An automatic identification method for obese patients based on gait analysis and its application

Abstract

The present invention belongs to the field of medical image analysis and intelligent diagnosis technology, and specifically relates to a method and application for automatically identifying obese patients based on gait analysis. The method uses YOLOv11 for pedestrian detection and BOT‑SORT for tracking; uses MMPose to extract human key points, and calculates gait and facial features based on the extracted human key points; uses a decision tree classifier to establish an obesity classification model, inputs gait and facial features into the obesity classification model, and outputs a prediction result of whether the individual is an obese patient. The present invention uses computer vision technology and deep learning methods to detect the target individual, extract skeletal key points, analyze gait and facial features in a non-contact and wearable device manner, and automatically classifies the target individual through a machine learning model to achieve efficient and accurate obesity screening.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical image analysis and intelligent diagnosis, and specifically relates to a method and application for automatically identifying obese patients based on gait analysis. Background Art

[0002] Obesity is a chronic disease that seriously affects global public health and is closely related to a variety of health problems, including cardiovascular disease, diabetes, and bone and joint diseases. Traditional obesity diagnosis methods mainly rely on biometric indicators such as body mass index or body fat analysis through imaging. However, these methods have certain limitations, including:

[0003] Limitations of the BMI indicator: BMI is calculated solely as the ratio of weight to height and cannot accurately distinguish between muscle mass and fat mass, potentially leading to misclassification. For example, individuals with high muscle mass may be misclassified as obese, while individuals with high body fat percentage but normal BMI may be missed.

[0004] Imaging methods are costly and difficult to perform on a large scale: Although imaging methods such as MRI, CT, and DXA can provide accurate information on body fat distribution, they are expensive, complex to operate, and limited by the availability of equipment and professionals, making them unsuitable for large-scale screening.

[0005] In recent years, gait analysis, as a non-invasive, remote, and scalable diagnostic tool, has garnered increasing attention in the healthcare field. Studies have shown that individuals with obesity exhibit specific gait patterns, such as decreased gait speed, shorter stride length, abnormal gait angles, and knee varus or valgus. Therefore, gait analysis can aid in determining whether an individual is obese.

[0006] Currently, gait-based health analysis primarily relies on wearable sensors or computer vision technology. The former requires users to wear additional equipment, which interferes with their natural gait and poses challenges with data synchronization and sensor wearability. Computer vision-based methods have been widely used in gait analysis, but existing research still suffers from the following shortcomings:

[0007] (1) Lack of in-depth exploration of complex gait features: Existing methods focus more on basic gait parameters such as speed, frequency, and stride length, but pay less attention to finer-grained gait features such as joint angles, center of mass swing, and knee valgus, which affects the accuracy of obesity identification.

[0008] (2) Difficulty in accurate detection in multi-person crowd environments: Traditional gait analysis methods are usually targeted at single-person videos. However, in multi-person environments, how to accurately extract the gait information of the target individual remains a challenge.

[0009] (3) Insufficient fusion of gait and facial features: Obese patients may also have significant features in facial morphology (such as facial width-to-height ratio, changes in jaw line, etc.), but existing gait-based obesity recognition methods often fail to effectively combine gait and facial information, resulting in limited diagnostic accuracy. Summary of the Invention

[0010] To address the above technical issues, the present invention provides a method and application for automatically identifying obese individuals based on gait analysis. This method utilizes computer vision technology and deep learning methods to detect target individuals, extract skeletal key points, analyze gait and facial features, and automatically classify them using a machine learning model, achieving efficient and accurate obesity screening.

[0011] The present invention is achieved through the following technical solutions:

[0012] A method for automatically identifying obese patients based on gait analysis, the method comprising:

[0013] (1) Data acquisition and processing: Collect videos containing the subject's gait data and preprocess the collected videos;

[0014] (2) Target detection and tracking: For the video collected and processed in step (1), YOLOv11 is used to detect pedestrians, obtain all pedestrians in the video frame, and output the coordinates of the pedestrian detection box; BOT-SORT is used for multi-target tracking to extract the complete gait sequence of the target individual;

[0015] (3) Human key point detection and gait and facial feature extraction: MM-Pose is used to extract human key points, including body joint key points, hand key points, foot key points, and facial key points, and gait and facial features are calculated based on the extracted human key points;

[0016] The gait and facial features include cadence, gait angle, knee angle, hip angle, ankle angle, center of mass swing amplitude, knee valgus angle, facial width-to-height ratio, and jawline angle;

[0017] (4) Establishing an obesity classification model: A decision tree classifier is used to establish an obesity classification model. Gait and facial features are input into the obesity classification model, and the output is a prediction result of whether the individual is an obese patient.

