Underwater treadmill underwater motion risk identification method, device and equipment

Abstract

The present invention discloses a method, device, and equipment for identifying underwater exercise risks for an underwater treadmill. The method comprises: obtaining a trainee's current exercise state data on the underwater treadmill and the current water flow velocity data of the underwater treadmill; the exercise state data comprises: an image of the trainee's exercise posture and heart rate; performing feature extraction on the exercise posture image and the heart rate, respectively, to obtain corresponding exercise posture features and heart rate features of the trainee underwater, and extracting water flow velocity features from the water flow velocity data; based on the exercise posture features, the heart rate features, and the water flow velocity features, and using a pre-trained underwater exercise risk identification model, identifying the underwater exercise risk type of the trainee on the underwater treadmill. The present invention can improve the accuracy of underwater exercise risk assessment for underwater treadmills.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, device and equipment for identifying underwater motion risks of an underwater treadmill. Background Art

[0002] In the field of underwater sports, especially when using underwater treadmills, athletes face various potential risks, such as muscle strain caused by incorrect posture, excessive fatigue, discomfort caused by rapid water flow, and even the risk of drowning. Currently, some underwater sports monitoring systems on the market mainly focus on monitoring a single indicator, such as monitoring the athlete's heart rate through a heart rate belt or built-in sensor to determine whether the exercise intensity is too high, but lack consideration of exercise posture and water flow environment. Alternatively, they measure whether the water flow speed around the underwater treadmill is greater than the safety threshold to determine whether there is an underwater sports risk. Although these methods can provide certain exercise data, they have limitations in identifying underwater sports risks for underwater treadmills and cannot comprehensively evaluate underwater sports risks by integrating multiple factors. Summary of the Invention

[0003] Embodiments of the present invention provide a method, device, and equipment for identifying underwater motion risks of an underwater treadmill, which can improve the accuracy of underwater motion risk assessment of an underwater treadmill.

[0004] An embodiment of the present invention provides a method for identifying underwater exercise risks of an underwater treadmill, comprising:

[0005] Acquiring the trainee's current motion state data on the underwater treadmill and the current water flow speed data of the underwater treadmill; the motion state data includes: the trainee's motion posture image and heart rate;

[0006] Performing feature extraction on the motion posture image and the heart rate respectively to obtain corresponding motion posture features and heart rate features of the trainee underwater, and extracting water flow velocity features from the water flow velocity data;

[0007] Based on the motion posture characteristics, the heart rate characteristics and the water flow velocity characteristics, and through a pre-trained underwater motion risk identification model, the underwater motion risk type of the trainee on the underwater treadmill is identified.

[0008] As an improvement to the above solution, the method of identifying the underwater exercise risk type of the trainee on the underwater treadmill based on the exercise posture characteristics, the heart rate characteristics, and the water flow velocity characteristics and using a pre-trained underwater exercise risk identification model includes:

[0009] A spatiotemporal attention mechanism is introduced to perform temporal attention calculation on the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series, and a modal attention calculation on the correlation between the three features;

[0010] Performing feature fusion on the motion posture features, the heart rate features, and the water flow velocity features that have been subjected to attention calculation and modal attention calculation to obtain a fused underwater motion risk feature;

[0011] Based on the fused underwater motion risk features and through a pre-trained underwater motion risk identification model, the underwater motion risk type of the trainee on the underwater treadmill is identified.

[0012] As an improvement to the above solution, the spatiotemporal attention mechanism is introduced to perform temporal attention calculation on the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series and modal attention calculation on the correlation between the three, including:

[0013] The time attention calculation formula below is used to calculate the correlation between the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series:

[0014] ,in, For the time step The attention weight reflects the importance of the feature at this time step; is a learnable temporal attention weight vector used to calculate the temporal attention weight; For the time step The LSTM hidden state on , which is the internal state of the LSTM unit at that point in time, contains the cumulative effect of all previous time steps; Is an index variable used to traverse the length of the entire time series ;

[0015] The modal attention calculation formula is used to calculate the modal attention of the correlation between the three:

[0016] ,in, The first of the three The attention weight of a modality reflects the importance of the modality in the entire recognition process; is a learnable modal attention weight vector used to calculate the modal attention weight; The first of the three The attention vector of the modality is obtained after the modal features of the three are linearly transformed and is used to calculate the modal attention; is an index variable used to traverse the three; the features of the three modes of the motion posture feature, the heart rate feature and the water flow velocity feature are respectively expressed as , and .

