Athletic training system and method of generating training program

Abstract

To provide an athletic training system.SOLUTION: An athletic training system includes an athlete data repository, an adaptive algorithm, and at least one training unit. The athlete data repository includes biometric data and performance data of athletes. The adaptive algorithm has access to the athlete data repository and is trained on biometric data and performance data to generate athlete-specific training programs for use with the training unit. Each athlete-specific training program is adapted to biometric data and performance data of an individual athlete. The training unit includes adjustable athletic resistance equipment and a controller.SELECTED DRAWING: Figure 5

Description

[Technical field]

[0001] The present invention relates generally to athletic training systems and methods, and more particularly to sport-specific training and performance improvement programs, where applicable position-specific and athlete-specific training and performance improvement programs SYSTEM, METHOD, APPARATUS, AND COMPUTER PROGRAM PRODUCT ADAPTED FOR PROVIDING A SYSTEM - Patent application Regarding.

[0002] [Related Applications] This application is a continuation of U.S. Provisional Patent Application No. 62 / 877,594, filed July 23, 2019. No. 6,393,636, the entire disclosure of which is incorporated herein by reference. Let us assume that. [Background technology]

[0003] Known athletic training units are used by a user, athlete, coach, or trainer. The force is provided by resistance from weights, bungee cords, hydraulic fluid, or air compression. It is necessary to judge the resistance and adjust the resistance for athletic training. Teams, athletes and trainers should use their own personal know-how and non-sport specific The athlete then performs a so-called Routines based on unquantifiable interpretations of athletes' strength by experts and tacit knowledge Athletes often receive a program that includes vertical, agility, and drills. Training without establishing baseline measures in the areas of lateralization, speed, and horizontality Coaches and trainers often develop and improve aspiring athletes in these four areas. Without setting a baseline measure of leadership, athletes can Train your athlete. Athletic gains in four areas: improvement, progress, Progress (advancement) is the validation of observed improvements using previously established metrics. Since it is not possible to do so, it becomes subjective.

[0004] Generalized training programs created by coaches and trainers for athletes Because these routines are designed to be an all-purpose training program, These programs are optimized to provide The athletic improvements or gains that result are unpredictable and random, and therefore They are not guaranteed because they are not necessarily measured and verified for a particular athlete. Individualized to biological and anatomical profile and athletic abilities or goals. Creating individualized plans for athletes is a challenge for many coaches and trainers. This is time-consuming, costly and inefficient for coaches and trainers. The company does not have access to a lot of data from athletes and therefore By leveraging the overall availability of data, It is not possible to extrapolate the data.

[0005] Athletes can achieve gains through traditional training programs However, the benefits are not always consistent with each specific need. An athlete who is trying to improve his speed may want to improve his speed. Monst to carry out the "Leap and Burst" training program We have contracted with specialized training centers such as VASH Vertical Power. Achieve gains in all four areas (vertical, agility, speed, and horizontal). However, this athlete believes that verticality affects his performance as a hockey player. Verticality is not a big consideration since it does not have a significant effect on the performance. In this case, the coach will create a training program specifically for this athlete. Obviously, the latest enabling technologies in machine learning and artificial intelligence Without technological automation, this would not be possible for large numbers of athletes.

[0006] Additionally, coaches and trainers in clubs, schools, universities and professional fitness centers Pre-planned training programs using the Athletes have no way to optimally calibrate the resistance they use based on their current training program. Resistance is individualized according to the program, but if resistance is too high or too low, When using resistance training, the resistance level must be adjusted manually. This is done for each athlete. Resistance must be adjusted manually for each athlete by the coach, but this is still based on trial and error. Therefore, it may still be an inaccurate process.

[0007] Similarly, it is used to extract resistance for athletes while performing drills. The accessories include the athlete's age, height, weight, gender, grip and hand size, waist and This should be optimized based on the dimensions of the shoes and seams, but this can be adjusted for each individual athlete. Coaches must rely on trial and error to determine the appropriate accessories for each individual athlete. Fine-tune and manually optimize drills and exercises to meet each athlete's specific goals and needs. The science of optimizing sequences of resisted and non-resisted drills, rest periods, and attachments The mathematical approach significantly improves the results of the training process and allows more Allows Sleet to train at any given time. Summary of the Invention

[0008] The athletic training system will have an athlete data repository and adaptive algorithms. The adaptive algorithm includes at least one training unit. The Athlete Data Repository contains athlete biometric and performance data. The fitting algorithm accesses the athlete data repository and Regarding biometric and performance data of the type stored in the Leiden Data Repository Multiple athletes are trained and used with the training unit. Generate a unique training program for each athlete. , adapted to each individual athlete's biometric and performance data.

[0009] The training unit includes electronically adjustable athletic resistance equipment and a control. The controller includes a training program specific to the plurality of athletes. receiving at least one athlete-specific training program from The resistance level of the adjustable resistance equipment can be adjusted according to the athlete's specific training program. The controller is also configured to adjust the received athletics signal. The athlete's use of the training unit in accordance with the athlete's specific training program The adaptive algorithm is configured to record performance data of the individual access biometric and performance data of athletes to Generate a sleet-specific training program.

[0010] The training unit is adapted to interface with at least one mobile device of the user. The mobile device may further comprise a training interface for the athlete. a training program specific to the athlete from the adaptive algorithm; The mobile device communicates a program to the training unit. receiving the performance data recorded by the controller, The adaptive algorithm may communicate the recorded puff Performance data may be communicated to the athlete data repository.

