breakage caused by the lead pulling during the movement of the test subject.

[0064] When the subject's foot rests on the ankle joint fixation brace 1, the pressure from the subject's foot acts on the flexible pressure sensor array 21 in the heel area and the flexible pressure sensor array 22 in the forefoot area. The pressure-sensitive layer of the flexible pressure sensor deforms, causing a change in the sensor's resistance. The resistance value changes non-linearly with the magnitude, direction, and location of the applied pressure, thus converting the foot pressure into a resistance signal. The flexible pressure sensor array 21 in the heel area and the flexible pressure sensor array 22 in the forefoot area are fixed to the heel of the brace using an embedded limiting groove and a medical double-sided adhesive layer, ensuring consistent position and adhesion even after multiple uses or when changing subjects. Each flexible sensor unit is connected to a signal lead via a conductive flexible cable. The signal lead is located on the outside of the brace and is electrically connected to the processor's data acquisition module via a shielded wire. To prevent signal interference or wire traction during the subject's movement, the outer layer of the wire is covered with insulating tape. The aforementioned fixing and connection structure not only ensures the reusability and signal stability of the flexible sensor array, but also facilitates disassembly, replacement, and maintenance in clinical or home rehabilitation settings. Through this structural design, the flexible sensor array can stably conform to the subject's foot for extended periods, enabling continuous and reliable acquisition of foot pressure distribution, providing high-quality input signals for angle prediction and rehabilitation assessment.

[0065] This structure enables stable acquisition of plantar pressure signals and automatic angle calibration, avoiding the dependence on external optical systems in traditional angle measurement methods.

[0066] 3) Ruler limiting structure and adjustable tilting platform.

[0067] In this embodiment, the ruler limiting structure 3 is fixedly installed on the outer heel position of the ankle joint fixation brace 1 with tape. After installation, its upper edge is at the same level as the sole of the test subject's foot and is used as the angle trigger reference for the heel raise movement.

[0068] In this embodiment, the adjustable tilting platform 4 is positioned on the ground. The tilting surface 41 of the platform forms a variable angle with the platform base via an adjustment knob, and the tilt angle is continuously adjustable. (Refer to...) Figure 2 As shown, the angle meter 42 is fixed to the inclined surface 41 by magnetic adsorption, and its measuring plane is in complete contact with the inclined surface to display the angle between the inclined surface and the horizontal plane in real time.

[0069] Reference Figure 3As shown, the upper edge of the ruler limiting structure 3 in this embodiment will collide with the inclined surface 41 of the adjustable tilting platform 4 when the test subject performs lower limb movements. By setting the angle between the inclined surface of the tilting platform and the horizontal plane, when the test subject raises his heels to the support and the ruler contacts the tilting platform, it indicates that the action angle has reached the set value, and it can be determined that each lower limb movement such as raising the heels can reach the preset angle.

[0070] 4) Processor.

[0071] The processor in this embodiment includes: a data acquisition module, a data processing module, an angle prediction module, and a rehabilitation assessment module.

[0072] The data acquisition module is used to simultaneously receive signals from multiple sensors at different locations on the sole of the foot. Connected to the plantar pressure acquisition sensor via wires, it synchronously acquires resistance change signals from eight channels of the flexible pressure sensor array 21 in the heel area and the flexible pressure sensor array 22 in the forefoot area.

[0073] This embodiment employs a multi-channel synchronous acquisition method, ensuring that the resistance signals of the flexible pressure sensor array 21 in the heel area and the flexible pressure sensor array 22 in the forefoot area are recorded synchronously under the same time reference, thereby maintaining temporal consistency. Through multi-channel acquisition, the spatial distribution characteristics and temporal variation patterns of the plantar force throughout the entire movement process can be fully captured, providing fundamental data for subsequent feature extraction. During acquisition, the system automatically records the original resistance change curves of each channel and establishes a timestamp index to ensure the traceability of subsequent signal analysis.

[0074] The data processing module is used to filter the acquired signals, perform initial resistance calibration and normalization, and extract the resistance change rate, time gradient and statistical characteristics.

