Method and device for determining risk of knee injury, wearable device and storage medium

Abstract

The application discloses a knee joint injury risk determination method and device, a wearable device and a storage medium. The method comprises the following steps: acquiring X-direction signals, Y-direction signals and Z-direction signals collected by an inertial measurement unit worn on the lower limbs of a user; extracting signals between the ground contact time and the take-off time from the X-direction signals, the Y-direction signals and the Z-direction signals to obtain first signals, second signals and third signals; extracting features from the first signals, the second signals and the third signals to obtain first feature maps, second feature maps and third feature maps, and converting the first feature maps, the second feature maps and the third feature maps into feature vectors; and inputting the feature vectors into a target classification model to obtain a knee joint injury risk result of the user. Through the application, the problem that the judgment result is inaccurate in the related art by analyzing the influence of the ground reaction force on the knee joint to determine the knee joint injury risk is solved.

Description

Technical Field

[0001] This application relates to the field of sports injury detection technology, and more specifically, to a method, apparatus, wearable device, and storage medium for determining the risk of knee joint injury. Background Technology

[0002] Running is a popular sport with a high participation rate. Studies have reported that the injury rate among amateur runners is as high as 62.06%, with the knee and ankle being the most common injury sites, accounting for 31.07% and 24.21% of all injuries, respectively.

[0003] The knee joint is the largest weight-bearing joint in the human body, bearing immense pressure during exercise. Improper posture, excessive intensity, and inadequate preparation can all lead to knee injuries. The significant and repetitive impact on the knee joint during running increases the risk of overuse injuries. Knee pain and injuries are common problems faced by both professional and amateur runners. One powerful method for preventing and treating sports injuries is motion monitoring to understand the condition of the joint during daily activities or exercise, providing crucial information for training and injury treatment. Continuous motion monitoring and risk analysis are essential for protecting and restoring knee function during running.

[0004] In related technologies, force analysis is performed on the foot when it strikes the ground during running, mainly by analyzing the impact of the ground reaction force on the knee joint to determine the risk of knee injury. However, relying solely on the ground reaction force results in low accuracy of force analysis and thus low accuracy in determining the risk of knee injury.

[0005] There is currently no effective solution to the problem that the results of analyzing the impact of ground reaction force on the knee joint to determine the risk of knee injury are inaccurate in related technologies. Summary of the Invention

[0006] This application provides a method, device, wearable device, and storage medium for determining the risk of knee joint injury, in order to solve the problem that the judgment results of related technologies, which rely on analyzing the impact of ground reaction force on the knee joint to determine the magnitude of the risk of knee joint injury, are inaccurate.

[0007] According to one aspect of this application, a method for determining the risk of knee joint injury is provided. The method includes: acquiring X-direction, Y-direction, and Z-direction signals collected by an inertial measurement unit worn by a user on their lower limb; extracting signals between the ground contact time and the ground departure time from the X-direction, Y-direction, and Z-direction signals respectively, to obtain a first signal, a second signal, and a third signal; extracting features from the first signal, the second signal, and the third signal respectively, to obtain a first feature map, a second feature map, and a third feature map, and converting the first feature map, the second feature map, and the third feature map into feature vectors; inputting the feature vectors into a target classification model, and processing to obtain the user's knee joint injury risk result, wherein the target classification model is trained based on multiple sets of sample data, each set of sample data including a sample feature vector and a knee joint injury label.

[0008] Optionally, extracting features from the first signal, the second signal, and the third signal respectively to obtain a first feature map, a second feature map, and a third feature map, and converting the first feature map, the second feature map, and the third feature map into feature vectors includes: performing short-time Fourier transform on the first signal, the second signal, and the third signal respectively to obtain a first spectrum map, a second spectrum map, and a third spectrum map; performing convolutional feature extraction on the first spectrum map, the second spectrum map, and the third spectrum map respectively through a convolutional neural network to obtain a first feature map, a second feature map, and a third feature map; concatenating the first feature map, the second feature map, and the third feature map to obtain a target feature map, and determining the feature vector based on the target feature map.

[0009] Optionally, the first, second, and third spectrograms each contain two-dimensional spectrograms on the R, G, and B channels. The height of the two-dimensional spectrogram of each channel represents the amplitude of the energy density, and the width of the two-dimensional spectrogram of each channel represents the time. Convolutional feature extraction is performed on the first, second, and third spectrograms to obtain the first, second, and third feature maps, respectively. This includes: performing dimensionality compression on the features in the width direction and the features in the length direction of the first, second, and third spectrograms, respectively, and performing dimensionality expansion on the features of the R, G, and B channels in the length direction to obtain the first, second, and third feature maps.