[0017] As an improvement to the above solution, the motion posture features, the heart rate features, and the water flow velocity features that have been calculated through attention and modal attention are fused to obtain the fused underwater motion risk features, including:

[0018] The motion posture features, the heart rate features, and the water velocity features that have been calculated through attention and modal attention are input into the following feature fusion formula to obtain the fused underwater motion risk features:

[0019] ,in, is the fused feature vector; is the mode among the three at time step The eigenvector of and They are respectively the temporal attention weight and modal attention weight calculated by the three.

[0020] As an improvement to the above solution, the underwater motion risk type of the trainee on the underwater treadmill is identified based on the fused underwater motion risk features and a pre-trained underwater motion risk identification model, including:

[0021] The fused underwater motion risk features are obtained through the fully connected layer of the underwater motion risk recognition model which is a pre-trained deep learning model. Perform linear transformation to obtain linear transformation output. The linear transformation formula of the fully connected layer is: ;in, is a Matrix of is the number of output nodes, i.e. the number of risk categories; is the dimension of the fused feature vector; is a dimensional bias vector, each element corresponds to the bias value of an output node; is the linear transformation output, which is a dimensional vector;

[0022] The linear transformation output is calculated using the Softmax function of the underwater exercise risk identification model to obtain the underwater exercise risk probability of the trainee on the underwater treadmill; the Softmax function is: ;in, It is The predicted probability of each category; yes The corresponding vector Elements of the category; Refers to the number of all possible risk types; is an index variable used to traverse all risk categories;

[0023] The underwater sports risk probability is predicted into a risk type through the output layer of the underwater sports risk identification model to obtain the underwater sports risk type of the trainee on the underwater treadmill; wherein the risk type predicted by the output layer of the underwater sports risk identification model is the category corresponding to the maximum probability.

[0024] Another embodiment of the present invention provides an underwater exercise risk identification device for an underwater treadmill, comprising:

[0025] An acquisition module is used to acquire the current motion state data of the trainee on the underwater treadmill and the current water flow speed data of the underwater treadmill; the motion state data includes: the motion posture image and heart rate of the trainee;

[0026] a feature extraction module, configured to extract features from the motion posture image and the heart rate, respectively, to obtain the motion posture features and heart rate features of the trainee underwater, and to extract water flow velocity features from the water flow velocity data;

[0027] The risk identification module is used to identify the underwater exercise risk type of the trainee on the underwater treadmill based on the movement posture characteristics, the heart rate characteristics and the water flow speed characteristics, and through a pre-trained underwater exercise risk identification model.

[0028] As an improvement to the above solution, the risk identification module includes:

[0029] a calculation submodule, configured to introduce a spatiotemporal attention mechanism to perform temporal attention calculation on the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series, and to perform modal attention calculation on the correlation between the three features;

[0030] a fusion submodule, configured to perform feature fusion on the motion posture features, the heart rate features, and the water flow velocity features obtained through attention calculation and modal attention calculation to obtain a fused underwater motion risk feature;

[0031] The identification submodule is used to identify the underwater motion risk type of the trainee on the underwater treadmill based on the fused underwater motion risk features and a pre-trained underwater motion risk identification model.

[0032] As an improvement to the above solution, the calculation submodule is specifically configured to:

[0033] The time attention calculation formula below is used to calculate the correlation between the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series:

[0034] ,in, For the time step The attention weight reflects the importance of the feature at this time step; is a learnable temporal attention weight vector used to calculate the temporal attention weight; For the time step The LSTM hidden state on , which is the internal state of the LSTM unit at that point in time, contains the cumulative effect of all previous time steps; Is an index variable used to traverse the length of the entire time series ;

[0035] The modal attention calculation formula is used to calculate the modal attention of the correlation between the three:

[0036] ,in, The first of the three The attention weight of a modality reflects the importance of the modality in the entire recognition process; is a learnable modal attention weight vector used to calculate the modal attention weight; The first of the three The attention vector of the modality is obtained after the modal features of the three are linearly transformed and is used to calculate the modal attention; is an index variable used to traverse the three; the features of the three modes of the motion posture feature, the heart rate feature and the water flow velocity feature are respectively expressed as , and .

