A multifunctional treadmill

Abstract

The present invention belongs to the technical field of treadmills and video game entertainment equipment, and specifically relates to a multifunctional treadmill, comprising a running belt, a motor, a slope adjustment device, a front-end large screen, a human-computer interaction device, a multi-user interaction module, and a deep learning module; the front-end large screen is used to display 3D virtual reality game scenes and the virtual characters of participating users, and is used to generate user motion data; the multi-user interaction module includes a user management module, a social interaction module, a real-time multi-person synchronization module, a real-time competition module, and a network communication module that enables communication between the treadmill body and the front-end large screen and a network server; the deep learning module is used to receive and process user motion data and automatically adjust the running belt speed based on the processing results. The present invention utilizes virtual reality technology to support real-time competition and contests between multiple users. The running belt is adaptively adjusted, and the slope adjustment is self-locking, which helps ensure that users have an immersive running competition experience.

Description

Technical Field

[0001] The invention belongs to the technical field of treadmills and video game entertainment equipment, and in particular relates to a multifunctional treadmill. Background Art

[0002] The information disclosed in the background of the invention is only intended to enhance understanding of the overall background of the invention and should not necessarily be regarded as an admission or any form of suggestion that the information constitutes the prior art already known to a person skilled in the art.

[0003] Treadmills are an essential fitness tool, widely used in gyms, homes, and offices. However, traditional treadmills only provide physical exercise, and the running process is often monotonous and boring, resulting in a suboptimal user experience. To address this issue, some modern treadmills have begun to enhance the user experience by adding entertainment elements, such as playing videos or music.

[0004] For example, existing patent CN 209612090 U discloses a multifunctional shock-absorbing treadmill, comprising a support assembly, a folding plate for dancing, a control assembly for controlling the operation of the treadmill, a running belt assembly, a shock-absorbing assembly, and a fitness assembly. The support assembly includes treadmill armrests, a base frame, and a running deck. The treadmill armrests and running deck are fixed to the base frame, and the fitness assembly is fixed to the treadmill armrests. The folding plate saves space and can be opened to be used for dancing, yoga, plank support, etc. The fitness assembly provided on the treadmill armrests allows users to perform waist twisting, supine and prone exercises, and can also provide massage after the exercise, making the treadmill's exercise content more diverse and rich. However, these enhanced functions still cannot change the monotony of the running process or provide an immersive experience.

[0005] In recent years, the development of virtual reality (VR) technology has opened up new possibilities for improving the treadmill user experience. VR technology can create a three-dimensional, immersive environment in which users can interact with virtual objects. Using VR technology, users can exercise in virtual environments such as mountains, beaches, or city streets, which undoubtedly greatly enhances the user's exercise experience. However, existing VR-based treadmill systems generally lack multi-user interaction and do not provide a multi-user competitive experience.

[0006] In addition, existing treadmill systems based on VR technology usually only rely on VR helmets to provide a virtual environment. Such a design may not only cause users to feel dizzy, but the weight and heat of the VR helmet may also affect the user's running experience. Summary of the Invention

[0007] In view of the above-mentioned deficiencies in the prior art, the present invention provides a multifunctional treadmill.

[0008] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0009] A multifunctional treadmill comprises a treadmill body, wherein the treadmill body comprises a running belt, a motor, and a slope adjustment device;

[0010] It also includes a front-end large screen, human-computer interaction equipment, multi-user interaction module and deep learning module;

[0011] The front-end large screen is used to display the 3D virtual reality game scene and the virtual characters of the participating users, and to generate user movement data;

[0012] The human-computer interaction device includes VR equipment, which is wirelessly connected to the treadmill body;

[0013] The multi-user interaction module includes a user management module, a social interaction module, a real-time multi-person synchronization module, a real-time competition module, and a network communication module that enables the treadmill body and the front-end large screen to communicate with the network server;

[0014] The network server is used to manage and distribute data;

[0015] The network communication module is used for transmitting network data;

[0016] The user management module is used to manage user login, registration, and information update;

[0017] The social interaction module provides an interaction window between users;

[0018] The real-time multi-person synchronization module is used to ensure that all users' 3D virtual characters can run synchronously in real time in the same 3D virtual scene;

[0019] The real-time competition module is used to support real-time competitions between users and to manage competition rules, scores and rankings;

[0020] The deep learning module is used to receive and process user motion data and automatically adjust the treadmill speed based on the processing results.