[0018] Furthermore, in step (1), a TL-IPC48GW binocular zoom surveillance camera was used for video acquisition. In the acquired video, the subject walked along a fixed route for more than 3 meters to ensure the integrity of the gait data.

[0019] Furthermore, in step (1), the preprocessing of the captured video includes: removing blurry frames, removing redundant frames, color normalization and background interference filtering.

[0020] Furthermore, in step (3), 133 human key points are extracted using MMPose, including 17 body joint key points, 42 hand key points, 6 foot key points and 68 facial key points.

[0021] Furthermore, in step (4), the obesity classification model is a binary decision tree, and the decision tree is used to perform binary classification of obesity, and the decision tree is a five-layer decision tree;

[0022] The five-layer decision tree achieved an accuracy rate of 91%, a precision rate of 0.88, a recall rate of 1.0, an F1-score of 0.94, and a 5-fold cross-validation screening accuracy rate of 87%.

[0023] Furthermore, in step (4), the gait and facial features calculated in step (3) are used as input, and the classification and recognition of the individual's health status is achieved through layer-by-layer determination.

[0024] Furthermore, the layer-by-layer determination method specifically includes:

[0025] (4.1) Root node judgment condition: Determine whether the average mandibular angle is less than or equal to 133.445°;

[0026] If the average mandibular angle is ≤ 133.445°, proceed to step (4.2) to determine the left subtree;

[0027] If the average mandibular angle is greater than 133.445°, proceed to step (4.3) to determine the right subtree;

[0028] (4.2) Judgment condition: Determine whether the maximum mandibular angle is less than or equal to 132.265°;

[0029] If the maximum mandibular angle is ≤132.265°, it is considered healthy;

[0030] If the maximum mandibular angle is greater than 132.265°, proceed to step (4.21) for the next judgment;

[0031] (4.21) Judgment condition: Determine whether the average value of knee valgus is less than or equal to 1.665°;

[0032] If the average knee valgus value is ≤1.665°, it is judged as healthy;

[0033] If the average value of knee valgus is greater than 1.665°, proceed to step (4.22) for the next judgment;

[0034] (4.22) Judgment condition: Determine whether the average gait angle is less than or equal to 11.27°;

[0035] If the average gait angle is ≤ 11.27°, proceed to step (4.23) for the next judgment;

[0036] If the average gait angle is >11.27°, it is considered obese;

[0037] (4.23) Judgment condition: Determine whether the mandibular angle is less than or equal to 17.475°;

[0038] If the mandibular angle range is ≤ 17.475°, proceed to step (4.24) for the next judgment;

[0039] If the mandibular angle range is greater than 17.475°, it is considered obese;

[0040] (4.24) Judgment condition: Determine whether the maximum center of mass swing is less than or equal to 15.265°;

[0041] If the maximum center of mass swing is ≤ 15.265°, it is judged to be healthy;

[0042] If the maximum center of mass swing is greater than 15.265°, the person is judged to be obese.

[0043] (4.3) Judgment condition: Determine whether the minimum gait angle is less than or equal to 6.535°;

[0044] If the minimum gait angle is greater than 6.535°, proceed to step (4.31) for the next judgment; if the minimum gait angle is less than or equal to 6.535°, proceed to step (4.32) for the next judgment;

[0045] (4.31) Judgment condition: Determine whether the knee joint angle is less than or equal to 6.47°;

[0046] If the knee angle is ≤6.47°, it is considered healthy.

[0047] If the knee joint angle is greater than 6.47°, the patient is considered obese.

[0048] (4.32) Judgment condition: Determine whether the maximum center of mass swing is less than or equal to 16.655°;

[0049] If the maximum center of mass swing is ≤16.655°, proceed to step (4.33) for the next judgment;

[0050] If the maximum center of mass swing is greater than 16.655°, proceed to step (4.34) for the next judgment;

[0051] (4.33) Judgment condition: Determine whether the average gait angle is less than or equal to 17.26°;

[0052] If the average gait angle is ≤17.26°, it is considered healthy.

[0053] If the average gait angle was >17.26°, it was considered obese;

[0054] (4.34) Judgment condition: Determine whether the average ankle joint angle is less than or equal to 161.17°;

[0055] If the average ankle angle is ≤ 161.17°, proceed to step (4.35) for the next judgment;

[0056] If the average ankle angle is greater than 161.17°, it is considered obese;

[0057] (4.35) Determine whether the average hip joint angle is less than or equal to 175.215°;

[0058] If the average hip joint angle is ≤175.215°, it is considered healthy;

[0059] If the average hip joint angle is greater than 175.215°, it is judged as obesity.

[0060] Furthermore, the method further includes step (5) result output: automatically generating a recognition report, wherein the recognition report includes individual gait and facial feature analysis and obesity screening classification results.

