Gait evaluation system and method based on wearable sensor network
By collecting gait data through wearable sensor networks and combining random forest algorithm and Stacked-LSTM model, the subjectivity and measurement limitations of existing gait evaluation methods are solved, achieving objective and accurate gait evaluation and improving the sensitivity and comprehensiveness of the evaluation.
Patent Information
- Application Number
- CN202410502569.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-25
- Publication Date
- 2026-05-15
- Estimated Expiration
- 2044-04-25
AI Technical Summary
Existing gait assessment methods are highly subjective, lack sensitivity, and have measurement limitations, failing to fully reflect an individual's physical function, especially in complex movements where they cannot be accurately assessed.
A system based on wearable sensor networks is adopted to collect gait data through inertial and pressure sensors. By combining random forest algorithm and stacked-LSTM network model, gait features are extracted and scored to achieve objective and accurate gait assessment.
It enables objective and accurate assessment of gait, avoids subjectivity, improves the sensitivity and comprehensiveness of the assessment, and can better reflect the body function in complex movements.
Smart Images
Figure CN118633929B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of gait anomaly assessment, and more specifically, to a gait assessment system and method based on wearable sensor networks. Background Technology
[0002] Lower limb motor function is a complex motor function of the human body, and gait characteristics of lower limb movement are a direct element and important indicator for assessing human health. Gait assessment is divided into qualitative gait assessment and quantitative gait assessment: qualitative gait assessment is used to evaluate whether the subject's gait function is abnormal, while quantitative gait assessment directly scores the gait characteristics of the lower limbs. Quantitative gait assessment plays an important role in displaying gait ability, comparing the effects of different rehabilitation programs for patients, and formulating rehabilitation plans.
[0003] In the past, gait assessment was mainly based on the direct experience of physicians and internationally advanced scales. However, the assessment process clearly exhibited the following problems:
[0004] 1. Subjectivity: Direct experience and rating scales rely on the test taker's subjective observation and rating, which may lead to differences in ratings among different test takers. This subjectivity may affect the repeatability and reliability of the test.
[0005] 2. Lack of sensitivity: The scale may lack sufficient sensitivity for some individuals with milder physical disabilities because it focuses primarily on more basic functional measurements. Some people may perform normally on these basic tasks but have problems on more complex tasks.
[0006] 3. Limitations of Measurement: The scale method mainly focuses on simple movements such as standing, walking, and getting up from a chair, and cannot comprehensively reflect an individual's physical function. It does not cover other important daily living functions, such as climbing stairs, carrying heavy objects, and bending over. Summary of the Invention
[0007] This invention provides a gait evaluation system, method, and apparatus based on wearable sensor networks, which at least addresses the problem of insufficient objectivity in gait evaluation in the prior art.
[0008] According to an embodiment of the present invention, a gait assessment system based on a wearable sensor network is provided, comprising the following steps:
[0009] The data acquisition module is used to collect users' gait data;
[0010] The data classification module is used to classify gait data, including normal gait data and abnormal gait data.
[0011] The scoring module is used to score the categorized gait data.
[0012] In one embodiment, the data acquisition module includes:
[0013] Several inertial sensors are installed and distributed around the user's waist, thighs, calves, and ankles to collect data.
[0014] Several pressure sensors are installed and distributed at the user's feet to collect data.
[0015] In one embodiment, the data classification module includes:
[0016] The first labeling unit is used to label normal gait data;
[0017] The second labeling unit is used to label abnormal gait data.
[0018] In one embodiment, the scoring module includes:
[0019] The gait segmentation unit is used to divide the user's gait cycle into four gait phases, which include a support phase, a right-side single support phase, a left-side double support phase, and a left-side single support phase.
[0020] The gait feature extraction unit is used to extract gait features in each gait phase. Gait features include maximum pressure and average pressure, pressure center trajectory, and gait asymmetry parameters.