[0037] As an improvement to the above solution, the fusion submodule is specifically used for:

[0038] The motion posture features, the heart rate features, and the water velocity features that have been calculated through attention and modal attention are input into the following feature fusion formula to obtain the fused underwater motion risk features:

[0039] ,in, is the fused feature vector; is the mode among the three at time step The eigenvector of and They are respectively the temporal attention weight and modal attention weight calculated by the three.

[0040] Another embodiment of the present invention provides an underwater motion risk identification device for an underwater treadmill, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the underwater motion risk identification method for the underwater treadmill described in the above-mentioned embodiment of the invention.

[0041] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0042] The embodiment of the present invention first collects multimodal data of the trainee on the underwater treadmill, including motion posture images, heart rate and water flow speed. Subsequently, through advanced feature extraction technology, the motion posture features, heart rate features and water flow speed features are respectively parsed, and these features together constitute a composite motion state description. Finally, with the help of a pre-trained deep learning model, these features are fused and analyzed to identify potential underwater motion risk types, such as improper posture, excessive fatigue or too fast water flow. It can be seen that the embodiment of the present invention significantly improves the accuracy and real-time performance of underwater motion risk assessment by integrating motion posture, physiological indicators and environmental parameters, and performing risk identification through an underwater motion risk identification model, effectively making up for the shortcomings of traditional monitoring systems in underwater motion risk assessment, and improving the safety and exercise experience of trainees when using underwater treadmills. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a flow chart of a method for identifying underwater exercise risks of an underwater treadmill provided by one embodiment of the present invention;

[0044] Figure 2 This is a schematic structural diagram of an underwater exercise risk identification device for an underwater treadmill provided by one embodiment of the present invention;

[0045] Figure 3 The present invention is a schematic structural diagram of an underwater sports risk identification device for an underwater treadmill provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0046] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0047] See also Figure 1 , is a flow chart of a method for identifying underwater exercise risks for an underwater treadmill provided by one embodiment of the present invention. The method for identifying underwater exercise risks for an underwater treadmill comprises steps S10 to S12:

[0048] S10, obtaining the trainee's current motion state data on the underwater treadmill and the current water flow speed data of the underwater treadmill; the motion state data includes: the trainee's motion posture image and heart rate;

[0049] S11, performing feature extraction on the motion posture image and the heart rate, respectively, to obtain corresponding motion posture features and heart rate features of the trainee underwater, and extracting water flow velocity features from the water flow velocity data;

[0050] S12, based on the movement posture characteristics, the heart rate characteristics and the water flow speed characteristics, and through a pre-trained underwater movement risk identification model, identify the underwater movement risk type of the trainee on the underwater treadmill.

[0051] The embodiment of the present invention first collects multimodal data of the trainee on the underwater treadmill, including motion posture images, heart rate and water flow speed. Subsequently, through advanced feature extraction technology, the motion posture features, heart rate features and water flow speed features are respectively parsed, and these features together constitute a composite motion state description. Finally, with the help of a pre-trained deep learning model, these features are fused and analyzed to identify potential underwater motion risk types, such as improper posture, excessive fatigue or too fast water flow. It can be seen that the embodiment of the present invention significantly improves the accuracy and real-time performance of underwater motion risk assessment by integrating motion posture, physiological indicators and environmental parameters, and performing risk identification through an underwater motion risk identification model, effectively making up for the shortcomings of traditional monitoring systems in underwater motion risk assessment, and improving the safety and exercise experience of trainees when using underwater treadmills.

[0052] It is understandable that the feature extraction algorithm adopted for extracting the features of the motion posture image, the heart rate and the water flow velocity can refer to the existing technology and will not be described in detail here. In addition, the motion posture image of the trainee can be obtained by underwater shooting with a camera installed on the underwater treadmill, and the heart rate of the trainee can be detected and sent by a smart device or heart rate monitoring device worn by the trainee. The current water flow velocity data of the underwater treadmill can be detected by a water flow meter placed at the water outlet of the underwater treadmill.

[0053] Among them, the risk types of underwater sports can be excessive fatigue, improper posture, uncomfortable water flow, etc.

