Personalized badminton physical training scheme recommendation method and system

Abstract

According to the personalized badminton physical training scheme recommendation method and system provided by the invention, precise personalized adaptation of the physical training scheme is realized, and the training pertinence and efficiency are effectively improved. According to the method, the badminton player physical ability data example containing various physical ability elements and constraint boundaries is obtained, different local queue factors are mined based on the feature queues to obtain two training type description queues, and the difference of physical ability features of the individual players and the internal association between the physical ability elements can be fully captured. Through feature description matching of a specific athlete, a main example and a secondary example, the physical ability weakness of the athlete and the adaptive training direction can be accurately positioned, the subjectivity and one-sidedness of a traditional experience driven or general standardized training mode are avoided, the recommended training scheme better meets the individual demand of the athlete, and the training effect is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of training scheme recommendation, in particular to a personalized badminton physical training scheme recommendation method and system. BACKGROUND

[0002] As a competitive sport requiring speed, strength, agility and endurance, the physical level of athletes directly determines their competitive performance, tactical execution effect and risk of sports injury. With the continuous development of competitive sports, personalized physical training has become a key link to enhance the core competitiveness of athletes, and its core requirement is to accurately match the individual physical characteristics of athletes and develop targeted and efficient training schemes.

[0003] Currently, the development of badminton physical training schemes still generally relies on the experience of coaches or uses standardized training modes based on general physical indicators. Such traditional methods have many limitations: on the one hand, experience-driven scheme development is highly subjective and difficult to fully cover individual differences among athletes. For example, different athletes have significant differences in short boards in explosive power, cardiorespiratory endurance, core stability and other physical elements, and general schemes cannot achieve precise reinforcement. On the other hand, standardized training modes do not fully exploit the internal relationships between physical data, only consider physical indicators in a single dimension, ignore the impact of the constraint relationship between physical elements on training effectiveness, and result in insufficient scientificity and effectiveness of training schemes.

[0004] To break through the limitations of traditional methods, physical training scheme recommendation methods combining data collection and analysis have gradually appeared in existing technologies. Such methods collect physical data of athletes (such as speed, strength, endurance test data, etc.), and determine training directions with simple data comparison or statistical analysis. However, existing data-driven methods still have obvious defects: first, the feature extraction of physical data is not deep enough, and only the numerical value of a single physical indicator is concerned, without forming a feature system that can fully represent the physical state of athletes; second, there is a lack of effective data mining and analysis mechanism, which cannot accurately mine key information hidden in physical data and is difficult to achieve precise matching of training types; third, the complementarity of multi-dimensional data analysis results is not fully utilized, and only a single analysis dimension is used to output training schemes, resulting in the need to improve the individual adaptation of the scheme.

[0005] In addition, the correlation relationship (i.e., constraint boundary) between different physical elements in the prior art is not effectively valued, and a systematic training scheme recommendation logic cannot be constructed based on such a correlation relationship, resulting in a problem that the recommended training scheme may be inconsistent or even conflicting between training items, affecting the maximization of training effect. Meanwhile, the existing method does not clearly distinguish between primary and secondary influencing factors when screening a matched training example for a specific athlete, making it difficult to achieve precise optimization of the training scheme.

[0006] Therefore, how to construct a badminton physical training scheme recommendation method capable of fully mining physical data characteristics of athletes, accurately utilizing constraint relationships between physical elements, and realizing individualized adaptation combined with multi-dimensional data analysis results has become a technical problem to be solved in the current field. SUMMARY

[0007] To improve the technical problems existing in the related art, the present application provides a personalized badminton physical training scheme recommendation method and system.

[0008] In a first aspect, a personalized badminton physical training scheme recommendation method is provided, the method comprising:

[0009] obtaining a badminton athlete physical data example, the badminton athlete physical data example comprising not less than two physical elements and a constraint boundary for connecting the not less than two physical elements;

[0010] extracting a feature queue of the badminton athlete physical data example, each feature in the feature queue corresponding to a feature description of a physical element in the badminton athlete physical data example;

[0011] based on the feature queue, obtaining a first physical training category description queue and a second physical training category description queue, the first physical training category description queue and the second physical training category description queue being used to represent the feature queue after a local queue factor is mined, the mined queue factors in the first physical training category description queue and the second physical training category description queue being inconsistent;

[0012] based on the first physical training category description queue and the second physical training category description queue, recommending a personalized athlete physical training scheme.

[0013] In the present application, the physical training analysis network comprises a first data analysis thread and a second data analysis thread, the first data analysis thread and the second data analysis thread being used to load feature descriptions in the feature queue into an AI vector space for analysis, the first data analysis thread and the second data analysis thread being data analysis threads with the same direction but inconsistent coefficients;

[0014] The first physical training category description queue and the second physical training category description queue are obtained based on the feature queue, and the obtaining comprises:

[0015] The feature queue is loaded into the first data analysis thread to obtain a first analysis queue, and the feature queue is loaded into the second data analysis thread to obtain a second analysis queue;

[0016] The first analysis queue is filtered to obtain a first possibility queue, and the second analysis queue is filtered to obtain a second possibility queue; the first possibility queue and the second possibility queue are used to represent the possibility of mining the local queue factor in the feature queue;

[0017] The first physical training category description queue is obtained based on the feature queue and the first possibility queue, and the second physical training category description queue is obtained based on the feature queue and the second possibility queue.

[0018] In the present application, the first physical training category description queue is obtained based on the feature queue and the first possibility queue, and the obtaining comprises:

[0019] A function processing result of the feature queue and the first possibility queue is calculated to obtain the first physical training category description queue;

[0020] The second physical training category description queue is obtained based on the feature queue and the second possibility queue, and the obtaining comprises:

[0021] A function processing result of the feature queue and the second possibility queue is calculated to obtain the second physical training category description queue.