[0011] In some embodiments, the adaptive algorithm comprises a plurality of adaptive algorithms. For example, the adaptive algorithm may include a plurality of machine learning models. The learning model is based on the biometric data of a given athlete and the performance data of other athletes. a rating prediction model that receives the data as input and generates a rating training program; The plurality of machine learning models may be adapted to evaluate the computer during use of an evaluation training program. A drill routine that receives as input performance data recorded by the troller The drill routine model may further include a drill routine model. During use of an athlete-specific training program previously generated by The performance data recorded by the controller may also be received as an input.

[0012] In some embodiments, the training unit includes a vertical jump station, a water In some embodiments, the training station includes a speed station, a horizontal agility station, or both. The training unit has multiple vertical jump stations and multiple horizontal agility and speed stations. The training unit further comprises a training unit responsive to the controller. The adjustable device may further include a motor for adjusting the resistance level.

[0013] A program to generate athlete-specific training programs adapted to individual athletes A computer-implemented method is also provided, the method comprising the steps of: Individual athletes and additional athletes with comparable biometric and performance data Biodata and performance data from athletes are stored in an athlete data repository In a second step, the individual athletes are The training unit includes an electronically adjustable athletic resistance and a controller, the controller comprising: Receive an athlete-specific training program from the training program and Adjust the resistance level of the adjustable resistance equipment according to your specific training program and training the athlete in accordance with the athlete's specific training program. The device is configured to record performance data of the use of the sleet.

[0014] The third step is to identify the individual athlete and the tracker being accessed. The adaptive algorithm provides an identifier of the processing unit. The system uses biometric and performance data of the type stored in the athlete data repository. Multiple units are trained on the sensor data and used with the training units. In the fourth step, a specific training program for the athlete is generated. The adaptive algorithm accesses the athlete data repository and At least one athletics program that incorporates Street's biological and performance data. In a fifth step, the athlete generates a specific training program. A unique training program is provided to the training unit.

[0015] The training unit is an interface for at least one mobile device of a user. In this case, the training unit may further include a user interface. The steps include: having the individual athlete connect a mobile device to the training unit; The training program is then combined with the training unit to provide the athlete with a specific training program. The step of providing a unique training record to a user on the mobile device includes: receiving a training program from the adaptive algorithm and The method may further include communicating a training program to the training unit.

[0016] The method further comprises the step of: the mobile device receiving the puff recorded by the controller; and communicating said performance data to said adaptive algorithm. The method may further include the step of:

[0017] The adaptive algorithm may include multiple adaptive algorithms, such as multiple trained machine learning models. The adaptive algorithm may include a biometric algorithm for the individual athlete. access information and / or performance data for that athlete Generate training programs for

[0018] The present invention provides a method for training athletes in at least four athletic disciplines, including vertical, speed, and horizontal. Adaptive athletics including hardware and software adapted to enhance gain in The adaptive athletic training system includes A sports training unit with several stations, each station comprising: It can accommodate one athlete at a time, allowing multiple athletes to train simultaneously. Some stations allow training at least vertically. While the remaining stations include a jumping platform for skiing, the remaining stations are focused on at least three other disciplines: training horizontality, agility, and speed. The idea is not limited to these areas, but can also be used for general fitness goals and endurance, etc. It can be expanded to train other physical attributes. Each station has a Athlete Partner for designed and personalized training routines and drills The performance-based resistance components are attached to the sports training unit. It includes a rechargeable battery and a cradle to hold several interfacing IoT devices. The sports training unit may include Bluetooth or Internet The data collected by the sports training unit is connected to the database via the Internet. Then, artificial training is performed to adapt the training of each individual athlete according to the sport. AI, machine learning (ML), or other computer software The sports training unit can store data that is relevant to the user's daily activities. can.

[0019] The sports training unit includes various sensors and sensors such as accelerometers, proxies, and gyroscope, ultrasonic, LiDAR, camera, computer software It can include software, databases and mobile applications to help athletes improve their sports. Recommend training programs specific to the skill and position. Athletic Training Systems is a combination of biological performance testing and sports specific Access and measure performance tests to help athletes improve their performance in sport. A unique athlete-based training program to achieve athletic gains You can customize the ram. Sports training units can be customized based on age, height, Gender, ethnicity, sport, position in sport, body mass index, weight, oxygen level Captures biometric information such as bell, height, wingspan and finger height to measure athlete performance Athlete- and sport-specific training plans to improve performance egimen) and generate scales. [Brief description of the drawings]

[0020] [Figure 1] FIG. 1 illustrates connectivity of a mobile device, a training unit, and an athlete data repository according to an example embodiment. [Diagram 2] FIG. 13 illustrates a single athlete adaptation of a training unit according to an exemplary embodiment. [Diagram 3] FIG. 1 illustrates a process for using a training unit according to one example embodiment. [Figure 4a] A diagram illustrating an arrangement of a jump platform 400 as part of a V station according to one example embodiment. [Figure 4b] FIG. 4b shows some components of the jump ramp 400 of FIG. 4a in more detail. [Diagram 5] FIG. 1 illustrates a training unit when intended for multi-athlete use according to an example embodiment. [Figure 6] FIG. 1 illustrates an example database structure and data therein that can be used by computer software to predict the best training plan for an athlete, according to one example embodiment. [Figure 7] FIG. 1 illustrates a HAS station according to an example embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0021] One exemplary embodiment of the adaptive athletic training system 10 includes at least one an athletic training unit 100 and an athlete data repository 105; Access this data repository to develop sport- and athlete-specific training programs a processing system 125 for generating a gram and at least one mobile device 110. The mobile device 110 includes a training unit 100 and computer software. A mobile app (application software) that communicates with the software 125 is installed. It has been installed.

