WO2025040789A1

WO2025040789A1 - Method and system for creating an exercise program using ai

Info

Publication number: WO2025040789A1
Application number: PCT/EP2024/073690
Authority: WO
Inventors: Peter MØLLER HANSEN; Brian OVERKÆR; Anders WILLEMOES HANSEN
Original assignee: Intelligent Training Group Aps
Current assignee: Intelligent Training Group Aps
Priority date: 2023-08-24
Filing date: 2024-08-23
Publication date: 2025-02-27
Anticipated expiration: 2026-02-24

Abstract

A computer implemented method for generating and playing back user-specific training sessions on a computer-controlled training system comprising a display system, wherein said user-specific training session comprises an exercise program describing training instructions to be played back visually via the display to a user. The method comprises the steps of receiving an input data comprising exercise criteria such as training intensity or type of training defining user-specific criteria of the exercise program. Then creating a supplemented input data set by combining the input data set, historical data from a training session database comprising historical data from previous user-specific training sessions of the user, and feedback data related to the historical data. A user-specific training session is generated, wherein the generation is based on the supplemented input data set using an intelligent system. The intelligent system has been trained on a preparation dataset comprising exercise programs for multiple users. The generated training session is played back on the computer-controlled training system by displaying the training instructions on the display.

Description

Method and system for creating an exercise program using Al

Field of the invention

The present disclosure relates to a method for generating and playing back user specific training sessions on a computer-controlled training system.

Background

In recent years, the fitness industry has seen a significant shift towards personalized training programs. These programs are designed to meet the specific needs and goals of each individual, taking into account factors such as their current fitness level, desired intensity, and preferred type of exercise. However, creating such personalized programs can be a complex and time-consuming process, often requiring the expertise of a professional trainer.

Furthermore, traditional methods of creating exercise programs often fail to take into account the individual's historical exercise data. This data, which includes information such as previous exercise durations, intensities, types, and frequencies, can provide valuable insights into the individual's fitness habits and preferences. By incorporating this data into the program creation process, it is possible to create a more effective and personalized exercise program.

Further, generating training sessions and using a musical playlist as background while performing the training session has become increasingly popular in various domains, such as fitness. When making these music playlists, it is often the aim to provide organized playlists of music that align with specific workout routines. The idea is to enhance the experience and optimize performance by synchronizing the music with the desired exercise, intensity, or energy level.

Existing systems in this domain employ various algorithms and techniques to generate playlists that suit specific user needs. For instance, some systems utilize collaborative filtering methods, analyzing user preferences and behavior to recommend songs and create personalized playlists. Other systems employ music information retrieval techniques, analyzing audio features such as tempo, key, and dynamics to match songs with desired characteristics. One example of a system in the field of generating workout plans is described in the patent titled "Generating optimized workout plans using genetic and physiological data" (US2021050086). The patent outlines a method for generating personalized training sessions based on data sets including data of genetics and physiological aspect of the user.

Despite the advancements in playlist generation systems, there are still challenges and limitations that need to be addressed to improve their effectiveness and the user experience. These challenges include limited personalization, lack of contextual understanding, restricted music selection, static playlist generation, and evaluation and feedback difficulties.

Addressing these challenges is crucial for the development of more effective and usercentric systems for generating training sessions supported by musical playlists.

SUMMARY

Embodiments of the present disclosure may include a computer implemented method for generating and playing back user specific training sessions on a computer-controlled training system, the method may include the steps of receiving an input data set from the user. Embodiments may also include creating a supplemented input data set by identifying the data in the input data set and adding historical data from a training session database including historical data from previous training sessions performed on the training system and combining these data with the received input data set.

Embodiments may also include generating the user specific training session ensuring the exercise program and the music playlist match. In some embodiments, the generation may be based on the supplemented input data set using an intelligent system that may include an artificial intelligence component having instruments to learn via one or more machine learning and data analysis algorithms. Embodiments may also include playing back the user specific training session on the computer operated training system by simultaneously playing back and displaying the training instructions on the display and playing back the playlist audially on the sound system.

In some embodiments, the step of generating the user specific training session may include generating respectively the exercise program and the playlist simultaneously in matching segments until the complete exercise program may be generated according to the supplemented input data set. The exercise program may be divided into multiple segments which essential is intervals, sections, sequences, or parts of the exercise program. The terms will be used interchangeably throughout the application. A sequence may be defined by having at least one constant characteristic that is by at least one exercise parameter being constant throughout the sequence e.g. constant intensity, constant target heart rate, or constant pace, such that the exercise program may be varied by comprising a series of sequences with different characteristics.

By a program is essentially meant a training program or an exercise program.

In some embodiments, the step of generating the user specific training session may include generating the complete exercise program according to the supplemented input data set and thereafter generating a matching playlist according to the supplemented input data set.

In some embodiments, the step of generating the user specific training session may include generating the complete playlist according to the supplemented input data set and thereafter generating a matching exercise program according to the supplemented input data set. In some embodiments, the historical data in the training session database may include data from previous training sessions including previous exercise programs and playlists.

In some embodiments, the method may include receiving a feedback request. In some embodiments, the user can rate the perceived match between the exercise program and the playlist with a user specific match rating value. In some embodiments, the user specific match rating value may be saved as historical data in the training session database.

Embodiments may also include generating a user specific exercise program and music playlist that matches which involves using user specific match rating values stored from previous training sessions. Embodiments may also include playing back the training instructions on the display including presenting a graphical illustration of a virtual world. In some embodiments, the user appears to be moving through the virtual world during playback.

