Fitness action energy consumption assessment method, device, equipment and storage medium
Through multi-dimensional sensor group and correlation conversion model, the equipment bulkiness and insufficient real-time performance of existing fitness movement energy consumption assessment methods are solved, and more accurate and personalized fitness evaluation and suggestions are achieved, improving fitness effects and motivation.
Patent Information
- Application Number
- CN202311758388.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-20
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2043-12-20
AI Technical Summary
The existing energy consumption evaluation methods for fitness movements have problems such as bulky equipment, inconvenient use, difficulty in comprehensively capturing biological signals, insufficient accuracy of evaluation results and insufficient real-time performance, which limits the personalization and dynamic nature of fitness training.
A multi-dimensional sensor group, including biological signals such as heart rate, electromyography and body temperature, is used to establish signal energy correlation conversion models and strategy analysis models to analyze and predict users' biological signal data in real time, and provide personalized fitness suggestions and evaluations.
It improves the accuracy and real-time performance of energy consumption evaluation of fitness movements, enables users to instantly understand fitness effects, enhance fitness motivation, improve training efficiency and personalize planning.
Smart Images

Figure CN117672458B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of deep learning technology, and in particular to a method, device, equipment and storage medium for evaluating energy consumption of fitness movements. Background Art
[0002] With the popularity of fitness and sports, people's demand for scientific evaluation of fitness effects is increasing. Traditional methods for evaluating energy consumption of fitness movements mainly rely on equipment and sensors. Although these methods can provide a certain degree of accuracy, they face many problems. These devices are usually bulky and inconvenient to use, which limits the user's freedom in actual exercise. Secondly, due to the limitations of sensors, existing methods often find it difficult to fully and accurately capture the user's biological signals, resulting in insufficient accuracy of the evaluation results. Most notably, traditional methods have bottlenecks in real time, making it difficult to provide users with timely feedback during exercise, which restricts the personalization and dynamism of fitness training. This leads to the low accuracy of existing technologies. Summary of the invention
[0003] The present application provides a method, device, equipment and storage medium for evaluating energy consumption of fitness movements, which are used to improve the accuracy of energy consumption evaluation of fitness movements.
[0004] In a first aspect, the present application provides a method for evaluating energy consumption of a fitness action, the method comprising:
[0005] Collecting biosignal data of a target user through a preset multi-dimensional sensor group to obtain a biosignal data set;
[0006] Performing energy consumption modeling on the target user through the biosignal data set to obtain a first signal energy correlation conversion model of the target user;
[0007] Calculating energy consumption prediction data on the biological signal data set based on the first signal energy association conversion model to obtain energy consumption prediction data;
[0008] Acquiring the target user's exercise time to obtain a set of exercise time data of the target user, and analyzing the target user's fitness movements through the set of exercise time data to obtain fitness movement recognition data;
[0009] Performing fitness action influencing factor analysis on the biosignal data set through the fitness action recognition data to obtain a fitness action influencing coefficient;
[0010] Based on the fitness action influence coefficient, extracting energy consumption features from the energy consumption prediction data to obtain an energy consumption feature set of the target user;
[0011] The energy consumption feature set is input into a preset strategy analysis model for model optimization strategy analysis to obtain a target model optimization strategy, and the first signal energy association conversion model is optimized through the target model optimization strategy to obtain a second signal energy association conversion model.
[0012] In a second aspect, the present application provides a fitness action energy consumption assessment device, the fitness action energy consumption assessment device comprising:
[0013] A collection module, used to collect biological signal data of a target user through a preset multi-dimensional sensor group to obtain a biological signal data set;
[0014] A modeling module, configured to perform energy consumption modeling on the target user through the biosignal data set to obtain a first signal energy correlation conversion model of the target user;
[0015] A calculation module, configured to calculate energy consumption prediction data for the biological signal data set based on the first signal energy association conversion model to obtain energy consumption prediction data;
[0016] An acquisition module, used for acquiring the exercise time of the target user to obtain the exercise time data set of the target user, and performing fitness action analysis on the target user through the exercise time data set to obtain fitness action recognition data;
[0017] An analysis module, used for analyzing the fitness action influencing factors of the biological signal data set through the fitness action recognition data to obtain a fitness action influencing coefficient;
[0018] An extraction module, configured to extract energy consumption features from the energy consumption prediction data based on the fitness action influence coefficient, to obtain an energy consumption feature set of the target user;
[0019] The optimization module is used to input the energy consumption feature set into a preset strategy analysis model to perform model optimization strategy analysis to obtain a target model optimization strategy, and to perform association model optimization on the first signal energy association conversion model through the target model optimization strategy to obtain a second signal energy association conversion model.
[0020] The third aspect of the present application provides a fitness action energy consumption assessment device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the fitness action energy consumption assessment device executes the above-mentioned fitness action energy consumption assessment method.
[0021] A fourth aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned fitness action energy consumption assessment method.