[0022] Internet of Things (IoT) devices and sports training An example of the training unit and its connectivity to the athlete data repository is shown in Figure 1. The user can connect his / her mobile device 110 to a wireless network using Bluetooth, Wi-Fi, The training unit 100 is connected to the training unit 100 via a digital interface, cable, or other digital communication means. The data collected by the training unit for that user is Anything that a user enters into their mobile app is sent to the user's mobile device. The mobile app also collects metadata about the user's activities. 125, which in turn stores the athlete data in the athlete data repository. Although FIG. 1 shows some data paths as an example, the present invention However, it is not limited to such. Those skilled in the art will appreciate that the mobile device 1 shown in FIG. 10 and the training unit 100 access the data repository via the Internet 115. It will be readily apparent that the present invention can be implemented in a variety of ways, including by directly communicating with the server 105.

[0023] When the user wants to complete another workout, the user simply taps their mobile device. The mobile device then connects to the athlete data repository. Using computer software processing system 125 data from the current Create training programs that are individualized to the needs of your athletes The mobile device may provide an athlete identifier and a training unit identifier. In some embodiments, the computer software 125 may include one or more at least one trained machine learning model or other suitable artificial intelligence technique; The computer software 125 includes an adaptive algorithm that is adaptive to the current user and other users. The data from the user can be used as a training set. Any data collected by the training unit in this way will be used to determine the training effort, training To monitor the duration of the training session and measure the gains made by the user Data from the athlete data repository is collected and sent to a database for The progress, gains, and other metrics are then provided to a web application via the Internet 115. The web application can be accessed through the application 120. , it may be possible that the training unit is not interfaced with the training unit.

[0024] 200 single-unit training units used by a single athlete at a time. The shape of the single use unit is shown in Figure 2. The shape of the station may be similar to that of the lead unit 500 (FIG. 5). Meanwhile, the remaining stations are This is the front of the training unit 200. 200 is a single station that combines V and HAS stations. The vertical component of the device (described in more detail in Figures 4a and 4b) The jump platform 205 and three V-nodes 210, 215, and 220 are included in this example. In an embodiment, each V node is connected to a cable connected to a resistive source, which may be an electric motor. Each V-node has a user A cable is provided which connects to a spool inside the knit and to a ring on the exterior of the unit.

[0025] This unit consists of lateral, agility and speed training. One HAS track 235 facing the athlete and used for the gymnastics It also includes a camp platform 205. The HAS station is described in more detail with respect to FIG. The example HAS track 235 includes two HAS nodes 225 and 230. Nodes 225, 230 are used to adjust resistance in training drills The HAS node is connected by a cable that can reach the electric motor resistor. A ring 7 is attached to the end of a cable that is connected to the athlete while performing the drill. 20 (FIG. 7). The HAS nodes 225, 230 move along the HAS track 235. The HAS track 235 can move the nodes to provide a greater range of motion. It can also be rotated up to 90 degrees to level horizontally or vertically, thus further Provides a large range of motion.

[0026] The Single Athlete Device may be a mobile device or other type of smart device. Alternatively, the device may include a single cradle 240 that can accommodate an IoT device. The device is powered by a rechargeable battery to eliminate the need to always have the device plugged in. It can also include 245.

[0027] A flow process using the training units 200 (FIG. 2) and 500 (FIG. 5) is shown. This process is shown in Figure 3. The user connects his / her mobile device to the training unit. It all starts with connecting to the Knit 300. The system is designed to help athletes The program is checked to see if it has been used before (305). In this case, computer software 125 is used for unit learning programs, AI programs, or using another type of computer program to conduct pre-training evaluation drills 325 Users can create a mobile app by uploading their metadata and biometric data to the mobile app. Input via application (310a) or web application (310b) Athletes will be asked to use the units, BMI scale and oximeter. For conditioning gains in addition to VASH gains, use with related accessories such as Obtain data for the design of individualized training programs, including rest periods. The data and biometric data can be queried by the computer software 125. This metadata and biometric data is stored in the athlete data repository 315 as follows: , athlete identifier, gender, age, BMI, the position the athlete plays The sport, if applicable, the area the athlete wants to improve, height, weight, finger height, maximum reach Includes, but is not limited to, points, wingspan, and sport-specific professional association tests Second, new athletes should use a baseline of their current athletic status. The biometric data entered and the combined base ID must be set (320). Trainer data tracks your training with personalized resistance levels and drills The training program is used to design a step-by-step program (325). The training unit 200, 500 is then sent to the training unit 200, 500 in step 330. Sleet uses the training unit 330 to Complete the training suggested by (335). If you are a first-time user, you can Data collected from the training can be used to suggest routines. The results of the training are collected by the training unit and used in the athlete data collection. The data is stored in the repository 315 .

[0028] If the user has previously used athletic training system 10 The computer software queries the athlete data repository 315 for data. Specify the resistance level, drills, and training routines for that athlete. The data used by the computer software is based on past workouts from the athlete. This can include crowd-sourced data from workouts by athletes or other athletes. Computer software125 generates specific training programs for athletes The athlete then performs the drill (335) and the training unit 330 stores the data The athlete performance is collected and recorded (340) in the athlete data repository 315. If there are any improvements in performance, those improvements are also logged.

[0029] The computer software 125 generates a training program based on several factors. The resistors are trained on the unit, including the metadata entered manually. Sport-specific profiles for that athlete and all athletes in total and athlete specific. Data is collected from current users, It is also crowd-sourced user data. The data collected from the athlete includes type 2a muscle fiber signals, resistance levels, cable These may include, but are not limited to, tension from the cable or acceleration of the cable. The program will be tailored to the athlete's current and predicted future condition. Training with and without mats, resistance and non-resistance, and rest periods are included in the program. Resistance in training plans can be proposed as a concentric muscle movement. Simulating the opposing force of muscle movement from eccentric to eccentric muscle movements Resistance can be measured by athletes or sports-based athletic test scores. Rest in a training plan can also be extracted between drills by a It can change over time based on the oxygen levels in the sleet.