[0075] In this embodiment, after acquiring the original signal, an adaptive filtering algorithm is first used to smooth the original resistance signal, performing multi-layer filtering. The filtering method is adaptively selected according to the signal length: when the number of sampling points is less than n, median filtering is used; when the number of points is large, Savitzky-Golay polynomial smoothing is used to remove high-frequency noise caused by environmental interference, mechanical vibration, or electrical contact jitter, while preserving the true trend of resistance change.

[0076] Subsequently, the reference initial resistance of each sensor was calculated by detecting the resistance value at the beginning of each action segment and using either the two-point averaging method or the first-value selection method. Abnormal drift segments are automatically marked and processed. The calculated initial resistance value is used to eliminate individual differences and attitude baseline offset, and based on this, differential correction is performed on subsequent signals to eliminate individual differences between different subjects and sensitivity deviations between sensors, achieving cross-cycle resistance feature alignment.

[0077] This embodiment uses The process normalizes the resistance changes, ensuring that the outputs of different sensors are within a uniform scale range, thus guaranteeing the comparability of feature inputs and the stability of model calculations. After this stage of processing, the original signal is converted into a standardized time-series signal, which clearly reflects the temporal relationship and relative intensity distribution of pressure changes in different regions.

[0078] The data processing module in this embodiment also performs multi-dimensional feature analysis on the preprocessed time-series signal to extract key indicators that reflect movement characteristics and plantar force patterns. The feature extraction process mainly includes resistance change rate features, time-series statistical features (time gradient), and dynamic features. The resistance change rate feature is calculated based on the ratio of the calculated resistance to the initial resistance to reflect the magnitude of pressure change. Time-series statistical features include mean, variance, extreme values, and the number of peaks, used to characterize the overall force pattern during movement. Dynamic features are calculated based on the derivative of the resistance change to reflect the speed, rhythm, and smoothness of pressure changes. The data processing module ultimately outputs a multi-dimensional feature vector that integrates spatial and temporal information, which serves as input data for the machine learning model.

[0079] The angle prediction module receives the feature vector output by the data processing module, performs nonlinear fitting using a multi-layer residual random forest regression model, corrects the prediction residual layer by layer, and outputs the predicted angle value of the lower limb heel raise movement.

[0080] In this embodiment, before feeding the feature vectors into the regression model, the features are first subjected to a unified numerical transformation and standardization process to improve the stability and generalization ability of the model training. The embodiment employs a standardization transformation based on the sample mean and standard deviation: the mean of each feature is calculated on the training set. with standard deviation This method is used in the training / validation / prediction phases. , Perform on features Transformation.

[0081] Secondly, regarding the multi-layer residual random forest regression model, this embodiment uses a three-layer residual random forest regression structure to regress and predict the angle of ankle calf raises or tilting movements. This embodiment consists of three sequentially connected regression sub-models, each constructed using an ensemble tree method, i.e., based on a random forest regressor.

[0082] In each layer, the random forest regressor consists of multiple regression decision trees. Each decision tree randomly samples a subset of data from the training samples using a bootstrap sampling method, and randomly selects a subset of features from all features for splitting decisions at tree nodes, thus forming a differentiated tree structure. Each decision tree is built based on the Classification and Regression Tree (CART) algorithm, using minimizing the variance of the split nodes as the optimal splitting criterion. Leaf nodes output the mean angle value of the data samples falling into that node. The output of this layer of the random forest is the average of the predictions from all decision trees.

[0083] To improve the modeling ability of the nonlinear relationship between plantar pressure changes and ankle joint angle, this embodiment uses a three-layer residual structure for stepwise approximation. The specific training process is as follows:

[0084] The first-layer regressor takes the extracted plantar pressure features as input. To predict the target, directly predict the coarse angle value. And obtain the first residual. .

[0085] The goal of the second-layer regressor is to learn the first residual. The mapping relationship between the input features and the output is the residual compensation amount. And calculate the second residual. .

[0086] The third-layer regressor continues to fit the second residual. Compensation prediction .