[0022] In the present application, the first possibility queue is obtained by filtering the first analysis queue, and the filtering comprises:

[0023] A first physical index value is obtained from a physical data set;

[0024] A first data group and a second data group are calculated, the first data group is a data group with the smallest natural number in the first physical index value, and the second data group is a data group with the smallest natural number in a first comparison result, the first comparison result being a difference between one and the first physical index value;

[0025] calculate a first possibility that a queue factor of an a-th physical element in the characteristic queue in a b-th direction is excavated, the first possibility being a possibility obtained after a first function calculation value is loaded to an activation function, the first function calculation value being a calculation result of a first transition value and a first vector, the first transition value being a value obtained after a first data set minus a second data set plus a factor value in the a-th row and the b-th column of a first analysis queue, a and b being integers greater than zero;

[0026] filter the second analysis queue to obtain a second possibility queue, including:

[0027] obtain a second physical index value from the physical data set;

[0028] calculate a third data set and a fourth data set, the third data set being a data set with the smallest natural number in the second physical index value, the second data set being a data set with the smallest natural number in a second comparison result, the second comparison result being a difference between one and the second physical index value;

[0029] calculate a second possibility that a queue factor of an a-th physical element in the characteristic queue in a b-th direction is excavated, the second possibility being a possibility obtained after a second function calculation value is loaded to an activation function, the second function calculation value being a calculation result of a second transition value and a second vector, the second transition value being a value obtained after the fourth data set minus the third data set plus a factor value in the a-th row and the b-th column of a second analysis queue, a and b being integers greater than zero.

[0030] In the present application, the first physical training category description queue and the second physical training category description queue are used to recommend a personalized athlete physical training scheme, including:

[0031] obtain a characteristic description of a specific athlete from the first physical training category description queue, the specific athlete being a physical element corresponding to an a-th row in the first physical training category description queue, a being an integer greater than zero;

[0032] obtain a characteristic description of a main example from the second physical training category description queue, the main example being a physical element corresponding to an a-th row in the second physical training category description queue, the main example and the specific athlete corresponding to a same physical element in the badminton athlete physical data example;

[0033] obtain a characteristic description of a secondary example from the first physical training category description queue or the second physical training category description queue, the secondary example being a physical element in the badminton athlete physical data example that is not adjacent to the specific athlete and the main example.

[0034] recommending a personalized athlete physical training plan based on the characteristic description of the specific athlete, the characteristic description of the primary example, and the characteristic description of the secondary example.

[0035] In the present application, the physical training analysis network further comprises a third data analysis thread, the coefficients of the first data analysis thread and the second data analysis thread being different from the coefficients of the third data analysis thread; the method further comprises:

[0036] loading the first physical training category description queue into the third data analysis thread to obtain a third physical training category description queue;

[0037] recommending a personalized athlete physical training plan based on the first physical training category description queue and the second physical training category description queue, comprising:

[0038] recommending a personalized athlete physical training plan based on the third physical training category description queue and the second physical training category description queue.

[0039] In the present application, the recommending a personalized athlete physical training plan based on the third physical training category description queue and the second physical training category description queue comprises:

[0040] obtaining a characteristic description of a specific athlete from the third physical training category description queue, the specific athlete being a physical element corresponding to the a-th row in the third physical training category description queue, a being an integer greater than zero;

[0041] obtaining a characteristic description of a primary example from the second physical training category description queue, the primary example being a physical element corresponding to the a-th row in the second physical training category description queue, the primary example and the specific athlete corresponding to the same physical element in the badminton athlete physical data example;

[0042] obtaining a characteristic description of a secondary example from the third physical training category description queue or the second physical training category description queue, the secondary example being a physical element in the badminton athlete physical data example that is not adjacent to the specific athlete and the primary example;

[0043] recommending a personalized athlete physical training plan based on the characteristic description of the specific athlete, the characteristic description of the primary example, and the characteristic description of the secondary example.

[0044] In the present application, the recommending a personalized athlete physical training plan based on the characteristic description of the specific athlete, the characteristic description of the primary example, and the characteristic description of the secondary example comprises:

[0045] a first commonality score between the characteristic description of the specific athlete and the characteristic description of the primary example; and a second commonality score between the characteristic description of the specific athlete and the characteristic description of the secondary example;

[0046] a quantitative indicator of the physical training analysis network according to the first commonality score and the second commonality score;

[0047] recommending a personalized physical training scheme for the athlete according to the quantitative indicator of the physical training analysis network.

[0048] In the present application, the badminton athlete physical data examples constitute a database, the physical elements represent user personalized information, and the constraint boundaries represent the association relationship between the user personalized information;

[0049] The badminton athlete physical data examples are clusters of athlete physical indicators, the physical elements represent user personalized information and athlete physical indicator information, and the constraint boundaries represent the relationship between the user personalized information and the athlete physical indicator information and the relationship between user personalized information and user personalized information.

[0050] The badminton athlete physical data examples are reference data, the physical elements represent standard data, and the constraint boundaries represent the association relationship between the standard data.

[0051] In a second aspect, a personalized badminton physical training scheme recommendation system is provided, which includes a processor and a memory in communication with each other, and the processor is configured to read a computer program from the memory and execute the computer program to implement the method described above.