[0030] An example of a jump platform 400 included as part of a V-station is shown in FIG. The V station is at least a useful tool in vertical training to increase the height of the vertical jump. In the illustrated example, each jump platform 400 includes three nodes. The number of nodes does not have to be limited to three. The three nodes in this example are the left hip node, The nodes are the right hip node 421, the left seam node 405, the right seam node 411, and the right hip node 421. Cables 403, 410, 420 and rings 4 are attached to the athlete performing the drill. Includes 02, 409, and 419. Each unit is attached to the main body of the training unit. A swing-down flap arm having a mounting bracket 429 for fixing the swing-down flap arm This exemplary embodiment also includes a flap arm for each individual cable spool. Shown are hip nodes 405, 421 and seam node 411 connected to unit 426. Referring to FIG. 4b, the spool unit includes a spool coupled to an electric motor 428. The spool 430 connects to an electric motor in the base of the training device. The HAS node has a cable connection port for connecting to the HAS node.

[0031] The electric motor 428 in the illustrated example may be adapted to adaptively calibrate the resistance of the athlete. drills and the repetitions and effort or, if applicable, depending on the athlete in each drill The device can be controlled or programmed to provide varying levels of resistance. Data is collected before, during and after the training drill. It can be programmed to collect data to improve future recommended training plans. An embedded system that controls the electric motor 428 is 00 via a mobile device 110. Data from the unit 100 may be stored as results in a data repository 105. can.

[0032] Another exemplary embodiment of the present invention is a training unit 500 intended for multi-athlete use. An embodiment is shown in Figure 5. This exemplary embodiment has eight stations, namely Although four V stations and four HAS stations are shown, the present invention uses eight The present invention is not limited to these stations or to each of the four station types. The V stations in are V Station 1 505, V Station 3 515, V Station 5 525, and Station 7 535. All of the programs include a jump platform 400 as described with reference to FIG. Each jump platform is equipped with multiple nodes. FIG. 5 shows three nodes on each jump platform: right hip node 42 1, seam node 411, and left hip node 405. In this embodiment, Each node is a link that the athlete can hang on to while performing the drill. In this example, the cable runs from the left hip node to Extends from the right hip node to the seam node. Inside the base is at least one electric motor. The motor is connected to the cables between each node. The motor is controlled by computer software and a training unit. It can be automated to increase or decrease resistance based on recommendations provided by The V Station is mounted on the main body of the training unit, facing the jump platform. It can also contain a set of rotation nodes that rotate the object through 90 degrees. It can also move along the rails.

[0033] There is a HAS station between V stations. Figure 5 shows HAS station 2 51 0, HAS Station 4 520, HAS Station 6 530, and HAS Station In this figure, HAS station 6 is the training station. It is located at the back of the unit and cannot be seen in Figure 5. The HAS station is Use the training unit to at least train the athlete's horizontality, agility, and speed. Each HAS station is used to synchronize the These nodes include rings and Each node or combination of nodes is designed for a specific purpose in a training drill. These nodes can move up or down along the rails. The rails can rotate up to 90 degrees to allow for a greater range of motion. Within the training unit, there is a resistance training station that changes for athletes using the HAS station. There may be an electric motor programmed to provide a resistance level. This is described in further detail below with respect to FIG.

[0034] The body of the training unit is at least one display that displays training drills or other information. The unit may also include a display console 545. The number of training stations can be adjusted to accommodate the number of mobile devices or other IoT devices. The embodiment also includes several cradles 550-560 for storing data. However, a training unit with eight stations uses this device. At least eight mobile devices will be held by each of the eight athletes using them simultaneously. The device may have one cradle and one power outlet. It can also include a rechargeable battery so that plugging in is not always required. .

[0035] Figure 7 can be used to improve horizontality, agility, speed or another training area. 7 shows an exemplary embodiment of a HAS station 700 for use in 700 can be used for HAS stations 510, 520, 530, 540 The HAS station illustrated in this example embodiment has two nodes 705 and 71 0. The two nodes are electrical motors that change the resistance level during training drills. Each node is connected to the HAS cable 725, which can extend to the Rings 715 and 720 are attached to the ends of the cables to be used by the athletes performing the training drills. The cable has various tensions depending on the level of resistance provided by the electric motor. You can unwind and rewind in a level. Nodes are HAS rails. and moved along rail 730 to provide increased mobility to the athlete performing the drill. The rail can rotate up to 90 degrees (735) to provide athlete The range of motion of the rail after a 90 degree turn can also be increased. and two nodes are logically shown.

[0036] Machine learning algorithms to design optimal training plans and predict athlete gains An exemplary database structure that can be used by the algorithm, and the data therein. The data is shown in Figure 6. The athlete data repository 105 collects data from different sources. It may contain several tables or collections of different data collected. The data in the data repository is used by athletes in conjunction with their training units. Used by computer software to predict the best possible training plan for The database structure can be used in cloud, SQL Server, relational The data contained in the database may include a database structure, a Data can be made anonymous and available only to the user to whom it belongs. Cut.

[0037] This example shows four data collection bodies, but other data sources may be used to plan future training. These data sources do not need to be kept separate. The first data source shown in the figure is the Loud source data 605. This may be generated manually by a person or by other training units. This can include training results collected by the training unit. This may also include data collected by the knit from other athletes. For example, Training results from athletes fall into this category. This data is similar to By comparing the results of other athletes of the same build, gender, age, sport, etc. It can be used to improve training plans for current athletes. The algorithm looks at the data collected from every athlete to determine the best gains. Increase the accuracy and effectiveness of planning for athletes with similar profiles to the athlete selected It can be developed like this.