[0061] An application of an automatic identification method for obese patients based on gait analysis, wherein the automatic identification method for obese patients based on gait analysis is used for the identification of obesity in hospitals, health examination centers, telemedicine and fitness rehabilitation scenarios.

[0062] Beneficial technical effects of the present invention:

[0063] (1) Improving the comprehensiveness of gait feature extraction: Existing gait analysis methods do not pay enough attention to obesity-related gait features. The method provided by the present invention detects key points of the human body and then extracts multi-dimensional gait feature parameters (including step frequency, gait angle, knee joint angle, hip joint angle, ankle joint angle, center of mass swing amplitude, and knee valgus angle), characterizing the gait characteristics of obese patients from multiple angles and improving the accuracy of recognition.

[0064] (2) Optimizing target detection and tracking in multi-person environments: Traditional gait analysis is difficult to accurately obtain data on target individuals in multi-person scenarios. This invention combines target detection and tracking technology (YOLO + multi-target tracking), using Ultralytics YOLO for detection and BOT-SORT for tracking, to achieve accurate target recognition in multi-person environments, stable tracking of target individuals in complex environments, ensuring the accuracy and integrity of gait data and improving robustness.

[0065] (3) Integrating gait and facial features to improve recognition effect: In addition to gait features, the present invention introduces facial key point information to obtain facial features (including facial width-to-height ratio and jaw line angle), and combines gait features and facial features to classify obesity, thereby improving the accuracy of obesity recognition. Compared with the single feature detection method, the recognition accuracy is improved and the algorithm's distinguishing ability is enhanced.

[0066] (4) Constructing an efficient classification model to achieve automatic screening: The present invention uses machine learning methods such as decision trees to classify the extracted gait and facial features, realize automatic identification of obese patients, improve screening efficiency, and reduce dependence on manual diagnosis.

[0067] (5) Non-contact screening: The method provided by the present invention does not require wearing equipment and can complete the detection only through a camera, thereby improving screening efficiency, being suitable for large-scale health screening, and reducing the consumption of medical resources.

[0068] (6) Strong scalability: The method provided by the present invention is applicable to scenarios such as hospitals, physical examination centers, telemedicine, and gyms. It can be combined with mHealth for health management and support remote monitoring and personalized analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 Schematic diagram of facial key points in an embodiment of the present invention;

[0070] Figure 2 Schematic diagram of key points of body joints in an embodiment of the present invention;

[0071] Figure 3 Schematic diagram of key points of the hand in an embodiment of the present invention;

[0072] Figure 4 Schematic diagram of key points of the foot in an embodiment of the present invention;

[0073] Figure 5 Schematic diagram of gait angle calculation in an embodiment of the present invention;

[0074] Figure 6 This is a schematic diagram of knee joint angle calculation in an embodiment of the present invention;

[0075] Figure 7 This is a schematic diagram of hip joint angle calculation in an embodiment of the present invention;

[0076] Figure 8 Schematic diagram of ankle joint angle calculation in an embodiment of the present invention;

[0077] Figure 9 Schematic diagram of calculation of center of mass swing amplitude in an embodiment of the present invention;

[0078] Figure 10 Schematic diagram of knee valgus angle calculation in an embodiment of the present invention;

[0079] Figure 11 Schematic diagram of mandibular line angle calculation in an embodiment of the present invention;

[0080] Figure 12 Schematic diagram of facial aspect ratio calculation in an embodiment of the present invention. DETAILED DESCRIPTION

[0081] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0082] On the contrary, the present invention covers any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention as defined by the claims. Furthermore, to facilitate a better understanding of the present invention, certain specific details are described in detail below in the detailed description of the present invention. Those skilled in the art will be able to fully understand the present invention without these details.

[0083] In response to the problem that gait health analysis in the existing technology mainly relies on wearable sensors or computer vision technology for research, the present invention combines deep learning target detection, human posture estimation, and gait and facial feature extraction to realize automatic analysis and classification of gait data, and uses a decision tree algorithm to automatically classify obese patients to achieve non-contact obesity screening; it can be widely used in medical health monitoring, disease screening, intelligent rehabilitation assessment and other fields.

[0084] The present invention provides an embodiment of a method for automatically identifying obese patients based on gait analysis, the method comprising:

[0085] (1) Data acquisition and processing: Collect videos containing the subject's gait data and preprocess the collected videos. Specifically, this step uses a TL-IPC48GW binocular zoom surveillance camera for video acquisition. In the collected videos, the subject walks along a fixed route for more than 3 meters to ensure the integrity of the gait data.