[0021] The scoring unit is used to calculate the importance of features based on the extracted gait features. It uses a random forest to evaluate the contribution of each gait feature to each tree in the random forest. Based on the contribution, the scores of different gait features of the user are calculated in turn, and the scores are combined to obtain the total score.
[0022] A gait evaluation method based on wearable sensor networks, the method comprising:
[0023] Collect user gait data;
[0024] Gait data is classified into normal gait data and abnormal gait data.
[0025] The gait data after classification are scored.
[0026] In one embodiment, collecting the user's gait data specifically includes:
[0027] Data is collected from the user's waist, thighs, calves, and ankles;
[0028] Data was collected for the location of the foot.
[0029] In one embodiment, scoring the classified gait data specifically includes:
[0030] The user's gait cycle is divided into four gait phases, which include the support phase, the right single support phase, the left double support phase, and the left single support phase.
[0031] Gait features are extracted from each gait phase, including maximum pressure and average pressure, pressure center trajectory, and gait asymmetry parameters;
[0032] The importance of features is calculated based on the extracted gait features. Random forest is used to evaluate the contribution of each gait feature to each tree in the random forest. Based on the contribution, the scores of different gait features of the user are calculated in turn, and the scores are combined to obtain the total score.
[0033] In one implementation, the contribution is quantified using the Gini coefficient formula;
[0034] The formula for the Gini coefficient is:
[0035]
[0036] Among them, GI m P represents the Gini coefficient of node m in a decision tree; K represents the number of feature categories extracted in a decision tree in the random forest algorithm; k represents the current k-th category; P mk This represents the probability value of the k-th feature category in node m; the Gini coefficient of a leaf node in a random forest is the contribution of that node's feature.
[0037] A computer-readable medium, a computer-readable storage medium storing one or more programs, which can be executed by one or more processors to implement the steps in the gait evaluation method based on wearable sensor networks as described in any of the preceding claims.
[0038] A terminal device includes: a processor, a memory, and a communication bus; the memory stores a computer-readable program that can be executed by the processor;
[0039] The communication bus enables communication between the processor and memory;
[0040] When the processor executes a computer-readable program, it implements the steps in any of the above-described gait evaluation methods based on wearable sensor networks.
[0041] The gait assessment system and method based on wearable sensor networks in this invention collects users' gait data; classifies the gait data into normal and abnormal gait data; and scores the classified gait data. The system provides an objective and accurate assessment of gait using an instrument. Lower limb gait data is collected through a data acquisition module, and this data is then processed to perform qualitative and quantitative assessments of gait. This avoids the reliance on subjective observation and scoring by the test-taker using experience and scales in existing technologies, making the assessment more objective. Attached Figure Description
[0042] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:
[0043] Figure 1 This is a block diagram of the gait evaluation system based on wearable sensor networks of the present invention;
[0044] Figure 2 This is a schematic diagram of the gait evaluation scheme of the present invention.
[0045] Figure 3 This is a schematic diagram of the inertial sensor network of the present invention;
[0046] Figure 4 This is a schematic diagram of the pressure sensing network of the present invention;
[0047] Figure 5 This is a schematic diagram of gait period division according to the present invention;
[0048] Figure 6 This is a schematic diagram illustrating the calculation of the gait asymmetry coefficient in this invention;
[0049] Figure 7 This is a schematic diagram illustrating the convergence speed of the Stacked-LSTM model of this invention;
[0050] Figure 8 This is a comparison chart of the evaluation and scoring system of the present invention under different features and the Mini-BESTest test scores;
[0051] Figure 9 This is a linear regression diagram between the distance score and the Mini-BESTest test score of this invention;
[0052] Figure 10 This is a flowchart of the gait evaluation method based on wearable sensor networks of the present invention;
[0053] Figure 11 This is the terminal device of the present invention. Detailed Implementation
[0054] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.
[0055] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0056] Example 1
[0057] According to one embodiment of the present invention, a gait evaluation system based on wearable sensor networks is provided, see [link to relevant documentation]. Figure 1 ,include:
[0058] Data acquisition module 100 is used to collect the user's gait data;
[0059] The data classification module 200 is used to classify gait data, including normal gait data and abnormal gait data.