[0038] The next source of data is current athlete metadata and biometric data 610. , gender, age, BMI, the sport the athlete plays, and the area the athlete wants to improve. , height, weight, finger height, maximum reach, wingspan, and sport-specific professional associations This data may include, but is not limited to, crowd-sourced data. Use this in conjunction with your current user's metadata to compare with other athletes who have similar metadata. Training results can be compared.

[0039] The third source of data in this example is the baseline testing of the current athlete. Baseline tests may be performed, for example, on the initial vertical jump, long jump, or training unit. This may include other athletic measurements that may be collected by the It is used as a benchmark to measure any gains or to suggest areas for improvement.

[0040] The final data source in this example is the current athlete's training results 620. These results measure improvements in athletes' performance and provide a guide for further improvement. Baseline testing of athletes is collected to suggest areas of expertise. Can be combined with training results.

[0041] As will be described in more detail below, the four data sources in this exemplary embodiment and other The data comes from the future training of athletes using the Training Unit 100. Computer software such as AI or machine learning algorithms to forecast plans1 25 These four sources of information are included in the Athlete Data Report. Data can be collected and stored in the data repository, but each athlete can only access their own data. It is not possible to

[0042] The machine learning / artificial intelligence components of the adaptive athletic training system are known to those of skill in the art. This can be done in many ways that are well known. The following machine learning example shows one way of doing machine learning: Although provided as an illustrative example, the invention is not limited to this particular example.

[0043] In the examples herein, multiple machine learning models (ML models) are used to generate adaptive athletic The training system is developed for different purposes. This is based on the specific needs of each athlete. To meet your individual needs / goals and achieve maximum athletic gains, Provide each athlete with the exact specifications for the routines. These routines are based on the drills used. resistance levels and accessories, each optimized and tailored for a given individual athlete. and individualized.

[0044] The combined goal of the machine learning model is to improve performance for each athlete and each sport. Predict the optimal drill routine and resistance level for each drill to maximize performance gains. Each focuses on a different aspect of the personal training program. Multiple machine learning models can be used.

[0045] In some embodiments, the ML model engages in an adaptive training program. It is developed to create an individualized training routine for each athlete. , training sessions over the duration of the athlete's chosen training period. Drill routines are designed to be adapted to each individual athlete's sport. should be individualized to improve athlete performance in the four VASH domains Routine customization is, first, the areas that users prioritize; areas important to the user's sport, and finally, weaknesses identified by the user evaluation sessions. Areas of weakness should be focused on improving. and more generally, VASH areas where users received lower scores than expected. The routine for each training session will include: Resisted sets with concentric and eccentric resistance levels and non-resistance sets during the time the athlete is disconnected and during rest periods. The adaptive training program consists of a sequence of drills and Coil and recoil based on training performance and heart rate ) resistance and rest time are calibrated.

[0046] In some embodiments, the ML model is adapted to perform the following functions according to the athlete's biometric data and the evaluation results: Each drill is individually developed with a resistance level that is optimized for the verticality test. Athletes who score lower than expected based on their biometric data should be considered for vertical drills. The system experiences less initial resistance at the athlete's After collecting the training data, the system will compare the athlete's performance the last time they did the drill. Since we know what the performance was like, we can determine the resistance level for a particular drill. can be optimized.

[0047] In some embodiments, the ML model is adapted to estimate the athlete's ... Some athletes may have a certain height of play, but the accessories are developed to be individualized for that player. Some people are not tall enough to use lyometric boxes or hurdles. Some people may not have enough strength to lift a heavier medicine ball. Therefore, training routines are individualized for the athlete with applicable drills. First, the model will include the athlete's test results and athlete biodata. Based on the data, we predict which accessory the athlete should use. Similar to the resistance prediction, the model will adjust accessory regulations as it has more training data on athletes. and optimize how the athlete operates with the accessory. In this way, it is possible to have a better modeling of how the

[0048] In one example, five ML models are used. The models described herein are , can be combined or further specialized into more models. More or fewer ML models can be used, and the user can As you start using your training unit, more training data will become available. In addition, ML models can be added. Each of the modules can be used individually or in combination with each other. Cut.

[0049] The more data the adaptive athletic training system 10 collects, the The learning becomes more powerful.

[0050] The first ML model is the Predict Combine Scores model. This model is based on user metadata and biometric data ("profile data"). The output of the model is the predicted score and the actual score. can be used to identify areas of weakness of the user.

[0051] The score prediction combination model is based on age, sex, height, weight, BMI, sports, and position. Enter user profile data including the application and test for which the results are predicted. The input of the model includes the biometric information of the user. The output of the model is the current user's biometric information during the input test. Optionally, to better predict the outcome, you can select the type of evaluation (evaluation can include as inputs: evaluate, appraise, verify The output of this model can have several useful applications. First, it can be used to test the user's By predicting test results and comparing them to actual results, the system works in four areas: Areas of weakness in (VASH) can be identified. These areas of weakness are These can then be input into a second model that recommends drills. Second, the four areas (VA The identification of weak points in the SH was used as input to a third model that predicts the optimum resistance level for the drill. Can be used as a force.

[0052] The second ML model is the Drill Routine Prediction model. This model combines the user's chosen field of interest with the first score prediction model. Weak areas identified from user evaluation scores in the rule and maximized user gains in the rule Predict the drills you will need to do.

[0053] The drill routine prediction model inputs the user's evaluation results and Model 1 "Score prediction group" Use the “Align” function to input the predicted evaluation results. In addition, the athlete selects the The actual results are below the predicted results in each of the four measurement areas ( These percentages are converted to average percentages that are greater than or equal to the user's This will be used along with the chosen goals and the importance of each area to the athlete's sport. You can predict the drills of the session.