[0054] As one embodiment, the identifying of the underwater exercise risk type of the trainee on the underwater treadmill based on the exercise posture feature, the heart rate feature, and the water flow velocity feature and using a pre-trained underwater exercise risk identification model includes:

[0055] A spatiotemporal attention mechanism is introduced to perform temporal attention calculation on the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series, and a modal attention calculation on the correlation between the three features;

[0056] Performing feature fusion on the motion posture features, the heart rate features, and the water flow velocity features that have been subjected to attention calculation and modal attention calculation to obtain a fused underwater motion risk feature;

[0057] Based on the fused underwater motion risk features and through a pre-trained underwater motion risk identification model, the underwater motion risk type of the trainee on the underwater treadmill is identified.

[0058] In this embodiment, temporal attention weights for motion posture, heart rate, and water velocity features are first independently calculated for each moment in the time series. This helps the model identify which features during the motion process are crucial for risk assessment. Next, modal attention calculations are used to further analyze the interactions between the three features, emphasizing the relative importance of different data sources in overall risk identification and ensuring that the model understands the inherent connections between modal features and their contributions to risk types. After the attention mechanism is calculated, the weighted features are fused to generate a comprehensive underwater motion risk feature representation. This representation integrates key information from the time series and the correlations between multimodal data, enhancing the model's recognition and generalization capabilities. Finally, the fused features are processed using a pre-trained underwater motion risk identification model to identify the risk type of the exerciser on the underwater treadmill, achieving accurate early warning of potential motion risks. Therefore, by introducing a spatiotemporal attention mechanism and feature fusion, this embodiment improves the accuracy and real-time response capabilities of underwater motion risk identification, effectively avoiding the false positives and false negatives often associated with single indicators or simple threshold judgments in traditional monitoring systems. In addition, it can provide personalized risk assessments based on the characteristics of different athletes and sports situations, thereby providing strong technical support for the safety management of underwater sports.

[0059] As one embodiment, the introduction of the spatiotemporal attention mechanism performs temporal attention calculation on the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series and modal attention calculation on the correlation between the three, including:

[0060] The time attention calculation formula below is used to calculate the correlation between the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series:

[0061] ,in, For the time step The attention weight reflects the importance of the feature at this time step; is a learnable temporal attention weight vector used to calculate the temporal attention weight; For the time step The LSTM hidden state on , which is the internal state of the LSTM unit at that point in time, contains the cumulative effect of all previous time steps; Is an index variable used to traverse the length of the entire time series ;

[0062] The modal attention calculation formula is used to calculate the modal attention of the correlation between the three:

[0063] ,in, The first of the three The attention weight of a modality reflects the importance of the modality in the entire recognition process; is a learnable modal attention weight vector used to calculate the modal attention weight; The first of the three The attention vector of the modality is obtained after the modal features of the three are linearly transformed and is used to calculate the modal attention; is an index variable used to traverse the three; the features of the three modes of the motion posture feature, the heart rate feature and the water flow velocity feature are respectively expressed as , and .

[0064] In this embodiment, a spatiotemporal attention mechanism is used to finely weight time series data from different modalities (movement posture, heart rate, and water velocity) to improve the accuracy and reliability of underwater sports risk identification. Specifically, a temporal attention formula is first used to weight the LSTM hidden state at each time step. The LSTM (Long Short-Term Memory) hidden state contains the accumulated information of historical data, while the temporal attention weight reflects the contribution of the features at a specific time point to the overall prediction task. By learning the temporal attention weight vector, the model can dynamically adjust the weights at different time points, focusing more on those moments that are critical for risk identification. Subsequently, a modal attention formula is used to quantify the relative importance of different data sources (i.e., modalities) in the overall risk assessment. The features of each modality are linearly transformed to form an attention vector. These vectors are multiplied by a learnable modal attention weight vector to generate modal attention weights, which reveal which modal data is more critical in underwater sports risk identification. This allows the model to not only consider the strength of a single modality but also capture the correlations between different modalities, enhancing its ability to understand complex scenarios. From the above analysis, we can see that this embodiment improves the accuracy and robustness of the model in underwater motion risk identification through a sophisticated temporal and modal attention mechanism. This enables the model to automatically focus on the most relevant spatiotemporal features, reduce the impact of redundant information, and ensure comprehensive consideration of multivariate data. It can maintain stable recognition performance even in the face of highly variable and uncertain underwater environments.

