CN116421953A - Tennis training method and system based on deep learning - Google Patents
Tennis training method and system based on deep learning Download PDFInfo
- Publication number
- CN116421953A CN116421953A CN202310708358.9A CN202310708358A CN116421953A CN 116421953 A CN116421953 A CN 116421953A CN 202310708358 A CN202310708358 A CN 202310708358A CN 116421953 A CN116421953 A CN 116421953A
- Authority
- CN
- China
- Prior art keywords
- training
- data
- tennis
- neural network
- athlete
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012549 training Methods 0.000 title claims abstract description 158
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000013135 deep learning Methods 0.000 title claims abstract description 22
- 238000013528 artificial neural network Methods 0.000 claims abstract description 81
- 125000004122 cyclic group Chemical group 0.000 claims abstract description 41
- 230000033001 locomotion Effects 0.000 claims abstract description 13
- 230000007613 environmental effect Effects 0.000 claims abstract description 7
- 230000000306 recurrent effect Effects 0.000 claims description 29
- 230000015654 memory Effects 0.000 claims description 20
- 238000012360 testing method Methods 0.000 claims description 20
- 230000006870 function Effects 0.000 claims description 14
- 238000012545 processing Methods 0.000 claims description 14
- 238000003860 storage Methods 0.000 claims description 14
- 238000012544 monitoring process Methods 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 11
- 238000004140 cleaning Methods 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 10
- 238000012795 verification Methods 0.000 claims description 10
- 230000004913 activation Effects 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 230000011218 segmentation Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 8
- 238000010586 diagram Methods 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 206010011971 Decreased interest Diseases 0.000 description 1
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 238000009533 lab test Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 230000036314 physical performance Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000033764 rhythmic process Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B69/00—Training appliances or apparatus for special sports
- A63B69/38—Training appliances or apparatus for special sports for tennis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B2214/00—Training methods
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B2220/00—Measuring of physical parameters relating to sporting activity
- A63B2220/30—Speed
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B2230/00—Measuring physiological parameters of the user
- A63B2230/04—Measuring physiological parameters of the user heartbeat characteristics, e.g. ECG, blood pressure modulations
- A63B2230/06—Measuring physiological parameters of the user heartbeat characteristics, e.g. ECG, blood pressure modulations heartbeat rate only
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Physical Education & Sports Medicine (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of tennis training, in particular to a deep learning-based tennis training method which can automatically train individual differences of athletes without manual instruction of a coach and can effectively improve training effects and tennis skills of the athletes; the method comprises the following steps: before training, training an athlete to obtain athlete training data; determining the technical level of the athlete according to the athlete training data through a first cyclic neural network trained in advance; determining training intensity according to the athlete skill level; in training, acquiring tennis motion state and field environmental factors of a player after batting in real time; determining the track of tennis through a pre-trained second circulating neural network according to the tennis state and the field environmental factors; and controlling the ball returning machine to return the ball by combining the track and training intensity of the tennis ball.
Description
Technical Field
The invention relates to the technical field of tennis training, in particular to a tennis training method and system based on deep learning.
Background
The current tennis training mode is mostly artificial for supplying the ball to the player, in addition, the ball is supplied to the player by using a ball returning machine, and the problem that the individual difference of the player is difficult to train by using the ball returning machine is solved.
Parameters such as a ball supply mode and a speed of the ball returning machine are fixed and cannot be adjusted according to athletes of different technical levels, so that the existing tennis training by using the ball returning machine can only be used as an auxiliary training means, and cannot completely replace the guiding training of an artificial coach.
Disclosure of Invention
In order to solve the technical problems, the invention provides the deep learning-based tennis training method which can automatically train the individual difference of the player without manual instruction of a coach and can effectively improve the training effect of the player and the tennis skill.
In a first aspect, the present invention provides a tennis training method based on deep learning, the method comprising:
before training, training an athlete to obtain athlete training data;
determining the technical level of the athlete according to the athlete training data through a first cyclic neural network trained in advance;
determining training intensity according to the athlete skill level;
in training, acquiring tennis motion state and field environmental factors of a player after batting in real time;
determining the track of tennis through a pre-trained second circulating neural network according to the tennis state and the field environmental factors;
and controlling the ball returning machine to return the ball by combining the track and training intensity of the tennis ball.