[0086] The camera needs to be installed in the subject's regular gait path, such as a hospital corridor, a health check center walkway, or an exercise area in a rehabilitation center.

[0087] Camera requirements: Maximum image size: 3840×2160, i.e. 4K resolution, frame rate 15FPS;

[0088] Camera angle: ensure that the individual can be fully photographed in the walking area;

[0089] Video acquisition requirements: The acquisition range can accommodate multiple people walking at the same time; the subjects need to walk along a fixed route for at least 3 meters to ensure the integrity of the gait data.

[0090] Preprocessing the captured video includes removing blurry frames (using motion blur detection to eliminate low-quality frames) and removing redundant frames (cropping irrelevant video sections to retain only the target individual's gait data). This preprocessing can improve the robustness of subsequent detection.

[0091] (2) Target detection and tracking: For the video collected and processed in step (1), YOLOv11 is used to detect pedestrians, obtain all pedestrians in the video frame, and output the coordinates of the pedestrian detection box; BOT-SORT is used for multi-target tracking to extract the complete gait sequence of the target individual;

[0092] Specifically, because the acquisition range can accommodate multiple people walking simultaneously, it is necessary to first detect and track the target individual in the video frame to ensure that subsequent steps can accurately extract the target individual's gait data. To accurately detect the target individual in multi-person scenes, this paper uses Ultralytics YOLO (YOLOv11) for pedestrian detection and obtains the target individual's position in the video frame; this model has the characteristics of high precision and low latency, and can accurately detect human outlines in multi-person environments (adaptive to multi-person environments).

[0093] Ultralytics YOLO is used to obtain all pedestrians in the video frame and output the coordinates of the detection boxes. Ultralytics YOLO input: video frames; output: all pedestrian detection boxes (x, y, w, h) and confidence scores in each frame;

[0094] The specific process is as follows: read each frame of the gait video; use YOLO to detect people and output the bounding box of each individual; filter out targets with too small detection boxes or low confidence to reduce false detections.

[0095] The present invention adopts BOT-SORT for multi-target tracking to ensure stable tracking of target individuals in a multi-person environment and extract complete gait sequences.

[0096] In this invention, BOT-SORT mainly includes trajectory prediction, data association, and identity assignment;

[0097] Trajectory prediction: Kalman filtering is used to predict the next frame position of the target;

[0098] Data association: Calculate the IoU between the current frame target and the previous frame trajectory, and combine it with the Bernoulli Bayesian update mechanism to improve the recovery ability under short-term occlusion;

[0099] Identity assignment: Assign a unique ID to each detection target to ensure target consistency.

[0100] (3) Human key point detection and gait feature extraction: Utilize MMPose to extract human key points, including body joint key points, hand key points, foot key points, and facial key points, and calculate gait and facial features based on the extracted human key points;

[0101] The gait and facial features include cadence, gait angle, knee angle, hip angle, ankle angle, center of mass swing amplitude, knee valgus angle, facial width-to-height ratio, and jawline angle;

[0102] (4) Establishing an obesity classification model: A decision tree classifier is used to establish an obesity classification model. Gait and facial features are input into the obesity classification model, and the output is a prediction result of whether the individual is an obese patient.

[0103] In step (3) of this embodiment, 133 key points of the human body are extracted using MMPose, such as Figures 1-4 As shown, it includes 17 body joint key points, 42 hand key points, 6 foot key points and 68 facial key points.

[0104] Specifically, the 17 body joint key points include 5 facial contour key points, 2 shoulder key points, 2 elbow key points, 2 wrist key points, 2 hip joint key points, 2 knee joint key points, and 2 ankle joint key points.

[0105] The 42 hand key points are symmetrically distributed on the two hands, each hand including 2 wrist key points and 19 knuckle key points (3 on the thumb and 4 on each of the remaining fingers).

[0106] The six foot key points are symmetrically distributed on both feet, each foot including one big toe key point, one heel key point and one fifth toe key point;

[0107] The 68 facial key points include 17 key points distributed on the facial contour, 10 key points distributed on the brow bones (symmetrically distributed on both sides of the brow bones), 14 key points distributed on both eyes (symmetrically distributed on both sides of the eyes), 9 key points distributed on the nose (4 on the bridge of the nose and 5 under the nostrils), 1 key point distributed between the eyebrows and 17 key points distributed on the edge of the lips.

[0108] In this embodiment, gait and facial features are calculated based on the extracted key points of the human body, where: cadence is the basic gait parameter, gait angle, knee joint angle, hip joint angle, and ankle joint angle are joint angle features, center of mass swing amplitude and knee valgus angle are center of mass stability features, and facial width-to-height ratio and jaw line angle are facial features.