[0021] Preferably, the deep learning module is equipped with a user motion prediction deep learning model to predict the user's running speed, cadence, and stride, thereby adaptively adjusting the treadmill speed. The user motion prediction deep learning model is as follows:

[0022] Includes input gate, forget gate and output gate;

[0023] Forget gate: f_t = σ(W_f [h_{t-1}, x_t] + b_f);

[0024] Input gate: i_t = σ(W_i [h_{t-1}, x_t] + b_i);

[0025] Output gate: o_t = σ(W_o [h_{t-1}, x_t] + b_o);

[0026] Cell state: C_t = f_t * C_{t-1} + i_t * tanh(W_C · [h_{t-1}, x_t] + b_C);

[0027] Hidden state: h_t = o_t * tanh(C_t);

[0028] Where σ represents the sigmoid function, * represents element-wise multiplication, and tanh is the hyperbolic tangent function;

[0029] Where x_t is the input at time t, h_{t-1} is the hidden state at time t-1, C_{t-1} is the cell state at time t-1, W is the weight matrix to be learned, and b is the bias parameter.

[0030] Preferably, the treadmill body further comprises a first speed sensor for measuring the speed of the running belt;

[0031] The VR equipment includes VR gloves and a VR head display, and the front large screen is a high-definition touch screen.

[0032] The VR equipment is equipped with an acceleration sensor and a second speed sensor for detecting the user's running speed.

[0033] Preferably, the training of the user motion prediction deep learning model includes the following steps:

[0034] Data Collection: First, user motion data for training needs to be collected. This includes the user's treadmill belt speed, cadence, and stride length, as well as relevant data provided by the accelerometer and secondary velocity sensor on the VR device.

[0035] Data preprocessing: The collected raw data needs to be preprocessed, including data cleaning, standardization, and normalization operations to ensure data quality and usability;

[0036] Construct a training dataset: Divide the preprocessed data into input features and target labels. The input features are the user's historical data within a certain time window, including running speed, cadence, and stride length information within the previous few seconds. The target label is the user's expected treadmill speed at the next time step.

[0037] Constructing an LSTM model: Based on the user motion prediction deep learning model, construct an LSTM model including an input gate, a forget gate, an output gate, and corresponding weight parameters and bias parameters;

[0038] Define the loss function: Choose an appropriate loss function to measure the difference between the predicted value and the true value, including mean squared error (MSE) or mean absolute error (MAE);

[0039] Model training: Use the constructed training dataset to train the LSTM model; use the backpropagation algorithm and optimizer to update the model's weights and biases and minimize the loss function, enabling the model to more accurately predict the user's running speed;

[0040] Model validation and tuning: The trained LSTM model is evaluated on the validation set. Based on the model's performance on the validation set, hyperparameter tuning or model structure adjustment is performed using the learning rate, the number of LSTM units, and the number of layers to improve the performance of the LSTM model.

[0041] Model deployment: After meeting expected performance on the validation set, the trained LSTM model was deployed to the deep learning module of the multi-function treadmill to enable real-time user motion prediction and belt speed adjustment.

[0042] Real-time prediction and adjustment: During the use of the multi-function treadmill, the deep learning module will receive the user's motion data in real time, make predictions based on the LSTM model, and then automatically adjust the running belt speed to meet the user's running needs.

[0043] Preferably, the network communication module includes a WebSocket protocol interface for connecting the treadmill body with the front large screen and the network server.

[0044] Preferably, the treadbelt is mounted on a belt roller, and the belt roller is rotatably mounted on the bottom of the frame. The slope adjustment device includes: the frame is provided with an arc-shaped mounting hole, and the central axis of the belt roller is cooperated and mounted in the arc-shaped mounting hole. An eccentric driving part is fixed to the end of the central axis, and the end of the central axis is eccentrically connected to the slope adjustment electric cylinder through the eccentric driving part. The eccentric end of the eccentric driving part is hingedly connected to the output end of the slope adjustment electric cylinder, and the mounting end of the slope adjustment electric cylinder is hingedly mounted on the frame. The eccentric driving part is provided with a self-locking part on the side surface on the upper side away from the slope adjustment electric cylinder and close to the central axis.

[0045] The beneficial effects of the present invention are as follows:

[0046] This treadmill solution combines virtual reality technology with a treadmill, displaying 3D virtual reality game scenes and the user's virtual character through a large front screen, providing users with a richer and more interesting running experience.

