Dance training load prediction method and system based on training data

Abstract

The invention discloses a dance training load prediction method and system based on training data. The method comprises the following steps: S1, acquiring a basic heart rate of a dancer; s2, calculating the metabolic equivalent of the dancer for calculating a heart rate load value, and carrying out endurance training to obtain the maximum heart rate value in the training process; s3, calculating a heart rate load prediction index of the dancer according to the basic heart rate; s4, performing standardization processing on the obtained heart rate load prediction indexes; s5, predicting a heart rate load value through a linear regression equation; and S6, calculating energy consumption according to the heart rate data in the exercise process. According to the dance training load prediction method and system based on the training data, the heart rate data of the dancer can be calculated for subsequent prediction of the heart rate load of the dancer and prediction of the energy consumption of the dancer, and personalized training plans can be made for different dancers.

Description

Technical Field

[0001] The invention belongs to the field of dance training load prediction, and in particular relates to a dance training load prediction method and system based on training data. Background Art

[0002] Dance is an art of comprehensive expression of body and emotion. Since ancient times, it has been an important carrier of cultural and artistic inheritance, carrying the mission of conveying the essence of dance and cultural connotation. In dance teaching, teachers will teach basic movements, techniques, emotional expression and dance choreography methods to ensure that students can accurately copy and reproduce excellent dance works. It is an art form that uses the body to complete various elegant or difficult movements, usually with musical accompaniment, and uses rhythmic movements as the main means of expression, generally with the help of music and other props. When teaching dance, trainers are required to conduct physical training for dancers.

[0003] Dance includes dance training, dance performance, and dance creation. Dance training is the basic component of the dance education and teaching system, and is also an important means of cultivating professional dance talents. Its scientific level not only affects the effect of dance education, but also determines the quality and level of professional dance talent training. The training of professional dance talents often requires a long-term, systematic dance training process that lasts for many years. This process not only enables the quantity and quality of the movement techniques and skills they learn and practice to be continuously improved, but also enables their body shape, function, and athletic quality to be continuously improved, thereby meeting the high requirements of dance performance practice activities. At present, the public's appreciation of dance performances and dance movement techniques is increasing, and higher and higher requirements are placed on the dancers' physical abilities and movement techniques, which easily leads to higher and higher training intensity for dancers.

[0004] Training load refers to the limit that the human body can withstand when doing various exercises, which is usually related to the training intensity. Therefore, the training intensity of dancers in training is very critical: if the training intensity is too low, the load intensity will be reduced, and the performer's body functions will not be effectively stimulated, thereby failing to achieve the ideal training effect and the training standard; on the contrary, if the training intensity is too high, the dancer's body will find it difficult to withstand the training load, which will have a counter-effect on the performer's physical health, making it difficult to achieve the expected effect and even causing accidents. Therefore, it is necessary to predict the training load based on the dancer's training data. The present invention solves this technical problem. Summary of the invention

[0005] The present invention provides a dance training load prediction method and system based on training data, which can calculate the dancer's heart rate data for subsequent prediction of their heart rate load and energy consumption, and help to formulate personalized training plans for different dancers.

[0006] A dance training load prediction method based on training data, comprising the following steps: S1. Obtain the basic heart rate of the dancer; , where s represents the basic heart rate, t represents the heartbeat time, and n represents the number of heartbeats; S2. Calculate the metabolic equivalent of the dancer, which is used to calculate the heart rate load value, and conduct endurance training to obtain the maximum heart rate value during the training; S3. Calculate the heart rate load prediction index of the dancer according to the basic heart rate; S4. Standardize the obtained heart rate load prediction index; S5. Predict the heart rate load value through a linear regression equation; S6. Calculate the energy consumption according to the heart rate data during the exercise.

[0007] Furthermore, the process of step S2 is as follows: S21. After obtaining the basic heart rate, the dancer makes up-and-down jumping movements for a certain period of time to obtain the jumping height H; S22. Calculate the metabolic equivalent of the dancer according to the following formula: , where MET represents the metabolic equivalent, and F represents the average heart rate during the exercise; S23. Conduct aerobic endurance training, and record the heart rate data and respiratory rate of the dancer for subsequent prediction of the heart rate load through linear regression analysis.

[0008] Furthermore, the process of step S3 is as follows: S31. Estimate the heart rate reserve value of the dancer according to the recorded data, specifically calculated by the following formula: , where SRF represents the heart rate reserve value of the dancer, and u represents the maximum heart rate value during the test; S32. Estimate the target heart rate according to the maximum heart rate value of the dancer during the test, specifically calculated by the following formula: , where THR represents the target heart rate of the dancer, k represents the age of the dancer, p represents the heart rate reference value of the dancer, and g represents the maximum number of breaths per unit time of the dancer during the exercise; S33. Calculate the aerobic heart rate threshold according to the age of the dancer and the minimum heart rate during the training, specifically calculated by the following formula: , Among them, OKL represents the aerobic heart rate threshold, and d represents the minimum heart rate of the dancer during training.

[0009] Further, the normalization process in step S4 is specifically represented by the following formula: , where x represents the heart rate load prediction index of the dancer's aerobic endurance training after normalization, represents the average value, represents the minimum value of the heart rate load prediction index, represents the maximum value of the heart rate load prediction index.