[0065] As an improvement to the above solution, the motion posture features, the heart rate features, and the water flow velocity features that have been calculated through attention and modal attention are fused to obtain the fused underwater motion risk features, including:

[0066] The motion posture features, the heart rate features, and the water velocity features that have been calculated through attention and modal attention are input into the following feature fusion formula to obtain the fused underwater motion risk features:

[0067] ,in, is the fused feature vector; is the mode among the three at time step The eigenvector of and They are respectively the temporal attention weight and modal attention weight calculated by the three.

[0068] In this embodiment, the innovative feature fusion strategy described above integrates motion posture features, heart rate features, and water velocity features optimized through the attention mechanism into a unified, more expressive underwater motion risk signature. Specifically, an innovative weighted fusion approach is employed, combining the feature vectors of each modality at different time points with their corresponding temporal attention weights and modal attention weights. This ensures that the contribution of each feature in the final fused signature reflects both its dynamic importance in the time series and its global value as a specific type of information. This feature fusion process effectively aggregates the advantages of multimodal data, overcoming the limitations and noise effects of individual features. The resulting fused feature vector more comprehensively and accurately characterizes the underwater motion risk status. Furthermore, by utilizing temporal attention weights, the model can highlight feature changes at key time points, while modal attention weights ensure that different types of monitoring data are appropriately weighted in the overall risk assessment, avoiding the over-dominance of any one feature or the omission of information. Therefore, the fused underwater motion risk characteristics not only capture the instantaneous changes of individual motion behaviors, physiological reactions and environmental conditions, but also fully consider the interactions between them, thereby improving the accuracy and reliability of risk prediction.

[0069] As an improvement to the above solution, the underwater motion risk type of the trainee on the underwater treadmill is identified based on the fused underwater motion risk features and a pre-trained underwater motion risk identification model, including:

[0070] The fused underwater motion risk features are obtained through the fully connected layer of the underwater motion risk recognition model which is a pre-trained deep learning model. Perform linear transformation to obtain linear transformation output. The linear transformation formula of the fully connected layer is: ;in, is a Matrix of is the number of output nodes, i.e. the number of risk categories; is the dimension of the fused feature vector; is a dimensional bias vector, each element corresponds to the bias value of an output node; is the linear transformation output, which is a dimensional vector;

[0071] The linear transformation output is calculated using the Softmax function of the underwater exercise risk identification model to obtain the underwater exercise risk probability of the trainee on the underwater treadmill; the Softmax function is: ;in, It is The predicted probability of each category; yes The corresponding vector Elements of the category; Refers to the number of all possible risk types; is an index variable used to traverse all risk categories;

[0072] The underwater sports risk probability is predicted into a risk type through the output layer of the underwater sports risk identification model to obtain the underwater sports risk type of the trainee on the underwater treadmill; wherein the risk type predicted by the output layer of the underwater sports risk identification model is the category corresponding to the maximum probability.

[0073] In this embodiment, by utilizing a pre-trained deep learning model—the underwater sports risk identification model—the fused underwater sports risk features are efficiently parsed and risk types identified. The fused features are linearly transformed through the model's fully connected layer. This process maps the complex, high-dimensional feature space into a lower-dimensional decision space, where each dimension corresponds to a potential underwater sports risk category. Through matrix multiplication and the addition of bias terms, the model can learn the association between different feature combinations and specific risk types, forming a linear representation that is crucial for risk identification. Furthermore, the linear output after the fully connected layer conversion is further processed by the Softmax function, converting the linearly transformed values ​​into a normalized probability distribution that intuitively reflects the likelihood of each risk type occurring. The Softmax function ensures that the sum of the output probabilities is 1, making it easier to select the most likely risk type from multiple possible risk types. Ultimately, the model's output layer determines the specific risk type of the trainee on the underwater treadmill based on the principle of maximum probability.

[0074] For example, use the Softmax function to transform the linear transformation output Z to obtain the predicted probability of each risk type. Assume that the four elements of Z are [2.0, 1.5, 1.0, 0.5]. The Softmax function outputs vector P. Find the index corresponding to the element with the highest probability, which is the predicted risk type. For example, if the Softmax output vector P is [0.45, 0.25, 0.20, 0.10], the maximum probability is 0.45, and the corresponding index is 0. Therefore, the model predicts the risk type as excessive fatigue.