[0087] The final angle prediction result is a linear superposition of the three layer outputs:

[0088] By using the three-layer residual accumulation method described above, the first layer learns the main trend between pressure changes and angle, while the second and third layers further compensate for prediction errors, enabling the model to reduce residuals layer by layer and achieve higher fitting accuracy. Since each regressor layer adopts a random forest structure, the entire model possesses high stability, good noise resistance, and a strong ability to interpret pressure change characteristics.

[0089] In this embodiment, during the model training phase, the optimal model structure is selected and saved by comparing validation metrics (e.g., Mean Absolute Error, MAE) under different layers or different training set partitions. When used across cycles or scenarios, if a drift in the input signal distribution is detected, an incremental calibration or retraining process is triggered: first, distribution matching calibration (linear or CDF mapping) is performed on the new cycle data, and then the current model is evaluated using the calibrated data. (Refer to...) Figure 4The figure shows the heel rise angle prediction curve output by the three-layer residual random forest regression model. The horizontal axis represents the number of samples, and the vertical axis represents the heel rise angle value. Figure 4 This image shows a comparison between the actual and predicted angles during the subject's heel raise movement. `seg-strong` represents the prediction curve after enhanced smoothing of specific segmented samples during the validation phase. Here, `seg` indicates that the validation samples were divided into several consecutive time periods, and `strong` represents the prediction result after processing with the enhanced residual smoothing strategy. (Refer to...) Figure 5 The figure shows the heel rise angle prediction curves output by the two-layer residual random forest regression model under different periods. The horizontal axis represents the number of samples, and the vertical axis represents the heel rise angle value. Figure 5 This shows a comparison between the actual and predicted angles during the calf raise movement of the test subject in two cycles.

[0090] The rehabilitation assessment module is used to calculate corresponding indicators based on predicted angle values ​​and their changing trends, and to divide rehabilitation stages according to preset thresholds to generate quantitative assessment results.

[0091] In this embodiment, the angle prediction results output by the multi-layer residual random forest regression model are further analyzed by the rehabilitation assessment module to calculate multiple rehabilitation indicators, such as maximum heel raise angle, movement stability, and angle change rate. The maximum heel raise angle is obtained based on the maximum value in the predicted angle value sequence; movement stability is defined as the mean of the absolute values ​​of the first-order differences of the predicted angle sequence throughout the entire heel raise process, i.e.: In the formula, N represents the total number of sampling points contained in the angle prediction sequence. This represents the predicted angle value at the i-th sampling point during the heel raise movement, reflecting the ankle joint angle estimated by the system at that moment. The angle change rate is the absolute value of the change in angle between two adjacent sampling points, used to measure the degree of fluctuation in the action within that time interval; the angle change rate is defined as the average rate of increase during the predicted angle growth phase, i.e. In the formula Indicates the angle at the start of the calf raise movement. , This indicates the point in time when the calf raise begins. This indicates the time point when the maximum angle is reached. This embodiment automatically divides the predicted angle into three stages: early rehabilitation, middle rehabilitation, and late rehabilitation, corresponding to different muscle control abilities in patients. This embodiment also visualizes these indicators, generating rehabilitation stage curves and trend charts so that rehabilitation physicians or patients can intuitively understand the training progress.

[0092] The standard for the initial stage of rehabilitation is the maximum calf raise angle. 10°, stability of movement <0.15, rate of change of angle <5° / s.

[0093] The standard for the mid-stage of rehabilitation is a maximum calf raise angle of 10°. 20°, motion stability 0.15 <0.25, angular change rate 5° / s <10° / s.

[0094] The standard for the later stage of rehabilitation is the maximum calf raise angle. 20°, stability of movement 0.25, rate of change of angle 10° / s.

[0095] When the predicted angle reaches the set standard range, it indicates that the rehabilitation stage goal has been achieved.

[0096] This invention enables real-time prediction of lower limb movement angles and assessment of rehabilitation status during ankle joint rehabilitation training. Compared to traditional solutions relying on video recognition or external goniometers, this device has the advantages of simple structure, stable signal, high prediction accuracy, and strong adaptability, providing objective quantitative references for clinical and home rehabilitation training.