[0060] The scoring module 300 is used to score the gait data after classification.
[0061] The instrument provides an objective and accurate assessment of gait. Lower limb gait data is collected through a data acquisition module, and this data is then processed to perform qualitative and quantitative assessments of gait. This avoids the reliance on subjective observation and scoring by the test-taker through experience and scales found in existing technologies, resulting in a more objective assessment.
[0062] In one embodiment, the data acquisition module includes:
[0063] Several inertial sensors are installed and distributed around the user's waist, thighs, calves, and ankles to collect data.
[0064] Several pressure sensors are installed and distributed at the user's feet to collect data.
[0065] Specifically, sensor nodes were placed at seven locations on the subject's lower limbs: the waist, thigh, calf, and ankle of each leg. Figure 3 As shown, eight flexible pressure sensors are deployed in the sensing insole of each smart shoe to collect data during walking at a frequency of 100Hz. A wireless transmission module is installed inside the shoe to transmit the data to a mobile application in real time, such as... Figure 4 As shown.
[0066] In one embodiment, the data classification module includes:
[0067] The first labeling unit is used to label normal gait data;
[0068] The second labeling unit is used to label abnormal gait data.
[0069] In one embodiment, the scoring module includes:
[0070] The gait segmentation unit is used to divide the user's gait cycle into four gait phases, which include a support phase, a right-side single support phase, a left-side double support phase, and a left-side single support phase.
[0071] The gait feature extraction unit is used to extract gait features in each gait phase. Gait features include maximum pressure and average pressure, pressure center trajectory, and gait asymmetry parameters.
[0072] The scoring unit is used to calculate the importance of features based on the extracted gait features. It uses a random forest to evaluate the contribution of each gait feature to each tree in the random forest. Based on the contribution, the scores of different gait features of the user are calculated in turn, and the scores are combined to obtain the total score.
[0073] The following is an overall schematic diagram of the system of this application, and a general description of the system is provided:
[0074] refer to Figure 2 The system of this application includes a hardware structure and a software analysis structure. The entire hardware structure consists of a sensor network, which includes an inertial sensor network and a plantar pressure sensor network. The inertial sensor network includes several inertial sensors, and the plantar pressure sensor network includes several pressure sensors.
[0075] Each inertial sensor collects data and, under the control of the main controller, transmits the data to the PC communication interface via Bluetooth / WIFI. When the host computer (software analysis architecture) receives the data from the inertial sensor network, it first processes the data and then performs gait anomaly classification. This classification uses a pre-trained Stacked-LSTM (Stacked Multilayer Long Short-Term Memory) network model. This Stacked-LSTM network model is trained on the collected gait anomaly dataset to separate the abnormal gait data. Then, the gait data is scored by calculating important gait features and then combining these selected gait features with the random forest algorithm to construct multiple decision trees. Different decision trees yield different scores, and the average of the scores is used to obtain the final score. This score is derived from the Mini-BESTest test results and has certain clinical value. The overall design of the sensor network is described below.
[0076] Inertial sensor network distribution: Sensor nodes are placed at seven locations on the subject's lower limbs, namely the waist, thigh, calf, and ankle of each leg. Figure 3 As shown.
[0077] Pressure sensor network distribution: Eight flexible pressure sensors are deployed in the sensing insole of each smart shoe to collect data during walking at a frequency of 100Hz. A wireless transmission module is installed inside the shoe to transmit the data to a mobile application in real time. (Reference) Figure 4 As shown.
[0078] The following is a description of the hardware implementation and structure of the sensor network node:
[0079] First, the STMicroelectronics STM32L051C8T6 low-power chip is used. This chip can output current as low as 5-10mA at a 100Hz sampling rate. The inertial sensor is the ICM42588 from the ICM series, which is known for its high accuracy and low power consumption. Bluetooth is achieved using an ESP chip, which highly integrates a TCP / IP protocol stack, offering extremely low power consumption and low cost. Finally, an external crystal oscillator is added to acquire data at different sampling rates and accuracies. Decoupling capacitors are applied to the power supply to eliminate instability. This completes the setup of a single inertial sensor node.