[0010] Optionally, the first feature map, the second feature map, and the third feature map are concatenated to obtain a target feature map. Determining the feature vector based on the target feature map includes: extracting the spectral energy intensity of each pixel from the target feature map, determining the spectral energy weight of each pixel, and performing a weighted summation of the spectral energy intensities of each pixel based on the spectral energy weight to obtain a spectral energy intensity feature value; extracting the temporal information of each pixel from the target feature map, determining the temporal weight of each pixel, and performing a weighted summation of the temporal information of each pixel based on the temporal weight to obtain a temporal information feature value; and extracting feature values ​​in the length direction from the target feature map to obtain multiple channel feature values. The feature vector is composed of the multiple channel feature values, the spectral energy intensity feature value, and the temporal information feature value.

[0011] Optionally, the target classification model is a distributed gradient boosting model, which is determined as follows: Sample feature vectors and knee injury labels of multiple sample users participating in running training are obtained from a user database. The sample feature vectors are determined based on signals collected by inertial measurement units worn by the sample users' lower limbs, and the knee injury labels include injury labels and non-injury labels. Each sample user's sample feature vector and knee injury label are used as a single sample data point, resulting in multiple sample data points. A pre-defined distributed gradient boosting model is trained based on these multiple sample data points to obtain the target classification model.

[0012] Optionally, obtaining sample feature vectors and knee injury labels from multiple sample users participating in running training from the user database includes: obtaining the X-direction signal, Y-direction signal, and Z-direction signal collected by the inertial measurement unit of each sample user, generating a sample feature vector based on the signal between the ground contact time and the ground departure time in the X-direction signal, Y-direction signal, and Z-direction signal; obtaining the interruption reason information of each sample user's running training plan, determining the knee injury data from the interruption reason information, and generating a knee injury label based on the knee injury data.

[0013] Optionally, the inertial measurement unit can be worn on the user's knee or the back of their foot.

[0014] According to another aspect of this application, a wearable device is provided, including an inertial measurement unit for acquiring X-direction signals, Y-direction signals, and Z-direction signals; and a controller connected to the inertial measurement unit for acquiring the signals acquired by the inertial measurement unit and performing a method for determining the risk of knee joint injury.

[0015] According to another aspect of this application, a device for determining the risk of knee joint injury is provided. The device includes: an acquisition unit for acquiring X-direction signals, Y-direction signals, and Z-direction signals collected by an inertial measurement unit worn by a user on their lower limb; an extraction unit for extracting signals between the ground contact time and the ground departure time from the X-direction signals, Y-direction signals, and Z-direction signals, respectively, to obtain a first signal, a second signal, and a third signal; a conversion unit for extracting features from the first signal, the second signal, and the third signal, respectively, to obtain a first feature map, a second feature map, and a third feature map, and converting the first feature map, the second feature map, and the third feature map into feature vectors; and a processing unit for inputting the feature vectors into a target classification model to process and obtain the user's knee joint injury risk result, wherein the target classification model is trained based on multiple sets of sample data, each set of sample data including a sample feature vector and a knee joint injury label.

[0016] According to another aspect of the present invention, a computer storage medium is also provided for storing a program, wherein the program, when running, controls the device where the non-volatile storage medium is located to execute a method for determining the risk of knee joint injury.

[0017] According to another aspect of the present invention, an electronic device is also provided, comprising a processor and a memory; the memory stores computer-readable instructions, and the processor is configured to execute the computer-readable instructions, wherein the computer-readable instructions, when executed, perform a method for determining the risk of knee joint injury.

[0018] This application employs the following steps: acquiring X-direction, Y-direction, and Z-direction signals collected by an inertial measurement unit worn by the user on their lower limbs; extracting signals between the ground contact time and the ground departure time from the X-direction, Y-direction, and Z-direction signals respectively, obtaining a first signal, a second signal, and a third signal; extracting features from the first signal, the second signal, and the third signal respectively, obtaining a first feature map, a second feature map, and a third feature map, and converting the first feature map, the second feature map, and the third feature map into feature vectors; inputting the feature vectors into a target classification model, processing to obtain the user's knee joint injury risk result. The target classification model is trained based on multiple sets of sample data, each set of sample data including a sample feature vector and a knee joint injury label. This solves the problem in related technologies where the judgment of knee joint injury risk is inaccurate when analyzing the impact of ground reaction force on the knee joint. By extracting features from the X-direction, Y-direction, and Z-direction signals collected by the inertial measurement unit worn by the user on their lower limbs to construct feature vectors, and determining the knee joint injury risk result based on the feature vectors and a pre-trained model, the effect of accurately judging the magnitude of knee joint injury risk is achieved. Attached Figure Description

[0019] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0020] Figure 1 This is a flowchart of a method for determining the risk of knee joint injury according to an embodiment of this application;

[0021] Figure 2 This is a schematic diagram of the forces acting during running according to an embodiment of this application;

[0022] Figure 3 This is a force curve diagram of the X, Y, and Z directions during running according to an embodiment of this application;

[0023] Figure 4 This is a schematic diagram of a wearable device according to an embodiment of this application;

[0024] Figure 5 This is a schematic diagram of a device for determining the risk of knee joint injury according to an embodiment of this application;

[0025] Figure 6 This is a schematic diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0026] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0027] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0028] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application 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 for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover 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.