In another aspect, the present application also provides a deep learning-based tennis training system, the system comprising:
the data monitoring module is used for acquiring test training data of a sportsman before training and real-time tennis state data and site environment factor data in training through the sensor and the high-speed camera and sending the test training data and the real-time tennis state data and the site environment factor data;
the level judging module is used for receiving the training data sent by the data monitoring module, analyzing and calculating the training data by utilizing a first cyclic neural network stored in advance, determining the technical level of the athlete and sending the training data;
the intensity adjusting module is used for receiving the athlete technical level sent by the level judging module, setting training intensity corresponding to the athlete technical level according to the athlete technical level, and manually setting the training intensity according to actual conditions;
the track prejudging module is used for receiving the real-time tennis state data and the site environment factor data in the training sent by the data monitoring module, determining the track of the tennis ball according to the service data and the site environment factor by utilizing a pre-stored second circulating neural network, simultaneously reading the training intensity set in the intensity adjusting module, generating a ball returning machine control signal according to the training intensity and the tennis ball track, and sending the ball returning machine control signal;
the ball returning machine is used for receiving the control signal sent by the track prejudging module and returning tennis balls according to the control signal.
In a third aspect, the present application provides an electronic device comprising a bus, a transceiver, a memory, a processor and a computer program stored on the memory and executable on the processor, the transceiver, the memory and the processor being connected by the bus, the computer program implementing the steps of any of the methods described above when executed by the processor.
In a fourth aspect, the present application also provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of any of the methods described above.
Further, the training data includes a serve speed, a heart rate, and a movement speed as input data of the first recurrent neural network.
Further, the construction and training method of the first cyclic neural network is as follows:
obtaining match data of athletes at different levels and under different match scenes, and taking the match data as training data of a neural network;
data cleaning is carried out on training data;
sequentially segmenting the training data, segmenting the training data into a plurality of time windows according to time sequence, and dividing a plurality of segmented groups of data into a training set, a verification set and a test set;
selecting a first recurrent neural network suitable for judging the technical level of the athlete;
model training, verification and testing are performed on the first recurrent neural network using the training set, the verification set and the test set, respectively.
Further, the cleaning of the training data includes checking and processing data missing values, checking and deleting data outliers, checking and converting data types, checking and deleting data repetition values.
Further, the acquired training data are subjected to data cleaning and sequential segmentation, and then substituted into a first cyclic neural network, and the technical level of the athlete is output by the first cyclic neural network.
Further, the operation formula of the first recurrent neural network is as follows:
；/>
wherein x is t Is an input layer composed of a ball serving speed, a heart rate and a moving speed in a set time window, o t Is the skill level guidance vector of the athlete, U is the weight matrix from the input layer to the hidden layer, V is the weight matrix from the hidden layer to the output layer, W is the weight matrix with the last hidden layer value as the input of this time, s t Is the value of the hidden layer of the first recurrent neural network, g and f are the activation functions.
Further, the first cyclic neural network and the second cyclic neural network adopt a basic cyclic neural network, a long-short-time memory network or a gate-controlled cyclic unit network.
Compared with the prior art, the invention has the beneficial effects that: before formal training, the technical level of the athlete can be accurately judged through the first cyclic neural network, so that the subsequent training intensity is determined; in the training process, the track pre-judging can be carried out on tennis balls sent by athletes through the second circulating neural network, and the ball returning machine is controlled to return by combining with the set training intensity; the automatic training can be performed aiming at individual differences of athletes without manual coach guidance, and the training effect and tennis skills of the athletes can be effectively improved.
Drawings
FIG. 1 is a logic flow diagram of the present invention;
FIG. 2 is a flow chart of the construction of a first recurrent neural network;
FIG. 3 is a schematic diagram of a first recurrent neural network;
FIG. 4 is a structural development of the first recurrent neural network;
fig. 5 is a flow chart of the construction of the second recurrent neural network.