[0080] The following is an explanation of gait abnormality screening:
[0081] Gait abnormality classification employs a Stacked-LSTM (Stacked Multilayer Long Short-Term Memory) network to screen subjects for gait abnormalities, selecting individuals with abnormal gait for scoring in the next step. The entire classification process consists of five steps, using publicly available datasets for normal gait, hemiplegic gait, Parkinsonian gait, ataxia gait, etc.
[0082] First, data is labeled to indicate normal and abnormal gait, typically with 1 and 0 respectively. Next, the dataset is divided into training and test sets, usually using random or time-series partitioning to ensure randomness and representativeness. A sliding window can be used to expand the dataset. After obtaining the labeled raw data, it is aggregated into multidimensional data and then used to build a classification model using a Stacked-LSTM neural network. The Stacked-LSTM neural network is a suitable neural network structure for processing time-series data, capable of capturing temporal dependencies. The input data is a time series, and the output layer is typically a classification layer used to identify normal gait and different types of abnormal gait. The LSTM model is trained using the labeled training set, including forward propagation, backpropagation, and weight updates. A loss function can be used to measure model performance, typically cross-entropy loss or other appropriate loss functions. Once the Stacked-LSTM neural network classification model is obtained, it can be directly used for gait abnormality detection. Finally, a validation set can be used to evaluate the model's performance. Performance metrics include accuracy and recall; if the performance is good enough, the model can be used to detect gait abnormalities in subjects.
[0083] The following is an explanation of the quantitative assessment of gait abnormalities:
[0084] Quantitative assessment of gait abnormalities includes gait feature calculation and extraction, and gait feature scoring, among which:
[0085] Gait feature calculation and extraction: Before gait extraction, a gait cycle is first divided into four gait phases, as shown in the reference. Figure 5 Gait cycle, encompassing the movement of the same foot from heel-off to heel-back, is a fundamental aspect of normal human movement. However, hemiplegic patients exhibit gait deviations, including a lack of heel-first or heel-off movements. To facilitate gait pattern analysis, such as... Figure 5 As shown, the gait cycle is divided into four phases based on total force. Specifically, the gait phases include the correct support cycle, the right-side single support cycle, the left-side double support cycle, and the left-side single support cycle. Gait division is typically achieved by detecting the onset and deviation of plantar pressure. However, due to the inherent variability of the data, false onsets and deviations may be detected erroneously. To address this issue, gait rule standards can be used to identify and eliminate abnormal take-offs and deviations.
[0086] After dividing the gait cycle, features are extracted for each cycle. These extracted gait features include maximum and average pressure, pressure center trajectory, and gait asymmetry parameters. In practical research, pressure feature calculation focuses on features extracted at each gait phase, rather than considering the entire gait cycle; gait asymmetry parameter calculation considers the entire gait cycle. The plantar pressure sensor meticulously divides a single gait cycle, allowing extraction of the time proportion of each gait phase, pressure distribution at each phase, short COP (center of pressure) position after foot strike, temporary COP position before foot liftoff, and the length of COP change, among other features. Simultaneously, inertial data from the IMU can be used to calculate gait asymmetry parameters.
[0087] The COP calculation formula is as follows:
[0088]
[0089]
[0090] Among them, (X) k ,Y k p(X) is the coordinate. k ,Y k ) represents the pressure value of the pressure sensor at this coordinate position, n represents the number of plantar pressure points, and k represents the currently calculated plantar pressure point.