[0022] In the technical solution provided by the present application, by adopting a multidimensional sensor group, including multiple biological signals such as heart rate, electromyographic signal and body temperature, the physiological state of the user can be captured comprehensively and comprehensively, thereby improving the accuracy of the energy consumption assessment of fitness movements. The diversity of sensors helps to reduce errors and make the evaluation results more reliable. By establishing an association conversion model and a strategy analysis model, the user's biological signal data can be analyzed and predicted in real time, and the energy consumption of fitness movements can be quickly obtained. This real-time advantage enables users to instantly understand their fitness effects and improve the effects of fitness activities. Based on the user's biological signal characteristics, the system can provide personalized fitness advice. By deeply analyzing the user's physiological state, more targeted exercise advice is provided to the user, so that the fitness plan is more in line with individual differences and increases the user's fitness motivation. Through the analysis of the exercise time data set and the fitness action recognition data, the system can fully understand the user's exercise time, action start time, and number of interruptions. This helps users better plan and adjust their fitness plans and improve training effects. By extracting the fitness action influence coefficient, the system can deeply understand the impact of different fitness actions on energy consumption. This helps users choose fitness actions that are more suitable for their goals and improve training efficiency. Through the optimization of the strategy analysis model and the association model, the system can continuously improve the accuracy of prediction and analysis. The continuous optimization of the model will make the system more adaptable and intelligent, and better serve the fitness needs of users. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.
[0024] Figure 1 This is a schematic diagram of an embodiment of a method for evaluating energy consumption of fitness movements in an embodiment of the present application;
[0025] Figure 2 This is a schematic diagram of an embodiment of a fitness movement energy consumption assessment device in an embodiment of the present application. DETAILED DESCRIPTION
[0026] The embodiments of the present application provide a method, device, equipment and storage medium for assessing energy consumption of fitness actions. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0027] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 , an embodiment of the method for evaluating energy consumption of fitness movements in the embodiment of the present application includes:
[0028] Step S101: collecting biosignal data of a target user through a preset multi-dimensional sensor group to obtain a biosignal data set;
[0029] It is understandable that the execution subject of the present application may be a fitness action energy consumption assessment device, or a terminal or a server, which is not specifically limited here. The present application embodiment is described by taking a server as the execution subject as an example.
[0030] Specifically, the multidimensional sensor group usually includes an accelerometer, a heart rate monitor, a skin galvanic sensor, a temperature sensor, etc. These sensors work together to monitor and record various biosignals of the user during fitness, such as heart rate, skin galvanic response, body temperature changes, and body movement intensity. In this process, the accelerometer can be used to capture the user's exercise intensity and exercise pattern, the heart rate monitor is used to record the heart rate and heart rate changes, the skin galvanic sensor can monitor the tiny current changes caused by emotions or physical activities, and the temperature sensor measures the user's body temperature. These data together constitute a comprehensive biosignal data set. These biosignal data are transmitted to a central processing unit such as a smartphone, tablet computer or a dedicated server via wireless or wired means. During the transmission process, appropriate data encoding and compression techniques are applied to ensure the integrity of the data and reduce data loss during transmission. After arriving at the central processing unit, the data will be decoded and decompressed, and then subjected to preliminary processing such as filtering, denoising, normalization, and time series analysis to improve the data quality and prepare for subsequent energy consumption analysis. In this way, the user's biosignal data during fitness can be effectively collected from multiple dimensions.
[0031] Step S102: Modeling the energy consumption of the target user through the biosignal data set to obtain a first signal energy correlation conversion model of the target user;
[0032] Specifically, the biosignal data set is classified to obtain the user's heart rate data, electromyographic signal data and body temperature data. The comprehensively collected biosignal data is separated into independent subsets through data processing technology, each of which corresponds to a specific type of signal. The signal parameter change trend analysis is performed on each type of data. Time series analysis is performed on data such as heart rate, electromyographic signal and body temperature to identify the trend of these biosignals changing over time. This analysis process usually includes steps such as trend line drawing, seasonal adjustment, and smoothing of the data, with the aim of revealing the change pattern and law of each signal during the fitness process. The analysis of these trends helps to understand the relationship between biosignals and the user's exercise intensity and energy consumption. The change amplitude value of each signal parameter change trend is extracted to quantify the degree of change of the signal parameter. The change amplitude of heart rate, electromyographic signal and body temperature data during the fitness process is extracted through mathematical and statistical methods, such as peak analysis and volatility calculation. The energy consumption mapping relationship analysis is performed through the signal parameter change amplitude value corresponding to each signal parameter change trend. The relationship between the change amplitude of different signal parameters and energy consumption is explored through data association analysis techniques, such as regression analysis and correlation analysis. Through this analysis, a set of energy consumption mapping relationships can be established, which can explain how changes in biosignal parameters such as heart rate, electromyographic signals, and body temperature are associated with energy consumption. Based on these mapping relationships, the energy consumption of the target user is modeled to obtain the first signal energy correlation conversion model. This model reflects the direct relationship between biosignal data and energy consumption.
[0033] Step S103, calculating energy consumption prediction data for the biological signal data set based on the first signal energy correlation conversion model to obtain energy consumption prediction data;
[0034] Specifically, the energy consumption mapping relationship is matched for the biosignal data set through the first signal energy association conversion model. The model parses and understands the biosignal data collected in real time to find the mapping relationship between each signal (such as heart rate, electromyographic signal, body temperature, etc.) and energy consumption. The model matches the change trend of each biosignal with a specific energy consumption pattern according to the previously trained parameters and algorithms, thereby obtaining the target mapping relationship matching result. According to these target mapping relationship matching results, the biosignal data set is converted into energy consumption data. By analyzing the correlation between each biosignal change and energy consumption, the biosignal data is converted into a specific energy consumption value. For example, a rapid increase in heart rate indicates a higher energy consumption, while a change in body temperature reflects a change in metabolism. Through this conversion, the change of each biosignal is quantified into an initial energy consumption prediction data. These initial consumption prediction data are subjected to data standardization to ensure the consistency and comparability of the energy consumption prediction data. Standardization can eliminate the scale differences between different data, making the results more accurate and reliable. Through methods such as minimum-maximum standardization and Z-score standardization, the data can be compared and analyzed under the same standard. The resulting energy expenditure prediction data not only reflects the user's energy expenditure pattern in a specific fitness activity, but also maintains comparability with other users or other activities.