Embodiments may also include parameters that may be analyzed by the artificial intelligence engine to ensure a match between the playlist and the exercise program including ensuring a timed match between intensity of training instructions and intensity of music elements. Embodiments may also include parameters that may be analyzed by the artificial intelligence engine to ensure a match between the playlist and the exercise program including ensuring a match between training length and length of playlist. Embodiments may also include creating a supplemented input data set for generating the user specific training session, which may include using user specific historical data from the training session database.

Embodiments of the present disclosure may also include a computer-controlled training system for generating and playing back user specific training sessions including a display system. Embodiments may also include a sound system. In some embodiments, the user specific training session may include an exercise program describing training instructions to be played back visually via the display of the computer-controlled training system. Embodiments may also include a music playlist including a sequence of musical elements to be played back audially via the sound system of the computer-controlled training system. In some embodiments, the exercise program and the music playlist may be adapted to be played back simultaneously. In some embodiments, the computer- controlled training system may further be communicatively connected to a training device for receiving training data from the device during the training session.

Embodiments of the present disclosure may also include an input data set from the user including exercise criteria defining user specific criteria of the exercise program. In some embodiments, these criteria may include at least training intensity. Embodiments may also include number of training sequences.

Embodiments may also include training length. Embodiments may also include type of training. Embodiments may also include playlist criteria defining user specific criteria of the playlist. Embodiments may also include music element category. Embodiments may also include music element intensity.

In one embodiment, a method for generating an exercise program for a software- controlled training system includes receiving a subgroup of exercise program data from a user, supplementing the subgroup of exercise program data with historical exercise data from a database to generate a complete exercise program data set, and using artificial intelligence to generate the exercise program based on the complete exercise program data set.

In some embodiments, the software-controlled training system is a cycling trainer. In some embodiments, the software-controlled training system is a treadmill.

In some embodiments, the subgroup of exercise program data includes at least one of the following: a desired exercise duration, a desired exercise intensity, a desired exercise type, and a desired exercise frequency.

In some embodiments, the historical exercise data includes at least one of the following: previous exercise durations, previous exercise intensities, previous exercise types, and previous exercise frequencies.

In some embodiments, the artificial intelligence includes at least one of the following: a machine learning algorithm, a neural network, and a deep learning algorithm.

In some embodiments, the artificial intelligence of the intelligent system is trained on historical data including human generated and/or machine generated exercise programs and optionally on user provided feedback and/or on sensor provided feedback.

In some embodiments, the artificial intelligence of the intelligent system is trained on a preparation data set including human generated and/or machine generated exercise programs for multiple users and optionally on user provided feedback and/or on sensor provided feedback.

The intelligent system may further be trained on the accumulated data of subsequent training programs generated for the same user and feedback specific to that user, such that the artificial intelligence may learn user preferences. In some embodiments, the artificial intelligence may compare the preferences of multiple users to learn which preferences may be common to multiple users and thus improve the matching of training programs and/or playlists.

The intelligent system may further be fine-tuned, retrained or trained on the accumulated data of subsequent training programs generated for a specific user preferrably in combination with feedback specific to that user, such that the artificial intelligence may learn user preferences. In some embodiments, the artificial intelligence may compare the preferences of multiple users to learn which preferences may be common to multiple users and thus improve the generation of training programs and/or matching playlists for a first user based on preferences correlating with other users.

In another embodiment, a software-controlled training system includes one or more processors, a memory storing a database of historical exercise data, a user interface configured to receive a subgroup of exercise program data from a user, and the one or more processors configured to supplement the subgroup of exercise program data with the historical exercise data from the database to generate a complete exercise program data set, and to use artificial intelligence to generate an exercise program based on the complete exercise program data set.

In yet another embodiment, a non-transitory computer-readable medium stores instructions that, when executed by one or more processors, cause the one or more processors to perform a method for generating an exercise program for a software- controlled training system, the method including receiving a subgroup of exercise program data from a user, supplementing the subgroup of exercise program data with historical exercise data from a database to generate a complete exercise program data set, and using artificial intelligence to generate the exercise program based on the complete exercise program data set.

Throughout the application the terms user specific training session and user-specific training session will be used interchangeably as is the case with user-specific criteria and user specific criteria and with user provided feedback and user-provided feedback.

In another aspect of the invention a computer implemented method for generating and playing back user-specific training sessions on a computer-controlled training system comprising a display system. The user-specific training session comprises:

- An exercise program describing training instructions to be played back visually via the display of the computer-controlled training system to a user.

The method comprises the steps of: o Receiving an input data set from the user comprising: o Exercise criteria defining user-specific criteria of the exercise program.

The exercise criteria comprise at least one of the following criteria: ■ Training intensity.

■ Number of training sequences.

■ Training length.

■ Type of training. o Creating a supplemented input data set by combining: o The input data set. o Historical data from a training session database comprising historical data from previous user-specific training sessions of the user performed on the training system. o Feedback data related to the historical data. o Generating the user-specific training session, wherein the generation is based on the supplemented input data set using an intelligent system, wherein the intelligent system has been trained on a preparation dataset comprising exercise programs for multiple users. o Playing back said user-specific training session on said computer-controlled training system by displaying said training instructions on said display.