[0091] The DWT (Dynamic Time Warping) algorithm is as follows: Figure 6 As shown, the gait feature calculation method is as follows. Assuming the gait IMU inertial data of the left and right feet are X and Y, the similarity between X and Y is calculated as follows:
[0092] X = [x1, x2, ... x m ];
[0093] Y = [y1, y2, ... y n ];
[0094] Where x1, x2, and x3 are the gait inertia data of the left foot, y1, y2, and y3 are the gait inertia data of the right foot, and m and n are the number of gait data points.
[0095] Assuming each sequence contains m and n samples, the 2x2 cost matrix D(m,n) is created by the following transition equation, which utilizes the algorithmic idea of dynamic programming:
[0096]
[0097] Where D(i,j) is the transition matrix, and Dist(i,j) is the distance from Xi to Yj.
[0098] Based on this transition equation, the total path W is calculated from (1, 1) to (n, m);
[0099] W = [ω1, ω2, ..., ω] i ω k ];
[0100] Where wi is the path length of the i-th segment.
[0101] The path length of each segment is calculated based on the path length of the previous segment. The final path D(m, n) is the asymmetry coefficient of the left and right feet obtained by the gait normalization algorithm.
[0102] DTW(X,Y)=D(n,m);
[0103] Where DWT(X,Y) is the asymmetry coefficient.
[0104] Improved DTW algorithm: WDTW (Weighted Dynamic Time Warping) adds a weight when calculating the Euclidean distance between two points on two sequences. This weight is related to the distance on the X-axis between the two points, thus solving the "Singularities" problem in time series data matching.
[0105] At this point, the gait characteristics, including maximum and average pressure, pressure center trajectory, and gait asymmetry under different time periods, have been calculated. The following section will use these gait characteristics for scoring:
[0106] Gait Feature Scoring: This application proposes a novel gait scoring method based on extracted gait features and the scoring results of the Mini-BESTest test. The Mini-BESTest test is a widely used evaluation tool with a scoring range of 0 to 28. Based on the Mini-BESTest test scores, users participating in the walk are categorized into four levels: 0-10, 11-18, 19-26, and 27-28. Subsequently, a random forest classifier is trained using the generated features, and feature importance is calculated. A random forest is a classifier containing multiple decision trees, where the classifier's output class is determined by the pattern of the class output by individual trees. The concept of using random forests for feature importance evaluation involves assessing the contribution of each feature to each tree in the random forest, calculating the average, and then comparing the contributions of different features. The Gini coefficient can be used to quantify the contribution.
[0107]
[0108] After obtaining the contribution of different features, the scores of different temporal features of the subject can be calculated sequentially by combining the contribution scores. Finally, these scores are combined to obtain the total score, as shown in the following formula:
[0109]
[0110] Where ρi represents the contribution of the i-th gait feature, X i Then X represents the value of the i-th feature of the subject. y This represents the value of the i-th characteristic of the gait-normal group. The calculated result, dist, is the final gait assessment score. The lower the score, the healthier the gait.
[0111] The gait evaluation system based on wearable sensor networks in this application has the following advantages:
[0112] 1. A sensor network structure combining an inertial sensor and a foot pressure sensor is proposed.
[0113] 2. Several important gait characteristic parameters are proposed as the basis for gait scoring, such as maximum pressure and average pressure, center of pressure (COP) location, and gait asymmetry parameters, and calculation methods are given.
[0114] 3. A method for further scoring gait parameters is proposed by combining the random forest algorithm and the Mini-BESTest scale method.
[0115] The following is a specific simulation experiment:
[0116] The gait anomaly classification verification is as follows: (Reference) Figure 7 As can be seen, the Stacked-LSTM model quickly reaches convergence as the training cycle increases, verifying that the model can be used for gait anomaly classification.
[0117] The gait assessment scoring system was validated as follows: as the number of selected features increases, the linear phase relationship first increases and then decreases, indicating that appropriate feature selection can improve the accuracy of gait assessment. (Reference) Figure 8 , represents the relationship between the number of selected features and the correlation between the weighted Euclidean distance score and the variation of the Mini-BESTest test score.