[0075] See also Figure 2 , is a schematic diagram of the structure of an underwater exercise risk identification device for an underwater treadmill provided by one embodiment of the present invention. The underwater exercise risk identification device for an underwater treadmill comprises:

[0076] The acquisition module 10 is used to acquire the current motion state data of the trainee on the underwater treadmill and the current water flow speed data of the underwater treadmill; the motion state data includes: the motion posture image and heart rate of the trainee;

[0077] a feature extraction module 11 for extracting features from the motion posture image and the heart rate, respectively, to obtain the motion posture features and heart rate features of the trainee underwater, and extracting water flow velocity features from the water flow velocity data;

[0078] The risk identification module 12 is used to identify the underwater exercise risk type of the trainee on the underwater treadmill based on the movement posture characteristics, the heart rate characteristics and the water flow speed characteristics, and through a pre-trained underwater exercise risk identification model.

[0079] The embodiment of the present invention first collects multimodal data of the trainee on the underwater treadmill, including motion posture images, heart rate and water flow speed. Subsequently, through advanced feature extraction technology, the motion posture features, heart rate features and water flow speed features are respectively parsed, and these features together constitute a composite motion state description. Finally, with the help of a pre-trained deep learning model, these features are fused and analyzed to identify potential underwater motion risk types, such as improper posture, excessive fatigue or too fast water flow. It can be seen that the embodiment of the present invention significantly improves the accuracy and real-time performance of underwater motion risk assessment by integrating motion posture, physiological indicators and environmental parameters, and performing risk identification through an underwater motion risk identification model, effectively making up for the shortcomings of traditional monitoring systems in underwater motion risk assessment, and improving the safety and exercise experience of trainees when using underwater treadmills.

[0080] As an improvement to the above solution, the risk identification module includes:

[0081] a calculation submodule, configured to introduce a spatiotemporal attention mechanism to perform temporal attention calculation on the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series, and to perform modal attention calculation on the correlation between the three features;

[0082] a fusion submodule, configured to perform feature fusion on the motion posture features, the heart rate features, and the water flow velocity features obtained through attention calculation and modal attention calculation to obtain a fused underwater motion risk feature;

[0083] The identification submodule is used to identify the underwater motion risk type of the trainee on the underwater treadmill based on the fused underwater motion risk features and a pre-trained underwater motion risk identification model.

[0084] As an improvement to the above solution, the calculation submodule is specifically configured to:

[0085] The time attention calculation formula below is used to calculate the correlation between the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series:

[0086] ,in, For the time step The attention weight reflects the importance of the feature at this time step; is a learnable temporal attention weight vector used to calculate the temporal attention weight; For the time step The LSTM hidden state on , which is the internal state of the LSTM unit at that point in time, contains the cumulative effect of all previous time steps; Is an index variable used to traverse the length of the entire time series ;

[0087] The modal attention calculation formula is used to calculate the modal attention of the correlation between the three:

[0088] ,in, The first of the three The attention weight of a modality reflects the importance of the modality in the entire recognition process; is a learnable modal attention weight vector used to calculate the modal attention weight; The first of the three The attention vector of the modality is obtained after the modal features of the three are linearly transformed and is used to calculate the modal attention; is an index variable used to traverse the three; the features of the three modes of the motion posture feature, the heart rate feature and the water flow velocity feature are respectively expressed as , and .

[0089] As an improvement to the above solution, the fusion submodule is specifically used for:

[0090] The motion posture features, the heart rate features, and the water velocity features that have been calculated through attention and modal attention are input into the following feature fusion formula to obtain the fused underwater motion risk features:

[0091] ,in, is the fused feature vector; is the mode among the three at time step The eigenvector of and They are respectively the temporal attention weight and modal attention weight calculated by the three.

[0092] See also Figure 3 , is a schematic diagram of an underwater exercise risk identification device for an underwater treadmill provided in one embodiment of the present invention. The underwater exercise risk identification device for an underwater treadmill in this embodiment includes: a processor 100, a memory 101, and a computer program stored in the memory 101 and executable on the processor 100, such as an underwater exercise risk identification program for an underwater treadmill. When the processor 100 executes the computer program, it implements the steps of each of the aforementioned underwater exercise risk identification method embodiments for an underwater treadmill. Alternatively, when the processor 100 executes the computer program, it implements the functions of each module / unit in each of the aforementioned device embodiments.