[0109] Among them, cadence (steps / second) refers to the number of steps taken per second; gait angle (°) is used to calculate the average, maximum, minimum and range of gait angles within the gait cycle; knee angle (°) is used to detect the maximum and minimum angles and range of change of the knee during the gait cycle; hip angle (°) is used to calculate the average angle change of the hip joint; ankle angle (°) is used to detect the range of motion of the ankle joint, reflecting gait stability. The center of mass swing amplitude is used to analyze the lateral and longitudinal displacement of the center of mass during walking; knee valgus is used to calculate the knee valgus angle and range of change. Obese patients usually show a larger knee valgus angle. The facial aspect ratio is calculated by extracting the facial contour through MMPose; the jawline angle and range are calculated by calculating the change in the inclination angle of the jawline to assist in identifying obesity characteristics.

[0110] like Figure 5 Schematic diagram for gait angle calculation: In gait analysis, the gait angle can reflect the symmetry of a person's gait and the distribution of their steps. If the angle is too large or too small, it may indicate a gait abnormality. It is generally believed that the gait angle of obese patients is larger than that of normal people when standing upright with their legs together, and then smaller than that of normal people when walking. Figure 5 In the figure, the dots represent key points used to calculate gait angles. These key points are located at the sacrum, left ankle joint, and right ankle joint. Black lines represent the connections between key points used to calculate gait angles. Two lines connect to the left and right ankle joints, forming the two edges of the gait angle.

[0111] Figure 6 This is a schematic diagram for calculating the knee joint angle; the increased load on the knee joint in obese patients may lead to a decrease in the knee flexion and extension angle. Figure 6In the figure, dots represent key points used to calculate the knee joint angle. These are located at the hip, knee, and ankle joints. Black lines represent the connections between key points used to calculate the knee joint angle. For example, the left knee joint angle calculation uses the left knee joint as a vertex, connected to the left hip joint and left ankle joint, forming two edges of the knee joint angle.

[0112] Figure 7 This is a schematic diagram for calculating the hip joint angle. The range of motion of the hip joint may be limited in obese patients, and the angle change may be smaller. Figure 7 In the figure, dots represent key points used to calculate the hip joint angle, located at the left hip, left knee, and left shoulder. Black lines represent the connections between key points used to calculate the hip joint angle. For example, the left hip joint angle calculation is performed by connecting the left shoulder and left knee, with the left hip joint as the vertex, to form two edges representing the hip joint angle.

[0113] Figure 8 This is a schematic diagram for ankle joint angle calculation; the range of ankle joint motion in obese patients may be limited, and the angle change may be smaller. Figure 8 The middle dots represent the key points used to calculate the ankle joint angle, located at the left knee, left ankle, and left big toe. Black lines represent the connections between key points used to calculate the ankle joint angle. For example, the left ankle joint angle calculation uses the left ankle joint as a vertex, connecting the left knee and left big toe to form the two edges of the ankle joint angle.

[0114] Figure 9 This is a diagram for calculating the center of mass swing amplitude; the center of mass swing amplitude = |∠1-∠2|. The center of mass swing amplitude may be larger in obese patients. Figure 9 The middle dots represent the key points used to calculate the center of mass sway amplitude. The three key points are located at the sacrum, left ankle joint, and right ankle joint. The black lines represent the connection between the key points, which are used to quantify the center of mass sway amplitude. Two angles are calculated. The first angle is the angle formed by the line connecting the left ankle joint and the sacrum and the midline of the body, called ∠1. The second angle is the angle formed by the line connecting the right ankle joint and the sacrum and the midline of the body, called ∠2. |∠1-∠2| is calculated. Healthy volunteers do not sway their bodies too much when walking, and |∠1-∠2| is small. Obese patients sway their bodies more when walking, and ∠1 and ∠2 vary greatly, resulting in a larger |∠1-∠2|.

[0115] Figure 10 Schematic diagram for knee valgus angle calculation: The dots in the figure represent the key points used to calculate the knee valgus angle, located at the right knee joint and right hip joint respectively. Black lines: represent the connection between the key points used to calculate the knee valgus angle. Taking the knee valgus angle on the right side of the body as an example, the knee valgus angle is the angle formed by the line connecting the right knee joint and the right hip joint and the midline of the body.

[0116] Figure 11 This diagram illustrates the jawline angle calculation: Measuring the jawline angle can help identify changes in facial morphology in obese patients. The dots in the diagram represent key points used to calculate the jawline angle. These three key points are located at the turning point of the right mandibular contour, the right ear, and the chin. The black lines represent the connections between these key points used to calculate the jawline angle. For example, the right jawline angle calculation is performed by connecting the right ear and chin, using the turning point of the right mandibular contour as an example.