[0118] Next, correlation analysis will be performed: (Reference) Figure 9 Feature importance weighting can limit the impact of irrelevant features on the evaluation results. Under the conditions of the first nine features and their importance weights, the highest linear correlation with the Mini-BESTest is -0.9413. The following is a linear regression between the distance score and the Mini-BESTest test score.
[0119] As can be seen, by selecting appropriate gait features, these features can be highly linearly correlated with the Mini-BESTest test score, which increases the clinical practical value of the scoring system.
[0120] Example 2
[0121] According to another embodiment of the present invention, a gait evaluation method based on wearable sensor networks is provided, see [link to relevant documentation]. Figure 10 The methods include:
[0122] S100: Collects user gait data;
[0123] S200: Classify gait data, including normal gait data and abnormal gait data;
[0124] S300: Scoring the gait data after classification.
[0125] This instrument provides an objective and accurate assessment of gait. It collects lower limb gait data through a data acquisition module, processes the data, and performs qualitative and quantitative assessments of gait. This avoids the reliance on subjective observation and scoring by the test-taker using experience and scales, as is common in existing technologies, making the assessment more objective.
[0126] The gait assessment method based on wearable sensor networks in this application can be implemented through the above system. Its specific implementation method has been described in detail in the above system implementation, and will not be repeated here.
[0127] Example 3
[0128] Based on the above-described gait evaluation method based on wearable sensor networks, this embodiment provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps in the gait evaluation method based on wearable sensor networks as described in the above embodiment.
[0129] Example 4
[0130] A terminal device includes: a processor, a memory, and a communication bus; the memory stores a computer-readable program that can be executed by the processor; the communication bus enables communication between the processor and the memory; when the processor executes the computer-readable program, it implements the steps in the above-described gait evaluation method based on wearable sensor networks.
[0131] Based on the above-mentioned gait evaluation method based on wearable sensor networks, this application provides a terminal device, such as... Figure 11As shown, it includes at least one processor 20; a display screen 21; and a memory 22, and may also include a communications interface 23 and a bus 24. The processor 20, display screen 21, memory 22, and communications interface 23 can communicate with each other via the bus 24. The display screen 21 is configured to display a preset user guide interface in the initial setup mode. The communications interface 23 can transmit information. The processor 20 can invoke logical instructions in the memory 22 to execute the methods described in the above embodiments.
[0132] Furthermore, the logical instructions in the aforementioned memory 22 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.
[0133] The memory 22, as a computer-readable storage medium, can be configured to store software programs, computer-executable programs, such as program instructions or modules corresponding to the methods in the embodiments of this disclosure. The processor 20 executes functional applications and data processing by running the software programs, instructions, or modules stored in the memory 22, thereby implementing the methods in the above embodiments.
[0134] The memory 22 may include a program storage area and a data storage area. The program storage area may store the operating system and application programs required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 22 may include high-speed random access memory (RAM) and non-volatile memory. Examples include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, as well as transient storage media.
[0135] Furthermore, the specific process of loading and executing multiple instruction processors in the aforementioned storage medium and terminal device has been described in detail in the above method, and will not be repeated here.