[0010] Further, the process in step S5 is as follows: S51. Substitute the normalized indicators into the heart rate load regression equation to calculate the heart rate load value; S52. The heart rate load regression equation is represented by the following formula: , where y represents the heart rate load value of aerobic endurance training, b represents the constant term, , , , respectively represent the regression coefficients of each prediction index; S53. To ensure the prediction accuracy, the prediction value is corrected for error by the empirical coefficient method, which is represented by the following formula: , where, represents the heart rate load value of aerobic endurance training after error correction, represents the empirical coefficient; S54. Optimize the obtained heart rate load regression equation.

[0011] Further, the process in step S6 is as follows: S61. Collect the heart rate of the dancer during training; S62. Calculate the percentage of the heart rate during training within the heart rate reserve value, which is represented by the following formula: , S63. Estimate the energy consumption of the dancer per minute through the metabolic equivalent and the heart rate percentage, so as to specify the corresponding training plan according to the training goal and adjust the energy consumption.

[0012] Further, the calculation method in step S63 is represented by the following formula: , where, is a conversion coefficient used to convert the calculation result from kilocalories to kilocalories per minute.

[0013] Further, step S54 includes the following: S541: Vectorize the heart rate load prediction indicators in the heart rate load regression equation; S542: Update the weight values in the heart rate load regression equation through an iterative method; S543: Introduce interaction terms and non-linear terms into the heart rate load regression equation to capture the interaction and non-linear relationship between different heart rate load prediction indicators; S544: Introduce a regularization term into the heart rate load regression equation to prevent overfitting; S545: Introduce time series features into the heart rate load regression equation to reflect the change trend of the response variable; S546: Use cross-validation to evaluate the performance of the heart rate load regression equation and ensure its reliability in the application process.

[0014] To achieve dance training load prediction, this solution also provides a dance training load prediction system based on training data. Based on the above dance training load prediction method based on training data, it includes a collection module for collecting heart rate, a heart rate index calculation module for calculating heart rate load prediction indicators, a prediction module for predicting heart rate load, and an energy consumption calculation module for calculating the energy consumption of dancers; The data collected by the collection module is uploaded to the heart rate index calculation module, and the data calculated by the heart rate index calculation module is uploaded to the prediction module and the energy consumption calculation module.

[0015] The technical effects of the present invention are as follows: (1) By collecting data such as the heart rate of dancers, this invention calculates various heart rate load prediction indicators, improves the accuracy of heart rate load prediction indicators through standardization processing, and then substitutes the values of various heart rate load prediction indicators into the mental load regression equation to achieve the prediction of dancers' heart rate load and perform error correction. Thus, by predicting the heart rate load, the prediction of dance training load is realized, which helps to avoid accidents during training and can formulate corresponding training plans for dancers according to the above method; (2) Based on the prediction of heart rate load, this solution can also calculate the energy consumption of dancers, be able to calculate the energy consumption of dancers during training, and thus can plan and adjust the training cycle and training intensity of dancers according to the energy consumption, which helps to formulate a more standardized training plan; (3) The heart rate load regression equation in this solution includes the heart rate reserve value, target heart rate, and heart rate threshold of the dancer respectively, enabling the heart rate load prediction of this solution to combine data from multiple aspects, making the predicted data more accurate. And through data standardization processing and error correction, the accuracy of the data can be further improved; (4) This solution can not only predict the load of the dancer through heart rate load prediction, but also obtain the energy consumption of the dancer by calculating the energy consumption. When the heart rate of the dancer is about to reach the predicted value or the energy consumption value is large, it can be considered that the training load is large, so that the load level of the dancer can be understood from multiple aspects, further ensuring the safety of the dancer; (5) By optimizing the heart rate load regression equation, this solution can make the optimized heart rate load regression equation applicable to the prediction of heart rate load values during the long-term training of dancers. And the calculation efficiency can be improved by vectorizing the data. The accuracy of the heart rate load regression equation prediction can also be improved by updating weights, constraining weights, introducing time series, etc., thereby improving the practicability of this solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flowchart of the present invention.

[0017] Figure 2 is a structural connection block diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The technical solution of the present invention will be described clearly and completely below in conjunction with specific embodiments and the accompanying drawings.

[0019] See Figure 1 , a dance training load prediction method based on training data, including the following steps: Embodiment 1:

[0020] S1. Obtain the basic heart rate of the dancer; , where s represents the basic heart rate, i.e., the resting heart rate, t represents the heart beat time, and n represents the number of heart beats; S2. Calculate the metabolic equivalent of the dancer, which is used to calculate the heart rate load value and perform endurance training to obtain the maximum heart rate value during the training process; S3. Calculate the heart rate load prediction index of the dancer according to the basic heart rate; S4. Perform standardization processing on the obtained heart rate load prediction index; S5. Predict the heart rate load value through a linear regression equation; S6. Calculate the energy consumption according to the heart rate data during the exercise process.

[0021] The training data in this solution is specifically heart rate data, and subsequent predictions are achieved through the collection and calculation of heart rate data.

[0022] Furthermore, the process of step S2 is as follows: S21. After obtaining the basic heart rate, the dancer makes up-and-down jumping movements for a certain period of time to obtain the jumping height H. S22. Calculate the metabolic equivalent of the dancer according to the following formula: , where MET represents the metabolic equivalent, F represents the average heart rate during exercise, and 0.2 are the empirical coefficients in the formula, used to adjust the influence of height on the metabolic equivalent, 3.5 represents the oxygen consumption at rest, and 1 MET = 3.5 mL / kg / min; S23. Conduct aerobic endurance training, and record the heart rate data and respiratory rate of the dancer for subsequent prediction of heart rate load through linear regression analysis.