[0035] Step S104, acquiring the target user's exercise time to obtain the target user's exercise time data set, and performing fitness action analysis on the target user through the exercise time data set to obtain fitness action recognition data;
[0036] Specifically, the target user's exercise time is acquired through a wearable smart device or fitness tracker, which can record the start and end time of the user's activity. The obtained exercise time data set includes the specific time point of each user's fitness. The exercise time data is analyzed for the start time of fitness to determine the exact time point when the user starts fitness. By analyzing the start time of fitness, the user's daily fitness habits can be better understood. The number of fitness interruptions in the exercise time data set is collected to identify the number of interruptions that occur in a complete fitness activity of the user, which are caused by rest or other external interference. The data on the number of interruptions helps to analyze the user's fitness continuity and the integrity of the activity. By combining the fitness start time and the number of fitness interruptions, the exercise time data set is analyzed for exercise time series. This analysis can reveal the distribution and continuity of the user's exercise time in a fitness cycle. According to the results of the exercise time series analysis, the target user's fitness action analysis is performed to obtain fitness action recognition data. Through pattern recognition and machine learning algorithms, specific fitness actions are identified from the user's exercise time data.
[0037] Step S105, analyzing the fitness action influencing factors on the biological signal data set through the fitness action recognition data to obtain the fitness action influencing coefficient;
[0038] Specifically, first. The fitness action recognition data is calibrated for action state types, and the original fitness action data captured by sensors and other devices are converted into specific action state types. Because different fitness action types have different effects on the body and energy consumption patterns. Based on the action state type, the fitness action recognition data is feature extracted to obtain the fitness action feature set of the target user. Key features such as the duration, intensity and frequency of the action are extracted from the fitness action through data analysis and pattern recognition technology. Through this feature extraction, the user's performance in a specific fitness action can be more accurately described, such as the degree of increase in heart rate, the change pattern of electromyographic signals, etc. These feature sets provide the necessary basis for understanding the specific impact of fitness actions on biological signals. The Pearson correlation coefficient is used to analyze the influencing factors of the fitness action feature set and the biological signal data set, so as to obtain the fitness action influence coefficient. The Pearson correlation coefficient is a statistical method used to measure the degree of linear correlation between two variables. The correlation between fitness action features and biological signals (such as heart rate, electromyographic signals, body temperature, etc.) is analyzed through the Pearson correlation coefficient. Through this analysis, it can be determined which fitness movement characteristics have a significant correlation with changes in specific biosignal parameters, thereby deriving the quantitative relationship between the impact of fitness movements on biosignals, namely the fitness movement impact coefficient.
[0039] Step S106: extracting energy consumption characteristics from the energy consumption prediction data based on the fitness action influence coefficient to obtain an energy consumption characteristic set of the target user;
[0040] Specifically, the energy consumption prediction data is matched with time series data to obtain a time series data set of the energy consumption prediction data. The continuous energy consumption data is arranged and organized in chronological order so that the data is ordered and easy to analyze in the time dimension. This matching process is to ensure that the subsequent analysis can accurately capture the trend and pattern of energy consumption over time, such as the peak and trough of energy consumption of users during fitness. Based on the time series data set, the energy consumption prediction data is extracted from abnormal energy consumption time series points. Abnormal energy consumption time series points refer to those energy consumption data points that are significantly different from the normal pattern and are caused by certain specific fitness actions or external factors. Through at least one abnormal energy consumption time series point, the energy consumption prediction data is cleaned. The purpose of data cleaning is to remove abnormal data points caused by errors, deviations or atypical behaviors to ensure that the remaining data sets are more accurate and reliable. Based on the fitness action influence coefficient, these cleaned initial consumption feature sets are subjected to feature weighting processing. Feature weighting processing refers to giving different weights according to the degree of influence of each fitness action on energy consumption, so as to more accurately reflect the actual contribution of different fitness actions to energy consumption. This weighted processing can help the system better understand and quantify the specific effects of different fitness movements on energy consumption, thereby obtaining a more comprehensive and detailed set of energy consumption characteristics.
[0041] Step S107: input the energy consumption feature set into a preset strategy analysis model to perform model optimization strategy analysis to obtain a target model optimization strategy, and optimize the first signal energy association conversion model through the target model optimization strategy to obtain a second signal energy association conversion model.
[0042] Specifically, the energy consumption feature set is input into a preset strategy analysis model, which includes an encoding network, a first attention mechanism layer, a decoding network, and a second attention mechanism layer. The encoding network uses a two-layer GRU (Gated Recurrent Unit) network to process the energy consumption feature set and extracts deep hidden feature vectors representing these features. The GRU network can capture the dynamic characteristics of data changing over time. The first attention mechanism layer performs attention mechanism analysis on these target hidden feature vectors to identify and emphasize the most important features for predicting energy consumption of fitness movements. By giving these key features more "attention", the accuracy and efficiency of the model predicting energy consumption can be improved. The single-layer GRU network in the decoding network performs further feature analysis on the first attention feature vector to generate a target feature prediction vector, and decodes the most critical information for energy consumption prediction from the weighted attention features. The single-layer GRU network here plays a role in simplifying and clarifying the prediction vector. The second attention mechanism layer performs model optimization strategy analysis on the target feature prediction vector and generates the final target model optimization strategy based on the information in the prediction vector. This strategy will guide how to adjust and optimize the first signal energy association conversion model so that it can more accurately and efficiently reflect the relationship between fitness actions and energy consumption. By applying this target model optimization strategy, the first signal energy association conversion model is optimized to generate the second signal energy association conversion model. This optimization process is not only based on the framework and logic of the original model, but also incorporates the information extracted from the energy consumption feature set, making the second signal energy association conversion model more accurate and adaptable in predicting energy consumption.