By a supplemented input data set is meant a data set for input to the intelligent system. The supplemented input data set comprises data of the user input and historical data specific to one or more of the user’s previous training sessions. In some preferred variants the supplemented input data may further comprise feedback data from one or more of the user’s previous training sessions. Feedback data may for example be a user evaluation of a training session and/or sensor data such as heart rate, pace, cadence, average watts or similar sensor data from the one or more previous training sessions. The supplemented input data set is provided as input to the intelligent system for the generation of a training program adapted for the user.

By sensor data is meant data obtained using a sensor such as a heart rate sensor, a watt sensor, a GPS, a magnetic sensor, an accelerometer and so on. The intelligent system is essentially an artificial intelligence (Al) trained on a preparation data set comprising data of exercise programs for multiple users. The preparation data set may comprise all programs performed on a similar training system in order to create an Al that has capabilities and knowledge within exercise programs and training. The Al may additionally be trained on data not associated with training to provide an Al with additional capabilities e.g. fluency in multiple languages for the output and input.

One advantage of the method is that exercise programs that do not fit a user based on the user feedback data may be adjusted to better fit the user in following exercises. In that way the training may be adapted to the user with the possibility of the user stating an exercise goal e.g., being able to run faster, bike faster, efficient calory burning, or cardio training for the Al to take into account when generating an exercise program. Moreover, the user does not need a coach to analyze each training session, and to create or adjust the program for next training session as this task is performed by the intelligent system and the described method. Another advantage is that users with no or little knowledge of training may be able to get an exercise program for training suited for them, e.g. by the intensity fitting their fitness level while the user does not put too much strain on themselves thereby minimizing the risk of injury. Thus, the method prevents injuries of the users and improves safety.

In some embodiments the feedback data comprise user-provided feedback data provided by said user during and/or after performing a user specific exercise session.

This ensures that more data will be added to the supplemented input data set that is provided to the Al and the intelligent system may ensure that the next exercise program is adjusted accordingly.

In some embodiments the feedback data comprise sensor data.

In some embodiments the feedback data comprise sensor data from one or more previous exercises of the user.

By using sensor data and measuring physical parameters of the user performance during exercise session, the supplemented input data set comprises precise data for user performance during that exercise according to this parameter. This information is used by the intelligent system to generate an exercise program fitting the user’s capabilities based on factual data. Furthermore, by collecting sensor data it is possible for the intelligent system to track whether the exercise sessions generated by the present method have had the desired effects on improving user performance. In a preferred variant the method comprises receiving feedback data related to multiple previous training sessions of the user whereby the progression of user performance may be accounted for in the generation of the training session. Thereby the training session may be generated adjusting the performance to a projected capability of the user based on previous progress.

In some embodiments the sensor data may be supplemented by user feedback. A benefit of combining sensor data with user-provided feedback is that the intelligent system may learn to correlate the user’s experience and preferences with the measurable data, thereby making the intelligent system capable of taking into account the user’s preferences and comfort during exercise in the generation of the user specific training session. Generating a user specific training session based on both physical performance and preference of the exercise experience increases the engagement and likelihood of continued exercise.

In some embodiments said computer-controlled training system further being communicatively connected to a training device, for receiving sensor data from said training device during said training session.

That the computer-controlled training system is connected to and can receive data from the training device on which the user is performing the user specific training session provides the possibility of the intelligent system to automatically collect data. Feedback data received from the exercise device may be provided to the intelligent system which may adjust the exercise program during the training session accordingly.

In some embodiments segments of the exercise program are adapted by the intelligent system using feedback data from the user obtained during previous segments of the exercise program.

In some embodiments one or more sequences/segments of said exercise program are adapted by the intelligent system based on feedback data obtained during one or more previous sequences of the user-specific exercise program whereby the exercise program is an adaptive exercise program. Adding data, preferably feedback data, from the ongoing exercise to the supplemented input data set provides an intelligent system capable of evaluating the user’s performance and comparing the user’s performance to the remaining segments of the exercise program. If the performance of the user during the performed segments of the user specific training sessions are not carried out in accordance with the training program, the intelligent system may adjust one or more of the remaining segments accordingly. By adjusting the exercise program based on the user performance of the current training sessions it may be avoided that too much strain or too little strain is put on the user thereby ensuring that the user specific training session is in accordance with the exercise criteria.

For example the training instructions of the exercise program may be generated by the intelligent system based on the exercise criteria to generate a number of segments of a training session having instructions meant to lead to user to exercise with a specific training intensity for each segment. If the sensor data collected by the sensors during the first segments of the training session indicates that the user does not perform with the intended intensity, the intelligent system may update one or more of the remaining segments of the training session to take into account the performance of the user during the current training session to better obtain performance results measured by the sensors being in accordance with the target of the training session.

In a variant of the method for generating a training session the intelligent system may generate a first segment, and the user may start the training session without a complete exercise program. During the first segment the intelligent system may use feedback sensor data from the first segment to generate the a second segment and/or another subsequent segment. In this way the exercise program may be generated during the training session.

In some embodiments one or more sequences of the exercise program are adapted by the intelligent system if collected feedback data of a previous sequence of the current training session deviate from a target value by more than a predetermined threshold deviation.

Having a threshold the user or the intelligent system may have a method for determining when a training session is deviating enough such that one or more of the remaining segments should be adjusted to the current performance of the user. In some embodiments the preparation dataset for training the intelligent system includes human generated and/or machine generated exercise programs.