[0029] To address the inaccuracy of related technologies that rely on analyzing the impact of ground reaction force on the knee joint to determine the risk of knee injury, the following methods have emerged in these technologies:

[0030] To improve the accuracy of force analysis, related technologies also assign different weights in the X, Y, and Z directions to determine the impact of forces in these three directions on the knee joint. However, fixed weights cannot be adapted to different people, different postures, and different scenarios.

[0031] Therefore, this application aims to provide a solution that can solve the above-mentioned technical problems, the details of which will be described in subsequent embodiments.

[0032] It should be noted that all information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) involved in this disclosure are information and data authorized by the user or fully authorized by all parties.

[0033] According to an embodiment of this application, a method for determining the risk of knee joint injury is provided.

[0034] Figure 1 This is a flowchart of a method for determining the risk of knee joint injury according to an embodiment of this application. Figure 1 As shown, the method includes the following steps:

[0035] Step S102: Acquire the X-direction signal, Y-direction signal, and Z-direction signal collected by the inertial measurement unit worn by the user on the lower limbs.

[0036] Specifically, users can have an inertial measurement unit (IMU) worn on the lower limbs of a running user. The IMU collects triaxial signals, including X-axis, Y-axis, and Z-axis signals, which can reflect the influence of forces in the Z, X, and Y directions on the user during running. Acquiring the triaxial signals collected by the IMU lays a data foundation for improving the accuracy of lower limb force analysis.

[0037] Since the force on the knee and instep is more reflective of the force on the lower limbs during running, thereby improving the accuracy of accurately judging the risk of knee joint injury, optionally, in the method for determining the risk of knee joint injury provided in the embodiments of this application, the inertial measurement unit is worn on the user's knee or instep.

[0038] Figure 2 This is a schematic diagram of the forces acting during running according to an embodiment of this application, such as... Figure 2As shown, a user runs on a foot pressure treadmill, with an inertial measurement unit (IMU) worn on the instep of the user. The triaxial signals collected by the IMU can reflect the influence of forces in the Z, X, and Y directions on the user during the running process.

[0039] Step S104: Extract the signals between the ground contact time and the ground departure time from the X-direction signal, Y-direction signal and Z-direction signal respectively to obtain the first signal, the second signal and the third signal.

[0040] Figure 3 This is a force curve diagram in the X, Y, and Z directions during running according to an embodiment of this application, such as... Figure 3 As shown, in running activities, each running cycle of alternating feet includes the landing moment, the standing phase, and the takeoff moment. At the landing moment and the takeoff moment, the signal strength of the X-direction, Y-direction, and Z-direction signals is the lowest, reaching the trough value in each force curve. During the standing phase, the signal strength of the X-direction, Y-direction, and Z-direction signals increases, and they are between two adjacent trough values ​​in each force curve.

[0041] Therefore, in the curve corresponding to the X-direction signal, the signal between two adjacent valley values ​​is extracted to obtain the first signal between the ground contact time and the ground departure time. In the curve corresponding to the Y-direction signal, the signal between two adjacent valley values ​​is extracted to obtain the second signal between the ground contact time and the ground departure time. In the curve corresponding to the Z-direction signal, the signal between two adjacent valley values ​​is extracted to obtain the third signal between the ground contact time and the ground departure time.

[0042] Step S106: Extract features from the first signal, the second signal and the third signal respectively to obtain the first feature map, the second feature map and the third feature map, and convert the first feature map, the second feature map and the third feature map into feature vectors.

[0043] Specifically, the first feature map, the second feature map, and the third feature map can be concatenated into a target feature map. Feature vectors can be extracted from the target feature map. The extracted feature vectors can represent the force situation after fusion in the X, Y, and Z directions.

[0044] Optionally, in the method for determining the risk of knee joint injury provided in this application embodiment, extracting features from the first signal, the second signal, and the third signal respectively to obtain a first feature map, a second feature map, and a third feature map, and converting the first feature map, the second feature map, and the third feature map into feature vectors includes: performing short-time Fourier transform on the first signal, the second signal, and the third signal respectively to obtain a first spectrum map, a second spectrum map, and a third spectrum map; performing convolutional feature extraction on the first spectrum map, the second spectrum map, and the third spectrum map respectively through a convolutional neural network to obtain a first feature map, a second feature map, and a third feature map; concatenating the first feature map, the second feature map, and the third feature map to obtain a target feature map, and determining the feature vector based on the target feature map.