Detailed Description
In the description of the present application, those skilled in the art will appreciate that the present application may be embodied as methods, apparatuses, electronic devices, and computer-readable storage media. Accordingly, the present application may be embodied in the following forms: complete hardware, complete software (including firmware, resident software, micro-code, etc.), a combination of hardware and software. Furthermore, in some embodiments, the present application may also be embodied in the form of a computer program product in one or more computer-readable storage media, which contain computer program code.
Any combination of one or more computer-readable storage media may be employed by the computer-readable storage media described above. The computer-readable storage medium includes: an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of the computer readable storage medium include the following: portable computer magnetic disks, hard disks, random access memories, read-only memories, erasable programmable read-only memories, flash memories, optical fibers, optical disk read-only memories, optical storage devices, magnetic storage devices, or any combination thereof. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, device.
The technical scheme of the application is that the acquisition, storage, use, processing and the like of the data meet the relevant regulations of national laws.
The present application describes methods, apparatus, and electronic devices provided by the flowchart and/or block diagram.
It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer-readable program instructions. These computer-readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks.
These computer readable program instructions may also be stored in a computer readable storage medium that can cause a computer or other programmable data processing apparatus to function in a particular manner. Thus, instructions stored in a computer-readable storage medium produce an instruction means which implement the functions/acts specified in the flowchart and/or block diagram block or blocks.
The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks.
The present application is described below with reference to the drawings in the present application.
Example 1
As shown in fig. 1 to 5, the deep learning-based tennis training method of the present invention comprises the steps of:
s1, training an athlete before training, and acquiring athlete training data;
the training data comprise indexes capable of reflecting physical performance and technology of tennis players, such as service speed, heart rate, moving speed and the like of the tennis players in a certain time;
specifically, through letting the sportsman carry out the match training in the certain time with coach or other accompanying and training personnel, monitor sportsman's rhythm of the heart and travel speed respectively through heart rate monitor and the motion sensor that the sportsman worn, wherein heart rate monitor and motion sensor adopt intelligent bracelet or intelligent wrist-watch, can also adopt intelligent shoe-pad to come the monitoring to travel speed simultaneously. In the competition process, the monitoring of the initial tennis speed of the player at each serving depends on a high-speed camera arranged on one side of the player; meanwhile, the specially marked tennis ball can be used, and the marked tennis ball is subjected to initial speed monitoring by using a radar velocimeter.
S2, determining the technical level of the athlete according to the training data of the athlete through a first cyclic neural network trained in advance;
in the application of judging the technical level of the tennis player, the first recurrent neural network can take the data such as the ball serving speed, the heart rate, the moving speed and the like of the tennis player in a period of time as input, and obtain an output result through processing and learning of a plurality of time steps for judging the technical level of the player. Specifically, the real-time data at a certain moment is used for judging that the technical level of the player is opposite to that of the player, and the first circulating neural network can better judge the performance and the technical level of the player by learning and memorizing the historical data of the tennis player.
Specifically, before the first recurrent neural network is applied to an actual scene, the first recurrent neural network needs to be trained, and the first recurrent neural network can accurately judge the technical level of a tennis player by repeatedly adjusting model parameters and an optimization algorithm, and the specific first recurrent neural network training steps are as follows:
s21, acquiring training data; training the first recurrent neural network requires a large amount of training data, and specifically, the data is obtained by the following method:
laboratory experiments: the tennis player can be subjected to experiments in a laboratory environment, and the data such as the service speed, the heart rate, the moving speed and the like of the tennis player are recorded and used as training data;
a field match: the data acquisition can be carried out on players in the tennis match scene, and the data such as the serving speed, the heart rate, the moving speed and the like of the players are recorded and used as training data;
open dataset: an existing open dataset, such as the dataset in UCI Machine Learning Repository, may be used as training data;
simulation data: virtual tennis match data may be generated using simulation software as training data;
it should be noted that the training data should be representative, covering different levels of tennis players and different game scenarios.