In some embodiments the preparation dataset comprising feedback data associated with the exercise programs of the preparation dataset, the feedback data being user provided feedback data and/or sensor-provided feedback data.

Using exercise programs and corresponding feedback data as preparation data to train the intelligent system provides an intelligent system generating more precise exercise programs.

In some embodiments the preparation dataset for training the intelligent system comprising exercise programs for multiple types of training devices.

Examples of multiple types of training devices is training bicycles, treadmills and/or rowing machines. When the intelligent system is trained on preparation data comprising exercise programs for multiple types of training devices meta-knowledge such as the change in parameters relevant to all types of training, such as heart rate or wattage, may be used to improve the training sessions.

In a preferred variant the preparation dataset comprises at least 100 exercise programs for each type of training device in relation to which the intelligent system is trained. Hence, if the intelligent system is trained to understand and generate training sessions solely for bicycle training at least 100 exercise programs for cycling have been used for the training of the intelligent system. More preferably the intelligent system is trained on at least 500 exercise programs such as at least 1000 exercise programs.

In a preferred variant at least a subset of the preparation data is exercise programs related to the type of training to which the user specific training pertains. For example, if the intelligent system is to generate training sessions for users running on a treadmill at least a subset of the preparation data is exercise programs related to running on treadmills while another subset of the preparation data may be related to other forms of exercise such as trail running and track running. In some embodiments the method further comprising the computer-controlled training system controlling the training device settings for each training sequence of the exercise program, such that the training device is automatically operated to follow the exercise program.

Hereby a user is directly interacting with the intelligent system based on the feedback data the user creates. The training system is thereby controlled by the intelligent system such that the exercise program is followed and the exercise program may be adapted by the intelligent system.

In some embodiments the computer-controlled training system further comprises a sound system. The user-specific training session further comprises a music playlist comprising a sequence of musical elements to be played back audially via the sound system. The exercise program and the music playlist are adapted to be played back simultaneously and the received input data set from the user further comprises: o Playlist criteria defining user specific criteria of the playlist, wherein these criteria comprise at least one of the following criteria:

• Music element category,

• Music element intensity.

The step of generating the user specific training session ensures that the exercise program and the music playlist match and the step of playing back the user-specific training session includes simultaneously playing back and displaying the training instructions on the display and playing back the playlist on the sound system.

In another aspect of the invention a computer implemented method for generating and playing back user specific training sessions on a computer-controlled training system comprising a display system. The user specific training session comprises:

The method comprises the steps of: o Receiving an input data set from the user comprising: o Exercise criteria defining user specific criteria of the exercise program.

These criteria comprise at least one of the following criteria:

■ training intensity, number of training sequences, training length, type of training. o Creating a supplemented input data set by combining: o the input data set, o historical data from a training session database comprising historical data from previous user specific training sessions of the user performed on the training system. o Generating the user specific training session, wherein the generation is based on the supplemented input data set using an intelligent system, wherein the intelligent system has been trained on a preparation dataset comprising exercise programs for multiple users. o Playing back the user specific training session on the computer operated training system by displaying the training instructions on the display.

Having a method using historical data from previous exercises performed by the user that is generating an exercise program using the method an exercise program taking into account one or more earlier exercise programs performed by the user is obtained. Such a system can take into account a users goal for the training and provide an exercise program building on top of the previous program for the user to get closer to a goal such as running faster or longer.

The method comprises the steps of: o Receiving an input data set from the user comprising: o Exercise criteria defining user specific criteria of the exercise program. These criteria comprise at least one of the following criteria:

■ training intensity,

■ number of training sequences, ■ training length,

■ type of training. o Creating a supplemented input data set by combining: o the input data set, o historical data from a training session database comprising historical data from previous training sessions performed on the training system of other users having similar characteristics to said user. o Generating the user specific training session, wherein the generation is based on the supplemented input data set using an intelligent system, wherein the intelligent system has been trained on a preparation dataset comprising exercise programs for multiple users. o Playing back the user specific training session on the computer operated training system by displaying the training instructions on the display.

In embodiments the user may provide a user ID to the computer controlled training system in order to store feedback data from the user such that it is coupled to the user. The User ID may comprise characteristics of the user such as average pace, average training length, BMI, fitness rating and so on. This information may be given as input to the intelligent system such that the intelligent system can take that into account when generating an exercise program for the user. Additional historical data may be provided as input and feedback related to the historical data as described by other aspects of the invention.

Using such a system the intelligent system can provide a customized exercise program to a user having performed no training on the training system.

Detailed description of embodiments

In the following specific examples according to aspects of the present disclosure will be explained in more detail with reference to the accompanying drawings. The present disclosure may, however, be embodied in different forms than depicted below, and should not be construed as limited to any examples set forth herein. Rather, any examples are provided so that the disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Like reference numerals refer to like elements throughout. Like elements will, thus, not be described in detail with respect to the description of each figure.

Summary of figures

FIG. 1 is a flowchart illustrating a method for generating and playing back user specific training sessions, according to some embodiments of the present disclosure.

FIG. 2 illustrates a system for generating and playing back a training session with a playlist and an exercise program , according to some embodiments of the present disclosure.

FIG. 3 is a block diagram illustrating a computer-controlled training system, according to some embodiments of the present disclosure.

FIG. 4 is a block diagram illustrating an input data set, according to some embodiments of the present disclosure.