[0043] In the embodiment of the present application, by adopting a multidimensional sensor group, including multiple biological signals such as heart rate, electromyographic signal and body temperature, the physiological state of the user can be captured comprehensively and comprehensively, thereby improving the accuracy of the energy consumption assessment of fitness movements. The diversity of sensors helps to reduce errors and make the assessment results more reliable. By establishing an association conversion model and a strategy analysis model, the user's biological signal data can be analyzed and predicted in real time, and the energy consumption of fitness movements can be quickly obtained. This real-time advantage enables users to instantly understand their fitness effects and improve the effects of fitness activities. Based on the user's biological signal characteristics, the system can provide personalized fitness advice. By deeply analyzing the user's physiological state, more targeted exercise advice is provided to the user, so that the fitness plan is more in line with individual differences and increases the user's fitness motivation. Through the analysis of the exercise time data set and the fitness action recognition data, the system can fully understand the user's exercise time, action start time, and interruption times. This helps users better plan and adjust their fitness plans and improve training effects. By extracting the fitness action influence coefficient, the system can deeply understand the impact of different fitness actions on energy consumption. This helps users choose fitness actions that are more suitable for their goals and improve training efficiency. Through the optimization of the strategy analysis model and the association model, the system can continuously improve the accuracy of prediction and analysis. The continuous optimization of the model will make the system more adaptable and intelligent, and better serve the fitness needs of users.
[0044] In a specific embodiment, the process of executing step S102 may specifically include the following steps:
[0045] (1) Classify the biosignal data set to obtain user heart rate data, electromyographic signal data, and user body temperature data;
[0046] (2) Analyze the change trends of signal parameters of the user's heart rate data, electromyographic signal data, and body temperature data to obtain the change trends of multiple signal parameters;
[0047] (3) Extract the variation amplitude value of each signal parameter variation trend respectively, and obtain the signal parameter variation amplitude value corresponding to each signal parameter variation trend;
[0048] (4) Analyze the energy consumption mapping relationship through the signal parameter change amplitude value corresponding to each signal parameter change trend to obtain an energy consumption mapping relationship set;
[0049] (5) Energy consumption modeling of the target user is performed according to the energy consumption mapping relationship set to obtain the first signal energy correlation conversion model of the target user.
[0050] Specifically, the biosignal data set is classified, which involves data collected using a multi-dimensional sensor group, such as a heart rate monitor, an electromyography (EMG) device, and a body temperature sensor. These devices record the user's heart rate data, electromyography data, and body temperature data, respectively. The accurate classification of these data is to ensure the pertinence and effectiveness of subsequent analysis. Heart rate monitors usually capture heartbeats through optical sensors or electrodes, electromyography devices capture electrical signals generated by muscle activity through sensors on the surface of the skin, and body temperature sensors measure surface or internal body temperature through contact or non-contact methods. The classified data is subjected to a signal parameter change trend analysis to reveal the change patterns of heart rate, electromyography, and body temperature over time. For example, the heart rate data analysis shows a sharp increase or decrease in heart rate during certain specific fitness movements, the electromyography signal analysis reveals the intensity and pattern of muscle activity, and the body temperature data analysis reflects changes in metabolic rate. The analysis of these signal parameter change trends helps to understand the impact of fitness movements on biosignals. The change amplitude value of each signal parameter change trend is extracted to quantify the degree of change of heart rate, electromyographic signal and body temperature, such as the maximum and minimum values of heart rate, the peak value of electromyographic signal and the average change rate of body temperature. The energy consumption mapping relationship analysis is performed through the change amplitude values corresponding to the change trends of these signal parameters. The relationship between the change of biological signals and energy consumption is quantified, such as analyzing the relationship between increased heart rate and increased energy consumption, or exploring the relationship between the change pattern of electromyographic signals and different types of exercise loads. Through this analysis, a set of energy consumption mapping relationships can be established, which illustrates how the changes in different biological signal parameters are mapped to energy consumption. According to this energy consumption mapping relationship set, the energy consumption of the target user is modeled to obtain the first signal energy association conversion model of the target user. This model not only reflects the direct relationship between biological signal data and energy consumption, but also can provide personalized energy consumption prediction based on the specific data of individual users.
[0051] In a specific embodiment, the process of executing step S103 may specifically include the following steps:
[0052] (1) performing energy consumption mapping relationship matching on the biological signal data set through the first signal energy association conversion model to obtain a target mapping relationship matching result;
[0053] (2) Convert the biosignal data set into energy consumption data according to the target mapping relationship matching results to obtain the corresponding initial consumption prediction data;
[0054] (3) Perform data standardization on the initial consumption forecast data to obtain energy consumption forecast data.