[0052] The personalized badminton physical training scheme recommendation method and system provided by the embodiments of the present application can achieve precise personalized adaptation of physical training schemes and effectively improve the training pertinence and efficiency. The present application can fully capture the differences in individual physical characteristics of athletes and the internal relationships between various physical elements by obtaining badminton athlete physical data examples containing various physical elements and constraint boundaries, and obtaining two types of training category description queues based on feature queue mining of different local queue factors. Subsequently, through matching of the characteristic descriptions of the specific athlete, the primary example and the secondary example, the athlete's physical short board can be accurately positioned and the training direction can be adapted, avoiding the subjectivity and one-sidedness of traditional experience-driven or general standardized training modes, so that the recommended training scheme is more suitable for the individual needs of athletes, and the training effect is significantly improved.

[0053] 2. The present invention optimizes the mining and utilization efficiency of physical fitness data and improves the scientificity of training program recommendation. The present invention adopts multi-threading (first, second and optional third data analysis threads) to perform hierarchical analysis on the feature queue, mines local queue factors of different dimensions through different coefficient data analysis threads, and realizes deep mining and multi-dimensional utilization of physical fitness data features by combining the filtering and function processing of the possibility queue. Compared with the existing simple data comparison or single-dimensional analysis method, the present invention can fully mine the key information implied in the physical fitness data, improve the comprehensiveness and accuracy of feature extraction, provide more reliable data support for the recommendation of training programs, and ensure the scientificity and rationality of the programs.

[0054] 3. The present invention defines the selection rules of specific athletes, main examples and secondary examples, calculates the quantitative indicators of the physical training analysis network by combining the commonality score, realizes the accurate selection of training examples and the quantitative optimization of programs, and further optimizes the accuracy of feature description by the third data analysis thread for the secondary processing of the first physical training category description queue, thereby reducing the errors caused by single analysis dimension. The design makes the recommendation logic of the training program clearer and traceable, improves the stability and reliability of the program recommendation, and reduces the deviation between the training program and the actual needs of the athletes.

[0055] 4. The present invention is suitable for various data scenarios and enhances the universality and practicality of the method. The badminton athlete physical fitness data examples in the present invention can adapt to various scenarios such as database, athlete physical index information cluster and reference data, and the definition of physical elements and constraint boundaries can flexibly match different data types such as user personalized information, athlete physical index information and standard data, so that the method can meet the needs of physical training program recommendation of different training scenarios (such as daily training, special improvement and pre-match preparation) and different levels of athletes (professional athletes and amateur enthusiasts), significantly improving the universality and practicality of the method and having a wide application prospect.

[0056] 5. The present invention reduces the professional threshold of training program development and improves the popularity of training guidance. The standardized data analysis process and automated program recommendation logic of the present invention reduce the over-reliance on coach experience, so that even non-senior professional coaches can use the method to develop scientific and personalized physical training programs for athletes. This helps to promote the popularity of personalized physical training guidance, enables more athletes to enjoy precise and scientific training guidance, and promotes the overall improvement of badminton training level. BRIEF DESCRIPTION OF DRAWINGS

[0057] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be regarded as a limitation to the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0058] Figure 1 A flow chart of a personalized badminton physical training scheme recommendation method provided by the embodiments of the present application. DETAILED DESCRIPTION

[0059] In order to better understand the above technical solutions, the technical solutions of the present application will be described in detail below through the drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solutions of the present application, and are not limitations to the technical solutions of the present application. In the case of no conflict, the technical features in the embodiments of the present application and the embodiments can be combined with each other.

[0060] Please refer to Figure 1 , which shows a personalized badminton physical training scheme recommendation method. The method can include the technical solutions described in the following steps 210-240.

[0061] Step 210: Obtain a badminton player physical data example, which includes no less than two physical elements and a constraint boundary for connecting the no less than two physical elements.

[0062] The constraint boundary represents the maximum allowable range of the physical element, such as: the standard exercise time is 3 hours, and the limit exercise time is 5 hours. The 5 hours is the constraint boundary.

[0063] The badminton athlete physical data example includes: first, a badminton athlete physical database is constructed, the database contains physical data examples of 50 athletes of the same age and the same project, each data example contains no less than two physical elements and corresponding constraint boundaries. Among them, the physical elements select 4 kinds of indexes related to badminton core: 1) explosive force (specifically, vertical jump height, standing long jump result); 2) cardiorespiratory endurance (specifically, 3000m running result, heart rate recovery speed); 3) agility (specifically, 5-10-5 agility running result, cross jump result); 4) core stability (specifically, plank support time, hanging leg times); the constraint boundary is the correlation threshold value between each physical element, for example, "the vertical jump height in explosive force is negatively correlated with the 5-10-5 agility running result in agility, the correlation threshold value is that when the vertical jump height is greater than or equal to 60cm, the 5-10-5 agility running result should be less than or equal to 5.8 seconds", "the 3000m running result in cardiorespiratory endurance is positively correlated with the plank support time in core stability, the correlation threshold value is that when the 3000m running result is less than or equal to 12 minutes, the plank support time should be greater than or equal to 90 seconds".

[0064] In this embodiment, the physical data example of the specific athlete (hereinafter referred to as "athlete A") to be recommended is as follows: explosive force (vertical jump 55cm, standing long jump 2.3m), cardiorespiratory endurance (3000m running 12 minutes 30 seconds, heart rate recovery speed 2 minutes to resting heart rate), agility (5-10-5 agility running 6.2 seconds, cross jump 30 times per minute), core stability (plank support 75 seconds, hanging leg 15 times), and the constraint boundary corresponding to each physical element is the same as the general correlation threshold value of the database.