[0045] Specifically, the first signal, the second signal, and the third signal are signals of force intensity changing with time. The first signal, the second signal, and the third signal can be converted into frequency domain and time domain spectrograms through short-time Fourier transform. The spectrogram represents the energy distribution of the signal at different frequencies, so that the time domain characteristics and frequency domain characteristics of the signal can be obtained simultaneously. The spectral characteristics can include frequency components, frequency distribution, and spectral density, etc.

[0046] Furthermore, using the spectrogram as the basis for signal processing and feature extraction, the first, second, and third spectrograms are expanded and compressed in dimension by a convolutional neural network to obtain the first, second, and third feature maps. The obtained features are then concatenated to obtain a target feature map that can integrate the force conditions in the X, Y, and Z directions, and feature vectors are then extracted from it.

[0047] Step S108: Input the feature vector into the target classification model and process it to obtain the user's knee joint injury risk result. The target classification model is trained based on multiple sets of sample data, and each set of sample data includes a sample feature vector and a knee joint injury label.

[0048] Specifically, the target classification model can be a distributed gradient enhancement model. The feature vectors that integrate the force conditions in the X, Y, and Z directions are input into the target classification model, and the classification results are obtained. If the classification result is 1, it means that the user's knee joint is at risk of injury; if the classification result is 0, it means that the user's knee joint is not at risk of injury.

[0049] The target classification model needs to be pre-trained. Optionally, in the method for determining knee joint injury risk provided in this application embodiment, the target classification model is a distributed gradient enhancement model. The target classification model is determined in the following way: obtaining sample feature vectors and knee joint injury labels of multiple sample users participating in running training from the user database, wherein the sample feature vectors are determined based on the signals collected by the inertial measurement units worn by the sample users' lower limbs, and the knee joint injury labels include injury labels and non-injury labels; determining the sample feature vector and knee joint injury label of each sample user as a sample data, obtaining multiple sample data; training the preset distributed gradient enhancement model based on the multiple sample data to obtain the target classification model.

[0050] The method for determining the sample feature vector is the same as the method for determining the feature vector of the user to be tested. However, the knee injury label needs to be determined based on the information fed back by the user during exercise. Optionally, in the method for determining the risk of knee injury provided in this application embodiment, obtaining the sample feature vectors and knee injury labels of multiple sample users participating in running training from the user database includes: obtaining the X-direction signal, Y-direction signal, and Z-direction signal collected by the inertial measurement unit of each sample user, generating a sample feature vector based on the signal between the ground contact time and the ground departure time in the X-direction signal, Y-direction signal, and Z-direction signal; obtaining the interruption reason information of each sample user's running training plan, determining the knee injury data from the interruption reason information, and generating a knee injury label based on the knee injury data.

[0051] The specific steps for obtaining the sample feature vector of a sample user include: extracting the signals between the ground contact time and the ground departure time from the X-direction signal, Y-direction signal, and Z-direction signal collected by the inertial measurement unit of each sample user; performing short-time Fourier transform on the three extracted signals to obtain three sample spectrograms; performing convolutional feature extraction on the three sample spectrograms using a convolutional neural network to obtain three sample feature maps; and concatenating and extracting features from the three sample feature maps to obtain the sample feature vector.

[0052] The specific steps for obtaining knee injury labels for sample users include: when a user is implementing a running plan using fitness software, if the user's running plan is interrupted, the user is prompted to provide the reason for the interruption. If the reason for the interruption is a knee injury, the knee injury label matched with the sample feature vector generated based on the inertial measurement signal of the sample user is 1. For users who do not interrupt their running plan, or whose reasons for interrupting their running plan do not include knee injury, the knee injury label matched with the sample feature vector generated based on the inertial measurement signal of the sample user is 0.

[0053] This embodiment can accurately determine the knee injury label matching the sample feature vector, laying a data foundation for accurately determining the risk of knee injury through the target model.

[0054] The method for determining knee joint injury risk provided in this application involves acquiring X-direction, Y-direction, and Z-direction signals collected by an inertial measurement unit worn by the user on the lower limb; extracting signals between the ground contact time and the ground departure time from the X-direction, Y-direction, and Z-direction signals respectively to obtain a first signal, a second signal, and a third signal; extracting features from the first signal, the second signal, and the third signal respectively to obtain a first feature map, a second feature map, and a third feature map, and converting the first feature map, the second feature map, and the third feature map into feature vectors; inputting the feature vectors into a target classification model, and processing to obtain the user's knee joint injury risk result. The target classification model is trained based on multiple sets of sample data, each set of sample data including a sample feature vector and a knee joint injury label. This solves the problem in related technologies where the judgment of knee joint injury risk by analyzing the impact of ground reaction force on the knee joint results in inaccurate results. Features are extracted from the X, Y, and Z direction signals collected by the inertial measurement unit worn by the user's lower limbs to form a feature vector. Based on the feature vector and the pre-trained model, the knee joint injury risk is determined, thereby achieving the effect of accurately judging the magnitude of the knee joint injury risk.