[0055] Specifically, the energy consumption mapping relationship is matched to the biosignal data set through the first signal energy association conversion model, and the specific parameters of these biosignals are mapped to energy consumption. The model considers the changing trends of heart rate, electromyographic signals and body temperature, and how these changes are associated with different types and intensities of exercise. For example, a rapid rise in heart rate is associated with high-intensity aerobic exercise, indicating higher energy consumption, while a stable body temperature indicates low-intensity exercise. According to the target mapping relationship matching result, the biosignal data set is converted into energy consumption data. Specific changes in the biosignal data are converted into estimated values of energy consumption. For example, if the model shows that the user's energy consumption is in a specific range at a specific heart rate level, then the user's energy consumption can be estimated based on the user's actual heart rate data. Similarly, specific patterns of electromyographic signals and body temperature can also be converted into predicted values of energy consumption. Thus, a set of initial energy consumption prediction data is obtained, which is based on the user's biosignal data and reflects their energy consumption in fitness activities. The initial consumption prediction data is subjected to data standardization to eliminate the deviation between different data sources or different measurement methods, making the data more comparable and consistent. Normalization ensures that data from different users, at different times, and under different conditions can be compared and analyzed under the same standards. This includes adjusting the scale of the data, applying a common unit of measurement, or converting the data to a common reference range.
[0056] In a specific embodiment, the process of executing step S104 may specifically include the following steps:
[0057] (1) Obtaining the target user’s exercise time to obtain the target user’s exercise time data set;
[0058] (2) Analyze the fitness start time of the exercise time data set to obtain the fitness start time of the target user;
[0059] (3) Collecting the number of fitness interruptions in the exercise time data set to obtain the corresponding number of fitness interruptions;
[0060] (4) Performing a sports time series analysis on the exercise time data set based on the fitness start time and the number of fitness interruptions to obtain the sports time series analysis results;
[0061] (5) Analyze the target user’s fitness movements based on the results of the motion time series analysis to obtain fitness movement recognition data.
[0062] Specifically, exercise time data is collected from the user's smart wearable device or fitness tracking application. This data includes the start and end time of each fitness activity, as well as any pauses or interruptions in the middle, to build a comprehensive exercise time data set that reflects the user's exercise habits, frequency and duration. Perform fitness start time analysis on the exercise time data set. Identify the specific time point when the user starts fitness in order to understand their exercise habits. Fitness start time analysis can reveal which time of the day the user is more inclined to exercise, which is helpful for personalized fitness recommendations and energy consumption assessment. Collect the number of fitness interruptions on the exercise time data set. Identify the number of interruptions that occur in a single fitness activity of the user. These interruptions are caused by rest or other external factors. Recording the number of fitness interruptions helps to evaluate the user's exercise continuity and exercise quality. For example, frequent interruptions indicate that the user fails to maintain a continuous exercise intensity during fitness. By comprehensively considering the fitness start time and the number of fitness interruptions, the exercise time series analysis is performed on the exercise time data set to provide an overview of the user's exercise behavior from a time series perspective, including exercise frequency, duration and interruption pattern. This analysis can help reveal the patterns and habits of users' fitness behaviors, as well as irregularities or abnormalities. According to the results of the exercise time series analysis, the fitness action of the target user is analyzed, and the specific fitness action is identified from the exercise time data. By using algorithms and pattern recognition technology to analyze the exercise time data, the specific type of exercise performed by the user can be inferred. For example, by analyzing the activity duration and interruption pattern of the user during fitness, it can be inferred whether the user is doing aerobic exercise or strength training. Furthermore, combining other data sources (such as heart rate or electromyography signal data) can improve the accuracy of action recognition.
[0063] In a specific embodiment, the process of executing step S105 may specifically include the following steps:
[0064] (1) Calibrate the action state type of the fitness action recognition data to determine the action state type corresponding to the fitness action recognition data;
[0065] (2) Based on the action state type, fitness action features are extracted from the fitness action recognition data to obtain the fitness action feature set of the target user;
[0066] (3) The Pearson correlation coefficient is used to analyze the influencing factors of fitness movements on the fitness movement feature set and the biological signal data set to obtain the fitness movement influence coefficient.
[0067] Specifically, the action state type of fitness action recognition data is calibrated. The original fitness action data captured by sensors, smart wearable devices or fitness applications are converted into specific action state types. Based on the action state type, the fitness action recognition data is feature extracted to obtain the fitness action feature set of the target user. Key features such as the duration, intensity, frequency and consistency of the action can be extracted from the fitness action. For example, for the running action, the duration, average speed and total distance of each run can be extracted. The fitness action influencing factors of the fitness action feature set and the biosignal data set are analyzed using the Pearson correlation coefficient to obtain the fitness action influence coefficient. The Pearson correlation coefficient is a statistical method used to measure the degree of linear correlation between two variables. It can help identify the correlation between fitness action features and biosignals (such as heart rate, electromyographic signals, body temperature, etc.). For example, through analysis, it can be found that the user's heart rate increases significantly when running at high intensity. This analysis can clarify which fitness action features are significantly correlated with changes in specific biosignal parameters, thereby deriving the quantitative relationship between the impact of fitness actions on biosignals, that is, the fitness action influence coefficient.
[0068] In a specific embodiment, the process of executing step S106 may specifically include the following steps:
[0069] (1) Performing time series data matching on the energy consumption prediction data to obtain a time series data set of the energy consumption prediction data;
[0070] (2) Based on the time series data set, extract abnormal energy consumption time series points from the energy consumption prediction data to obtain at least one abnormal energy consumption time series point;
[0071] (3) Performing data cleaning on the energy consumption prediction data through at least one abnormal energy consumption time series point to obtain a corresponding initial consumption feature set;
[0072] (4) Based on the fitness action influence coefficient, the initial consumption feature set is weighted to obtain the energy consumption feature set.