S22, data cleaning is carried out on the training data so as to ensure the accuracy and consistency of the data; the method specifically comprises the following steps: checking whether missing values exist in the data, and if so, processing by a method of filling the missing values or deleting the missing values; checking whether an abnormal value exists in the data, and if so, processing the data by deleting the abnormal value or replacing the abnormal value; checking whether the data type is correct, and if not, performing data type conversion, for example, converting text data into digital data; checking whether a repeated value exists in the data, and if so, processing the data by a method of deleting the repeated value; checking whether the data format is uniform, if not, performing format unification, for example unifying the date format into one format; checking the consistency of the data acquisition time, and ensuring that the format of the data acquisition time is consistent with the time zone; data cleaning is an important step of data analysis, can improve the quality and accuracy of data, and provides a reliable basis for subsequent data analysis.
S23, cutting the training data subjected to data cleaning into a plurality of time windows according to a time sequence, wherein each time window comprises data such as a ball serving speed, a heart rate, a moving speed and the like in a certain time, and dividing a plurality of cut data into a training set, a verification set and a test set, wherein the cut data are generally divided by adopting a ratio of 7:2:1; specifically, the length of each time window needs to be determined firstly, and can be determined according to the acquisition frequency and analysis requirement of the data, for example, the length of each time window can be 5 seconds, 10 seconds, 30 seconds and the like; determining the starting time of each time window according to the length of the time window; for example, the start time of the first time window may be the first point in time of data acquisition and the start time of the second time window may be the end time of the first time window; the training data is segmented into a plurality of time windows according to a time sequence, each time window comprises data such as a ball serving speed, a heart rate, a moving speed and the like in a certain time, and Python and other programming languages can be used for achieving segmentation of the data; for the data of each time window, statistical indexes such as average value, maximum value, minimum value, standard deviation and the like thereof can be calculated as the data of the time window.
S24, selecting a first circulating neural network suitable for judging the technical level of tennis players;
when the first cyclic neural network is selected, the aspects of network structure, activation function, loss function and the like of the first cyclic neural network need to be considered; the common first cyclic neural network comprises a basic cyclic neural network, a long-short-time memory network, a gate control cyclic unit network and the like; the long-short-time memory network and the gating circulation unit network have better memory and gradient disappearance preventing capacity, so that the long-time memory network and the gating circulation unit network perform better when processing long-sequence data;
according to the selected first cyclic neural network type, constructing a network structure, determining the quantity of neurons of an input layer, an output layer and a hidden layer, simultaneously, converting an output result into a probability value by using a softmax function as an activation function of the output layer, and selecting a loss function and an optimizer suitable for the task, wherein the common loss function comprises cross entropy, mean square error and the like, and the common optimizer comprises Adam, SGD and the like; finally, which first recurrent neural network to select needs to be evaluated and selected according to the specific data set and experimental results.
S25, performing model training, verification and test on the first circulating neural network selected in the S24 by using the training set, the verification set and the test set divided in the S23; training the network by using the training set, and continuously adjusting network parameters until a preset stopping condition is reached; verifying the trained model by using a verification set, and evaluating the performance and generalization capability of the model; testing the model by using a test set, and evaluating the actual effect of the model; the model can accurately judge the technical level of tennis players by repeatedly adjusting the model parameters and the optimization algorithm.
S26, performing the same processing as the S22 and the S23 on the training data acquired in the S1; then substituting the training parameters into the trained first circulating neural network, so that the technical level of tennis players can be accurately judged, and a foundation is laid for subsequent training.
Specifically, the first recurrent neural network is shown in fig. 3, wherein x is an input layer vector, namely a service speed, a heart rate and a moving speed in a set time window; the value s of the hidden layer of the first recurrent neural network depends not only on the current input x at this time, but also on the value s of the last hidden layer; the weight matrix W is the last value of the hidden layer as the input weight of this time, and the unfolding process of the first cyclic neural network is shown in fig. 4.
More specifically, after the first recurrent neural network receives the serving speed, the heart rate and the moving speed in the t-th time window, the specific calculation process is as follows:
equation 1; />
Equation 2, equation 1 is the meter of the output layerThe output layer is a full connection layer, namely, each node of the output layer is connected with each node of the hidden layer; v is the weight matrix of the output layer, g is the activation function;
equation 2 is the calculation formula for the hidden layer, which is a cyclic layer, U is the weight matrix of the input x, W is the weight matrix of the last hidden layer value as the input this time, and f is the activation function.