[0055] Convolutional feature extraction includes dimensional expansion and compression. Optionally, in the method for determining knee joint injury risk provided in this application embodiment, the first, second, and third spectrograms each contain two-dimensional spectrograms on the R, G, and B channels. The height of the two-dimensional spectrogram of each channel represents the amplitude of energy density, and the width of the two-dimensional spectrogram of each channel represents time. Convolutional feature extraction is performed on the first, second, and third spectrograms respectively to obtain the first, second, and third feature maps, including: dimensional compression of the features in the width direction and the features in the length direction of the first, second, and third spectrograms respectively, and dimensional expansion of the features in the R, G, and B channels in the length direction to obtain the first, second, and third feature maps; and concatenation of the first, second, and third feature maps includes: feature concatenation of the first, second, and third feature maps in the length direction.

[0056] A convolutional neural network can include an input layer, multiple sets of convolutional and pooling layers, and an output layer. It extracts features from the feature vector step by step through convolutional and pooling layers, and performs classification tasks through fully connected layers. Each set of convolutional and pooling layers can contain three convolutional layers, and each convolutional layer contains a 3*3 convolutional kernel. The convolutional layers perform convolution operations on the input feature map through the convolutional kernel, thereby extracting features at different locations. The pooling layers reduce the size of the feature map by downsampling, thereby reducing the number of model parameters and computational complexity.

[0057] For example, each of the first, second, and third spectrograms is a 3*512*512 feature map, where 3 refers to the three channels R, G, and B, and 512*512 represents the two-dimensional spectrogram representing the amplitude of energy density changing over time. The features of the R, G, and B channels in the length direction of the first, second, and third spectrograms are expanded dimensionally, for example, from 3 dimensions to 512 dimensions. The features in the width direction of the first, second, and third spectrograms are then compressed dimensionally, i.e., the corresponding two-dimensional spectrograms are compressed. For example, the 512*512 two-dimensional spectrogram can be compressed into a 32*32 feature map. Thus, the first, second, and third feature maps are all 512*32*32 feature maps.

[0058] In this embodiment, abstract features are gradually extracted from the first, second, and third spectrograms through multiple convolutional pooling layers in a convolutional neural network. The features in the width direction are compressed in dimension, while the features in the R, G, and B channels in the length direction are expanded in dimension. This enables an effective representation of complex data and lays a data foundation for accurately determining the risk of knee joint injury.

[0059] Optionally, in the method for determining the risk of knee joint injury provided in this application embodiment, the first feature map, the second feature map, and the third feature map are stitched together to obtain a target feature map. Determining the feature vector based on the target feature map includes: extracting the spectral energy intensity of each pixel from the target feature map, determining the spectral energy weight of each pixel, and performing a weighted summation of the spectral energy intensities of each pixel based on the spectral energy weight to obtain a spectral energy intensity feature value; extracting the temporal information of each pixel from the target feature map, determining the temporal weight of each pixel, and performing a weighted summation of the temporal information of each pixel based on the temporal weight to obtain a temporal information feature value; and extracting feature values ​​in the length direction from the target feature map to obtain multiple channel feature values, which together constitute a feature vector.

[0060] For example, if the first, second, and third feature maps are all 512*32*32, then concatenating them along their length results in a target feature map of 1536*32*32. Here, 1536 represents the number of feature values ​​along the length, which is the number of channel feature values ​​after expanding the R, G, and B channels of the three spectrograms, and 32*32 represents the size of the two-dimensional feature map.

[0061] The spectral energy intensity of each pixel is extracted from the 32*32 two-dimensional feature map. The spectral energy weight of each pixel is determined based on its intensity, with higher weights assigned to higher intensities and lower weights to lower intensities. The spectral energy intensities of all pixels are then weighted and summed to obtain a spectral energy intensity feature value. Temporal information of each pixel is also extracted from the 32*32 two-dimensional feature map. The temporal information weight of each pixel is determined based on its proximity to the target time (landing time, takeoff time, and standing time), with higher weights assigned to time information closer to the target time and lower weights to time information farther from the target time. The temporal information of all pixels is then weighted and summed to obtain a temporal information feature value.

[0062] Furthermore, a 1536*1*1 feature vector is constructed from 1536 channel feature values, spectral energy intensity feature values, and time information feature values. In this embodiment, the target feature map is processed using a self-attention mechanism to increase the weight of high-energy spectral information and indicate the weight of time information that is close to the target time. This allows the extracted feature vector to more accurately represent the force situation of the lower limbs after fusion in the X, Y, and Z directions during running.

[0063] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0064] This application also provides a wearable device. Figure 4 This is a schematic diagram of a wearable device according to an embodiment of this application, such as... Figure 4 As shown, the wearable device includes:

[0065] The inertial measurement unit 401 is used to acquire X-direction signals, Y-direction signals and Z-direction signals.