[0065] Extracting the feature queue of the badminton athlete physical data example

[0066] The above physical data example is extracted for features to construct a feature queue. Each feature in the feature queue corresponds to a feature description of a physical element, which is represented in the format of "physical element type-specific index-index value-achievement status", wherein the achievement status is based on the standard value of athletes of the same age and the same project (for example, the standard value of vertical jump is 60cm, and the achievement status of athlete A is "not up to standard" for 55cm). The finally constructed feature queue F is:

[0067] F = [f1, f2, f3, f4]

[0068] Wherein, f1 (explosive force) = "explosive force - vertical jump 55 cm - not up to standard; explosive force - standing long jump 2.3 m - up to standard"; f2 (cardiorespiratory endurance) = "cardiorespiratory endurance - 3000 m run 12 minutes 30 seconds - not up to standard; cardiorespiratory endurance - heart rate recovery speed 2 minutes - up to standard"; f3 (agility) = "agility - 5-10-5 agility run 6.2 seconds - not up to standard; agility - cross jump 30 times / minute - up to standard"; f4 (core stability) = "core stability - plank support 75 seconds - not up to standard; core stability - leg lifting 15 times - up to standard".

[0069] Step 220: Extract the feature queue of the badminton player physical fitness data example, and each feature in the feature queue corresponds to the feature description of a physical fitness element in the badminton player physical fitness data example.

[0070] Optionally, the feature queue of the badminton player physical fitness data example is extracted using a physical fitness training analysis network. The physical fitness training analysis network obtains the feature description of the first physical fitness element by aggregating the physical fitness element attributes of the first physical fitness element and at least one neighbor physical fitness element of the first physical fitness element.

[0071] Step 230: Based on the feature queue, obtain a first physical fitness training category description queue and a second physical fitness training category description queue, the first physical fitness training category description queue and the second physical fitness training category description queue are used to represent the feature queue after the local queue factor is mined, and the mined queue factors in the first physical fitness training category description queue and the second physical fitness training category description queue are inconsistent.

[0072] Wherein: based on the feature queue, obtain the first and second physical fitness training category description queues

[0073] This embodiment uses a physical fitness training analysis network for feature analysis, which includes a first data analysis thread T1 and a second data analysis thread T2, both of which have the analysis direction of "physical fitness element short board - training category matching", but have different coefficients (T1 coefficient is 0.6, focusing on explosive force and agility related training matching; T2 coefficient is 0.4, focusing on cardiorespiratory endurance and core stability related training matching).

[0074] 1. Generate the first and second analysis queues: load the feature queue F into T1 and T2 respectively for analysis. T1 outputs the first analysis queue A1, focusing on the training direction corresponding to the short board of explosive power and agility, specifically "A1 = [explosive power training matching degree, agility training matching degree, cardiorespiratory endurance training matching degree, core stability training matching degree] = [0.85, 0.82, 0.30, 0.25]"; T2 outputs the second analysis queue A2, focusing on the training direction corresponding to the short board of cardiorespiratory endurance and core stability, specifically "A2 = [explosive power training matching degree, agility training matching degree, cardiorespiratory endurance training matching degree, core stability training matching degree] = [0.32, 0.28, 0.88, 0.86]".

[0075] 2. Filter the first and second possibility queues: obtain the first physical index value (the explosive power and agility standard rate of athletes of the same age, 65% and 62% respectively) and the second physical index value (the cardiorespiratory endurance and core stability standard rate of athletes of the same age, 58% and 55% respectively) from the physical database.

[0076] For the calculation of the first possibility queue P1: calculate the first data set (the data set with the smallest natural number in the first physical index value, i.e. 62%) and the second data set (the data set with the smallest natural number in the first comparison result "1 - the first physical index value", 1 - 65% = 35%); take a = 1 (explosive power element) and b = 1 (training matching degree direction), calculate the first transition value = 62% - 35% + the factor value 0.85 in the first row and first column of A1 = 1.12; the first function calculation value is the product of the first transition value and the first vector (preset as [1.2]), i.e. 1.12 x 1.2 = 1.344; load it into the Sigmoid activation function to obtain the first possibility P 11 (explosive power element in the factor mining possibility of training matching degree direction) = 0.79. Similarly, calculate each element in P1, and finally P1 = [0.79, 0.77, 0.35, 0.30].

[0077] For the calculation of the second possibility queue P2: calculate the third data set (the data set with the smallest natural number in the second physical index value, i.e. 55%) and the fourth data set (the data set with the smallest natural number in the second comparison result "1 - the second physical index value", 1 - 58% = 42%); take a = 3 (cardiorespiratory endurance element) and b = 1 (training matching degree direction), calculate the second transition value = 42% - 55% + the factor value 0.88 in the third row and first column of A2 = 0.75; the second function calculation value is the product of the second transition value and the second vector (preset as [1.2]), i.e. 0.75 x 1.2 = 0.9; load it into the Sigmoid activation function to obtain the second possibility P23 The factor mining possibility of the cardiopulmonary endurance element in the training matching degree direction = 0.71. Similarly, the elements in P2 are calculated, and finally P2 = [0.36, 0.33, 0.71, 0.69].

[0078] 3. Generating the first and second physical training category description queues: calculating the function processing result (using dot multiplication operation) of the feature queue F and the possibility queue to obtain the first physical training category description queue Q1 and the second physical training category description queue Q2. Wherein Q1 focuses on the explosive strength and agility training category, specifically "Q1 = [weighted deep squat (explosive strength improvement), agility ladder shuttle run (agility improvement), interval run (auxiliary improvement), and plate support variation (auxiliary improvement)]"; Q2 focuses on the cardiopulmonary endurance and core stability training category, specifically "Q2 = [frogs jump (auxiliary improvement), jump rope (auxiliary improvement), long distance slow run (cardiopulmonary endurance improvement), and hanging leg lifting variation (core stability improvement)]".

[0079] Performing a random mining on the feature descriptions in the feature queue to obtain the first physical training category description queue; performing a random mining on the feature descriptions in the feature queue to obtain the second physical training category description queue.