The hidden layer has two inputs, the first is U and x t The product of vectors, second is the value s output by the previous hidden layer t-1 And W is equal to s of the previous calculation output t-1 Needs to be cached, and inputs x at this time t Together calculate, co-output the last o t 。
If equation 2 is iteratively brought into equation 1, we will get:
S3, determining training intensity according to the technical level of the athlete;
specifically, the training intensity should be adjusted according to the actual situation of the athlete to ensure that the training effect is maximized, and meanwhile, the athlete is prevented from being injured; for athletes with higher skill levels, training intensity can be increased appropriately, increasing training intensity and time to challenge their limits; for athletes with a lower skill level, the training intensity should be reduced appropriately to avoid injury or loss of interest from over-training.
Meanwhile, the training program should be adjusted according to the actual situation of the athlete to ensure that the training effect is maximized, so that the training strength can be manually set according to the self-evaluation of the athlete.
S4, acquiring tennis movement state data of the player after batting and site environment factor data in real time in training;
specifically, in order to achieve the same-level training of the tennis player by the pitching machine, the tennis track sent by the tennis player needs to be prejudged; there is a need to make a prediction of the course of tennis in connection with the context of the field and the tennis state of the player after striking the ball.
More specifically, site environmental factors include site hardness, site humidity, and real-time wind direction; tennis state refers to the spatial vector of tennis, including initial speed, tee point height, spin, initial height, and initial direction.
Wherein factors such as field hardness and humidity affect ball rebound and rolling, and further affect ball drop points and routes, and sensors and other devices can be used to monitor these factors; real-time wind direction can influence the flight track and air resistance of the ball, and then influence the drop point and the route of the ball, and wind direction and wind speed can be monitored by using equipment such as an anemometer and the like.
The motion state vectors such as the initial speed, the service point height, the rotation, the initial height, the initial direction and the like of the tennis ball can be monitored by a high-speed photography or other sensors; for example, a high-speed photographing may be used to record the movement track and rotation of the tennis ball, and then these parameters may be calculated by analyzing the image, or a sensor such as a radar or an accelerometer may be used to monitor the movement of the tennis ball.
S5, determining a tennis track according to the tennis state and the site factors through a pre-trained second circulating neural network;
the excellent performance of the cyclic neural network in the aspect of processing sequence data is achieved, and for the continuous transfer problem of tennis ball tracks, the second cyclic neural network can predict the tennis ball tracks by learning the relation between continuous frames in the tennis ball flight process.
Before the second cyclic neural network is applied to an actual scene, the second cyclic neural network needs to be trained, and the second cyclic neural network can accurately pre-judge the track of the tennis ball by repeatedly adjusting model parameters and an optimization algorithm, wherein the second cyclic neural network training steps are as follows:
s51, collecting past data of tennis sports states and site factors;
s52, preprocessing the data, such as data cleaning, smoothing, standardization and the like;
s53, dividing the data into a training data set and a test data set;
s54, selecting a second circulating neural network suitable for pre-judging the track of tennis;
s55, training and testing the second cyclic neural network model sequentially by using the training data set and the testing data set;
s56, substituting the data acquired in the S4 into a trained second cyclic neural network, calculating the track of the tennis ball through the second cyclic neural network by the data, and predicting the drop point position of the tennis ball.
The specific training method and structure of the second recurrent neural network are similar to those of the first recurrent neural network, and detailed description thereof is omitted.
On the other hand, the following method can be adopted to calculate the tennis track and the drop point:
kinematic model: based on the data acquired in the step S4, a tennis ball kinematic model can be established; tennis's motion can be considered as free-falling motion in a two-dimensional plane, but needs to take into account the effects of air resistance and rotation;
calculating a track: the track of tennis in the air can be calculated through a kinematic model, and the track is realized through a numerical calculation method, an analytic solution or a numerical simulation method and the like;
predicting a falling point: the position of the drop point of the tennis ball can be predicted through the calculated track; this can be achieved by computer vision techniques, for example by dividing the field into several grids, and then calculating the probability of a possible drop point of the tennis ball in each grid, the grid with the highest probability being the predicted tennis ball drop point position.