[0047] Through the multi-user interaction module, users can engage in social interaction, real-time multi-person synchronous running, and real-time competition; users can participate in competitions or interact with other users in virtual scenes, which increases the fun and challenge of running; such interactive functions can encourage users to compete with each other, share running results and motivate each other, thereby enhancing user participation and stickiness.

[0048] The solution's deep learning module incorporates a user motion prediction deep learning model that predicts a user's running speed, cadence, and stride length, automatically adjusting the treadmill speed based on the results. This eliminates the need for manual speed adjustments, allowing users to focus more on running, facilitates competitive running, and enhances the user experience.

[0049] The slope adjustment device can automatically adjust the slope of the treadmill according to user needs and preset rules, combined with 3D virtual reality game scenes, without affecting the tension of the running belt. Under the synergistic effect of the deep learning module to adjust the running belt speed, the loss of the running belt speed is avoided, ensuring the safety and realism of the treadmill race. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a schematic diagram of the main external structure of the present invention;

[0051] Figure 2 Schematic diagram of the main connection relationship of each module in the present invention;

[0052] Figure 3 This is a schematic diagram of the main internal structure of the slope adjustment device.

[0053] The reference numerals involved in the accompanying drawings are:

[0054] 1. Treadmill body; 11. Frame; 12. Running belt; 13. Belt roller; 14. Center axis; 15. Arc-shaped mounting hole; 16. Eccentric drive unit; 17. Self-locking unit; 18. Slope adjustment cylinder; 2. Large front screen. DETAILED DESCRIPTION

[0055] In order to enable those skilled in the art to better understand the present invention, the technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0056] like Figures 1 to 3A multifunctional treadmill is presented, comprising a treadmill body 1, wherein the treadmill body 1 comprises a running belt 12, a motor, a slope adjustment device and a control panel;

[0057] It also includes a front-end large screen 2, human-computer interaction equipment, multi-user interaction module and deep learning module;

[0058] The front-end large screen 2 is used to display the 3D virtual reality game scene and the virtual characters of the participating users, and to generate user movement data;

[0059] The human-computer interaction device includes VR equipment, which is wirelessly connected to the treadmill body 1;

[0060] The multi-user interaction module includes a user management module, a social interaction module, a real-time multi-person synchronization module, a real-time competition module, and a network communication module that enables the treadmill body 1 and the front large screen 2 to communicate with the network server;

[0061] The network server is used to manage and distribute data;

[0062] The network communication module is used for transmitting network data;

[0063] The user management module is used to manage user login, registration, and information update;

[0064] The social interaction module provides an interaction window between users;

[0065] The real-time multi-person synchronization module is used to ensure that all users' 3D virtual characters can run synchronously in real time in the same 3D virtual scene;

[0066] The real-time competition module is used to support real-time competitions between users and to manage competition rules, scores and rankings;

[0067] The deep learning module is used to receive and process user motion data and automatically adjust the speed of the treadmill 12 according to the processing results.

[0068] The deep learning module is equipped with a user motion prediction deep learning model to predict the user's running speed, cadence, and stride length, thereby adaptively adjusting the treadmill speed. The user motion prediction deep learning model is as follows:

[0069] Includes input gate, forget gate and output gate;

[0070] Forget gate: f_t = σ(W_f [h_{t-1}, x_t] + b_f);

[0071] Input gate: i_t = σ(W_i [h_{t-1}, x_t] + b_i);

[0072] Output gate: o_t = σ(W_o [h_{t-1}, x_t] + b_o);

[0073] Cell state: C_t = f_t * C_{t-1} + i_t * tanh(W_C · [h_{t-1}, x_t] + b_C);

[0074] Hidden state: h_t = o_t * tanh(C_t);

[0075] Where σ represents the sigmoid function, * represents element-wise multiplication, and tanh is the hyperbolic tangent function;

[0076] Where x_t is the input at time t, h_{t-1} is the hidden state at time t-1, C_{t-1} is the cell state at time t-1, W is the weight matrix to be learned, and b is the bias parameter.

[0077] Preferably, the treadmill body 1 further includes a first speed sensor for measuring the speed of the running belt 12;

[0078] The VR equipment includes VR gloves and a VR head display, and the front large screen 2 is a high-definition touch screen.

[0079] The VR equipment is equipped with an acceleration sensor and a second speed sensor for detecting the user's running speed.