[0023] The metabolic equivalent (MET) is an index used to measure the energy consumption of physical activities. It represents the energy consumption level during a certain activity relative to the quiet rest state. 1 MET is defined as the amount of oxygen consumed per kilogram of body weight per minute in the quiet state. This value is based on the amount of oxygen consumed per kilogram of body weight per minute by an average adult in the quiet state. MET can be used to quantify the exercise intensity and compare the energy consumption of different activities.

[0024] Furthermore, the process of step S3 is as follows: S31. Estimate the heart rate reserve value of the dancer according to the recorded data, specifically calculated by the following formula: , where SRF represents the heart rate reserve value of the dancer, and u represents the maximum heart rate value during the test; the heart rate reserve value is an index used to calculate the difference between an individual's maximum heart rate and resting heart rate, and is usually used to develop personalized exercise training plans, especially in aerobic exercise training; S32. Estimate the target heart rate according to the maximum heart rate value of the dancer during the test, specifically calculated by the following formula: , Among them, THR represents the target heart rate of the dancer, k represents the age of the dancer, p represents the reference heart rate value of the dancer, and g represents the maximum number of breaths per unit time during the dancer's exercise. Among them, the reference heart rate value p of the dancer is the maximum heart rate value u during the test. The target heart rate (abbreviated as THR) is the heart rate range that should be maintained during exercise to achieve a specific training effect. This range is usually expressed as a percentage of the maximum heart rate and is an important indicator for measuring exercise intensity. S33. Calculate the aerobic heart rate threshold based on the age of the dancer and the minimum heart rate during the training process, specifically calculated through the following formula: , Among them, OKL represents the aerobic heart rate threshold, and d represents the minimum heart rate of the dancer during the training process.

[0025] The aerobic heart rate threshold refers to the critical point at which lactic acid begins to accumulate in the blood during aerobic exercise. When the exercise intensity exceeds this threshold, the production rate of lactic acid will exceed its clearance rate, resulting in a gradual increase in the lactic acid concentration in the blood. In aerobic exercise and endurance training, the lactic acid threshold is a very important concept because it marks the transition point from aerobic metabolism to anaerobic metabolism. Once the heart rate corresponding to the lactic acid threshold is determined, this heart rate can be called the aerobic heart rate threshold or simply the aerobic heart rate threshold. This solution mainly targets the aerobic training part in dance training.

[0026] Furthermore, the normalization process in step S4 is specifically represented by the following formula: , Among them, x represents the heart rate load prediction index of the dancer's aerobic endurance training after normalization, represents the average value, represents the minimum value of the heart rate load prediction index, represents the maximum value of the heart rate load prediction index.

[0027] In order to achieve the normalization process, MET, SRF, THR, and OKL need to be substituted into the above formula for calculation respectively to calculate the heart rate load prediction index after normalization for each item. Taking MET as an example, since the height of each jump of the dancer during the jump process is different, multiple MET values will be obtained during the test. Arrange the MET values in ascending order, and represent the minimum value as , the maximum value is , and the average value is .

[0028] Furthermore, the process in step S5 is as follows: S51. Substitute the standardized indicators into the heart rate load regression equation to calculate the heart rate load value; S52. The heart rate load regression equation is expressed by the following formula: , where y represents the heart rate load value of aerobic endurance training, and b represents the constant term; , , , respectively represent the regression coefficients of each prediction index; in this embodiment, , , , are respectively 108.17, 0.521, 0.255, 0.338; S53. In order to ensure the prediction accuracy, the prediction value is corrected for error by the empirical coefficient method, which is expressed by the following formula: , where, represents the heart rate load value of aerobic endurance training after error correction, represents the empirical coefficient; S54. Optimize the obtained heart rate load regression equation.

[0029] Since y is the theoretically maximum heart rate load value, not the actual one, and when the heart rate of the dancer reaches the maximum load value during training, the actual heart rate value is very likely to be lower than the theoretically maximum heart rate load value. Therefore, it is necessary to adjust and correct the heart rate load value according to the teacher's experience and understanding of the dancer, so as to reduce the risk in terms of the safety of the dancer. The range of can be between 10% and 20%, and can be adjusted according to the actual situation specifically.

[0030] Furthermore, the process in step S6 is as follows: S61. Collect the heart rate of the dancer during training; S62. Calculate the percentage of the heart rate during training within the heart rate reserve value, which is expressed by the following formula: , S63. Estimate the energy consumption per minute of the dancer through the metabolic equivalent and the heart rate percentage, so as to formulate corresponding training plans according to the training objectives and adjust the energy consumption.

[0031] Furthermore, the calculation method in step S63 is expressed by the following formula: , where, is a conversion coefficient used to convert the calculation result from kilocalories to kilocalories per minute.

[0032] Monitoring energy consumption helps prevent overtraining. If a dancer's energy consumption is abnormally high, it may indicate that their training load is too heavy and needs adjustment. By measuring the energy consumption of dancers during training, the training load can be quantified. The amount of energy consumption can be used as an indicator of the training volume, helping the instructor evaluate whether the training plan is reasonable and adjust the training intensity and duration. Moreover, the measurement of energy consumption can help determine the nutritional needs and recovery strategies of dancers. Dancers need to consume sufficient energy and nutrients to supplement the energy consumption during training and promote recovery.

[0033] If the total time of dancers during dance training is too long, it will also impose a relatively large load on the body, thus having an adverse impact. Dance teachers should reasonably arrange intense exercises because, when the training intensity is relatively high, the longer the exercise time, the more energy is consumed, and the greater the training load. When reaching the limit that dancers can bear, it will have a certain impact on their bodies. Therefore, not only the heart rate load needs to be predicted, but also the energy consumption of dancers needs to be monitored.