[0080] Optionally, there is at least one mined feature description in the first physical training category description queue, and all or part of the queue factors in each mined feature description are mined. There is at least one mined feature description in the second physical training category description queue, and all or part of the queue factors in each mined feature description are mined.

[0081] Step 240: recommending a personalized athlete physical training program based on the first physical training category description queue and the second physical training category description queue.

[0082] The feature description of the similar physical element refers to the feature description of the same physical element in different physical training category description queues, or the feature description of the physical element and the feature description of the neighbor physical element; the feature description of the dissimilar physical element refers to the feature description of the physical element and the feature description of the non-neighbor physical element.

[0083] In some independently implemented embodiments, the physical training analysis network includes a first data analysis thread and a second data analysis thread, the first data analysis thread and the second data analysis thread are used to load the feature descriptions in the feature queue into the AI vector space for analysis, and the first data analysis thread and the second data analysis thread are data analysis threads with the same direction but different coefficients. Step 230 can be implemented as steps 231 to 233.

[0084] The first data analysis thread and the second data analysis thread are used to load the feature descriptions in the feature queue into the AI vector space for analysis. The first data analysis thread and the second data analysis thread are data analysis threads with the same direction but different coefficients. After the feature queue is loaded into the first data analysis thread and the second data analysis thread, the analysis queue obtained is located in different AI vector spaces. In other words, the first data analysis thread and the second data analysis thread are used to load the feature descriptions originally located in one AI vector space into two different AI vector spaces for analysis. It should be noted that the direction of the first data analysis thread and the second data analysis thread is the same as the direction of the aggregation network.

[0085] Step 231: loading the feature queue into the first data analysis thread to obtain a first analysis queue; and loading the feature queue into the second data analysis thread to obtain a second analysis queue.

[0086] The feature queue is loaded into the first data analysis thread to obtain the first analysis queue, that is, the feature descriptions corresponding to the badminton player physical data example are loaded from the first AI vector space to the second AI vector space. The first AI vector space refers to the AI vector space in which the feature descriptions corresponding to the badminton player physical data example are located after feature extraction is performed on the badminton player physical data example. The second AI vector space refers to the AI vector space corresponding to the first analysis queue. The feature queue is loaded into the second data analysis thread to obtain the second analysis queue, that is, the feature descriptions corresponding to the badminton player physical data example are loaded from the first AI vector space to the third AI vector space. The third AI vector space refers to the AI vector space corresponding to the second data analysis thread.

[0087] Step 232: filtering the first analysis queue to obtain a first possibility queue; and filtering the second analysis queue to obtain a second possibility queue; the first possibility queue and the second possibility queue are used to represent the possibility of mining the local queue factor in the feature queue.

[0088] The filtering process converts the random mining of the feature descriptions in the feature queue into a problem of finding the possibility of mining each feature description. The first possibility queue is used to represent the possibility of mining the first local queue factor in the feature queue. The second possibility queue is used to represent the possibility of mining the second local queue factor in the feature queue. Generally, the first local physical element is different from the second local physical element, and the first local feature description is different from the second local feature description.

[0089] Step 233: obtaining a first physical training category description queue based on the feature queue and the first possibility queue; and obtaining a second physical training category description queue based on the feature queue and the second possibility queue.

[0090] The feature queue is mined based on the possibility of each queue factor being mined indicated by the first possibility queue to obtain the first physical training category description queue; and the feature queue is mined based on the possibility of each queue factor being mined indicated by the second possibility queue to obtain the second physical training category description queue.

[0091] In some independently implemented embodiments, obtaining the first physical training category description queue based on the feature queue and the first possibility queue comprises: calculating a function processing result of the first analysis queue and the first possibility queue to obtain the first physical training category description queue.

[0092] In some independently implemented embodiments, obtaining the second physical training category description queue based on the feature queue and the second possibility queue comprises: calculating a function processing result of the second analysis queue and the second possibility queue to obtain the second physical training category description queue.

[0093] The parallel steps in steps 231 to 233 are independent of each other, such as steps 231-1 “loading the feature queue into the first data analysis thread to obtain the first analysis queue” and 231-2 “loading the feature queue into the second data analysis thread to obtain the second analysis queue”, steps 232-1 “filtering the first analysis queue to obtain the first possibility queue” and 232-2 “filtering the second analysis queue to obtain the second possibility queue”, and steps 233-1 “obtaining the first physical training category description queue based on the feature queue and the first possibility queue” and 233-2 “obtaining the second physical training category description queue based on the feature queue and the second possibility queue”.

[0094] In some independently implemented embodiments, step 240 can be implemented as steps 241 to 244.

[0095] Step 241: obtaining a feature description of a specific athlete from the first physical training category description queue, the specific athlete being a physical element corresponding to the a-th row in the first physical training category description queue, a being an integer greater than zero.

[0096] Optionally, the feature description of the specific athlete is obtained from the first physical training category description queue; or the feature description of the specific athlete is obtained from the second physical training category description queue. The specific athlete is a physical element corresponding to the a-th row in the first physical training category description queue or the second physical training category description queue.

[0097] Step 242: obtaining the feature description of the main example from the second physical training category description queue, the main example being the physical element corresponding to the a-th row in the second physical training category description queue, and the main example and the specific athlete corresponding to the same physical element in the badminton athlete physical data example.

[0098] Optionally, the feature description of the main example is obtained from the second physical training category description queue, or the feature description of the main example is obtained from the first physical training category description queue. The main example is the physical element corresponding to the a-th row in the first physical training category description queue or the second physical training category description queue.