[0093] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the underwater exercise risk identification device of the underwater treadmill.

[0094] The underwater treadmill's underwater exercise risk identification device can be a computing device such as a desktop computer, laptop, PDA, or cloud server. The underwater treadmill's underwater exercise risk identification device can include, but is not limited to, a processor and memory. Those skilled in the art will appreciate that the schematic diagram is merely an example of an underwater exercise risk identification device for an underwater treadmill and does not constitute a limitation on the underwater exercise risk identification device for an underwater treadmill. The underwater treadmill's underwater exercise risk identification device can include more or fewer components than shown, or a combination of certain components, or different components. For example, the underwater treadmill's underwater exercise risk identification device can also include input and output devices, network access devices, buses, and the like.

[0095] The processor may be a central processing unit (CPU), or other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the underwater sports risk identification device of the underwater treadmill, and utilizes various interfaces and lines to connect various parts of the underwater sports risk identification device of the entire underwater treadmill.

[0096] The memory can be used to store the computer programs and / or modules. The processor implements the various functions of the underwater treadmill's underwater exercise risk identification device by running or executing the computer programs and / or modules stored in the memory and accessing the data stored in the memory. The memory may primarily include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function (such as a sound playback function or an image playback function); the data storage area may store data generated based on the use of the mobile phone (such as audio data and a phone book). Furthermore, the memory may include high-speed random access memory (RAM) and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0097] If the module / unit integrated into the underwater treadmill's underwater exercise risk identification device is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention can also implement all or part of the process steps in the above-mentioned method embodiments by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a removable hard drive, a magnetic disk, an optical disk, computer memory, read-only memory (ROM), random access memory (RAM), an electrical carrier signal, a telecommunications signal, and software distribution media. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practices in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practices, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0098] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive effort.

[0099] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for identifying underwater exercise risks of an underwater treadmill, characterized in that: include: Acquiring the trainee's current motion state data on the underwater treadmill and the current water flow speed data of the underwater treadmill; The exercise state data includes: the exercise posture image and heart rate of the trainee; Performing feature extraction on the motion posture image and the heart rate respectively to obtain corresponding motion posture features and heart rate features of the trainee underwater, and extracting water flow velocity features from the water flow velocity data; Based on the motion posture characteristics, the heart rate characteristics, and the water flow velocity characteristics, and using a pre-trained underwater motion risk identification model, identifying the underwater motion risk type of the trainee on the underwater treadmill; The method of identifying the underwater exercise risk type of the trainee on the underwater treadmill based on the exercise posture characteristics, the heart rate characteristics, and the water flow speed characteristics and using a pre-trained underwater exercise risk identification model includes: A spatiotemporal attention mechanism is introduced to perform temporal attention calculation on the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series, and a modal attention calculation on the three features; Performing feature fusion on the motion posture feature, the heart rate feature, and the water flow velocity feature obtained through time attention calculation and modal attention calculation to obtain a fused underwater motion risk feature; Based on the fused underwater motion risk features and using a pre-trained underwater motion risk identification model, the underwater motion risk type of the trainee on the underwater treadmill is identified; The introduction of the spatiotemporal attention mechanism performs temporal attention calculation on the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series, and performs modal attention calculation on the three, including: The time attention calculation formula below is used to calculate the time attention of the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series: ,in, For the time step The attention weight reflects the importance of the feature at this time step; is a learnable temporal attention weight vector used to calculate the temporal attention weight; For the time step The LSTM hidden state on , which is the internal state of the LSTM unit at that point in time, contains the cumulative effect of all previous time steps; Is an index variable used to traverse the length of the entire time series ; The modal attention calculation formula is used to calculate the modal attention of the three: ,in, The first of the three The attention weight of a modality reflects the importance of the modality in the entire recognition process; is a learnable modal attention weight vector used to calculate the modal attention weight; The first of the three The attention vector of the modality is obtained after the modal features of the three are linearly transformed and is used to calculate the modal attention; is an index variable used to traverse the three; the modal features of the motion posture feature, the heart rate feature and the water flow velocity feature are respectively expressed as , and ; The step of performing feature fusion on the motion posture feature, the heart rate feature, and the water flow velocity feature obtained through time attention calculation and modal attention calculation to obtain the fused underwater motion risk feature includes: The temporal attention weight and modal attention weight calculated by the temporal attention calculation formula and the modal attention calculation formula are input into the following feature fusion formula to obtain the fused underwater motion risk feature: ,in, is the fused underwater motion risk characteristics; The first of the three modes at time step The eigenvector of and They are respectively the temporal attention weight and modal attention weight calculated by the three.