[0034] Since the physical qualities of different dancers are different, there will also be differences in the measured heart rate data, resulting in different predicted heart rate load values. Therefore, different individuals can be predicted in the manner described in this solution, which helps to formulate different training plans for different dancers, that is, personalized training plans.

[0035] The working process of this embodiment is as follows: First, dancers need to wear a wearable device that can collect body data such as heart rate. Then, let the dancers conduct short-term training to obtain corresponding body data. Then, process the obtained data to obtain a heart rate load prediction index and substitute it into the heart rate load regression equation to achieve the prediction of the heart rate load value.

[0036] Embodiment 2:

[0037] Based on Embodiment 1, this embodiment provides another embodiment aimed at measuring the heart rate load value of dancers during long-term training. Embodiment 1 is used to measure the heart rate load value of dancers in the current situation, so the method adopted is short-term training testing. The training time is short and the physical energy consumption is less, so the measurement result has a small error. However, during long-term training of dancers, the training time is long and the physical energy consumption is large. It is difficult for the heart rate load value calculated only by short-term training to conform to the heart rate load value of dancers during the entire long-term training. Therefore, a method that can continuously predict the heart rate load value of dancers is needed. In view of this situation, this application provides the following embodiment, which specifically includes the following content: Further, step S54 includes the following: S541. Vectorize the heart rate load prediction indicators in the heart rate load regression equation; S542. Update the weight values in the heart rate load regression equation through an iterative method; S543. Introduce interaction terms and non-linear terms into the heart rate load regression equation to capture the interaction and non-linear relationship between different heart rate load prediction indicators; S544. Introduce a regularization term into the heart rate load regression equation to prevent overfitting; S545. Introduce time series features into the heart rate load regression equation to reflect the change trend of variables; S546. Use cross-validation to evaluate the performance of the heart rate load regression equation and ensure its reliability during application.

[0038] For step S541, the present embodiment is implemented in the following manner: Vectorization of heart rate load prediction indicators: Since there may be multiple dancers during dance training, if the physical data of each dancer is calculated one by one, it will take a long time. Therefore, to save time, the data in the heart rate load indicators can be vectorized, and then the heart rate composite indicators of all dancers can be calculated at once through vectorized operations, thereby reducing the calculation process and calculation time and improving the calculation efficiency. The specific method is as follows: (1) Vectorize the metabolic equivalent of task (MET): Define the average heart rate vector F, , where represents the average heart rate of the i-th dancer; Define the height vector H, , where represents the height of the i-th dancer's jump; After that, replace the scalar operations in the formula with vector operations. At this time, MET in the original formula is a vector, representing the metabolic equivalent of task of all dancers. Thus, through vectorized operations, the metabolic equivalent of task of all dancers can be obtained at once. Through the above method, the metabolic equivalent of task of multiple dancers can be calculated at once using matrix operations, significantly improving the calculation efficiency. If more dancers need to be added, only the vector needs to be extended without modifying the formula, saving calculation time.

[0039] (2) Vectorize the target heart rate: If there are multiple dancers, the age k, heart rate reference value p, maximum breathing rate g, and resting heart rate s can be respectively formed into vectors, and then the target heart rate of all dancers can be calculated at once through vectorized operations. The specific process is as follows: Define the age vector k, k = , where represents the age of the \(i\)-th dancer; Define the heart rate reference value vector \(p\), \(p = , where represents the heart rate reference value of the \(i\)-th dancer; Define the maximum breathing rate vector \(g\), \(g = , where represents the maximum number of breaths per unit time of the \(i\)-th dancer during exercise; Define the resting heart rate vector \(s\), \(s = , where represents the resting heart rate of the \(i\)-th dancer; After that, replace the scalar operations in the original formula with vector operations. In the formula, THR is a vector representing the target heart rate of all dancers.

[0040] (3) For the heart rate reserve value SRF and the aerobic heart rate threshold OKL, also perform the above method to vectorize them, representing the heart rate reserve value of all dancers and the aerobic heart rate threshold of all dancers respectively. After that, substitute the vectorized heart rate load prediction index into the heart rate load regression equation for subsequent heart rate load calculation. Thus, through the above method, it is beneficial to calculate the heart rate load values of all dancers and improve the calculation efficiency.

[0041] For step S542, in this embodiment, the gradient descent method is used to update the weight values in the heart rate load regression equation, and all the weight values in the heart rate load regression equation are updated in this way. For the convenience of description, in the following content, the weight vector in the formula is represented by \(v\), \(v = , and the heart rate load prediction index is represented by \(z\); (1) Establish the parameter initialization formula: , ; where represents the initial value of the weight vector, represents the initial value of the constant term, represents a normal distribution with a mean of 0 and a covariance matrix of 0.01 times the identity matrix. Through the above method, the initial values of the weight values and the constant term are determined for subsequent updating of the weight values; (2) Establish the forward propagation formula: ; where represents the predicted value at the \(t\)-th moment; \(b\) represents the constant term; represents the linear part, which is used to capture the linear relationship between the features at the current moment and the target value. Specifically, represents the transpose of the weight vector. Since the weight vector v is a column vector, it can be transformed into a row vector through transpose. represents the feature vector input at time t. Since dancers consume a large amount of energy during long-term dance training and their body data will continuously change, it is necessary to collect body data in multiple time periods during the entire dance training process to continuously update the heart rate load index (i.e., the feature vector) in the formula, so that the predicted value can more conform to the current heart rate load value of the dancer; α represents the weight of the time series factor. represents the non-linear part, where represents the sequence feature, which is the feature set of the entire sequence. For example , where the represents the feature vector input at time t. Thus, it can be understood that contains the historical feature vectors of the past period of time, and then is input into the LSTM network, and the LSTM network will process and output the time-related feature representation , which is used to analyze the change relationship of the heart rate load prediction index over time, thereby improving the prediction ability of the model.