[0066] Specifically, the inertial measurement unit 401 may include a three-axis accelerometer and a three-axis gyroscope for measuring the acceleration and angular velocity of the object being measured in the X, Y, and Z directions. In scenarios with a risk of knee joint injury, the wearable device may be placed on the knee or instep of the running user, in which case the object being measured is the user's foot or knee.

[0067] The controller 402, connected to the inertial measurement unit 401, is used to acquire the signals collected by the inertial measurement unit 401 and execute the method for determining the risk of knee joint injury in any of the above embodiments.

[0068] Specifically, the controller 402 stores the model parameters of the pre-trained target classification model. The controller 402 extracts the signals between the ground contact time and the ground departure time from the X-direction signal, Y-direction signal, and Z-direction signal, respectively, to obtain the first signal, the second signal, and the third signal. Features are extracted from the first signal, the second signal, and the third signal to obtain the first feature map, the second feature map, and the third feature map. The first feature map, the second feature map, and the third feature map are converted into feature vectors. The feature vectors are then input into the target classification model. The user's knee joint injury risk is determined based on the classification result. If the classification result is 1, it means that the user's knee joint has an injury risk. If the classification result is 0, it means that the user's knee joint has no injury risk.

[0069] This application also provides a device for determining the risk of knee joint injury. It should be noted that this device can be used to execute the method for determining the risk of knee joint injury provided in this application. The following describes the device for determining the risk of knee joint injury provided in this application.

[0070] Figure 5 This is a schematic diagram of a device for determining the risk of knee joint injury according to an embodiment of this application. Figure 5 As shown, the device includes: an acquisition unit 502, an extraction unit 504, a conversion unit 506, and a processing unit 508.

[0071] The acquisition unit 502 is used to acquire the X-direction signal, Y-direction signal and Z-direction signal collected by the inertial measurement unit worn by the user on the lower limbs.

[0072] Extraction unit 504 is used to extract the signal between the ground contact time and the ground departure time from the X-direction signal, Y-direction signal and Z-direction signal respectively, to obtain the first signal, the second signal and the third signal.

[0073] The conversion unit 506 is used to extract features from the first signal, the second signal and the third signal respectively to obtain a first feature map, a second feature map and a third feature map, and convert the first feature map, the second feature map and the third feature map into feature vectors.

[0074] The processing unit 508 is used to input the feature vector into the target classification model and process it to obtain the user's knee joint injury risk result. The target classification model is trained based on multiple sets of sample data, and each set of sample data includes a sample feature vector and a knee joint injury label.

[0075] The knee joint injury risk determination device provided in this application embodiment acquires X-direction, Y-direction, and Z-direction signals collected by an inertial measurement unit worn by the user on the lower limb through an acquisition unit 502; an extraction unit 504 extracts signals between the ground contact time and the ground departure time from the X-direction, Y-direction, and Z-direction signals respectively to obtain a first signal, a second signal, and a third signal; a conversion unit 506 extracts features from the first signal, the second signal, and the third signal respectively to obtain a first feature map, a second feature map, and a third feature map, and converts the first feature map, the second feature map, and the third feature map into feature vectors; and a processing unit 508... The feature vector is input into the target classification model, which is then processed to obtain the user's knee joint injury risk result. The target classification model is trained based on multiple sets of sample data. Each set of sample data includes a sample feature vector and a knee joint injury label. This solves the problem of inaccurate judgment results in related technologies that judge the size of knee joint injury risk by analyzing the impact of ground reaction force on the knee joint. Features are extracted from the X-direction, Y-direction, and Z-direction signals collected by the inertial measurement unit worn by the user's lower limbs to form a feature vector. The knee joint injury risk result is determined based on the feature vector and the pre-trained model, thereby achieving the effect of accurately judging the size of the knee joint injury risk.

[0076] Optionally, in the knee joint injury risk determination device provided in this application embodiment, the conversion unit 506 includes: a transformation module, used to perform short-time Fourier transform on the first signal, the second signal, and the third signal respectively to obtain a first spectrum map, a second spectrum map, and a third spectrum map; an extraction module, used to perform convolutional feature extraction on the first spectrum map, the second spectrum map, and the third spectrum map respectively through a convolutional neural network to obtain a first feature map, a second feature map, and a third feature map; and a splicing module, used to splice the first feature map, the second feature map, and the third feature map to obtain a target feature map, and determine a feature vector based on the target feature map.

[0077] Optionally, in the knee joint injury risk determination device provided in the embodiments of this application, the first spectrum, the second spectrum, and the third spectrum all include two-dimensional spectrums on the R, G, and B channels. The height of the two-dimensional spectrum of each channel represents the amplitude of the energy density, and the width of the two-dimensional spectrum of each channel represents the time. The extraction module is used to: compress the features in the width direction and the features in the length direction of the first spectrum, the second spectrum, and the third spectrum, respectively, and expand the features in the R, G, and B channels in the length direction to obtain the first feature map, the second feature map, and the third feature map.