[0136] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A gait assessment system based on wearable sensor networks, characterized in that, include: The data acquisition module is used to collect users' gait data; A data classification module is used to classify the gait data, including normal gait data and abnormal gait data. The scoring module is used to score the gait data after classification; The scoring module includes: The gait segmentation unit is used to segment a user's gait cycle into four gait phases by detecting the onset and deviation of plantar pressure. The four gait phases include a support phase, a right-side single support phase, a left-side double support phase, and a left-side single support phase. Gait rule standards are used to identify and eliminate abnormal take-offs and deviations. A gait feature extraction unit is used to extract gait features in each gait phase, the gait features including maximum pressure and average pressure, pressure center trajectory, and gait asymmetry parameters; The scoring unit is used to calculate the feature importance based on the extracted gait features, evaluate the contribution of each gait feature to each tree in the random forest using random forest, calculate the scores of different gait features of the user in turn based on the contribution, and combine the scores to obtain the total score. The scoring unit uses extracted gait features and Mini-BESTest test scores, building upon the pre-scored Mini-BESTest test scoring sheet, and further scores based on the Mini-BESTest test results. BESTest's test scores categorize participating users into four levels: 0-10, 11-18, 19-26, and 27-28. It then uses the generated features to train a random forest classifier, calculating feature importance. The concept of using random forest for feature importance assessment involves evaluating the contribution of each feature to each tree in the random forest, calculating the average, and then comparing the contributions of different features. The contribution is quantified using the Gini coefficient formula. The formula for the Gini coefficient is: ; in, K represents the Gini coefficient of node m in a decision tree; K represents the number of feature categories extracted in a decision tree in the random forest algorithm; k represents the current k-th category; This represents the probability value of the k-th feature category in node m; the Gini coefficient of a leaf node in a random forest is the contribution of that node's feature. After obtaining the contribution of different features, the scores of different time-varying features of the user can be calculated sequentially by combining the contribution scores. Finally, these scores are combined to obtain the total score; the formula is shown below: ； in This represents the contribution of the i-th gait feature. This represents the value of the user's i-th feature. This represents the value of the i-th characteristic of the gait-normal group. The calculated result, dist, is the final gait assessment score. The lower the score, the healthier the gait.
2. The gait assessment system based on wearable sensor networks according to claim 1, characterized in that, The data acquisition module includes: Several inertial sensors are installed and distributed around the user's waist, thighs, calves, and ankles to collect data. Several pressure sensors are installed and distributed at the user's feet to collect data.
3. The gait assessment system based on wearable sensor networks according to claim 1, characterized in that, The data classification module includes: The first labeling unit is used to label normal gait data; The second labeling unit is used to label abnormal gait data.
4. A gait evaluation method based on wearable sensor networks, characterized in that, A gait evaluation system based on wearable sensor networks as described in any one of claims 1-3; the method includes: Collect user gait data; The gait data is classified into normal gait data and abnormal gait data. The gait data after classification is scored.
5. The gait evaluation method based on wearable sensor networks according to claim 4, characterized in that, The collection of user gait data specifically includes: Data is collected from the user's waist, thighs, calves, and ankles; Data was collected for the location of the foot.
6. The gait evaluation method based on wearable sensor networks according to claim 5, characterized in that, The specific steps of scoring the classified gait data include: The user's gait cycle is divided into four gait phases, which include a support phase, a right-side single support phase, a left-side double support phase, and a left-side single support phase. Gait features are extracted from each gait phase, including maximum pressure and average pressure, pressure center trajectory, and gait asymmetry parameters; Based on the extracted gait features, feature importance is calculated. Random forest is used to evaluate the contribution of each gait feature to each tree in the random forest. Based on the contribution, the scores of different gait features of the user are calculated in turn, and the scores are combined to obtain the total score.
7. The gait evaluation method based on wearable sensor networks according to claim 6, characterized in that, A novel scoring method based on extracted gait features and Mini-BES scoring results is adopted. This method trains a random forest classifier using generated features on the basis of a pre-scored Mini-BES scoring table and calculates the contribution of each feature. We quantify the contribution by using the Gini coefficient formula. The formula for the Gini coefficient is: ; in, K represents the Gini coefficient of node m in a decision tree; K represents the number of feature categories extracted in a decision tree in the random forest algorithm; k represents the current k-th category; This represents the probability value of the k-th feature category in node m; the Gini coefficient of a leaf node in a random forest is the contribution of that node's feature.
8. A computer-readable medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the steps in the gait evaluation method based on wearable sensor networks as described in any one of claims 4-7.
9. A terminal device, characterized in that, include: Processor, memory, and communication bus; The memory stores a computer-readable program that can be executed by the processor; The communication bus enables communication between the processor and the memory; When the processor executes the computer-readable program, it implements the steps in the gait evaluation method based on wearable sensor networks as described in any one of claims 4-7.