[0042] (3)Establish the loss function formula: , where is the mean square error term, which is used to measure the deviation between the predicted value and the true value. Among them, the represents the true value at time t, is used to simplify the subsequent gradient calculation; is the L2 regularization term, which is used to prevent overfitting and constrain the weight magnitude. The parameter therein is , The value of is usually taken as 0.01 - 0.1. For the true value therein, it can be obtained by asking professional physicians or through professional instruments, etc. Then, the obtained heart rate load value is used as the true value, the heart rate load value predicted by the model is used as the predicted value, and the true values and predicted values at multiple moments are used as samples (training set) to train the model to improve the prediction accuracy of the model.

[0043] (4)Perform gradient calculation and momentum update: The gradient of the loss function is calculated by the following formula: ; where represents the gradient of the loss function, represents the main gradient direction, represents the regularization gradient, which is used to shrink the weight towards zero; The calculation formula for momentum update is as follows: ; Among them, represents the change in the weight vector v at time t, represents the change in the weight vector at time t - 1, represents the momentum factor (usually 0.9), which is used to retain the historical gradient direction to accelerate convergence, represents the base learning rate (usually 0.001 - 0.1). By the above method, the change in the weight can be adjusted to reduce the error.

[0044] (5) Establish an adaptive learning rate formula: The base learning rate in the above formula can be adaptively adjusted through the following formula: ; Among them, represents the base learning rate at time t, represents the initial learning rate, represents the decay coefficient, t represents the number of iterations. Through the above formula, the formula can maintain a relatively large learning rate at the beginning of prediction to quickly approach the optimal solution. As time goes by and the number of iterations gradually increases, the base learning rate can gradually decrease with the increase of the number of iterations, which helps the model to finely adjust the parameters and approach the optimal solution, thus realizing the automatic adjustment of the learning rate and taking into account both the convergence speed and the accuracy.

[0045] Through the above gradient descent method, the weights can be iteratively adjusted over time to adjust the model parameters, enabling the model to gradually approach the optimal solution, thereby reducing the error between the predicted value and the true value of the model and significantly improving the accuracy of the model's prediction of the heart rate load value.

[0046] For step S543, in this embodiment, interaction terms and non - linear terms are introduced into the original heart rate load regression equation. The interaction term is the product of two variables (i.e., two different heart rate load prediction indicators), which is used to capture the synergistic effect. The non - linear term represents the quadratic term of the variable. Specifically, this embodiment is carried out in the following way: The heart rate load regression equation after introducing the interaction term and non - linear term is as follows: ; Among them, is the interaction term, representing the synergistic effect between MET and SRF. When , it means that when both MET and SRF increase, the promoting effect on y will be enhanced, that is, a positive impact is generated, increasing the value of y; when When it indicates that when both MET and SRF increase simultaneously, their effects may offset each other, that is, a negative impact is produced; through the above method, the impact on the predicted value y when both MET and SRF increase simultaneously can be analyzed. For whether it is greater than 0 or less than 0, it can be seen from the changes of MET and SRF in the original heart rate load regression equation. When an increase in MET or SRF causes an increase in y, that is, a positive impact is produced, otherwise it is a negative impact. The weight The specific value can be adjusted by using the above-mentioned gradient descent method; represents a non-linear term, which is used to capture the non-linear relationship between THR and y, such as a U-shaped or inverted U-shaped curve. The positive or negative of determines the opening direction of the curve. When the relationship between THR and y presents a U-shaped curve with an upward opening, indicating that the predicted value y is larger when THR is extremely low or extremely high; when the relationship between THR and y presents an inverted U-shaped curve with a downward opening, indicating that y is larger when THR is moderate. The weight The specific value can also be adjusted by using the above-mentioned gradient descent method.

[0047] Through the above method, the relationship between variables can be captured. The original heart rate load regression equation is a linear regression equation, which is used to assume a linear relationship between the independent variable (heart rate load prediction index) and the dependent variable (predicted heart rate load value). The interaction term and the non-linear term will allow the model to capture more complex relationships between the independent variable and the dependent variable, avoiding large errors in the model due to ignoring the interaction or non-linear effects between variables, thereby reducing the prediction error and making the predicted value more accurate; The above content provided in this embodiment is only one implementation manner. The variables in the interaction term can be MET and SRF, or other two variables. Similarly, for the non-linear term, the variable in the non-linear term in this embodiment is THR, or other variables. Therefore, the above content is only one embodiment of this solution and does not limit the variables in the formula.

[0048] For step S544, this embodiment introduces a regularization term into the original heart rate load regression equation to prevent overfitting. Overfitting refers to the phenomenon that the model performs well on the training data, that is, the error is very low, but performs poorly on the test data or actual data, that is, the prediction error is large. The essence of overfitting is that the model over-learns the noise and details in the training data, resulting in a decline in the generalization ability of the model. In response to this situation, this embodiment adopts the following method for improvement: Taking the original heart rate load regression equation as an example, after introducing the regularization term, it is as follows: ; Among them, is the regularization term, is the regularization coefficient, is the model coefficient, that is , 4 represents the number of model coefficients. By the above method, the value of the model coefficient (i.e., weight) can be constrained to prevent it from being too large, thereby preventing overfitting and improving the generalization ability of the model; Specifically, in this solution , in this solution, interaction terms, non-linear terms and regularization terms can be cited simultaneously. When regularization is required for all model coefficients (i.e., weights), 4 in the above formula should be replaced by the total number n of model coefficients.