[0117] Figure 12 The following diagram illustrates facial aspect ratio calculation. Obese individuals may have a larger facial aspect ratio, resulting in a rounder or wider face. The dots in the diagram mark key points for calculating facial width and height. These are located at the top of the head, the chin, and the left and right ears. Black lines represent the connections between key points used to calculate facial aspect ratio. Facial width is determined by the pixel difference between the left and right ears, while facial height is determined by the pixel difference from the top of the head to the chin. Facial aspect ratio = facial width / facial height.

[0118] In the present invention, the construction of the obesity classification model includes:

[0119] Feature dataset: 39 patients (20 obese and 19 healthy) were collected, and a total of 233 samples were extracted, including 145 obese samples and 88 healthy samples, for training the machine learning model.

[0120] Decision tree classification: A decision tree model is used for obesity classification, aiming to use gait characteristics to categorize individuals into two groups (obese / healthy). To improve the model's performance and generalization, in-depth exploration of feature engineering and hyperparameter optimization was conducted, including feature correlation analysis, feature importance assessment, and hyperparameter optimization. This ensured that the classifier accurately captured the gait characteristics of obese patients, improving the stability and reliability of predictions.

[0121] Feature importance evaluation: After screening out the key features, the importance score of each feature in the decision tree model is further calculated to ensure that the classifier mainly relies on the most discriminative features to make decisions without being disturbed by redundant information.

[0122] The feature importance of the decision tree in the present invention is calculated by the Gini index, and the specific steps are as follows:

[0123] Calculate the contribution of each feature to the target variable when splitting at each node;

[0124] Normalize the contribution of all features so that their sum is 1;

[0125] Select the features ranked top in terms of feature importance for visualization analysis to understand the basis for the model's decision-making.

[0126] The most important features included jawline angle and gait angle; the jawline angle was significantly more important than other features, indicating that the jawline angle is closely related to obesity identification. This difference in jawline angle may be due to changes in facial structure in obese individuals. The gait angle reflects the gait angle of leg movement during walking and is associated with factors such as gait stability and movement pattern. The gait angle of obese individuals may be affected by body mass, resulting in different gait patterns.

[0127] Other high-importance features include knee valgus, center of mass swing range, knee joint angle and hip joint angle; knee valgus describes the valgus angle of the knee joint, which may be related to the gait pattern and joint stress of obese individuals, affecting their walking stability; center of mass swing range: individuals with a larger center of mass swing range may show poor gait stability, and obese individuals may have different center of mass control patterns due to different mass distribution; knee joint angle: obese patients have increased knee load, which may lead to a decrease in knee flexion and extension angle; hip joint angle: obese patients may have a limited range of hip joint motion, and the angle change may be small.

[0128] 5-fold cross-validation optimizes model performance: To further evaluate the stability of the model, 5-fold cross-validation is used to evaluate the model: the dataset is randomly divided into 5 parts, 4 of which are used for training and 1 for testing, and repeated 5 times to ensure that the model performs evenly on different training / test sets.

[0129] The accuracy, precision, recall, and F1-score of the model under different folds were calculated to evaluate the overall performance of the model. The cross-validation results ensured the robustness of the model and avoided performance fluctuations caused by a single data split. Decision trees of different depths were compared, and the results are shown in Table 1:

[0130] Table 1 Comparison of decision tree results at different depths

[0131]

[0132] According to the comparison results, the present invention selects a 5-layer decision tree, which avoids the overfitting problem caused by an overly deep tree while ensuring high accuracy, and can effectively utilize gait features for classification.

[0133] In step (4), the gait features calculated in step (3) are used as input, and the classification and identification of the individual's health status is achieved through layer-by-layer determination. The layer-by-layer determination method specifically includes:

[0134] (4.1) Root node judgment condition: Determine whether the average mandibular angle is less than or equal to 133.445°;

[0135] If the average mandibular angle is ≤ 133.445°, proceed to step (4.2) to determine the left subtree;

[0136] If the average mandibular angle is greater than 133.445°, proceed to step (4.3) to determine the right subtree;

[0137] (4.2) Judgment condition: Determine whether the maximum mandibular angle is less than or equal to 132.265°;

[0138] If the maximum mandibular angle is ≤132.265°, it is considered healthy;

[0139] If the maximum mandibular angle is greater than 132.265°, proceed to step (4.21) for the next judgment;

[0140] (4.21) Judgment condition: Determine whether the average value of knee valgus is less than or equal to 1.665;

[0141] If the average value of knee valgus is ≤1.665, it is judged as healthy;

[0142] If the average value of knee valgus is greater than 1.665, proceed to step (4.22) for the next judgment;

[0143] (4.22) Judgment condition: Determine whether the average gait angle is less than or equal to 11.27°;