[0099] Specifically, the physical element corresponding to the a-th row feature description is obtained from the first physical training category description queue and the second physical training category description queue respectively, wherein one of the physical elements corresponding to the a-th row feature description is the specific athlete, and the other of the physical elements corresponding to the a-th row feature description is the main example. It should be noted that the embodiment of the present application takes the physical element corresponding to the a-th row feature description obtained from the first physical training category description queue as the specific athlete, and the physical element corresponding to the a-th row feature description obtained from the second physical training category description queue as the main example for example, but is not limited thereto, that is, the physical element corresponding to the a-th row feature description obtained from the first physical training category description queue can be taken as the main example, and the physical element corresponding to the a-th row feature description obtained from the second physical training category description queue can be taken as the specific athlete.

[0100] Step 243: obtaining the feature description of the secondary example from the first physical training category description queue or the second physical training category description queue, the secondary example being the physical element in the badminton athlete physical data example that is not adjacent to the specific athlete and the main example.

[0101] Wherein, the specific athlete and the main example correspond to the same physical element in the badminton athlete physical data example.

[0102] Step 244: recommending a personalized athlete physical training scheme based on the feature description of the specific athlete, the feature description of the main example, and the feature description of the secondary example.

[0103] Optionally, the body training analysis network is subjected to contrastive learning by taking the specific athlete and the major example as the first training sample and the specific athlete and the minor example as the second training sample, or by taking the specific athlete and the major example as the first training sample and the major example and the minor example as the second training sample. The contrastive learning is to make the difference in the first training sample as small as possible and the difference in the second training sample as large as possible. That is, the specific athlete is as similar as possible to the major example and as dissimilar as possible to the minor example.

[0104] In addition, the asymmetric structure is beneficial to preventing model collapse in the contrastive learning process, and therefore, a third data analysis thread can be added to the body training analysis network. That is, the body training analysis network further includes a third data analysis thread, and the coefficients of the first data analysis thread and the second data analysis thread are different from those of the third data analysis thread. The method further includes step 310, and step 240 can be implemented as step 320.

[0105] Step 310: loading the first body training category description queue to the third data analysis thread to obtain a third body training category description queue.

[0106] The third body training category description queue obtained through the third data analysis thread mapping has the same direction as the first body training category description queue, but the feature description (or local queue factor) is different.

[0107] Step 320: recommending a personalized athlete body training scheme based on the third body training category description queue and the second body training category description queue.

[0108] Step 321: obtaining the feature description of the specific athlete from the third body training category description queue, the specific athlete being the body element corresponding to the a-th row in the third body training category description queue, and a being an integer greater than zero.

[0109] Optionally, the feature description of the specific athlete is obtained from the third body training category description queue, or the feature description of the specific athlete is obtained from the second body training category description queue. The specific athlete is the body element corresponding to the a-th row in the third body training category description queue or the second body training category description queue.

[0110] Step 322: obtaining the feature description of the major example from the second body training category description queue, the major example being the body element corresponding to the a-th row in the second body training category description queue, and the major example and the specific athlete corresponding to the same body element in the badminton athlete body data example.

[0111] Optionally, the feature description of the main example is obtained from the second physical training category description queue; or the feature description of the main example is obtained from the third physical training category description queue. The main example is the physical element corresponding to the ath row in the third physical training category description queue or the second physical training category description queue.

[0112] Specifically, the physical element corresponding to the ath row of feature description is obtained from the third physical training category description queue and the second physical training category description queue respectively, wherein one physical element corresponding to the ath row of feature description is taken as the specific athlete, and the other physical element corresponding to the ath row of feature description is taken as the main example. It should be noted that the embodiments of the present application take the physical element corresponding to the ath row of feature description obtained from the third physical training category description queue as the specific athlete, and the physical element corresponding to the ath row of feature description obtained from the second physical training category description queue as the main example for illustration, but this is not limited thereto, that is, the physical element corresponding to the ath row of feature description obtained from the third physical training category description queue can be taken as the main example, and the physical element corresponding to the ath row of feature description obtained from the second physical training category description queue can be taken as the specific athlete.

[0113] Step 323: obtaining the feature description of the secondary example from the third physical training category description queue or the second physical training category description queue, wherein the secondary example is the physical element in the badminton athlete physical data example that is not adjacent to the specific athlete and the main example.

[0114] Step 324: recommending the personalized athlete physical training scheme based on the feature description of the specific athlete, the feature description of the main example, and the feature description of the secondary example.

[0115] Further, the way of obtaining the training sample in the asymmetric contrast learning is also shown, that is, obtaining the specific athlete, the main example, and the secondary example from the third physical training category description queue and the second physical training category description queue to obtain a set of training samples (including the first training sample composed of the specific athlete and the main example, and the second training sample composed of the specific athlete and the secondary example), and using the set of training samples to train the physical training analysis network, so as to realize the unsupervised training of the physical training analysis network (without manually pre-labeling the training sample, the physical training analysis network can extract the training sample by itself), reduce the labor cost in the training process of the physical training analysis network, and make the method also applicable to recommending the personalized athlete physical training scheme in the industrial field (a field requiring large-scale data training).

[0116] Next, it is shown how to recommend the personalized athlete physical training scheme according to the feature description of the specific athlete, the main example, and the secondary example. Step 244 or step 324 can be implemented as steps 410 to 430.

[0117] Step 410: calculating a first commonality score between the feature description of the specific athlete and the feature description of the primary example; and calculating a second commonality score between the feature description of the specific athlete and the feature description of the secondary example.

[0118] Step 420: calculating a quantitative index of the physical training analysis network according to the first commonality score and the second commonality score.

[0119] Step 430: recommending a personalized physical training scheme for the athlete according to the quantitative index of the physical training analysis network.

[0120] On the basis of the above, a personalized badminton physical training scheme recommendation system is shown, which comprises a processor and a memory in communication with each other, and the processor is used to read a computer program from the memory and execute it to realize the above method.