[0049] For step S545, this embodiment introduces time series features to analyze the change of the heart rate load prediction value over time, that is, the dependence on time, and to understand the change trend of the heart rate load prediction index over time, thereby improving the prediction ability of the model. Specifically, the method adopted in this embodiment is as follows: Introduce time series features into the original heart rate load regression equation. The specific formula is as follows: ; Among them, represents the predicted value at time t, represents the predicted value at the previous moment, represents the target heart rate change rate, that is , which is used to capture the change trend of the independent variable THR and can reflect the influence on through the short-term change of THR. Thus, the historical relationship between the independent variable (heart rate prediction index, specifically THR in this embodiment) and the dependent variable (predicted value) can be reflected through the above method, as well as the influence of the predicted heart rate load value at the previous moment on the predicted heart rate load value at the current moment, thereby improving the prediction accuracy of the model. The change rate in this embodiment is the target heart rate change rate. According to the actual situation, it can also be replaced by the metabolic equivalent change rate or other change rates, which is beneficial to reflecting the relationship between the changes of different heart rate load prediction indexes and the predicted value. Weights and can be adjusted and valued by the above gradient descent method, and a regularization term can be added to the above formula to prevent overfitting.

[0050] For step S546, this embodiment uses cross-validation to evaluate the performance of the model. Specifically, by dividing the data set into multiple subsets and training and validating the model repeatedly, the generalization ability of the model can be estimated more accurately. The main purpose is to evaluate the performance of the model on new data and avoid overfitting. The specific method is as follows: First, collect several groups of data to form a data set D. Each group of data includes various parameters contained in the content of steps S541 - S545, such as various heart rate load prediction indicators, corresponding heart rate load values, etc. Then, randomly divide the data set D into k subsets of equal size. , and use as the validation set, and the other k - 1 subsets as the training set for training the model. After training is completed, use the validation set to evaluate the model performance, which is used to judge the error between the predicted value and the true value of the model. This entire process will be repeated k times, each time using a different subset. as the validation set. After that, calculate the average error of k times.

[0051] Through the above method, it can be ensured that each subset (sample) can be used as both the validation set and the training set, thus making full use of the data. Moreover, through multiple trainings and validations, the performance of the model on different data can be evaluated more reliably, making the evaluation more comprehensive. And through the above method, the generalization ability of the model can be better reflected, reducing the risk of overfitting.

[0052] The working method of this embodiment is as follows: During the training process of the dancer, collect the body data of the dancer to calculate the corresponding heart rate load prediction indicators. Since the training time of the dancer is relatively long and the physical energy consumption is relatively large, it is necessary to collect data in real time and update the heart rate load prediction indicators regularly according to the collected data, so as to continuously predict the heart rate load value of the dancer in the current state. In order to achieve the above functions and make the predicted value more accurate, this solution is optimized based on the heart rate load regression equation. Specifically, in this embodiment, various heart rate load prediction indicators are vectorized for subsequent calculations. When multiple dancers are training, the body data of each dancer can be collected, calculated to obtain the heart rate load prediction indicators, and vectorized. Subsequently, through one calculation process, the heart rate load value of each dancer can be predicted, thus improving the prediction efficiency. Since the physical energy of the dancer will decline over time, if the weight value remains unchanged, the error may increase. Therefore, it is necessary to update the weight in the heart rate load regression equation. This embodiment uses the gradient descent method to update the weight in the equation, so as to reduce the error between the predicted value and the true value of the model and improve the accuracy of the heart rate load regression equation for predicting the heart rate load value. Since the above heart rate load regression equation is a linear regression equation, that is, it is assumed that there is a linear relationship between the heart rate load prediction index and the heart rate load value. Therefore, on this basis, in order to capture more complex relationships between variables, this embodiment introduces interaction terms and non-linear terms into the heart rate load regression equation, enabling the heart rate load regression equation to capture more complex relationships between independent variables and dependent variables, avoiding large errors in the model caused by ignoring the interaction or non-linear effects between variables, effectively reducing the prediction error, and making the predicted value more accurate; To avoid overfitting of the heart rate load regression equation, that is, having a small error for the training data but a large error for predicting new data, this embodiment introduces a regularization term to constrain the magnitude of the weights, avoiding large weights that may lead to large errors, thereby avoiding overfitting and improving the generalization ability of the heart rate load regression equation; Moreover, since dancers have a long training time, various physical data of dancers will change over time, and the data collected at different time periods have different magnitudes. Therefore, this embodiment introduces time series features into the heart rate load regression equation to analyze the variation of the heart rate load predicted value over time, that is, its dependence on time, and to analyze the variation trend of the heart rate load prediction index over time, thereby improving the prediction ability of the heart rate load regression equation; After that, it is necessary to evaluate the accuracy of the heart rate load regression equation prediction. This embodiment uses cross-validation to detect the error magnitude of the predicted value, can make full use of the data, and can further avoid overfitting.

[0053] Since the above content mentions multiple optimization methods, those skilled in the art can, according to the actual situation, introduce all of them into the heart rate load regression equation, or select several of them to introduce into the heart rate load regression equation. This solution does not limit the above methods.