[0080] The training of the user motion prediction deep learning model includes the following steps:

[0081] Data Collection: First, it is necessary to collect user motion data for training; the motion data includes the speed, cadence, and stride length of the treadmill belt 12 when the user is running, as well as relevant data provided by the acceleration sensor and the second speed sensor on the VR equipment;

[0082] Data preprocessing: The collected raw data needs to be preprocessed, including data cleaning, standardization, and normalization operations to ensure data quality and usability;

[0083] Construct a training dataset: Divide the preprocessed data into input features and target labels. The input features are the user's historical data within a certain time window, including the treadmill speed, cadence, and stride length information in the previous few seconds. The target label is the user's expected treadmill speed at the next time step.

[0084] Constructing an LSTM model: Based on the user motion prediction deep learning model, construct an LSTM model including an input gate, a forget gate, an output gate, and corresponding weight parameters and bias parameters;

[0085] Define the loss function: Choose an appropriate loss function to measure the difference between the predicted value and the true value, including mean squared error (MSE) or mean absolute error (MAE);

[0086] Model training: Use the constructed training dataset to train the LSTM model. Update the model's weights and biases through the backpropagation algorithm and an optimizer. The optimizer can choose SGD, which minimizes the loss function with gradient descent, so that the model can more accurately predict the user's running speed.

[0087] Model validation and tuning: The trained LSTM model is evaluated on the validation set. Based on the model's performance on the validation set, hyperparameter tuning or model structure adjustment is performed using the learning rate, the number of LSTM units, and the number of layers to improve the performance of the LSTM model.

[0088] Model deployment: After meeting expected performance on the validation set, the trained LSTM model was deployed to the deep learning module of the multi-function treadmill to enable real-time user motion prediction and 12-speed adjustment of the treadmill belt.

[0089] Real-time prediction and adjustment: During the use of the multi-function treadmill, the deep learning module will receive the user's motion data in real time, make predictions based on the LSTM model, and then automatically adjust the running belt speed to meet the user's running needs.

[0090] One of the tasks of this model is to predict the user's expected treadmill speed at the next time step. In the LSTM model, by inputting information such as the treadmill speed, cadence, and stride length of the current time step, as well as the hidden state h_{t-1} and cell state C_{t-1} of the previous time step, the model will output the hidden state h_t and cell state C_t of the current time step through the operation of the gating mechanism and weight matrix, and at the same time predict the user's treadmill speed at the next time step.

[0091] Specifically, the process of predicting the speed of the treadmill 12 at the next time step is as follows:

[0092] Input the input features x_t of the current time step (including treadmill speed, cadence, stride length, etc.) and the hidden state h_{t-1} of the previous time step into the LSTM model, and calculate the forget gate f_t, input gate i_t, output gate o_t and unit state C_t;

[0093] Using the calculated cell state C_t and output gate o_t, calculate the hidden state h_t of the current time step;

[0094] The hidden state h_t of the current time step is input into the fully connected layer or other network layers, and finally the predicted treadmill speed of the next time step is output.

[0095] When performing user motion prediction tasks, we need to define a loss function to measure the difference between the model's predictions and the true values. The choice of loss function is crucial for model training and performance; it guides the optimizer in adjusting the model's parameters to bring the predictions closer to the true values.

[0096] In this multi-function treadmill user motion prediction deep learning model, we can choose the following two commonly used loss functions: mean squared error (MSE) and mean absolute error (MAE).

[0097] Mean Squared Error (MSE):

[0098] MSE is one of the most commonly used loss functions. It measures the average quadratic error between the predicted value and the true value. For each example, the difference between the predicted value and the true value is calculated, then squared, and the average of the squared differences is taken over all examples.

[0099] The formula for MSE is as follows:

[0100] MSE = (1 / n) * Σ(y_true - y_pred)^2

[0101] Here, y_true represents the true value (target label), y_pred represents the predicted value of the model, and n is the number of samples.

[0102] Using MSE as a loss function penalizes large errors between predicted and true values, prompting the model to focus on samples with large differences. In the task of predicting user motion on a multi-function treadmill, MSE can be used to measure the model's prediction of the user's running speed.

[0103] Mean Absolute Error (MAE):

[0104] MAE is another commonly used loss function that measures the mean absolute error between the predicted value and the true value. For each example, the absolute difference between the predicted value and the true value is calculated, and then the average of the absolute differences of all examples is taken.

[0105] The formula for MAE is as follows:

[0106] MAE = (1 / n) * Σ|y_true - y_pred|

[0107] Using MAE as the loss function is less sensitive to outliers than MSE because it only considers the absolute difference between the predicted value and the true value, rather than the squared difference. Therefore, the mean square error is more effective.