Detailed description of the invention

FIG. 1 is a flowchart that illustrates a method for generating and playing back user specific training sessions and FIG. 2 illustrates the system performing this method according to some embodiments of the present disclosure. In some embodiments, at the step 110, the method may include receiving an input data set from the user. At the step 120, the method may include creating a supplemented input data set by identifying the data in the input data set and adding historical data from a training session database comprising historical data from previous training sessions performed on the training system and combining these data with the received input data set. The supplemented data set may be based on other data like other users training sessions, other user specific performances while a specific music piece is played, and/or the performance of the specific user during specific music.

The input data from the user may be data comprising information about exercise criteria and information about playlist criteria. The information about exercise criteria may describe one or more of the following characteristics: the length of the training session, what exercises should be included, the intensity of one or more exercises or part of exercises, the intensity of the whole training session, number of training sequences, type of training, cadence of training, tempo of training, and/or training discipline.

The information about the playlist may describe one or more of the following characteristics: music element category, music element intensity, music element genre, music element tempo, music element rhythm, music element cadence, music element length in time, music element instrumentation, music element mood/emotion, music element melody, music element harmony, music element form, music element dynamics, and/or music element lyrics. A music element may be a section of a music piece, a music piece or a collection of music pieces.

In some embodiments, at step 130, the method may include generating the user specific training session ensuring the exercise program and the music playlist match. By matching an exercise program and a music playlist is meant that one or more characteristics of the training session and music playlist match. That may as examples be the length of the training session and music playlist, the intensity, the tempo and so on. Characteristics like intensity or dynamics may vary throughout a music piece, and for such characteristics like intensity and dynamics the music playlist and training session may be denoted to match if the characteristics are similar or close to each other. This may be a logical consequence of music having specific characteristics and that an exercise program may not fit these characteristics completely and no music may be found having these characteristics. Thus, in the step 130 of generating the user specific training session, the best possible match of the training session program and the playlist music may be generated based on the received inputted data set and supplemented data set as the perfect match may not exist.

In the process of generating the user specific training session, an intelligent system is used. The intelligent system comprises an artificial intelligence component having instruments to learn via one or more machine learning and data analysis algorithms. The intelligent system may be stored on a memory and processed by a processor. The artificial component of the intelligent system may be considered an intelligent engine. The machine learning and data analysis algorithms may use freely available data from training sessions or exercise programs.

The machine learning and data analysis algorithms may use freely available data from training sessions or exercise programs. The machine learning and data analysis algorithms may use freely available data from training sessions or exercise programs such as exercise programs published online. This data may be comprised in the preparation data set. The preparation data may further comprise proprietary data such as exercise programs of professional training coaches.

The preparation data set may be specific for each training type such as biking and running. Thus, one intelligent system may be trained to provide biking programs and another system to provide running programs. Data including exercise programs from multiple sport may be used as data in the preparation data set to get a more complete intelligent system. For example, data relating to intensity measured by sensors during a training session may be correlated with user preferences based on user feedback across various types of exercise to teach the intelligent system to adjust the exercise program to the specific user. As another example, the intelligent system may learn overall correlations between various types of exercises and performance based on all user data such that the intelligent system may generate a training session for one type of training device, e.g. a bicycle, for a user related to whom feedback data only exists for a different type of training device, e.g. a treadmill.

The preparation data may furthermore comprise data of user feedback regarding the performed exercise program if an exercise program was completed. User feedback may include ratings from the users, sensor data recorded during performance of the exercise program and/or number of times a program has been performed. User ratings may comprise ratings being an overall score of the exercise program, or ratings of specific parameters of the exercise program such as the intensity of the exercise program, the length of the exercise program, or the difficulty of the exercise program, and/or it may be feedback on whether the exercise program was as the user expected or not.

The data may include music playlists and music pieces and analysis of the music. Likewise, the data may include analysis of the training sessions and exercises stored in the data. All described data may be part of the supplemented data set provided to the intelligent system. The data used for training the intelligent system may be stored in database.

The preparation data set may be freely available or obtained by user agreements. After training of the intelligent system an input to the intelligent system is needed to create a training program as output. The input data may vary depending on which exercise program is desired. The exercise criteria may include any of the following: the length of the training session,

- what exercises should be included, the intensity of one or more exercises or parts of exercises, the intensity of the whole training session, number of training sequences, type of training, cadence of training, tempo of training, training discipline.

The input may comprise feedback data from earlier exercise sessions and exercise programs performed by the user requesting the new exercise program. Using user feedback from earlier training sessions enables the intelligent system to generate an exercise program which is progressive and healthy for the user without the need of a personal trainer to create the training program.

The feedback data may comprise sensor data from at least one of the user’s previous training sessions based on an exercise program. This sensor data may e.g. be data from a watt meter, speed from a watch or a bike computer, heartrate and/or rotations per minute of the pedals. The data may be collected and processed as a function of time or as an average from the training session. Thus the intelligent system can use the feedback sensor data to adjust the generation of a new training program such that it is progressive and the user will still be able to complete the program.

The intelligent system may let the data obtained from the user outweigh data of other users saved in the data set or database. In one embodiment, the system may collect data from the user during a training session and use this in the generation of other training sessions. This may be data from sensors measuring the intensity, cadence, pulse and so on. Thus, the system may learn how a user reacts to specific music or exercises and may use that to generate a training session fitting the user’s desire as much as possible.