[0054] Since the actual maximum heart rate value of a dancer is lower than the theoretical value, the teacher needs to set an empirical coefficient according to the dancer's physical condition, and multiply the theoretical maximum heart rate value by the empirical coefficient to obtain an actual maximum heart rate value close to the actual situation. After the dancer undergoes training, their physical fitness will be stronger than before. At this time, the dancer can be tested again to obtain the current heart rate load prediction index, and substitute it into the linear regression equation for calculation, so as to calculate the theoretical and current maximum heart rate values. Then, calculate the obtained value with the empirical coefficient to obtain the current actual maximum heart rate value, thereby achieving prediction.

[0055] Heart rate testing is an effective way to monitor the load of dancers. By monitoring the heart rate, the training load can be directly understood. During the usual training process, dancers should pay attention to the practice of strength, flexibility, speed and coordination, continuously improve their physical fitness, and targeted training can achieve better results.

[0056] In fact, there are many factors affecting the load capacity of dancers. When teachers conduct training, they should adjust according to the actual situation of the dancers. In this process, teachers need to consider the physiological and psychological changes of the dancers, and combine the characteristics of the project to determine the load level of the whole class. When conducting a training, the intensity is from weak to strong, and then from strong to weak. In this process, according to the training principle of alternating movement and stillness, high and low combination, the effective development of the training is ensured.

[0057] See Figure 2 , in order to further implement the above solution, the present application also discloses a dance training load prediction system based on training data, based on the above-mentioned dance training load prediction method based on training data, including a collection module for collecting heart rate, a heart rate index calculation module for calculating heart rate load prediction indicators, a prediction module for predicting heart rate load, and an energy consumption calculation module for calculating the energy consumption of dancers; The data collected by the collection module is uploaded to the heart rate index calculation module, and the data calculated by the heart rate index calculation module is uploaded to the prediction module and the energy consumption calculation module.

[0058] The collection module in this embodiment is a wearable device, specifically a portable device such as a smart watch for collecting heart rate and jump height. The heart rate index calculation module, the prediction module and the energy consumption calculation module can be installed on a computer, and data analysis and calculation are performed through computer software. Preferably, the heart rate index calculation module, the prediction module and the energy consumption calculation module can be integrated in the same software, which is convenient for data processing and calculation. This is a conventional means that those skilled in the art can easily think of and will not be elaborated here.

[0059] The training load prediction in this solution is applied to dancers, specifically applicable to students majoring in dance performance, especially students who have not received professional dance skill training. The students are older and their physical bones have developed maturely. Their dance nature and dance sense do not meet the professional requirements. Therefore, during the training process, it is more necessary to predict the heart rate load through the content recorded in this solution, and teachers can formulate relevant training plans; Through training, the physical fitness of students learning dance can be improved and their movements can be more standard. During the training process, teachers must follow the laws of human physiological activities. When formulating training goals, they should set different load levels according to the actual situation of the students and in combination with this plan, flexibly adjust the training methods, scientifically formulate training plans, carry out training contents of different intensities, and ensure the effective progress of the courses; Moreover, there are obvious differences in the physical fitness of different students. During training, teachers should adopt differentiated teaching methods according to the actual situation of the students: they can group students with similar physical load tolerances and determine suitable sports projects and plans for each group. For example, for students with relatively weak physiques, they can adopt a way with a smaller exercise intensity and gradually extend the exercise time, and make the best use of the situation and promote step by step to finally reach the normal training standard; Furthermore, for students who have not received professional dance skill training as mentioned above, during the dance performance, there will also be inappropriate expressions in terms of body and emotion, and it is difficult to achieve the natural rhythm and professional requirements of dance performance. This is not only related to the external training plan, but also related to the internal emotion and the understanding of things; In terms of the emotional expression of dance, it can be divided into four elements: emotion understanding, emotion expression, emotion movement, and emotion regulation; Emotion understanding refers to the ability of an individual to accurately recognize and interpret their own and others' emotional states. In dance teaching, students need to deeply understand the deep emotions contained behind dance movements, such as sadness, joy, love, hatred, etc., to ensure that their performances can accurately interpret dance works. Teachers should assist students in this process through careful observation, in-depth discussion, and personal experience to enhance their feelings and understanding of the emotions in dance works.

[0060] Emotion expression emphasizes the ability of an individual to effectively convey their own emotional state. In dance practice, dancers use artistic means such as facial expressions, body language, and movements to convey and express their inner emotions, enabling the audience to deeply feel the emotional state of the dancers. Excellent emotion expression ability can greatly enhance the appeal of dance works and thus trigger the emotional resonance of the audience.

[0061] Emotion application refers to the ability of an individual to effectively use emotions to promote and guide the thinking and behavior process. For dancers, it is manifested in the creation and performance process that they can use emotions to stimulate creativity, or adjust their own performance styles under the guidance of emotions to adapt to the expressions of different dance works.

[0062] Emotional regulation is an individual's ability to manage and adjust their own emotional state. In dance learning, dancers need to learn to stay calm when facing stress and challenges, and at the same time be able to release and express emotions appropriately during performances. This ability is crucial for dancers to maintain a stable performance on stage and helps to avoid affecting the quality and effect of dance works due to emotional out-of-control.