[0054] The output of the Drill Routine Prediction Model is a user-selected training routine for the next training session. Currently, these drills are listed in the User Training The planning table will be stored in the database, but in future machine learning, the data will be stored on the user's mobile device. The mobile device is capable of interfacing with the training unit. The interface provides drill instructions to the machine and the user.

[0055] The third ML model is Initial Drill Resistance Prediction. The model is a model in which an athlete has drill results stored in an athlete data repository. If not, the system will notify you of the resistance level at which the athlete struggled or was successful. Therefore, we do not have data on the optimal initial resistance level. A learning model is provided.

[0056] This initial drill resistance prediction model is based on age, sex, height, weight, BMI, finger height, and sports Enter your profile data including the drill, position, and resistance you expect In addition, the drilling is linked to the measurement field (VASH), so the initial drilling resistance prediction model Dell uses the “score prediction combination” output of the first model to predict the user’s The percentage of the prediction that is below or above the expected value can then be determined. The initial drill resistance prediction model is based on a percentage below or above the The resistance can be increased or decreased appropriately and provided as an output. For example, the initial drill resistance Predictive ML models determine maximum jump (vertical drill) based on user profile data The user expects resistance of 6.75 kg (15 lbs). However, for example, On average, users scored 20% below the predicted results on the perpendicularity test. In that case, the initial drill resistance prediction ML model doubled the predicted resistance by 15 pounds. It can be lowered by 0% to 5.4 kg (12 lbs).

[0057] The output of the initial drill resistance prediction model is not necessarily used as input to other models. However, the results of training a user on the output of this model can be used as input to a new model. Can be used as a force.

[0058] The fourth ML model is the Predict Recursive Resistance model. After Sleet does his first drill, the system monitors how the user responds to the resistance level. The system uses this information to improve future Resistance levels can be better optimized.

[0059] For example, after an athlete completes a drill for the first time, the system will detect when resistance was too strong First, the system has more information about whether each iteration was too weak or too weak. The system can track the time between repetitions. If the repetitions are evenly spaced, the system If the user makes a long break between repetitions, it can be predicted that the resistance was too low. If you take a hit, the system can infer that the resistance is too high. If applicable, sensors are used to track the applied muscle forces. Both data sets were accurately analyzed using LOESS curves, giving a more accurate view of fatigue. The LOESS predictions of the last iteration are the inputs to the fourth machine learning model. Can be used as a force.

[0060] The output of this model is not just to more precisely optimize resistance levels for a particular drill, but also to Used to provide resistance levels to the device and mobile apps.

[0061] The fifth ML model is the Predict Accessory Configuration model. The system preferably includes a drill that requires an attachment (i.e., a 3-hurdle shuttle). We can predict the optimal attachment for athletes performing a full (three hurdle shuffle) Some athletes just don't have the right body type for some accessories. However, some tests have high enough scores that accessories can be used. For example, an athlete who is 1.2 m (4 ft) tall may not have a 0. Using a 6m (2ft) plyometric box is not advisable. The athlete is simply not tall enough. Athletes with a landing point less than 0.6m (2ft) may attempt to reach a 0.6m (2ft) plyometric It is also not advisable to use trick boxes.

[0062] The model includes age, sex, height, weight, wingspan, finger height, hand and palm size. Size, foot size, waist and seam length, BMI, sport, position, and accessories The user's profile data is entered including the relevant drill for which a prediction about the product is to be made. The model outputs a reasonable attachment configuration for the current athlete and drill. The force is the applicable drill attachment configuration. For example, the output is "6 lbs medicine ball". It can be written as "ru".

[0063] In some embodiments, drills based on the athlete's biometric data and test results; The machine learning model can then be adapted to fit the constraints set by the model itself, including constraints on resistance, and attachments. When outputting a result, the system outputs it so that the result remains within the constrained range of values. For example, the ML model can be adjusted to The system predicts resistance of 40 pounds for a 2-year-old athlete. However, the system predicts resistance of 40 pounds for a 12-year-old athlete. For athletes performing "ski jumps," limit resistance to 5 to 30 pounds. The system can be configured to reduce the resistance from 40 lbs to 30 lbs. Other restrictions may be implemented for other exercises, athlete categories, etc. can.

[0064] In some embodiments, a recurrent neural network is used to generate a machine learning model. This type of model can be implemented when the input is a mixture of numerical and categorical variables. Therefore, it is well suited for adaptive athletic training systems. Other algorithms, such as decision tree regression, may also be suitable. There are also many robust supervised regression models that can be used.

[0065] The variables input to predict a user's score on a given test are: Age, gender, ethnicity, height, weight, wrist height, wingspan, sports, sports positions Sports The position is the athlete's indicated at the start of their training program. This can be implemented as a single variable and is a sports position. and position cannot be implemented as two separate variables. Football wide receivers score differently than American football linemen. In addition, including only positions is not sufficient. So, there may be a difference between the center in basketball and the center in hockey. Instead, this variable is entered as a unique identifier for the sport and position. For example, id1=American football wide receiver, id2=American football id3=basketball center, id4=hockey center A drill is a drill or test where the system is predicting the athlete's outcome. The output of the model is the predicted outcome, i.e. the probability of the input data being used given the biometric information of the user. This is the result of a strike.

[0066] Here is one example of how to create an ML model that corresponds to the score prediction combination ML model: Using this and other techniques, we train a variety of ML models. It is possible.

[0067] To train a machine learning model, training data is typically generated by the machine learning algorithm. The related data is often compiled into a single table that is fed into the algorithm. These tables can be stored in different normalized database tables. The data can be combined by extracting related data and discarding non-related data.