[0144] If the average gait angle is ≤ 11.27°, proceed to step (4.23) for the next judgment;

[0145] If the average gait angle is >11.27°, it is considered obese;

[0146] (4.23) Judgment condition: Determine whether the mandibular angle is less than or equal to 17.475°;

[0147] If the mandibular angle range is ≤ 17.475°, proceed to step (4.24) for the next judgment;

[0148] If the mandibular angle range is greater than 17.475°, it is considered obese;

[0149] (4.24) Judgment condition: Determine whether the maximum center of mass swing is less than or equal to 15.265;

[0150] If the maximum center of mass swing is ≤ 15.265, it is judged to be healthy;

[0151] If the maximum center of mass swing is greater than 15.265, it is judged as obesity.

[0152] (4.3) Judgment condition: Determine whether the minimum gait angle is less than or equal to 6.535°;

[0153] If the minimum gait angle is greater than 6.535°, proceed to step (4.31) for the next judgment; if the minimum gait angle is less than or equal to 6.535°, proceed to step (4.32) for the next judgment;

[0154] (4.31) Judgment condition: Determine whether the knee joint angle is less than or equal to 6.47°;

[0155] If the knee joint angle is ≤6.47°, it is considered healthy.

[0156] If the knee joint angle is greater than 6.47°, the patient is considered obese.

[0157] (4.32) Judgment condition: Determine whether the maximum center of mass swing is less than or equal to 16.655;

[0158] If the maximum center of mass swing is ≤16.655, proceed to step (4.33) for the next judgment;

[0159] If the maximum center of mass swing is greater than 16.655, proceed to step (4.34) for the next judgment;

[0160] (4.33) Judgment condition: Determine whether the average gait angle is less than or equal to 17.26°;

[0161] If the average gait angle is ≤17.26°, it is considered healthy.

[0162] If the average gait angle was >17.26°, it was considered obese;

[0163] (4.34) Judgment condition: Determine whether the average ankle joint angle is less than or equal to 161.17°;

[0164] If the average ankle angle is ≤ 161.17°, proceed to step (4.35) for the next judgment;

[0165] If the average ankle angle is greater than 161.17°, it is considered obese;

[0166] (4.35) Determine whether the average hip joint angle is less than or equal to 175.215°;

[0167] If the average hip joint angle is ≤175.215°, it is considered healthy;

[0168] If the average hip joint angle is greater than 175.215°, it is judged as obesity.

[0169] The method further includes step (5) outputting results: automatically generating an identification report, wherein the identification report includes individual gait feature analysis and obesity screening classification results. Specifically, the obesity screening classification results include: a predicted label of the subject as "obese" or "healthy"; and individual gait feature analysis outputting the subject's gait features and a visual chart.

[0170] This paper improves the performance of decision tree models in obesity identification through feature correlation analysis, feature importance assessment, hyperparameter optimization, and cross-validation. Ultimately, it achieved an 87% screening accuracy, significantly outperforming traditional discrimination methods. Compared to traditional body mass index-based screening methods, this method can more accurately identify gait abnormalities, improving the reliability and applicability of screening, and providing an efficient and low-cost solution for intelligent healthcare.

[0171] Example 2: Application of an automatic identification method for obese patients based on gait analysis. The automatic identification method for obese patients based on gait analysis in Example 1 is used for the identification of obesity in hospitals, health examination centers, telemedicine and fitness rehabilitation scenarios.

[0172] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for automatically identifying obese patients based on gait analysis, characterized in that: The method comprises: (1) Data acquisition and processing: Collect videos containing the subject's gait data and preprocess the collected videos; (2) Target detection and tracking: For the video collected and processed in step (1), YOLOv11 is used to detect pedestrians, obtain all pedestrians in the video frame, and output the coordinates of the pedestrian detection box; BOT-SORT is used for multi-target tracking to extract the complete gait sequence of the target individual; (3) Human key point detection and gait and facial feature extraction: MM-Pose is used to extract human key points, including body joint key points, hand key points, foot key points, and facial key points, and gait and facial features are calculated based on the extracted human key points; The gait and facial features include cadence, gait angle, knee angle, hip angle, ankle angle, center of mass swing amplitude, knee valgus angle, facial width-to-height ratio, and jawline angle; (4) Establishing an obesity classification model: A decision tree classifier is used to establish an obesity classification model. Gait and facial features are input into the obesity classification model, and the output is a prediction result of whether the individual is an obese patient.

2. The method for automatically identifying obese patients based on gait analysis according to claim 1, characterized in that: In step (1), a TL-IPC48GW binocular zoom surveillance camera is used for video acquisition. In the acquired video, the subject walks along a fixed route for more than 3 meters to ensure the integrity of the gait data.