[0063] Therefore, based on this solution, teachers can scientifically awaken the inner emotions of dancers through dance training to achieve awakening teaching. Specifically, it is necessary to truly integrate dance into natural life to stimulate the transmission of dancers' natural emotions, awaken the natural and flexible state of the body and emotions from within, make the body more relaxed and the dance postures more flexible, and release self-emotions through dance postures, continuously stimulating the flexibility, agility, fluency and infectivity of dancers' performances, so as to achieve a new teaching concept including emotional inspiration, physical awakening and naturalness of all things. By combining with this solution, dance teaching can be made more scientific, systematic and professional, and thus dancers can reach the best state of performance in self-perception, self-release and self-integration.

[0064] The above are only exemplary embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent structural transformation made by using the content of the specification and drawings of the present invention under the technical concept of the present invention, or direct / indirect application in other related technical fields, is included in the patent protection scope of the present invention.

Claims

1. A dance training load prediction method based on training data, characterized in that It includes the following steps: S1. Obtain the basic heart rate of the dancer; ， Among them, s represents the basic heart rate, t represents the heartbeat time, and n represents the number of heartbeats; S2. Calculate the metabolic equivalent of the dancer, which is used to calculate the heart rate load value, and conduct endurance training to obtain the maximum heart rate value during the training process; S3. Calculate the heart rate load prediction index of the dancer based on the basic heart rate; S4. Perform standardization processing on the obtained heart rate load prediction index; S5. Predict the heart rate load value through a linear regression equation; S6. Calculate the energy consumption based on the heart rate data during the exercise process.

2. The dance training load prediction method based on training data according to claim 1, wherein The process of step S2 is as follows: S21. After obtaining the basic heart rate, the dancer makes up-and-down jumping movements for a certain period of time to obtain the jumping height H; S22. Calculate the metabolic equivalent of the dancer according to the following formula: ， Among them, MET represents the metabolic equivalent, and F represents the average heart rate during the exercise process; S23. Conduct aerobic endurance training, and record the heart rate data and respiratory rate of the dancer for subsequent prediction of the heart rate load through linear regression analysis.

3. The dance training load prediction method based on training data according to claim 1, wherein The process of step S3 is as follows: S31. Estimate the heart rate reserve value of the dancer based on the recorded data, specifically calculated by the following formula: ， Among them, SRF represents the heart rate reserve value of the dancer, and u represents the maximum heart rate value during the test process; S32. Estimate the target heart rate based on the maximum heart rate value of the dancer during the test process, specifically calculated by the following formula: ， Among them, THR represents the target heart rate of the dancer, k represents the age of the dancer, p represents the heart rate reference value of the dancer, and g represents the maximum number of breaths per unit time of the dancer during the exercise process; S33. Calculate the aerobic heart rate threshold based on the age of the dancer and the minimum heart rate during the training process, specifically calculated by the following formula: ， Among them, OKL represents the aerobic heart rate threshold, and d represents the minimum heart rate of the dancer during the training process.

4. The dance training load prediction method based on training data according to claim 1, wherein The standardization processing in step S4 is specifically represented by the following formula: ， Among them, x represents the predicted index of the heart rate load of the dancer's aerobic endurance training after standardization, represents the average value, represents the minimum value of the predicted index of the heart rate load, represents the maximum value of the predicted index of the heart rate load.

5. The dance training load prediction method based on training data according to claim 3, characterized in that The process in step S5 is as follows: S51. Substitute the standardized heart rate load prediction indexes into the heart rate load regression equation to calculate the heart rate load value; S52. The heart rate load regression equation is represented by the following formula: ， Among them, y represents the heart rate load value of aerobic endurance training, and b represents the constant term. 、 、 、 respectively represent the regression coefficients of each prediction index. S53. In order to ensure the prediction accuracy, the prediction value is corrected for errors through the empirical coefficient method, represented by the following formula: ， Among them, represents the aerobic endurance training heart rate load value after error correction, represents the empirical coefficient; S54. Optimize the obtained heart rate load regression equation.

6. The dance training load prediction method based on training data according to claim 1, wherein The process in step S6 is as follows: S61. Collect the heart rate of the dancer during the training process; S62. Calculate the percentage of the heart rate during training within the heart rate reserve value range, represented by the following formula: ， S63. Estimate the energy consumption per minute of the dancer through the metabolic equivalent and heart rate percentage, which is convenient for formulating corresponding training plans according to the training goals and adjusting the energy consumption.

7. The dance training load prediction method based on training data according to claim 6, wherein The calculation method in step S63 is represented by the following formula: ， wherein, is a conversion coefficient for converting the calculation result from kcal to kcal per minute.

8. The dance training load prediction method based on training data according to claim 5, wherein, Step S54 includes the following contents: S541. Vectorize the heart rate load prediction indexes in the heart rate load regression equation; S542. Update the weight values in the heart rate load regression equation through an iterative method; S543. Introduce interaction terms and non-linear terms into the heart rate load regression equation to capture the interaction and non-linear relationship between different heart rate load prediction indexes; S544. Introduce a regularization term into the heart rate load regression equation to prevent overfitting; S545. Introduce time series features into the heart rate load regression equation to reflect the change trend of variables S546. Use cross-validation to evaluate the performance of the heart rate load regression equation and ensure its reliability during application.

9. A dance training load prediction system based on training data, based on the dance training load prediction method based on training data as described in claims 1-8, characterized in that, It includes a collection module for collecting heart rate, a heart rate index calculation module for calculating heart rate load prediction indicators, a prediction module for predicting heart rate load, and an energy consumption calculation module for calculating the energy consumption of dancers; The data collected by the collection module is uploaded to the heart rate index calculation module, and the data calculated by the heart rate index calculation module is uploaded to the prediction module and the energy consumption calculation module.

Images