[0068] Once the training dataset is compiled, the relevant biometric data is then used to This is the athlete's GUID (Globally Unique Identifier). (used to identify a user across all tables) and use this to By filtering the dataset to extract only the evaluation results for that athlete, Then, you can measure your height, weight, finger height, wingspan, oxygen level, or body fat. All the evaluation results for that athlete are extracted, including stored biodata such as fat percentage. do.

[0069] The assessment test and results will be compared against the biodata closest to the date the current assessment test was conducted. For example, if an athlete completes the “long jump” on February 20, 2019, If an athlete has not measured their height on the date, the system will You can check other dates that were measured. In addition, the system will check the time between the two measurement dates. Biometric data can be interpolated. Various existing algorithms can be used to query the data. (also known as imputation) (This is the case.)

[0070] You can choose different algorithms for different variables based on the data type and function of the variables. In addition, for missing values ​​that may be dependent on other variables, Multivariate Imputation using Chained Equations (MICE) or Probabilistic cleaners such as Principal Component Analysis (PCA) A probabilistic cleaning algorithm can be used. For example, the finger height is It is reasonable to assume that the combination of height and wingspan is roughly correlated with Therefore, the system will use the current height and wingspan values, if they exist, to From this, it is possible to infer the missing height probabilistically.

[0071] The score prediction combination model requires a supervised machine learning algorithm. Machine learning is what happens when inputs are mapped to an output space, and the model is trained with labeled inputs and outputs, and the desired Specify the output to the model.

[0072] In addition, this requires a regression model rather than a classification model. Classification is done because the range of outputs is It is discrete and finite, and the only possible outputs are the labeled outputs for the model. On the other hand, regression can have a continuous output space. For example, Two athletes with scores of 2 and 3 seconds respectively for the 47 foot sprint. If there is a third user, the model can predict that the third user will have a score of 4 seconds. It is entirely possible. The difference is that classification deals with categorical outputs, whereas regression deals with numerical The first step is to handle the output.

[0073] When training an ML model to split a dataset into two, A dataset to train on and a test and score the model (validate its effectiveness) It is common to create a second dataset, a test set, and a second dataset, a test set, to see how well the model performs. In addition to providing a way to test whether a particular It also prevents "overfitting" of the model, where multiple data points are learned as rules by the model. do.

[0074] For example, a dataset may consist of 80% of rows for the training set and 20% for the test set. 0% of rows. If the dataset is large, a 70-30 split is appropriate. is a better split. The rows can be split by choosing the top 80% of the rows, or Alternatively, the selection may be made randomly.

[0075] When training a machine learning algorithm, the inputs are typically all weighted equally. For example, without scaling, a weight of 150 would be 1 has a bigger impact than an age of 2 because 150 is a much larger number. To solve this, each variable is scaled from 0 to 1. Various methods for scaling features, including scaling or binning features There is a law.

[0076] Each neural network is modeled by the number of hidden layers and hidden nodes, which are parameters in the algorithm. The number of hidden layers and hidden nodes is chosen by the user. For example, the first neural network is implemented with 12 hidden nodes. and the second neural network is implemented with 15 hidden nodes. They can function equally well. There are other neural network parameters that can be used. Different regression algorithms can be used. When used, the type of parameter may change. For example, "random forest" In linear regression there is no such thing as a hidden node.

[0077] The initial configuration, tuned according to a small training dataset, has the following parameters: The parameters are: hidden nodes - 100; learning rate - .005; learning iteration ret - 100; initial training weight - .15; and normalizer - min / max normalization. For the reasons stated above, this example is provided for the completeness of the disclosure and is not intended to be limiting. These parameters are intended to be indicative of the quality and quantity of data in the system. can be modified taking into account.

[0078] Depending on the type of model used, the model may be adjusted after it has been scored. For example, it may be necessary to use a model such as random forest regression. Once the model is adjusted, a process called "cross-validation" is completed, and the current statistical analysis is Predict parameters based on how they perform on an independent data set This process also reduces overfitting. However, is not applicable to some regression algorithms.

[0079] Once a model is trained, it is scored using a test dataset. The data from the test set (this In this example, the inputs are the age, gender, ethnicity, test, etc., and the output test result is determined by the model. Then you can evaluate how the model performed. A common way to judge the validity of a rule is the coefficient of determination, also known as R2, or the root mean square error. (RMSE: Root Mean Squared Error), or other statistics. These scores are If it falls short of the assumed case, remodel using different inputs or algorithm parameters. It may be necessary to retrain the

[0080] Once the model is trained, it will predict future outcomes based on biometric data collected from the athlete. To predict the results of athletes' tests, an application programming interface was developed. You can build and use an API (Application Programming Interface) The results of this model are stored along with other user information for further machine learning or predictive analysis. It can be used as an input to the model.

[0081] The training unit captures sport-specific performance tests and The test is based on the standing maximum reach, running maximum reach, Horizontal jump on both legs, horizontal jump on right leg and horizontal jump on left leg, side step jump on right leg and side step jump on left leg Right leg step jump, right leg step back jump and left leg step back jump, 4.55 meters (5 yards) right shuffle and left shuffle, 5 yards forward shuffle Includes sprints and backward sprints, 40-yard dash and 75-foot sprints. For example, in basketball, an athlete may perform a side step jump shot. The training unit will measure the athlete's optimal sidestep jump performance. This can include tests to measure the pre-training span of the training state. Jumps are the The training unit simulates the exact movements required for athletes' spa. The training unit measures the athlete's side step span. Require training routines that include both exercise and full motion components. After the training period, the athlete's entire customized training program After completion of the test, the athlete is retested to verify the gains.