In some embodiments, the intelligent system is provided with sensor input whereby training programs may subsequently be adapted based on measured user feedback. Measured user feedback may be more accurate than self-reported user feedback, thereby providing more efficient training data for the intelligent system. In some embodiments, collected sensor data may be part of the historical data from a training session database used in creating the supplemented input data set. In other embodiments, the supplemented input dataset may be dynamically updated based on the sensor input whereby the training program may be dynamically generated. By the training program being dynamically generated is understood that segments of the current training program following the currently executed segment are generated during execution of the current segment of the training program. This allows for further customization of the training program to the performance.

By the training program being dynamically generated is understood that segments of the current training program following the currently executed segment are generated during execution of the current segment of the training program. This allows for further customization of the training program to the performance of the user. Hereby the user interacts with the intelligent system during the training session in order to adapt the exercise program if needed such that the user/athlete gets a suitable program that is progressive and achievable. A users performance may vary from day to day and therefore the ability of the intelligent system to adapt to a user’s daily condition by using data feedback from the user while training will increase the value of the training for the user and the outcome in order for the user/athlete to progress and become faster, burn calories, or building endurance.

In embodiments wherein a playlist is created to match the training program, and the training program is created dynamically, matching segments of the training program and the playlist may be generated simultaneously.

In some embodiments, the intelligent system may compare sensor data with user- provided inputs, thereby allowing the intelligent system to learn user specific preferences for the exercise conditions. In some embodiments, the intelligent system may compare sensor data with the user- provided match rating value, thereby allowing the intelligent system to learn user specific preferences for the exercise conditions.

In some embodiments, the intelligent system may be taught to let the historical feedback data in the form of match rating values and/or sensor data outweigh the user input such that the training program and/or playlist is based on evaluated preference rather than the user’s own expectations of his preference.

At step 140, the method may include playing back the user specific training session on the computer operated training system by simultaneously playing back and displaying the training instructions on the display and playing back the playlist audially on the sound system.

In some embodiments the intelligent system may adapt or generate one or more segments of the exercise program based on the feedback data being sensor data received during previous segments of the same training session.

Furthermore, the intelligent system may control the exercise system, such as the home trainer or treadmill, by adjusting the system parameters, e.g. the resistance, speed, or inclination, according to changes in the exercise program. For example if the user was unable to keep the heart rate under a certain threshold/target during a first interval of the training session the speed or resistance of the next interval may be reduced to not stress the user’s body too much. In such a configuration the intelligent system may adjust one or more physical system parameters of the exercise device without the need for the user to do so.

For the steps 110 to 140 the generation may be based on the supplemented input data set using an intelligent system that comprises an artificial intelligence component having instruments to learn via one or more machine learning and data analysis algorithms.

In some embodiments, the step of generating the user specific training session comprises generating respectively the exercise program and the playlist simultaneously in matching segments until the complete exercise program may be generated according to the supplemented input data set. In some embodiments, the step of generating the user specific training session comprises generating the complete exercise program according to the supplemented input data set and thereafter generating a matching playlist according to the supplemented input data set. In some embodiments, the step of generating the user specific training session comprises generating the complete playlist according to the supplemented input data set and thereafter generating a matching exercise program according to the supplemented input data set. In some embodiments, the historical data in the training session database may comprise data from previous training sessions including previous exercise programs and playlists.

In some embodiments, the user can rate the perceived match between the exercise program and the playlist with a user specific match rating value. The user specific match rating value may be saved as historical data in the training session database. In some embodiments, generating a user specific exercise program and music playlist that matches which involves using user specific match rating values stored from previous training sessions.

In one embodiment, the intelligent system may generate the user specific training session by adding some pseudo-randomness to the training session. How much of pseudo-randomness is added to the playlist and training instructions may be related to the uncertainty of the intelligent system. The pseudo-randomness may be static. The randomness added to the training session may change the parameters of part of the exercise like the intensity, the tempo, dynamics, repetitions and the like. Adding this pseudo-randomness or uncertainty may test how the user reacts to music and exercises similar to what the user provides in the user input data set. Thus, the intelligent system may learn from these variations to be able to generate a training session the user rates even better.

In embodiments where match rating values are provided, the historical data of the rated sessions in combination with the match rating values may be used as training data for the intelligent system, such that the intelligent system is improved. The training may be user specific or may be global such that it may improve the matching of the generated training programs for multiple users.

In some embodiments, playing back the training instructions on the display may include presenting a graphical illustration of a virtual world. The user may appear to be moving through the virtual world during playback. In some embodiments, for the artificial intelligence engine to ensure a match between the playlist and the exercise program may include ensuring a timed match between intensity of training instructions and intensity of music elements. In some embodiments, the artificial intelligence engine may ensure a match between the playlist and the exercise program by ensuring a match between training length and length of playlist. In some embodiments, creating a supplemented input data set for generating the user specific training session comprises using user specific historical data from the training session database. In some embodiments, the intelligent system may include some but not all musical elements of a song stored in a music database to improve the match between the exercise program and the playlist. For example, a chorus may be added or removed from a known song in the playlist to better match the length of the exercise program or the intensity of subsequent music elements.

A system for performing a method of generating a training session with an exercise program and a playlist is illustrated in FIG. 2. A data set with inputs from a user is provided to an intelligent system. The intelligent system may comprise a processor 200 capable of running machine learning and data analysis algorithms. After a training session has been generated, the display system 212 and sound system 214 may play back the training session. The playlist and the training instructions are preferably played back simultaneously.