[0078] Optionally, in the knee joint injury risk determination device provided in this application embodiment, the stitching module includes: a first determination submodule, used to extract the spectral energy intensity of each pixel from the target feature map, determine the spectral energy weight of each pixel, and perform a weighted summation of the spectral energy intensity of each pixel according to the spectral energy weight to obtain a spectral energy intensity feature value; a second determination submodule, used to extract the temporal information of each pixel from the target feature map, determine the temporal weight of each pixel, and perform a weighted summation of the temporal information of each pixel according to the temporal weight to obtain a temporal information feature value; and a third determination submodule, used to extract feature values ​​in the length direction from the target feature map to obtain multiple channel feature values, and a feature vector is composed of multiple channel feature values, spectral energy intensity feature values, and temporal information feature values.

[0079] Optionally, in the knee joint injury risk determination device provided in this application embodiment, the target classification model is a distributed gradient enhancement model. The device further includes a model training unit, which includes: an acquisition module, used to acquire sample feature vectors and knee joint injury labels of multiple sample users participating in running training from a user database, wherein the sample feature vectors are determined based on signals collected by inertial measurement units worn by the lower limbs of the sample users, and the knee joint injury labels include injury labels and non-injury labels; a sample data determination module, used to determine the sample feature vector and knee joint injury label of each sample user as a sample data, thereby obtaining multiple sample data; and a training module, used to train a preset distributed gradient enhancement model based on the multiple sample data to obtain a target classification model.

[0080] Optionally, in the knee joint injury risk determination device provided in this application embodiment, the acquisition module includes: a first acquisition submodule, used to acquire the X-direction signal, Y-direction signal and Z-direction signal collected by the inertial measurement unit of each sample user, and generate a sample feature vector based on the signal between the ground contact time and the ground departure time in the X-direction signal, Y-direction signal and Z-direction signal; a second acquisition submodule, used to acquire the interruption reason information of each sample user's running training plan, determine the knee joint injury data from the interruption reason information, and generate a knee joint injury label based on the knee joint injury data.

[0081] Optionally, in the knee injury risk determination device provided in this application embodiment, the inertial measurement unit is worn at the user's knee or instep position.

[0082] The aforementioned device for determining the risk of knee joint injury includes a processor and a memory. The aforementioned acquisition unit 502, extraction unit 504, conversion unit 506, and processing unit 508 are all stored in the memory as program units. The processor executes the aforementioned program units stored in the memory to achieve the corresponding functions.

[0083] The processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured. By adjusting kernel parameters, the inaccurate results obtained in related technologies that rely on analyzing the impact of ground reaction force on the knee joint to determine the risk of knee injury can be addressed.

[0084] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0085] This application also provides a computer storage medium for storing a program, wherein the program, when running, controls the device where the non-volatile storage medium is located to execute a method for determining the risk of knee joint injury.

[0086] This application also provides an electronic device. Figure 6 This is a schematic diagram of an electronic device according to an embodiment of this application. The electronic device 60 includes a processor and a memory; the memory stores computer-readable instructions, and the processor is used to execute the computer-readable instructions, wherein the computer-readable instructions, when executed, perform a method for determining the risk of knee joint injury. The electronic device in this document may be a server, PC, PAD, mobile phone, etc.

[0087] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0088] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0089] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0090] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0091] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0092] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0093] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0094] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0095] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for determining the risk of knee joint injury, characterized in that, include: Acquire the X-direction, Y-direction, and Z-direction signals collected by the inertial measurement unit worn by the user on the lower limbs; The signals between the ground contact time and the ground departure time are extracted from the X-direction signal, the Y-direction signal, and the Z-direction signal, respectively, to obtain the first signal, the second signal, and the third signal; Features are extracted from the first signal, the second signal, and the third signal to obtain a first feature map, a second feature map, and a third feature map, respectively. These features are then converted into feature vectors. The process includes: performing short-time Fourier transforms on the first signal, the second signal, and the third signal to obtain a first spectrum map, a second spectrum map, and a third spectrum map; performing convolutional feature extraction on the first spectrum map, the second spectrum map, and the third spectrum map using a convolutional neural network to obtain the first feature map, the second feature map, and the third feature map; concatenating the first feature map, the second feature map, and the third feature map to obtain a target feature map; and determining the feature vector based on the target feature map. The first, second, and third spectrograms each contain two-dimensional spectrograms on the R, G, and B channels. The height of the two-dimensional spectrogram for each channel represents the amplitude of the energy density, and the width of the two-dimensional spectrogram for each channel represents time. Convolutional feature extraction is performed on the first, second, and third spectrograms to obtain the first, second, and third feature maps, respectively. This includes: performing dimensionality compression on the features in the width direction and the features in the length direction of the first, second, and third spectrograms, respectively; and performing dimensionality expansion on the features of the R, G, and B channels in the length direction, to obtain the first, second, and third feature maps. The feature vector is input into the target classification model and processed to obtain the user's knee joint injury risk result. The target classification model is trained based on multiple sets of sample data, each set of sample data including a sample feature vector and a knee joint injury label.