[0082] The training unit consists of at least four exercises including vertical, agility, speed, and horizontal. It is designed to improve athletic performance in three specific areas. The unit consists of standing vertical jump, right leg takeoff jump and left leg takeoff jump, parallel jump, Improve verticality using drills such as the long jump and two-foot takeoff. Use drills such as the right leg horizontal jump, left leg horizontal jump and two leg horizontal jump, also known as the right leg horizontal jump. The unit also allows for improved levelness using the left and right lateral thrusters. Provides measurable improvements in agility through drills such as lateral sprints. Afterwards, the unit can sprint forward or backward. print) to improve speed. Athletes will perform drills to improve their skills. Cables attached to the unit provide the specific resistance athletes need to improve their Perform these drills by hooking yourself onto the cable.

[0083] Between the V stations, there are HAS stations, and each HAS station has two or more The rail includes nodes on the top of the rail that guide the athlete during training. These nodes include rings and cables that are attached to the rails. Each node or combination of nodes represents a training drill. The unit body also includes at least one display console. The main body of the unit is a training pad for attaching a mobile device or other IoT device. The system also has a number of cradles corresponding to the number of training stations.

Claims

1. Athlete data repository containing athletes' biometric and performance data and, Accessing said athlete data repository and recording to said athlete data repository The system is trained on the biometric and performance data of the type stored in the system. Multiple athlete-specific training used with several training units An adaptive algorithm that generates a program that is specific to each athlete. The grams are adapted to the individual athlete's biological and performance data. Response algorithm and A training unit including adjustable athletic resistance equipment and a controller. The controller then selects one of the plurality of athlete-specific training programs. receiving a training program specific to an athlete; and adjusting the resistance level of the adjustable resistance equipment according to the training program of the user. The training unit performs a training program specific to the athlete. configured to record performance data of the athlete's use, The ing unit and An athletic training system comprising:

2. The training unit is adapted to interface with at least one mobile device of a user. and wherein the mobile device further comprises a training interface for displaying the athlete-specific training program. and receiving a training program specific to the athlete from the adaptive algorithm.

10. The athletic training system of claim 1, Tem.

3. The mobile device receives the performance data recorded by the controller.

3. The method of claim 2, further comprising: receiving performance data from the adaptive algorithm; and communicating the performance data to the adaptive algorithm.

13. An athletic training system as described herein.

4. The adaptive algorithm aggregates the performance data into the athlete data repository.

13. The athletic training system of claim 1, wherein the at least one

5. The athletic training device of claim 1 , wherein the adaptive algorithm comprises a plurality of adaptive algorithms. Training system.

6. 10. The athletic training system of claim 1, wherein the adaptive algorithm comprises a plurality of machine learning models. Training system.

7. The multiple machine learning models are based on the biometric data of a given athlete and the performance of other athletes. An evaluation system that receives performance data as input and generates an evaluation training program. The athletic training system of claim 6 including a predictive model.

8. The plurality of machine learning models are evaluated during use of a previously generated training program. A drill that receives as input performance data recorded by the controller.

8. The athletic training system of claim 7, further comprising a routine model.

9. The drill routine model includes an athlete generated by the drill routine model. The puffs recorded by the controller during use of a specific training program.

10. The athletic training system of claim 8, further comprising: Stem.

10. 10. The method of claim 1, wherein the training unit further comprises a vertical jump station. Athletic training system.

11. 4. The training unit according to claim 1 , further comprising a horizontal agility speed station.

2. The athletic training system according to claim 1.

12. The training unit includes a vertical jump station and a horizontal agility speed station.

10. The athletic training system of claim 1, further comprising:

13. The training unit includes a plurality of vertical jump stations and a plurality of horizontal agility speed stations.

10. The athletic training system of claim 1, further comprising a training station.

14. The training unit adjusts the resistance of the adjustable device in response to the controller.

10. The athletic training system of claim 1, further comprising a motor for adjusting the level. Tem.

15. The adaptive algorithm is adapted to a given athlete's biometric and performance data. access to generate an athlete-specific training program for that athlete, 2. The athletic training system of claim 1.

16. Generate athlete-specific training programs adapted to each individual athlete a computer-implemented method comprising: A tracking group with comparable biometric and performance data for each of these athletes Athlete data repository for biometric and performance data from participating athletes and storing the The athlete is provided with an electronically adjustable athletic resistance device and a controller. Accessing a training unit, the controller accessing a plurality of access units. Athlete-specific training programs Receive grams of resistance and adjust it according to the athlete's specific training program Adjusting the resistance level of the equipment and training the athlete according to the athlete's specific training program. Configured to record performance data for athletes' use of the training unit The steps and Identification of the individual athlete and the identity of the training unit being accessed providing the identifier to an adaptive algorithm, the adaptive algorithm The types of biometric and performance data stored in the Sleet Data Repository Multiple athletes trained on and used with the training unit generating a unique training program; The adaptive algorithm accesses the athlete data repository and At least one athletics program adapted to the athlete's biometric and performance data. generating a participant-specific training program; Providing said training unit with a training program specific to said athlete Steps and The method comprising:

17. The training unit includes an interface for at least one mobile device of a user. and the step of accessing the training unit further comprises: each athlete coupling a mobile device to said training unit; The method further comprises: a training program specific to the athlete received from the adaptive algorithm; 20. The athlete specific method of claim 16, further comprising the step of communicating to a training unit. How to generate a training program for.

18. The mobile device receives the performance data recorded by the controller. and communicating said performance data to said adaptive algorithm.

20. The method of generating an athlete specific training program of claim 17, further comprising: 。

19. 17. The athlete of claim 16, wherein the adaptive algorithm comprises a plurality of machine learning models. How to generate a unique training program.

20. The adaptive algorithm uses the individual athlete's biometric and performance data to generate an athlete-specific training program for that athlete 20. The method of generating an athlete specific training program according to claim 16.

Images