[0108] The network communication module includes a WebSocket protocol interface for connecting the treadmill body 1 with the front large screen 2 and a network server.

[0109] The treadmill 12 is mounted on a belt roller 13, and the belt roller 13 is rotatably mounted on the bottom of the frame 11. The slope adjustment device includes: an arc-shaped mounting hole 15 is opened in the frame 11, and the central axis 14 of the belt roller 13 is cooperated and mounted in the arc-shaped mounting hole 15. An eccentric driving part 16 is fixed to the end of the central axis 14, and the end of the central axis 14 is eccentrically connected to the slope adjustment electric cylinder 18 through the eccentric driving part 16. The eccentric end of the eccentric driving part 16 is hingedly connected to the output end of the slope adjustment electric cylinder 18, and the mounting end of the slope adjustment electric cylinder 18 is hingedly mounted on the frame 11. The eccentric driving part 16 is provided with a self-locking part 17 on the side surface on the upper side away from the slope adjustment electric cylinder 18 and close to the central axis 14.

[0110] This treadmill solution combines virtual reality technology with a treadmill, and displays 3D virtual reality game scenes and the user's virtual character through the front large screen 2, providing users with a richer and more interesting running experience.

[0111] Through the multi-user interaction module, users can engage in social interaction, real-time multi-person synchronous running, and real-time competition; users can participate in competitions or interact with other users in virtual scenes, which increases the fun and challenge of running; such interactive functions can encourage users to compete with each other, share running results and motivate each other, thereby enhancing user participation and stickiness.

[0112] The deep learning module is equipped with a deep learning model for user motion prediction. By learning the user's motion data, including speed, cadence, and stride length, it can predict the treadmill speed for the next step in real time. This allows the treadmill to automatically adjust the speed of the treadmill based on the user's motion behavior to meet their running needs. This adaptive adjustment function enhances the treadmill's intelligence and provides users with more comfort and convenience during running.

[0113] The slope adjustment device features a curved mounting hole 15, which allows the belt roller 13 to rotate flexibly to adjust the slope. Compared to traditional fixed slope adjustment devices, the curved mounting hole 15 is more flexible, allowing users to adjust the slope to suit their needs, adapting to different training goals and intensities. Users can choose the appropriate slope for their individual needs, whether climbing or running on flat ground, adding more challenge and fun to their runs.

[0114] In summary, the present invention utilizes virtual reality technology to support real-time competition and contest among multiple users. The treadmill is adaptively adjustable, and the slope adjustment is self-locking, which helps ensure that users have an immersive running competition experience.

[0115] The incline adjustment mechanism is equipped with a self-locking portion 17 to ensure stability and safety during incline adjustment. When the user adjusts the incline, the eccentric drive unit 16 locks in place, preventing the treadmill 12 from accidentally losing speed due to an unstable incline during exercise. This design provides a better user experience and ensures safe running.

[0116] The solution also includes a multi-user interaction module and a real-time competition module, which support real-time competition and competition among users. The virtual reality game scene and the participating users' avatars are displayed on the front-end large screen 2, and users can participate through VR equipment. The deep learning module predicts and adaptively adjusts the speed of each participating user's treadmill 12 to ensure fair competition and a realistic competitive experience. This competition function provides users with more motivation and interaction, making the treadmill more interesting and attractive.

[0117] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referenced to each other.

[0118] The terms "upper," "lower," "outer," "inner," and the like, if used in the present description and claims, and in the accompanying drawings, are used to distinguish relative positions and are not necessarily qualitative. It should be understood that the terms used in this manner are interchangeable where appropriate, such that the embodiments of the present invention described herein can be implemented in sequences other than those illustrated or described herein. Furthermore, the terms "including," "having," and any variations thereof, are intended to cover non-exclusive inclusions.