[0073] Specifically, the energy consumption prediction data is time-series matched to construct a time-series data set of the energy consumption prediction data. The energy consumption prediction data at different time points are arranged and organized in chronological order. The construction of this time-series data set is to be able to analyze the changing trends and patterns of energy consumption in the time dimension, so that the data analysis is more orderly and easy to understand. For example, a detailed time-series data set can be constructed by recording the estimated energy consumption per minute or per hour during each fitness activity. Based on the time-series data set, the energy consumption prediction data is subjected to abnormal energy consumption time-series point extraction. Abnormal energy consumption time-series points refer to those energy consumption data points that are significantly different from the normal pattern, which are caused by specific fitness actions, changes in exercise intensity or other external factors. Through at least one abnormal energy consumption time-series point, the energy consumption prediction data is subjected to data cleaning processing. Abnormal data points caused by measurement errors, data transmission problems or atypical user behaviors are removed to ensure that the remaining data sets are more accurate and reliable. Through data processing techniques such as filtering, removing outliers, data smoothing, etc., it is ensured that the information in the data set accurately and truly reflects the actual energy consumption of the user. Based on the fitness action influence coefficient, the cleaned initial consumption feature set is subjected to feature weighting processing. Feature weighting processing refers to giving different weights according to the degree of influence of each fitness action on energy consumption, so as to more accurately reflect the actual contribution of different fitness actions to energy consumption. For example, for some high-intensity exercises such as intermittent training, since they contribute more to energy consumption, they will be given higher weights in the feature weighting process. For low-intensity exercises such as walking, their contribution to energy consumption is relatively small, so they will be given lower weights in the weighting process. This weighted processing can help the system better understand and quantify the specific effects of different fitness actions on energy consumption, thereby obtaining a more comprehensive energy consumption feature set. These feature sets not only reflect the energy consumption patterns of users in various fitness actions, but also take into account the different degrees of influence of various actions on energy consumption.
[0074] In a specific embodiment, the process of executing step S107 may specifically include the following steps:
[0075] (1) Inputting the energy consumption feature set into a preset strategy analysis model, which includes: an encoding network, a first attention mechanism layer, a decoding network, and a second attention mechanism layer;
[0076] (2) The hidden features of the energy consumption feature set are extracted through the two-layer GRU network in the encoding network to obtain the target hidden feature vector;
[0077] (3) Performing attention mechanism analysis on the target hidden feature vector through the first attention mechanism layer to obtain the first attention feature vector;
[0078] (4) Perform feature analysis on the first attention feature vector through the single-layer GRU network in the decoding network to obtain the target feature prediction vector;
[0079] (5) Perform model optimization strategy analysis on the target feature prediction vector through the second attention mechanism layer to obtain the target model optimization strategy;
[0080] (6) The first signal energy correlation conversion model is optimized through the target model optimization strategy to obtain the second signal energy correlation conversion model.
[0081] Specifically, the energy consumption feature set is input into a preset strategy analysis model. The strategy analysis model includes an encoding network, a first attention mechanism layer, a decoding network, and a second attention mechanism layer. Hidden features of the energy consumption feature set are extracted through a double-layer GRU (gated recurrent unit) network in the encoding network. The double-layer GRU network is an efficient recurrent neural network structure that can capture long-term dependencies in time series data. In this process, the network will learn which features are most critical for energy consumption prediction, thereby extracting target hidden feature vectors. Attention mechanism analysis is performed on these target hidden feature vectors through the first attention mechanism layer. The function of the attention mechanism layer is to identify and emphasize those hidden features that are most important for energy consumption prediction, thereby generating a first attention feature vector. This can ensure that the model pays more attention to the key features that best reflect energy consumption in subsequent processing, thereby improving the accuracy and efficiency of prediction. The single-layer GRU network in the decoding network performs further feature analysis on the first attention feature vector to generate a target feature prediction vector. The single-layer GRU network can decode and refine the first attention feature vector to obtain more specific energy consumption prediction information. Then, the target feature prediction vector is analyzed for model optimization strategy through the second attention mechanism layer. The model generates a final target model optimization strategy based on the information in the prediction vector, which is designed to guide how to adjust and optimize the energy expenditure prediction model. For example, if it is found that the prediction accuracy of the model is not high under certain specific conditions, the optimization strategy will suggest adjusting the model parameters or introducing additional features to improve accuracy. Through the target model optimization strategy, the first signal energy association conversion model is optimized for association model optimization to obtain the second signal energy association conversion model. For example, if the optimization strategy indicates that the impact of heart rate changes on energy expenditure prediction during high-intensity intermittent training is greater than previously assumed, then the second signal energy association conversion model will pay more attention and weight to this point.