S6, controlling the ball returning machine to return the ball by combining the track of the tennis ball and the training intensity;
according to the training intensity given by the first cyclic neural network and the tennis track given by the second cyclic neural network, the ball returning machine can be controlled to return; specifically, the output of the first and second recurrent neural networks is used as the input of the ball returning machine, and then the parameters such as the ball sending speed, the ball sending direction, the ball sending rotation and the like of the ball returning machine are controlled according to the input, so that the ball returning machine can accurately return tennis balls according to the technical level of a player.
Of course, the above parameters are only typical part of the parameters, and other parameters that can be obtained and utilized are also within the scope of the present invention.
Example two
A deep learning-based tennis training system, the system comprising:
the data monitoring module is used for acquiring test training data of a sportsman before training and real-time tennis state data and site environment factor data in training through the sensor and the high-speed camera and sending the test training data and the real-time tennis state data and the site environment factor data;
the level judging module is used for receiving the training data sent by the data monitoring module, analyzing and calculating the training data by utilizing a first cyclic neural network stored in advance, determining the technical level of the athlete and sending the training data;
the intensity adjusting module is used for receiving the athlete technical level sent by the level judging module, setting the training intensity corresponding to the athlete technical level according to the athlete technical level, and manually setting the training intensity according to actual conditions;
the track prejudging module is used for receiving the real-time tennis state data and the site environment factor data in the training sent by the data monitoring module, determining the track of the tennis ball according to the service data and the site environment factor by utilizing a pre-stored second circulating neural network, simultaneously reading the training intensity set in the intensity adjusting module, generating a ball returning machine control signal according to the training intensity and the tennis ball track, and sending the ball returning machine control signal;
the ball returning machine is used for receiving the control signal sent by the track prejudging module and returning tennis balls according to the control signal.
The above-described various modifications and embodiments of the deep learning-based tennis training method in the first embodiment of fig. 1 are equally applicable to the deep learning-based tennis training system of this embodiment, and those skilled in the art will be aware of the implementation method of the deep learning-based tennis training system of this embodiment through the foregoing detailed description of the deep learning-based tennis training method, so that the detailed description thereof will not be repeated for brevity.
In addition, the application further provides an electronic device, which comprises a bus, a transceiver, a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the transceiver, the memory and the processor are respectively connected through the bus, and when the computer program is executed by the processor, the processes of the method embodiment for controlling output data are realized, and the same technical effects can be achieved, so that repetition is avoided and redundant description is omitted.
The foregoing is merely a preferred embodiment of the present invention, and it should be noted that it will be apparent to those skilled in the art that modifications and variations can be made without departing from the technical principles of the present invention, and these modifications and variations should also be regarded as the scope of the invention.
Claims (10)
1. A tennis training method based on deep learning, the method comprising:
before training, training an athlete to obtain athlete training data;
determining the technical level of the athlete according to the athlete training data through a first cyclic neural network trained in advance;
determining training intensity according to the athlete skill level;
in training, acquiring tennis motion state and field environmental factors of a player after batting in real time;
determining the track of tennis through a pre-trained second circulating neural network according to the tennis state and the field environmental factors;
and controlling the ball returning machine to return the ball by combining the track and training intensity of the tennis ball.
2. The deep learning-based tennis training method according to claim 1, wherein said trial data includes a serve speed, a heart rate, and a movement speed as input data of said first recurrent neural network.
3. The deep learning-based tennis training method according to claim 2, wherein the first recurrent neural network is constructed and trained as follows:
obtaining match data of athletes at different levels and under different match scenes, and taking the match data as training data of a neural network;
data cleaning is carried out on training data;
sequentially segmenting the training data, segmenting the training data into a plurality of time windows according to time sequence, and dividing a plurality of segmented groups of data into a training set, a verification set and a test set;
selecting a first recurrent neural network suitable for judging the technical level of the athlete;
model training, verification and testing are performed on the first recurrent neural network using the training set, the verification set and the test set, respectively.