[0121] On the basis of the above, a computer readable storage medium is also provided, and the computer program stored thereon realizes the above method when running.

[0122] In summary, based on the above scheme, 1. The precision of the physical training scheme is adapted, and the training pertinence and efficiency are effectively improved. The present application can fully capture the differences of individual physical characteristics of athletes and the internal relations between various physical elements by obtaining badminton athlete physical data examples containing various physical elements and constraint boundaries, and based on feature queue mining different local queue factors to obtain two types of training category description queues. Subsequently, through the feature description matching of the specific athlete, the primary example and the secondary example, the athlete's physical short board and the training direction can be accurately positioned, avoiding the subjectivity and one-sidedness of the traditional experience-driven or general standardized training mode, so that the recommended training scheme is more suitable for the individual needs of athletes, and the training effect is significantly improved.

[0123] 2. The efficiency of mining and utilization of physical data is optimized, and the scientificity of the training scheme recommendation is improved. The present application adopts multi-thread (first, second and optional third data analysis threads) to analyze the feature queue in layers, and mines local queue factors of different dimensions through different coefficient data analysis threads, and realizes the deep mining and multi-dimensional utilization of physical data features by combining the filtering and function processing of the possibility queue. Compared with the existing simple data comparison or single-dimensional analysis method, the present application can fully mine the key information hidden in the physical data, improve the comprehensiveness and accuracy of feature extraction, provide more reliable data support for the recommendation of training scheme, and guarantee the scientificity and rationality of the scheme.

[0124] 3. The screening logic of the explicit training examples improves the accuracy and stability of the program recommendation. The invention realizes the accurate screening of training examples and the quantitative optimization of the program by defining the selection rules of specific athletes, main examples and secondary examples, combining the common score to calculate the quantitative indicators of the physical training analysis network. At the same time, through the secondary processing of the third data analysis thread to the first physical training category description queue, the accuracy of feature description is further optimized, and the error caused by single analysis dimension is reduced. The design makes the recommendation logic of the training program clearer and traceable, improves the stability and reliability of the program recommendation, and reduces the deviation between the training program and the actual demand of the athletes.

[0125] 4. Adapt to various data scenarios to enhance the universality and practicality of the method. The badminton athlete physical data examples in the invention can adapt to various scenes such as database, athlete physical index information cluster and reference data. The definition of physical elements and constraint boundaries can flexibly match different data types such as user personalized information, athlete physical index information and standard data, so that the method can meet the physical training program recommendation needs of different training scenes (such as daily training, special improvement and pre-match preparation) and different levels of athletes (professional athletes and amateur enthusiasts), significantly improving the universality and practicality of the method, and having a wide application prospect.

[0126] 5. Reduce the professional threshold of training program development and improve the popularity of training guidance. The invention reduces the over-reliance on coach experience through standardized data analysis process and automated program recommendation logic, so that even non-professional coaches can use the method to develop scientific and personalized physical training programs for athletes. This helps to promote the popularity of personalized physical training guidance, so that more athletes can enjoy accurate and scientific training guidance, and promote the overall improvement of badminton training level.

[0127] It should be appreciated that the systems and modules thereof described above can be implemented in a number of ways. For example, in some embodiments, the systems and modules thereof can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using specialized logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated design hardware. Those skilled in the art will appreciate that the methods and systems described above can be implemented using computer-executable instructions and / or in processor control code, for example, provided on a carrier medium such as a disk, CD or DVD-ROM, programmable memory such as read-only memory (firmware), or data carrier such as an optical or electrical signal carrier. The systems and modules thereof of the present application can be implemented not only in hardware circuitry such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., but also in software, for example, executed by various types of processors, and in a combination of the above (e.g., firmware).

[0128] It should be noted that the beneficial effects produced by different embodiments can be different, and in different embodiments, the beneficial effects that can be produced can be any one or a combination of the above, or any other beneficial effects that can be obtained.

Claims

1. A method for recommending personalized badminton physical fitness training programs, characterized in that, The method includes: An example of obtaining physical fitness data of a badminton player, wherein the example of physical fitness data of a badminton player includes at least two physical fitness elements and constraint boundaries for connecting the at least two physical fitness elements; Extract the feature queue of the badminton player's physical fitness data example, where each feature in the feature queue corresponds to a feature description of a physical fitness element in the badminton player's physical fitness data example; Based on the feature queue, a first physical training category description queue and a second physical training category description queue are obtained. The first physical training category description queue and the second physical training category description queue are used to represent the feature queue after the local queue factors are mined. The queue factors mined in the first physical training category description queue and the second physical training category description queue are not consistent. Based on the first physical training type description queue and the second physical training type description queue, a personalized physical training program for athletes is recommended.

2. The method as described in claim 1, characterized in that, The physical training analysis network includes a first data analysis thread and a second data analysis thread. The first data analysis thread and the second data analysis thread are used to load the feature descriptions in the feature queue into the AI ​​vector space for parsing. The first data analysis thread and the second data analysis thread are data analysis threads with the same direction but different coefficients. The process of obtaining a first physical training category description queue and a second physical training category description queue based on the feature queue includes: The feature queue is loaded into the first data analysis thread to obtain a first analysis queue; and the feature queue is loaded into the second data analysis thread to obtain a second analysis queue; The first analysis queue is filtered to obtain a first probability queue; and the second analysis queue is filtered to obtain a second probability queue; the first probability queue and the second probability queue are used to represent the probability that the local queue factor in the feature queue will be mined; The first physical training category description queue is obtained based on the feature queue and the first possibility queue; and the second physical training category description queue is obtained based on the feature queue and the second possibility queue.