[0097] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0098] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A lower limb rehabilitation training and assessment device based on a multi-point pressure array on the sole of the foot, characterized in that, include: Ankle fixation brace, adapted to the contour of the human foot, is worn on the lower limb of the test subject and moves synchronously with the test subject's heel raise movement; A plantar pressure sensor is integrated into the medial plantar surface of the ankle joint fixation brace to detect the pressure distribution in different areas of the sole of the foot in real time during the subject's heel raise and output the corresponding sensing signal. The processor is communicatively connected to the plantar pressure acquisition sensor, and is used to receive and process the sensing signals, predict the heel raise angle based on a preset model, and quantitatively assess the lower limb function recovery status of the subject by combining the angle change trend, and output the rehabilitation assessment results. Also includes: The ruler limiting structure is fixedly installed on the heel position on the outside of the ankle joint fixation brace. Its upper edge is flush with the sole plane of the subject's foot when the subject is at rest, and serves as an angle trigger reference during the heel raise movement. An adjustable tilting platform, which is on the same horizontal plane as the ankle joint fixation brace, includes an angle-adjustable tilting surface and an angle meter; the angle meter is used to display the angle between the tilting surface and the horizontal plane in real time.

2. The apparatus as claimed in claim 1, characterized in that, The ankle joint fixation brace also includes straps and / or buckle structures on both sides.

3. The apparatus as described in claim 1, characterized in that, The plantar pressure sensor includes: The flexible pressure sensor array in the heel area consists of five flexible pressure sensor units arranged in a "five-point array" layout. One sensor unit is located in the center, and the other four are arranged above, below, left and right of the center, respectively, covering the key pressure areas of the heel. The flexible pressure sensor array in the forefoot area consists of three flexible pressure sensor units arranged in a "three points in a line" along the lateral direction, corresponding to the main force points in the metatarsal head area of ​​the forefoot. Each flexible pressure sensor unit is electrically connected to the data port via flexible leads and communicates with the processor.

4. The apparatus as claimed in claim 1, characterized in that, The angle gauge is a magnetic angle gauge, which is fixed to the inclined surface.

5. The apparatus as described in claim 3, characterized in that, The processor includes: The data acquisition module is used to simultaneously acquire resistance change signals from eight channels of the flexible pressure sensor array in the heel area and the flexible pressure sensor array in the forefoot area. The data processing module is used to filter the acquired signals, perform initial resistance calibration and normalization, and extract the resistance change rate, time gradient and dynamic features. Angle prediction module is used to receive the feature vector output by the data processing module, perform nonlinear fitting using a multi-layer residual random forest regression model, correct the prediction residual layer by layer, and output the predicted angle value of the lower limb heel raise movement. The rehabilitation assessment module is used to calculate corresponding indicators based on the predicted angle values ​​and their changing trends, and to divide the rehabilitation stages according to preset thresholds to generate quantitative assessment results.

6. The apparatus as claimed in claim 5, characterized in that, The multilayer residual random forest regression model includes a series of cascaded multilayer random forest regressors; each layer takes the original feature vector as input, and its training objective is the residual between the prediction result of the previous layer and the true angle, which is used by the next layer to learn and fit the prediction error of the previous layer.

7. The apparatus as claimed in claim 6, characterized in that, The random forest regressor consists of multiple decision regression trees. Each decision regression tree uses a bootstrap sampling method to randomly select a subset of all input features as candidate split features during the node splitting process of the tree, thereby enhancing the diversity of the model. The output of the leaf node of each tree is the mean of the target values ​​corresponding to all training samples falling into that leaf node. The final output of the random forest regressor in this layer is the arithmetic mean of the predictions from all decision trees in this layer.

8. The apparatus as claimed in claim 5, characterized in that, The rehabilitation assessment module, This is used to receive the sequence of predicted angle values ​​continuously output by the subject when performing calf raises within a training cycle, and to calculate the following indicators based on the sequence and its dynamic characteristics: the maximum calf raise angle in a single movement, the rate of angle change, and the movement smoothness index across movements. It is also used to divide the rehabilitation process into pre-rehabilitation, mid-rehabilitation and post-rehabilitation based on preset multi-level thresholds, and generate quantitative rehabilitation assessment results for the training cycle by combining the statistical characteristics and stability of the indicators throughout the whole cycle.

Images