4. A deep learning based tennis training method according to claim 3 wherein said cleaning of training data comprises checking and processing data missing values, checking and deleting data outliers, checking and converting data types, checking and deleting data duplicate values.
5. The deep learning-based tennis training method according to claim 3, wherein the acquired training data is subjected to data cleaning and sequential segmentation, and then substituted into a first cyclic neural network, and the first cyclic neural network outputs the skill level of the athlete.
6. The deep learning-based tennis training method according to claim 3, wherein the first recurrent neural network has an operation formula of:
；/>
wherein x is t Is an input layer composed of a ball serving speed, a heart rate and a moving speed in a set time window; o (o) t Is the skill level guidance vector of the athlete, U is the weight matrix from the input layer to the hidden layer, V is the weight matrix from the hidden layer to the output layer, W is the weight matrix with the last hidden layer value as the input of this time, s t Is the value of the hidden layer of the first recurrent neural network, g and f are the activation functions.
7. The deep learning-based tennis training method according to claim 1, wherein the first cyclic neural network and the second cyclic neural network are basic cyclic neural networks, long-short-term memory networks or gate cyclic unit networks.
8. A deep learning-based tennis training system, the system comprising:
the data monitoring module is used for acquiring test training data of a sportsman before training and real-time tennis state data and site environment factor data in training through the sensor and the high-speed camera and sending the test training data and the real-time tennis state data and the site environment factor data;
the level judging module is used for receiving the training data sent by the data monitoring module, analyzing and calculating the training data by utilizing a first cyclic neural network stored in advance, determining the technical level of the athlete and sending the training data;
the intensity adjusting module is used for receiving the athlete technical level sent by the level judging module, setting training intensity corresponding to the athlete technical level according to the athlete technical level, and manually setting the training intensity according to actual conditions;
the track prejudging module is used for receiving the real-time tennis state data and the site environment factor data in the training sent by the data monitoring module, determining the track of the tennis ball according to the service data and the site environment factor by utilizing a pre-stored second circulating neural network, simultaneously reading the training intensity set in the intensity adjusting module, generating a ball returning machine control signal according to the training intensity and the tennis ball track, and sending the ball returning machine control signal;
the ball returning machine is used for receiving the control signal sent by the track prejudging module and returning tennis balls according to the control signal.
9. A deep learning based tennis training electronic device comprising a bus, a transceiver, a memory, a processor and a computer program stored on said memory and executable on said processor, said transceiver, said memory and said processor being connected by said bus, characterized in that said computer program when executed by said processor implements the steps of the method according to any of claims 1-7.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310708358.9A CN116421953A (en) | 2023-06-15 | 2023-06-15 | Tennis training method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310708358.9A CN116421953A (en) | 2023-06-15 | 2023-06-15 | Tennis training method and system based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116421953A true CN116421953A (en) | 2023-07-14 |
Family
ID=87085917
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310708358.9A Pending CN116421953A (en) | 2023-06-15 | 2023-06-15 | Tennis training method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116421953A (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101034502A (en) * | 2006-03-10 | 2007-09-12 | 通用汽车环球科技运作公司 | Method and system for driver handling skill recognition through driver's steering behavior |
| CN104606867A (en) * | 2015-01-14 | 2015-05-13 | 伍润人 | Serving method and system, data acquisition device and serving machine |
| CN107803010A (en) * | 2016-09-08 | 2018-03-16 | 张镜如 | A kind of table tennis training system |