2. The method according to claim 1, characterized in that, The first feature map, the second feature map, and the third feature map are concatenated to obtain the target feature map. The feature vector is determined based on the target feature map, including: The first feature map, the second feature map, and the third feature map are concatenated along the length direction to obtain the target feature map; Extract the spectral energy intensity of each pixel from the target feature map, determine the spectral energy weight of each pixel, and perform a weighted sum of the spectral energy intensities of each pixel based on the spectral energy weight to obtain the spectral energy intensity feature value. The temporal information of each pixel is extracted from the target feature map, the temporal weight of each pixel is determined, and the temporal information of each pixel is weighted and summed according to the temporal weight to obtain the temporal information feature value. Feature values ​​in the length direction are extracted from the target feature map to obtain multiple channel feature values. The feature vector is composed of the multiple channel feature values, the spectral energy intensity feature values, and the time information feature values.

3. The method according to claim 1, characterized in that, The target classification model is a distributed gradient boosting model, and the target classification model is determined in the following way: Sample feature vectors and knee injury labels of multiple sample users participating in running training are obtained from the user database. The sample feature vectors are determined based on signals collected by inertial measurement units worn by the sample users' lower limbs, and the knee injury labels include injury labels and non-injury labels. Each sample user's sample feature vector and knee injury label are used to determine one sample data point, resulting in multiple sample data points. The target classification model is obtained by training a pre-defined distributed gradient enhancement model based on the multiple sample data.

4. The method according to claim 3, characterized in that, The sample feature vectors and knee injury labels of multiple users who participated in running training were obtained from the user database, including: The X-direction signal, Y-direction signal, and Z-direction signal collected by the inertial measurement unit of each sample user are acquired, and a sample feature vector is generated based on the signal between the ground contact time and the ground departure time in the X-direction signal, the Y-direction signal, and the Z-direction signal. Obtain information on the reasons for interruption of each sample user's running training plan, determine knee joint injury data from the interruption reasons information, and generate knee joint injury tags based on the knee joint injury data.

5. The method according to any one of claims 1 to 4, characterized in that, The inertial measurement unit is worn on the user's knee or the back of their foot.

6. A wearable device, characterized in that, include: An inertial measurement unit is used to acquire signals in the X, Y, and Z directions. A controller, connected to the inertial measurement unit, is used to acquire signals collected by the inertial measurement unit and execute the method for determining the risk of knee joint injury as described in any one of claims 1 to 5.

7. A device for determining the risk of knee joint injury, characterized in that, include: The acquisition unit is used to acquire the X-direction signal, Y-direction signal, and Z-direction signal collected by the inertial measurement unit worn by the user on the lower limbs; The extraction unit is used to extract the signal between the ground contact time and the ground departure time from the X-direction signal, the Y-direction signal and the Z-direction signal respectively, to obtain a first signal, a second signal and a third signal; The conversion unit is used to extract features from the first signal, the second signal, and the third signal respectively to obtain a first feature map, a second feature map, and a third feature map, and to convert the first feature map, the second feature map, and the third feature map into feature vectors; The processing unit is used to input the feature vector into the target classification model and process it to obtain the user's knee joint injury risk result. The target classification model is trained based on multiple sets of sample data, and each set of sample data includes a sample feature vector and a knee joint injury label. The conversion unit includes: a transformation module, used to perform short-time Fourier transform on the first signal, the second signal, and the third signal respectively to obtain a first spectrum map, a second spectrum map, and a third spectrum map; an extraction module, used to perform convolutional feature extraction on the first spectrum map, the second spectrum map, and the third spectrum map respectively through a convolutional neural network to obtain a first feature map, a second feature map, and a third feature map; and a concatenation module, used to concatenate the first feature map, the second feature map, and the third feature map to obtain a target feature map, and determine the feature vector based on the target feature map; The first, second, and third spectrograms each contain two-dimensional spectrograms on the R, G, and B channels. The height of the two-dimensional spectrogram of each channel represents the amplitude of the energy density, and the width of the two-dimensional spectrogram of each channel represents the time. The extraction module is used to: compress the features in the width direction and the features in the length direction of the first, second, and third spectrograms respectively, and expand the features of the R, G, and B channels in the length direction to obtain the first feature map, the second feature map, and the third feature map.

8. A computer storage medium, characterized in that, The computer storage medium is used to store a program, wherein when the program runs, it controls the device where the computer storage medium is located to execute the method for determining the risk of knee joint injury as described in any one of claims 1 to 5.

9. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to execute the method for determining the risk of knee injury as described in any one of claims 1 to 5 through the computer program.

Images