FIG. 3 is a block diagram that describes a computer-controlled training system 210, according to some embodiments of the present disclosure. In some embodiments, the computer-controlled training system 210 may include a display system 212 and a sound system 214. The user specific training session 220 may include an exercise program 222 describing training instructions to be played back visually via the display of the computer-controlled training system 210 and a music playlist 224. The music playlist 224 may include a sequence 226 of musical element to be played back audially via the sound system 214 of the computer-controlled training system 210. The exercise program 222 and the music playlist 224 may be adapted to be played back simultaneously. In some embodiments, the computer-controlled training system 210 may further be communicatively connected to a training device, for receiving training data from the device during the training session 220. The computer-controlled training system 210 may comprise the computer from which the system is controlled. The computer-controlled training system may further comprise a sensor to measure one or more parameters of the training and provide this to the supplemented data set used for generating a training session.

FIG. 3 is a block diagram that describes a computer-controlled training system 210, according to some embodiments of the present disclosure. In some embodiments, the computer-controlled training system 210 may include a display system 212 and a sound system 214. The user specific training session 220 may include an exercise program 222 describing training instructions to be played back visually via the display of the computer-controlled training system 210 and a music playlist 224. The music playlist 224 may include a sequence 226 of musical element to be played back audially via the sound system 214 of the computer-controlled training system 210. The exercise program 222 and the music playlist 224 may be adapted to be played back simultaneously. In some embodiments, the computer-controlled training system 210 may further be communicatively connected to a training device, for receiving sensor data from the device during the training session 220. The computer-controlled training system 210 may comprise the computer from which the system is controlled. The computer-controlled training system may further comprise a sensor to measure one or more parameters of the training and provide this to the supplemented data set used for generating a training session.

FIG. 4 is a block diagram that describes an input data set 300, according to some embodiments of the present disclosure. In some embodiments, the input data set 300 may include exercise criteria 310 defining user specific criteria of the exercise program, training intensity 320, number of training sequences 330, training length 340, type of training 350, playlist criteria 360 defining user specific criteria of the playlist, music element category 370, and music element intensity 380.

The above embodiments have described both an exercise program and a musical playlist being generated using Al. In a similar manner, the tool could be used only for generating the exercise program by an input data set describing criteria to the exercise program in 110 and then supplemented this with historical exercise program data in 120. In 130 just the exercise program will then be generated using Al ensuring that the exercise program fulfills the requests of the user.

In a system of this specific embodiment, the computer-controlled training system does not need a sound system, further, the user specific training session does not need a musical playlist and the input data does not include playlist criteria, music element category and music element intensity.

Now follows a set of items, which constitute aspects of the present disclosure which may be considered independently patentable and as such the following sets form basis for possible future sets of claims: 1. A computer implemented method for generating and playing back user specific training sessions on a computer-controlled training system comprising a display system, wherein said user specific training session comprises: an exercise program describing training instructions to be played back visually via said display of said computer-controlled training system, said method comprises the steps of: o receiving an input data set from the user comprising o exercise criteria defining user specific criteria of the exercise program, wherein these criteria comprise at least one of the following criteria:

■ training intensity,

■ number of training sequences,

■ training length,

■ type of training, o creating a supplemented input data set by identifying the data in said input data set and adding historical data from a training session database comprising historical data from previous training sessions performed on said training system and combining these data with the received input data set, o generating said user specific training session, wherein the generation is based on said supplemented input data set using an intelligent system that comprises an artificial intelligence component having instruments to learn via one or more machine learning and data analysis algorithms, o playing back said user specific training session on said computer operated training system by displaying said training instructions on said display.

2. A method according to item 1 , wherein the computer-controlled training system further comprises a sound system and wherein said user specific training session further comprises a music playlist comprising a sequence of musical elements to be played back audially via said sound system, wherein the exercise program and the music playlist are adapted to be played back simultaneously and the received input data set from the user further comprises: o playlist criteria’s defining user specific criteria of the playlist, wherein these criteria comprise at least one of the following criteria:

• music element category,

• music element intensity, wherein the step of generating said user specific training session ensures that the exercise program and the music playlist matches and the step of playing back the user specific training session includes simultaneously playing back and displaying said training instructions on said display and playing back said playlist on said sound system.

3. A method according to item 2, wherein the step of generating said user specific training session comprises generating respectively the exercise program and the playlist simultaneously in matching segments until the complete exercise program is generated according to the supplemented input data set.

4. A method according to items 2-3, wherein the step of generating said user specific training session comprises generating the complete exercise program according to the supplemented input data set and thereafter generating a matching playlist according to the supplemented input data set.

5. A method according to items 2-4, wherein the step of generating said user specific training session comprises generating the complete playlist according to the supplemented input data set and thereafter generating a matching exercise program according to the supplemented input data set.

6. A method according to items 2-5, wherein said historical data in said training session database comprises data from previous training sessions including previous exercise programs and playlists.

7. A method according to items 2-6, wherein the method further comprises a feedback request, wherein the user can rate the perceived match between the exercise program and the playlist with a user specific match rating value and wherein said user specific match rating value is saved as historical data in the training session database. 8. A method according to items 2-7, wherein generating a user specific exercise program and music playlist that matches which involves using user specific match rating values stored from previous training sessions.

9. A method according to items 1-8, wherein said computer-controlled training system further is communicatively connected to a training device, for receiving training data from said device during the training session.