[0082] The above describes the method for evaluating the energy consumption of fitness movements in the embodiment of the present application. The following describes the device for evaluating the energy consumption of fitness movements in the embodiment of the present application. Figure 2In the embodiment of the present application, an embodiment of the fitness action energy consumption assessment device includes:
[0083] The collection module 201 is used to collect biological signal data of the target user through a preset multi-dimensional sensor group to obtain a biological signal data set;
[0084] A modeling module 202, configured to perform energy consumption modeling on the target user through the biosignal data set to obtain a first signal energy correlation conversion model of the target user;
[0085] A calculation module 203, configured to calculate energy consumption prediction data for the biological signal data set based on the first signal energy correlation conversion model to obtain energy consumption prediction data;
[0086] The acquisition module 204 is used to acquire the exercise time of the target user to obtain the exercise time data set of the target user, and perform fitness action analysis on the target user through the exercise time data set to obtain fitness action recognition data;
[0087] An analysis module 205 is used to analyze the fitness action influencing factors of the biological signal data set through the fitness action recognition data to obtain a fitness action influencing coefficient;
[0088] An extraction module 206, configured to extract energy consumption features from the energy consumption prediction data based on the fitness action influence coefficient, to obtain an energy consumption feature set of the target user;
[0089] The optimization module 207 is used to input the energy consumption feature set into a preset strategy analysis model to perform model optimization strategy analysis to obtain a target model optimization strategy, and to perform association model optimization on the first signal energy association conversion model through the target model optimization strategy to obtain a second signal energy association conversion model.
[0090] Through the synergy of the above components, by adopting a multi-dimensional sensor group, including multiple biological signals such as heart rate, electromyographic signals and body temperature, the user's physiological state can be fully and comprehensively captured, thereby improving the accuracy of the energy consumption assessment of fitness movements. The diversity of sensors helps to reduce errors and make the assessment results more reliable. By establishing an association conversion model and a strategy analysis model, the user's biological signal data can be analyzed and predicted in real time, and the energy consumption of fitness movements can be quickly obtained. This real-time advantage enables users to instantly understand their fitness effects and improve the effectiveness of fitness activities. Based on the user's biological signal characteristics, the system can provide personalized fitness suggestions. Through in-depth analysis of the user's physiological state, more targeted exercise suggestions are provided to users, making fitness plans more in line with individual differences and increasing users' fitness motivation. Through the analysis of the exercise time data set and fitness action recognition data, the system can fully understand the user's exercise time, action start time, and number of interruptions. This helps users better plan and adjust their fitness plans and improve training results. By extracting the fitness action influence coefficient, the system can deeply understand the impact of different fitness actions on energy consumption. This helps users choose fitness actions that are more suitable for their goals and improve training efficiency. Through the optimization of the strategy analysis model and the association model, the system can continuously improve the accuracy of prediction and analysis. The continuous optimization of the model will make the system more adaptable and intelligent, and better serve the fitness needs of users.
[0091] The present application also provides a fitness action energy consumption assessment device, which includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor executes the steps of the fitness action energy consumption assessment method in the above-mentioned embodiments.
[0092] The present application also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein instructions are stored in the computer, and when the instructions are executed on a computer, the computer executes the steps of the fitness action energy consumption assessment method.
[0093] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0094] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.
[0095] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims
1. A method for evaluating energy consumption of fitness movements. It is characterized in that The fitness action energy consumption assessment method comprises: Collecting biosignal data of a target user through a preset multi-dimensional sensor group to obtain a biosignal data set; Performing energy consumption modeling on the target user through the biosignal data set to obtain a first signal energy correlation conversion model of the target user; Calculating energy consumption prediction data on the biological signal data set based on the first signal energy association conversion model to obtain energy consumption prediction data; Acquiring the target user's exercise time to obtain a set of exercise time data of the target user, and analyzing the target user's fitness movements through the set of exercise time data to obtain fitness movement recognition data; Performing fitness action influencing factor analysis on the biosignal data set through the fitness action recognition data to obtain a fitness action influencing coefficient; Based on the fitness action influence coefficient, energy consumption feature extraction is performed on the energy consumption prediction data to obtain the energy consumption feature set of the target user; specifically including: performing time series data matching on the energy consumption prediction data to obtain the time series data set of the energy consumption prediction data; based on the time series data set, abnormal energy consumption time series points are extracted on the energy consumption prediction data to obtain at least one abnormal energy consumption time series point; through at least one abnormal energy consumption time series point, data cleaning processing is performed on the energy consumption prediction data to obtain a corresponding initial consumption feature set; based on the fitness action influence coefficient, feature weighting processing is performed on the initial consumption feature set to obtain an energy consumption feature set; The energy consumption feature set is input into a preset strategy analysis model for model optimization strategy analysis to obtain a target model optimization strategy, and the first signal energy association conversion model is optimized through the target model optimization strategy to obtain a second signal energy association conversion model; specifically including: inputting the energy consumption feature set into a preset strategy analysis model, the strategy analysis model includes: an encoding network, a first attention mechanism layer, a decoding network and a second attention mechanism layer; performing hidden feature extraction on the energy consumption feature set through the double-layer GRU network in the encoding network to obtain a target hidden feature vector; performing attention mechanism analysis on the target hidden feature vector through the first attention mechanism layer to obtain a first attention feature vector; performing feature analysis on the first attention feature vector through the single-layer GRU network in the decoding network to obtain a target feature prediction vector; performing model optimization strategy analysis on the target feature prediction vector through the second attention mechanism layer to obtain a target model optimization strategy; performing association model optimization on the first signal energy association conversion model through the target model optimization strategy to obtain a second signal energy association conversion model.