3. The method as described in claim 2, characterized in that, The process of obtaining a first physical training category description queue based on the feature queue and the first probability queue includes: The function processing results of the feature queue and the first possibility queue are calculated to obtain the first physical training category description queue. The process of obtaining the second physical training category description queue based on the feature queue and the second probability queue includes: The function processing results of the feature queue and the second possibility queue are calculated to obtain the second physical training type description queue.

4. The method as described in claim 2, characterized in that, The step of filtering the first analysis queue to obtain the first probability queue includes: The first physical fitness index value is obtained from the physical fitness dataset; Calculate the first data set and the second data set. The first data set is the data set with the smallest natural number among the first physical fitness index values. The second data set is the data set with the smallest natural number in the first comparison result. The first comparison result is the difference between one and the first physical fitness index value. Calculate the first possibility, which is the probability that the queue factor of the a-th physical element in the b-th direction in the feature queue is mined. The first possibility is the probability obtained after loading the calculated value of the first function into the activation function. The calculated value of the first function is the result of the calculation of the first transition value and the first vector. The first transition value is the value obtained by subtracting the second data group from the first data group and adding the factor value of the a-th row and b-th column in the first analysis queue. Both a and b are integers greater than zero. The step of filtering the second analysis queue to obtain the second possibility queue includes: The second physical fitness index value is obtained from the physical fitness dataset; Calculate the third data group and the fourth data group, wherein the third data group is the data group with the smallest natural number among the second physical fitness index values, and the second data group is the data group with the smallest natural number among the second comparison results, and the second comparison result is the difference between one and the second physical fitness index value; Calculate the second possibility, which is the probability that the queue factor of the a-th physical element in the b-th direction in the feature queue is mined. The second possibility is the probability obtained after loading the calculated value of the second function into the activation function. The calculated value of the second function is the result of the calculation of the second transition value and the second vector. The second transition value is the value obtained by subtracting the third data group from the fourth data group and adding the factor value of the a-th row and b-th column in the second analysis queue. Both a and b are integers greater than zero.

5. The method according to any one of claims 1 to 4, characterized in that, The method of recommending personalized physical training programs for athletes based on the first physical training type description queue and the second physical training type description queue includes: The feature description of a specific athlete is obtained from the first physical training type description queue, wherein the specific athlete is the physical element corresponding to the a-th row in the first physical training type description queue, and a is an integer greater than zero; The feature description of the main example is obtained from the second physical training type description queue. The main example is the physical element corresponding to the a-th row in the second physical training type description queue. The main example and the specific athlete correspond to the same physical element in the badminton player physical fitness data example. The feature description of a secondary example is obtained from the first physical training type description queue or the second physical training type description queue. The secondary example is a physical element in the badminton player physical fitness data example that is not adjacent to the specific athlete and the main example. Based on the feature descriptions of the specific athlete, the feature descriptions of the primary example, and the feature descriptions of the secondary example, a personalized athlete physical training program is recommended.

6. The method according to any one of claims 1 to 4, characterized in that, The physical training analysis network also includes a third data analysis thread, the coefficients of which are different from those of the first and second data analysis threads; The method further includes: The first physical training category description queue is loaded into the third data analysis thread to obtain the third physical training category description queue; The method of recommending personalized physical training programs for athletes based on the first physical training type description queue and the second physical training type description queue includes: Based on the third physical training type description queue and the second physical training type description queue, a personalized physical training program for athletes is recommended.

7. The method as described in claim 6, characterized in that, The method of recommending personalized physical training programs for athletes based on the third physical training type description queue and the second physical training type description queue includes: The feature description of a specific athlete is obtained from the third physical training type description queue, wherein the specific athlete is the physical element corresponding to the a-th row in the third physical training type description queue, and a is an integer greater than zero; The feature description of the main example is obtained from the second physical training type description queue. The main example is the physical element corresponding to the a-th row in the second physical training type description queue. The main example and the specific athlete correspond to the same physical element in the badminton player physical fitness data example. The feature description of a secondary example is obtained from the third physical training type description queue or the second physical training type description queue. The secondary example is a physical element in the badminton player physical fitness data example that is not adjacent to the specific athlete and the main example. Based on the feature descriptions of the specific athlete, the feature descriptions of the primary example, and the feature descriptions of the secondary example, a personalized athlete physical training program is recommended.

8. The method as described in claim 5 or 7, characterized in that, The method of recommending personalized physical training programs for athletes based on the feature descriptions of the specific athlete, the feature descriptions of the primary examples, and the feature descriptions of the secondary examples includes: Calculate a first commonality score, which is the commonality score between the feature description of the specific athlete and the feature description of the primary example; and calculate a second commonality score, which is the commonality score between the feature description of the specific athlete and the feature description of the secondary example. Quantitative indicators of the physical training analysis network are calculated based on the first commonality score and the second commonality score. Based on the quantitative indicators of the aforementioned physical training analysis network, personalized physical training programs for athletes are recommended.

9. The method according to any one of claims 1 to 8, characterized in that, The badminton player's physical fitness data forms a database, the physical fitness elements represent user personalized information, and the constraint boundaries represent the correlation between the user personalized information. The badminton player's physical fitness data example is an athlete's physical fitness indicator information cluster, the physical fitness element represents user personalized information and athlete's physical fitness indicator information, and the constraint boundary represents the relationship between user personalized information and athlete's physical fitness indicator information, and the relationship between user personalized information and user personalized information. The badminton player's physical fitness data example is for reference, the physical fitness elements represent standard data, and the constraint boundaries indicate the correlation between the standard data.

10. A personalized badminton physical fitness training program recommendation system, characterized in that, The method includes a processor and a memory that communicate with each other, the processor being configured to read a computer program from the memory and execute it to implement the method of any one of claims 1-9.

Images