10. A method according to items 1-9, wherein playing back said training instructions on said display includes presenting a graphical illustration of a virtual world, wherein the user appears to be moving through said virtual world during playback.

11. A method according to items 2-10, wherein parameters that are being analyzed by the artificial intelligence engine to ensure a match between the playlist and the exercise program including ensuring a timed match between intensity of training instructions and intensity of music elements.

12. A method according to item 11 , wherein parameters that are being analyzed by the artificial intelligence engine to ensure a match between the playlist and the exercise program including ensuring a match between training length and length of playlist.

13. A method according to items 1-12, wherein creating a supplemented input data set for generating the user specific training session comprises using user specific historical data from said training session database.

14. A method according to items 1-13, wherein the intelligent system is trained on historical data including human generated and/or machine generated exercise programs and optionally on user provided feedback and/or on sensor provided feedback.

15. A computer-controlled training system comprising a display system and a sound system adapted to perform a method according to items 1-14.

16. A computer implemented method for generating and playing back user specific training sessions on a computer-controlled training system comprising a display system, wherein said user specific training session comprises: an exercise program describing training instructions to be played back visually via said display of said computer-controlled training system to a user, said method comprises the steps of: o receiving an input data set from said user comprising: o exercise criteria defining user specific criteria of the exercise program, wherein these criteria comprise at least one of the following criteria:

■ training intensity,

■ number of training sequences,

■ training length,

■ type of training, o creating a supplemented input data set by combining: o said input data set, o historical data from a training session database comprising historical data from previous user specific training sessions of said user performed on said training system, o generating said user specific training session, wherein the generation is based on said supplemented input data set using an intelligent system, wherein said intelligent system has been trained on a preparation dataset comprising exercise programs for multiple users, o playing back said user specific training session on said computer operated training system by displaying said training instructions on said display.

Claims

1. A computer implemented method for generating and playing back user-specific training sessions on a computer-controlled training system comprising a display system, wherein said user-specific training session comprises: an exercise program describing training instructions to be played back visually via said display of said computer-controlled training system to a user, said method comprises the steps of: o receiving an input data set from said user comprising: o exercise criteria defining user-specific criteria of the exercise program, wherein these criteria comprise at least one of the following criteria:

■ training intensity,

■ number of training sequences,

■ training length,

■ type of training, o creating a supplemented input data set by combining: o said input data set, o historical data from a training session database comprising historical data from previous user-specific training sessions of said user performed on said training system, o feedback data related to said historical data, o generating said user-specific training session, wherein the generation is based on said supplemented input data set using an intelligent system, wherein said intelligent system has been trained on a preparation dataset comprising exercise programs for multiple users, o playing back said user-specific training session on said computer-controlled training system by displaying said training instructions on said display.

2. A method according to claim 1 , wherein said feedback data comprise user-provided feedback data provided by said user during and/or after performing a user- specific exercise session.

3. A method according to any of the preceding claims, wherein said feedback data comprise sensor data.

4. A method according to any of the preceding claims, said computer-controlled training system further being communicatively connected to a training device, for receiving sensor data from said training device during said training session.

5. A method according to any of the preceding claims, wherein one or more sequences of said exercise program are adapted by said intelligent system based on feedback data obtained during one or more previous sequences of said user-specific exercise program whereby said exercise program is an adaptive exercise program.

6. A method according to claim 5, wherein one or more sequences of said exercise program are adapted by said intelligent system if collected feedback data of a previous sequence of the current training session deviate from a target value by more than a predetermined threshold deviation.

7. A method according to any of the preceding claims, said preparation dataset for training said intelligent system including human-generated and/or machine-generated exercise programs.

8. A method according to any of the preceding claims, said preparation dataset comprising feedback data associated with said exercise programs of said preparation dataset, said feedback data being user-provided feedback data and/or sensor-provided feedback data.

9. A method according to any of the preceding claims, said preparation dataset for training said intelligent system comprising exercise programs for multiple types of training devices.

10 A method according to any of the preceding claims, said method further comprising said computer-controlled training system controlling said training device settings for each training sequence of said exercise program, such that said training device is automatically operated to follow said exercise program.

11. A method according to any of the preceding claims, wherein the computer-controlled training system further comprises a sound system and wherein said user-specific training session further comprises a music playlist comprising a sequence of musical elements to be played back audially via said sound system, wherein the exercise program and the music playlist are adapted to be played back simultaneously and the received input data set from the user further comprises: o playlist criteria defining user-specific criteria of the playlist, wherein these criteria comprise at least one of the following criteria:

• music element category,

• music element intensity, wherein the step of generating said user-specific training session ensures that the exercise program and the music playlist match and the step of playing back the user-specific training session includes simultaneously playing back and displaying said training instructions on said display and playing back said playlist on said sound system.

12. A method according to claim 11 , wherein said historical data in said training session database comprise data from previous training sessions including previous exercise programs and playlists.

13. A method according to claims 11-12, wherein the method further comprises a feedback request, wherein the user can rate the perceived match between the exercise program and the playlist with a user-specific match rating value and wherein said user-specific match rating value is saved as historical data in the training session database.

14. A method according to any of the preceding claims, wherein playing back said training instructions on said display includes presenting a graphical illustration of a virtual world, wherein the user appears to be moving through said virtual world during playback.

15. A computer-controlled training system comprising a display system and a sound system adapted to perform a method according to any of the previous claims.