| CN109716444A (en) * | 2016-09-28 | 2019-05-03 | Bodbox股份有限公司 | The assessment and guidance of athletic performance |
| CN110314361A (en) * | 2019-05-10 | 2019-10-11 | 新华智云科技有限公司 | A kind of basketball goal score judgment method and system based on convolutional neural networks |
| CN112085761A (en) * | 2020-09-10 | 2020-12-15 | 上海庞勃特科技有限公司 | Table tennis track capturing and analyzing method and system |
| CN112494915A (en) * | 2020-12-14 | 2021-03-16 | 清华大学深圳国际研究生院 | Badminton robot and system and control method thereof |
| CN112613532A (en) * | 2020-11-26 | 2021-04-06 | 西安电子科技大学 | Moving target tracking method based on radar and recurrent neural network complete infrared fusion |
| CN114582195A (en) * | 2022-03-22 | 2022-06-03 | 上海创屹科技有限公司 | Intelligent table tennis teaching system and teaching method |
-
2023
- 2023-06-15 CN CN202310708358.9A patent/CN116421953A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101034502A (en) * | 2006-03-10 | 2007-09-12 | 通用汽车环球科技运作公司 | Method and system for driver handling skill recognition through driver's steering behavior |
| CN104606867A (en) * | 2015-01-14 | 2015-05-13 | 伍润人 | Serving method and system, data acquisition device and serving machine |
| CN107803010A (en) * | 2016-09-08 | 2018-03-16 | 张镜如 | A kind of table tennis training system |
| CN109716444A (en) * | 2016-09-28 | 2019-05-03 | Bodbox股份有限公司 | The assessment and guidance of athletic performance |
| CN110314361A (en) * | 2019-05-10 | 2019-10-11 | 新华智云科技有限公司 | A kind of basketball goal score judgment method and system based on convolutional neural networks |
| CN112085761A (en) * | 2020-09-10 | 2020-12-15 | 上海庞勃特科技有限公司 | Table tennis track capturing and analyzing method and system |
| CN112613532A (en) * | 2020-11-26 | 2021-04-06 | 西安电子科技大学 | Moving target tracking method based on radar and recurrent neural network complete infrared fusion |
| CN112494915A (en) * | 2020-12-14 | 2021-03-16 | 清华大学深圳国际研究生院 | Badminton robot and system and control method thereof |
| CN114582195A (en) * | 2022-03-22 | 2022-06-03 | 上海创屹科技有限公司 | Intelligent table tennis teaching system and teaching method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10748376B2 (en) | Real-time game tracking with a mobile device using artificial intelligence | |
| CN113269363B (en) | Trajectory prediction method, system, equipment and medium of hypersonic aircraft | |
| CN117009876B (en) | Motion state quantity evaluation method based on artificial intelligence | |
| CN112052948B (en) | Network model compression method and device, storage medium and electronic equipment | |
| CN110478883B (en) | Body-building action teaching and correcting system and method | |
| CN109800717A (en) | Activity recognition video frame sampling method and system based on intensified learning | |
| CN110413838A (en) | A kind of unsupervised video frequency abstract model and its method for building up | |
| Oughali et al. | Analysis of NBA players and shot prediction using random forest and XGBoost models | |
| CN116421953A (en) | Tennis training method and system based on deep learning | |
| US20220121920A1 (en) | Multi-agent coordination method and apparatus | |
| CN108898221A (en) | The combination learning method of feature and strategy based on state feature and subsequent feature | |
| KR102104007B1 (en) | Apparatus and method for predicting result of game using predictive model of game result | |
| Wiktorowicz et al. | Predictive modeling in race walking | |
| CN110420016B (en) | Athlete fatigue prediction method and system | |
| CN113457108A (en) | Cognitive characterization-based exercise performance improving method and device | |
| Wen et al. | Modern technology, artificial intelligence, machine learning and internet of things based revolution in sports by employing graph theory matrix approach | |
| KR20210047169A (en) | System and method for automatically recognizing virtual ball sports information | |
| WO2019230264A1 (en) | Model training device for action classification, action classification device, model training method for action classification, and program | |
| Fu et al. | Model-based reinforcement learning for playing Atari games | |
| CN117556681B (en) | Intelligent air combat decision method, system and electronic equipment | |
| LeCun | A path to ai | |
| CN115439510B (en) | Active target tracking method and system based on expert strategy guidance | |
| Hao | Research on the Applications of Artificial Intelligence in Golf | |
| Han et al. | Artificial Intelligence Technologies and Their Application for Reform and Development of Table Tennis Training in Complex Environments | |
| Achkar et al. | SmartCoach personal gym trainer: An Adaptive Modified Backpropagation approach |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20230714 |