[0119] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A multifunctional treadmill, comprising a treadmill body, wherein the treadmill body comprises a running belt, a motor, and a slope adjustment device; characterized in that: It also includes a front-end large screen, human-computer interaction equipment, multi-user interaction module and deep learning module; The front-end large screen is used to display the 3D virtual reality game scene and the virtual characters of the participating users, and to generate user movement data; The human-computer interaction device includes VR equipment, which is wirelessly connected to the treadmill body; The multi-user interaction module includes a user management module, a social interaction module, a real-time multi-person synchronization module, a real-time competition module, and a network communication module that enables the treadmill body and the front-end large screen to communicate with the network server; The network server is used to manage and distribute data; The network communication module is used for transmitting network data; The user management module is used to manage user login, registration, and information update; The social interaction module provides an interaction window between users; The real-time multi-person synchronization module is used to ensure that all users' 3D virtual characters can run synchronously in real time in the same 3D virtual scene; The real-time competition module is used to support real-time competitions between users and to manage competition rules, scores and rankings; The deep learning module is used to receive and process user motion data and automatically adjust the treadmill speed according to the processing results; The treadbelt is mounted on a belt roller, and the belt roller is rotatably mounted on the bottom of the frame. The slope adjustment device includes an arc-shaped mounting hole provided on the frame. The central axis of the belt roller is cooperatively mounted in the arc-shaped mounting hole. An eccentric driving part is fixed to the end of the central axis. The end of the central axis is eccentrically connected to the slope adjustment electric cylinder through the eccentric driving part. The eccentric end of the eccentric driving part is hingedly connected to the output end of the slope adjustment electric cylinder. The mounting end of the slope adjustment electric cylinder is hingedly mounted on the frame. The eccentric driving part is provided with a self-locking part on the side surface on the upper side away from the slope adjustment electric cylinder and close to the central axis.

2. The multifunctional treadmill according to claim 1, characterized in that: The deep learning module is equipped with a user motion prediction deep learning model to predict the user's running speed, cadence, and stride length, thereby adaptively adjusting the treadmill speed. The user motion prediction deep learning model is as follows: Includes input gate, forget gate and output gate; Forget gate: f_t = σ(W_f [h_{t-1}, x_t] + b_f); Input gate: i_t = σ(W_i [h_{t-1}, x_t] + b_i); Output gate: o_t = σ(W_o [h_{t-1}, x_t] + b_o); Cell state: C_t = f_t * C_{t-1} + i_t * tanh(W_C · [h_{t-1}, x_t] + b_C); Hidden state: h_t = o_t * tanh(C_t); Where σ represents the sigmoid function, * represents element-wise multiplication, and tanh is the hyperbolic tangent function; Where x_t is the input at time t, h_{t-1} is the hidden state at time t-1, C_{t-1} is the cell state at time t-1, W is the weight matrix to be learned, and b is the bias parameter.

3. The multifunctional treadmill according to claim 2, characterized in that: The treadmill body further includes a first speed sensor for measuring the speed of the running belt; The VR equipment includes VR gloves and a VR head display, and the front large screen is a high-definition touch screen; The VR equipment is equipped with an acceleration sensor and a second speed sensor for detecting the user's running speed.

4. The multifunctional treadmill according to claim 3, characterized in that: The training of the user motion prediction deep learning model includes the following steps: Data Collection: First, user motion data for training needs to be collected. This includes the user's treadmill belt speed, cadence, and stride length, as well as relevant data provided by the accelerometer and secondary velocity sensor on the VR device. Data preprocessing: The collected raw data needs to be preprocessed, including data cleaning, standardization, and normalization operations to ensure data quality and usability; Constructing a training dataset: Dividing the preprocessed data into input features and target labels; The input features are the user's historical data within a certain time window, including the running speed, cadence, and stride information in the previous few seconds; the target label is the user's expected treadmill speed in the next time step; Constructing an LSTM model: Based on the user motion prediction deep learning model, construct an LSTM model including an input gate, a forget gate, and an output gate, as well as corresponding weight parameters and bias parameters; Define the loss function: Choose an appropriate loss function to measure the difference between the predicted value and the true value, including mean squared error (MSE) or mean absolute error (MAE); Model training: Use the constructed training dataset to train the LSTM model; use the backpropagation algorithm and optimizer to update the model's weights and biases and minimize the loss function, enabling the model to more accurately predict the user's running speed; Model validation and tuning: The trained LSTM model is evaluated on the validation set. Based on the model's performance on the validation set, hyperparameter tuning or model structure adjustment is performed using the learning rate, the number of LSTM units, and the number of layers to improve the performance of the LSTM model. Model deployment: After meeting expected performance on the validation set, the trained LSTM model was deployed to the deep learning module of the multi-function treadmill to enable real-time user motion prediction and belt speed adjustment. Real-time prediction and adjustment: During the use of the multi-function treadmill, the deep learning module will receive the user's motion data in real time, make predictions based on the LSTM model, and then automatically adjust the running belt speed to meet the user's running needs.

5. The multifunctional treadmill according to claim 1, characterized in that: The network communication module includes a WebSocket protocol interface for connecting the treadmill body with the front large screen and the network server.

Images