2. The underwater exercise risk identification method for an underwater treadmill according to claim 1, characterized in that: The underwater motion risk type of the trainee on the underwater treadmill is identified based on the fused underwater motion risk features and using a pre-trained underwater motion risk identification model, including: The fused underwater motion risk features are obtained through the fully connected layer of the underwater motion risk recognition model which is a pre-trained deep learning model. Perform linear transformation to obtain linear transformation output. The linear transformation formula of the fully connected layer is: ;in, is a Matrix of In the matrix is the number of output nodes, i.e. the number of risk categories; is the dimension of the fused feature vector; is a dimensional bias vector, each element corresponds to the bias value of an output node; is the linear transformation output, which is a dimensional vector; The linear transformation output is calculated using the Softmax function of the underwater exercise risk identification model to obtain the underwater exercise risk probability of the trainee on the underwater treadmill; the Softmax function is: ;in, It is The predicted probability of each category; yes The corresponding vector Elements of the category; the Softmax function refers to the number of risk categories; is an index variable used to traverse all risk categories; The underwater sports risk probability is predicted into a risk type through the output layer of the underwater sports risk identification model to obtain the underwater sports risk type of the trainee on the underwater treadmill; wherein the risk type predicted by the output layer of the underwater sports risk identification model is the category corresponding to the maximum probability.

3. An underwater treadmill underwater sports risk identification device, characterized in that: include: An acquisition module is used to acquire the current motion state data of the trainee on the underwater treadmill and the current water flow speed data of the underwater treadmill; The exercise state data includes: the exercise posture image and heart rate of the trainee; a feature extraction module, configured to extract features from the motion posture image and the heart rate, respectively, to obtain the motion posture features and heart rate features of the trainee underwater, and to extract water flow velocity features from the water flow velocity data; a risk identification module, configured to identify the underwater exercise risk type of the trainee on the underwater treadmill based on the exercise posture characteristics, the heart rate characteristics, and the water flow velocity characteristics and using a pre-trained underwater exercise risk identification model; The risk identification module includes: a calculation submodule, configured to introduce a spatiotemporal attention mechanism to perform temporal attention calculation on the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series, and perform modal attention calculation on the three features; a fusion submodule, configured to perform feature fusion on the motion posture features, the heart rate features, and the water flow velocity features obtained through time attention calculation and modal attention calculation to obtain a fused underwater motion risk feature; an identification submodule, configured to identify the underwater motion risk type of the trainee on the underwater treadmill based on the fused underwater motion risk features and a pre-trained underwater motion risk identification model; The calculation submodule is specifically used for: The time attention calculation formula below is used to calculate the time attention of the motion posture feature, the heart rate feature, and the water flow velocity feature at each moment in the time series: ,in, For the time step The attention weight reflects the importance of the feature at this time step; is a learnable temporal attention weight vector used to calculate the temporal attention weight; For the time step The LSTM hidden state on , which is the internal state of the LSTM unit at that point in time, contains the cumulative effect of all previous time steps; Is an index variable used to traverse the length of the entire time series ; The modal attention calculation formula is used to calculate the modal attention of the three: ,in, The first of the three The attention weight of a modality reflects the importance of the modality in the entire recognition process; is a learnable modal attention weight vector used to calculate the modal attention weight; The first of the three The attention vector of the modality is obtained after the modal features of the three are linearly transformed and is used to calculate the modal attention; is an index variable used to traverse the three; the modal features of the motion posture feature, the heart rate feature and the water flow velocity feature are respectively expressed as , and ; The fusion submodule is specifically used for: The temporal attention weight and modal attention weight calculated by the temporal attention calculation formula and the modal attention calculation formula are input into the following feature fusion formula to obtain the fused underwater motion risk feature: ,in, is the fused underwater motion risk characteristics; The first of the three modes at time step The eigenvector of and They are respectively the temporal attention weight and modal attention weight calculated by the three.

4. An underwater treadmill underwater sports risk identification device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for identifying underwater exercise risks of an underwater treadmill according to any one of claims 1 to 2 is implemented.

Images