3. The method for automatically identifying obese patients based on gait analysis according to claim 1, characterized in that: In step (1), preprocessing the captured video includes removing blurry frames and removing redundant frames.

4. The method for automatically identifying obese patients based on gait analysis according to claim 1, characterized in that: In step (3), 133 human key points are extracted using MMPose, including 17 body joint key points, 42 hand key points, 6 foot key points and 68 facial key points.

5. The method for automatically identifying obese patients based on gait analysis according to claim 4, characterized in that: In step (4), the obesity classification model is a binary decision tree, and the decision tree is used to perform binary classification of obesity, and the decision tree is a five-layer decision tree; The five-layer decision tree achieved an accuracy of 91%, a precision of 0.88, a recall of 1.0, an F1-score of 0.94, and a 5-fold cross-validation screening accuracy of 87%.

6. The method for automatically identifying obese patients based on gait analysis according to claim 5, characterized in that: In step (4), the gait and facial features calculated in step (3) are used as input, and the classification and recognition of the individual's health status is achieved through layer-by-layer determination.

7. The method for automatically identifying obese patients based on gait analysis according to claim 6, characterized in that: The layer-by-layer determination method specifically includes: (4.1) Root node judgment condition: Determine whether the average mandibular angle is less than or equal to 133.445°; If the average mandibular angle is ≤ 133.445°, proceed to step (4.2) to determine the left subtree; If the average mandibular angle is greater than 133.445°, proceed to step (4.3) to determine the right subtree; (4.2) Judgment condition: Determine whether the maximum mandibular angle is less than or equal to 132.265°; If the maximum mandibular angle is ≤132.265°, it is considered healthy; If the maximum mandibular angle is greater than 132.265°, proceed to step (4.21) for the next judgment; (4.21) Judgment condition: Determine whether the average value of knee valgus is less than or equal to 1.665°; If the average knee valgus value is ≤1.665°, it is judged as healthy; If the average value of knee valgus is greater than 1.665°, proceed to step (4.22) for the next judgment; (4.22) Judgment condition: Determine whether the average gait angle is less than or equal to 11.27°; If the average gait angle is ≤ 11.27°, proceed to step (4.23) for the next judgment; If the average gait angle is >11.27°, it is considered obese; (4.23) Judgment condition: Determine whether the mandibular angle is less than or equal to 17.475°; If the mandibular angle range is ≤ 17.475°, proceed to step (4.24) for the next judgment; If the mandibular angle range is greater than 17.475°, it is considered obese; (4.24) Judgment condition: Determine whether the maximum center of mass swing is less than or equal to 15.265°; If the maximum center of mass swing is ≤ 15.265°, it is judged to be healthy; If the maximum center of mass swing is greater than 15.265°, it is judged as obesity; (4.3) Judgment condition: Determine whether the minimum gait angle is less than or equal to 6.535°; If the minimum gait angle is greater than 6.535°, proceed to step (4.31) for the next judgment; if the minimum gait angle is less than or equal to 6.535°, proceed to step (4.32) for the next judgment; (4.31) Judgment condition: Determine whether the knee joint angle is less than or equal to 6.47°; If the knee joint angle is ≤6.47°, it is considered healthy. If the knee angle is > 6.47°, it is considered obese; (4.32) Judgment condition: Determine whether the maximum center of mass swing is less than or equal to 16.655°; If the maximum center of mass swing is ≤16.655°, proceed to step (4.33) for the next judgment; If the maximum center of mass swing is greater than 16.655°, proceed to step (4.34) for the next judgment; (4.33) Judgment condition: Determine whether the average gait angle is less than or equal to 17.26°; If the average gait angle is ≤17.26°, it is considered healthy. If the average gait angle was >17.26°, it was considered obese; (4.34) Judgment condition: Determine whether the average ankle joint angle is less than or equal to 161.17°; If the average ankle angle is ≤ 161.17°, proceed to step (4.35) for the next judgment; If the average ankle angle is greater than 161.17°, it is considered obese; (4.35) Determine whether the average hip joint angle is less than or equal to 175.215°; If the average hip joint angle is ≤175.215°, it is considered healthy; If the average hip joint angle is greater than 175.215°, it is judged as obesity.

8. The method for automatically identifying obese patients based on gait analysis according to claim 1, characterized in that: The method further includes step (5) result output: automatically generating an identification report, wherein the identification report includes individual gait and facial feature analysis and obesity screening classification results.

9. An application of an automatic identification method for obese patients based on gait analysis, characterized in that: The automatic identification method of obese patients based on gait analysis as described in any one of claims 1 to 8 is used for the identification of obesity in hospitals, health examination centers, telemedicine and fitness rehabilitation scenarios.

Images