2. The method for evaluating energy consumption of fitness movements according to claim 1, It is characterized in that The step of performing energy consumption modeling on the target user through the biosignal data set to obtain a first signal energy correlation conversion model of the target user includes: Performing biosignal data classification on the biosignal data set to obtain user heart rate data, electromyographic signal data, and user body temperature data; Performing signal parameter change trend analysis on the user's heart rate data, the electromyographic signal data, and the user's body temperature data respectively to obtain multiple signal parameter change trends; Extract the variation amplitude value of each signal parameter variation trend respectively, and obtain the signal parameter variation amplitude value corresponding to each signal parameter variation trend; The energy consumption mapping relationship is analyzed by using the signal parameter change amplitude value corresponding to each signal parameter change trend to obtain an energy consumption mapping relationship set; Energy consumption modeling is performed on the target user according to the energy consumption mapping relationship set to obtain a first signal energy correlation conversion model of the target user.
3. The method for evaluating energy consumption of fitness movements according to claim 2, It is characterized in that The step of calculating energy consumption prediction data on the biological signal data set based on the first signal energy association conversion model to obtain energy consumption prediction data includes: Performing energy consumption mapping relationship matching on the biological signal data set through the first signal energy association conversion model to obtain a target mapping relationship matching result; According to the target mapping relationship matching result, the biosignal data set is converted into energy consumption data to obtain corresponding initial consumption prediction data; The initial consumption prediction data is subjected to data standardization processing to obtain energy consumption prediction data.
4. The method for evaluating energy consumption of fitness movements according to claim 2, It is characterized in that The step of acquiring the exercise time of the target user to obtain an exercise time data set of the target user, and performing fitness action analysis on the target user through the exercise time data set to obtain fitness action recognition data includes: Acquiring the exercise time of the target user to obtain an exercise time data set of the target user; Performing fitness start time analysis on the exercise time data set to obtain the fitness start time of the target user; Collecting the number of fitness interruptions from the exercise time data set to obtain a corresponding number of fitness interruptions; Performing a sports time series analysis on the sports time data set according to the fitness start time and the fitness interruption number to obtain a sports time series analysis result; The target user's fitness movements are analyzed according to the motion time series analysis result to obtain fitness movement recognition data.
5. The method for evaluating energy consumption of fitness movements according to claim 1, It is characterized in that The step of analyzing the fitness action influencing factors of the biological signal data set by using the fitness action recognition data to obtain the fitness action influencing coefficient includes: Performing action state type calibration on the fitness action recognition data to determine the action state type corresponding to the fitness action recognition data; Based on the action state type, extracting fitness action features from the fitness action recognition data to obtain a fitness action feature set of the target user; The Pearson correlation coefficient is used to analyze the fitness action influencing factors of the fitness action feature set and the biological signal data set to obtain the fitness action influencing coefficient.
6. A fitness movement energy consumption assessment device, It is characterized in that The fitness action energy consumption assessment device comprises: A collection module, used to collect biological signal data of a target user through a preset multi-dimensional sensor group to obtain a biological signal data set; A modeling module, configured to perform energy consumption modeling on the target user through the biosignal data set to obtain a first signal energy correlation conversion model of the target user; A calculation module, configured to calculate energy consumption prediction data for the biological signal data set based on the first signal energy association conversion model to obtain energy consumption prediction data; An acquisition module, used for acquiring the exercise time of the target user to obtain the exercise time data set of the target user, and performing fitness action analysis on the target user through the exercise time data set to obtain fitness action recognition data; An analysis module, used for analyzing the fitness action influencing factors of the biological signal data set through the fitness action recognition data to obtain a fitness action influencing coefficient; An extraction module is used to extract energy consumption features from the energy consumption prediction data based on the fitness action influence coefficient to obtain the energy consumption feature set of the target user; specifically comprising: performing time series data matching on the energy consumption prediction data to obtain the time series data set of the energy consumption prediction data; based on the time series data set, performing abnormal energy consumption time series point extraction on the energy consumption prediction data to obtain at least one abnormal energy consumption time series point; performing data cleaning processing on the energy consumption prediction data through at least one abnormal energy consumption time series point to obtain a corresponding initial consumption feature set; based on the fitness action influence coefficient, performing feature weighting processing on the initial consumption feature set to obtain an energy consumption feature set; An optimization module is used to input the energy consumption feature set into a preset strategy analysis model for model optimization strategy analysis to obtain a target model optimization strategy, and to optimize the first signal energy association conversion model through the target model optimization strategy to obtain a second signal energy association conversion model; specifically comprising: inputting the energy consumption feature set into a preset strategy analysis model, the strategy analysis model comprising: an encoding network, a first attention mechanism layer, a decoding network and a second attention mechanism layer; performing hidden feature extraction on the energy consumption feature set through a double-layer GRU network in the encoding network to obtain a target hidden feature vector; performing attention mechanism analysis on the target hidden feature vector through the first attention mechanism layer to obtain a first attention feature vector; performing feature analysis on the first attention feature vector through a single-layer GRU network in the decoding network to obtain a target feature prediction vector; performing model optimization strategy analysis on the target feature prediction vector through the second attention mechanism layer to obtain a target model optimization strategy; performing association model optimization on the first signal energy association conversion model through the target model optimization strategy to obtain a second signal energy association conversion model.
7. A fitness movement energy consumption assessment device, It is characterized in that The fitness action energy consumption assessment device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the fitness action energy expenditure assessment device to perform the fitness action energy expenditure assessment method according to any one of claims 1 to 5.
8. A computer-readable storage medium having instructions stored thereon, It is characterized in that When the instructions are executed by a processor, the method for evaluating energy consumption of a fitness action as described in any one of claims 1 to 5 is implemented.
Citation Information
Patent Citations
Method and device for calculating exercise energy consumption
CN105748039A
Model training method and device, use method and device, equipment and storage medium
CN117219181A
