CN119015671B - Underwater running platform intelligent speed regulation system and method based on deep learning - Google Patents

Underwater running platform intelligent speed regulation system and method based on deep learning Download PDF

Info

Publication number
CN119015671B
CN119015671B CN202410906937.9A CN202410906937A CN119015671B CN 119015671 B CN119015671 B CN 119015671B CN 202410906937 A CN202410906937 A CN 202410906937A CN 119015671 B CN119015671 B CN 119015671B
Authority
CN
China
Prior art keywords
heart rate
running platform
underwater running
layer
user
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202410906937.9A
Other languages
Chinese (zh)
Other versions
CN119015671A (en
Inventor
陈景裕
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangzhou J&j Sanitary Ware Co ltd
Original Assignee
Guangzhou J&j Sanitary Ware Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangzhou J&j Sanitary Ware Co ltd filed Critical Guangzhou J&j Sanitary Ware Co ltd
Priority to CN202410906937.9A priority Critical patent/CN119015671B/en
Publication of CN119015671A publication Critical patent/CN119015671A/en
Application granted granted Critical
Publication of CN119015671B publication Critical patent/CN119015671B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • A—HUMAN NECESSITIES
    • A63—SPORTS; GAMES; AMUSEMENTS
    • A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
    • A63B22/00—Exercising apparatus specially adapted for conditioning the cardio-vascular system, for training agility or co-ordination of movements
    • A63B22/02—Exercising apparatus specially adapted for conditioning the cardio-vascular system, for training agility or co-ordination of movements with movable endless bands, e.g. treadmills
    • A—HUMAN NECESSITIES
    • A63—SPORTS; GAMES; AMUSEMENTS
    • A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
    • A63B21/00—Exercising apparatus for developing or strengthening the muscles or joints of the body by working against a counterforce, with or without measuring devices
    • A63B21/008—Exercising apparatus for developing or strengthening the muscles or joints of the body by working against a counterforce, with or without measuring devices using hydraulic or pneumatic force-resisters
    • A63B21/0084—Exercising apparatus for developing or strengthening the muscles or joints of the body by working against a counterforce, with or without measuring devices using hydraulic or pneumatic force-resisters by moving the surrounding water
    • A—HUMAN NECESSITIES
    • A63—SPORTS; GAMES; AMUSEMENTS
    • A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
    • A63B24/00—Electric or electronic controls for exercising apparatus of preceding groups; Controlling or monitoring of exercises, sportive games, training or athletic performances
    • A63B24/0062—Monitoring athletic performances, e.g. for determining the work of a user on an exercise apparatus, the completed jogging or cycling distance
    • A—HUMAN NECESSITIES
    • A63—SPORTS; GAMES; AMUSEMENTS
    • A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
    • A63B24/00—Electric or electronic controls for exercising apparatus of preceding groups; Controlling or monitoring of exercises, sportive games, training or athletic performances
    • A63B24/0087—Electric or electronic controls for exercising apparatus of groups A63B21/00 - A63B23/00, e.g. controlling load
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • A—HUMAN NECESSITIES
    • A63—SPORTS; GAMES; AMUSEMENTS
    • A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
    • A63B2225/00—Miscellaneous features of sport apparatus, devices or equipment
    • A63B2225/60—Apparatus used in water
    • A—HUMAN NECESSITIES
    • A63—SPORTS; GAMES; AMUSEMENTS
    • A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
    • A63B2230/00—Measuring physiological parameters of the user
    • A63B2230/04—Measuring physiological parameters of the user heartbeat characteristics, e.g. ECG, blood pressure modulations
    • A63B2230/06—Measuring physiological parameters of the user heartbeat characteristics, e.g. ECG, blood pressure modulations heartbeat rate only
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/30—Assessment of water resources

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Mathematical Physics (AREA)
  • General Physics & Mathematics (AREA)
  • Physical Education & Sports Medicine (AREA)
  • Data Mining & Analysis (AREA)
  • Biophysics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computational Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • Orthopedic Medicine & Surgery (AREA)
  • Algebra (AREA)
  • Cardiology (AREA)
  • Vascular Medicine (AREA)
  • Image Analysis (AREA)

Abstract

本发明公开了一种基于深度学习的水下跑台智能调速系统及方法，所述方法包括：获取用户的心率和图像；使用图像处理算法对所述图像进行运动姿态分析，得到所述用户的运动姿态；将所述心率和所述运动姿态作为输入数据，输入到训练好的带注意力机制的多层感知机神经网络模型中，得到所述用户在水下跑台上的运动状态；将所述运动状态输入到训练好的水下跑台调速模型中，得到所述用户当前在水下跑台上的最佳训练水流速度；控制水下跑台当前水流速度调整至所述最佳训练水流速度。本发明能够根据运动员的实时生理状态和运动表现，动态调整水流速度，有效提升了水下运动的训练效果。

The present invention discloses an underwater treadmill intelligent speed regulation system and method based on deep learning, the method comprising: obtaining the heart rate and image of the user; using an image processing algorithm to perform motion posture analysis on the image to obtain the motion posture of the user; using the heart rate and the motion posture as input data, inputting them into a trained multi-layer perceptron neural network model with an attention mechanism to obtain the motion state of the user on the underwater treadmill; inputting the motion state into a trained underwater treadmill speed regulation model to obtain the optimal training water flow speed of the user on the underwater treadmill; and controlling the current water flow speed of the underwater treadmill to adjust to the optimal training water flow speed. The present invention can dynamically adjust the water flow speed according to the real-time physiological state and sports performance of the athlete, effectively improving the training effect of underwater sports.

Description

Underwater running platform intelligent speed regulation system and method based on deep learning
Technical Field
The invention relates to the technical field of data processing, in particular to an intelligent speed regulation system and method for an underwater running platform based on deep learning.
Background
In the field of modern body building and rehabilitation, the underwater running platform becomes a preferred tool for athlete training and rehabilitation due to the characteristics of low impact and high efficiency. However, the prior art has obvious limitations in realizing intelligent speed regulation of the underwater running platform, and is characterized in that most of the underwater running platforms adopt a preset training mode, and cannot respond to physiological state changes and athletic performance requirements of athletes in real time by manually regulating the water flow speed, so that individuation and intellectualization are lacking.
Therefore, the existing intelligent speed regulating system of the underwater running platform generally lacks the capabilities of real-time monitoring, depth analysis and self-adaptive speed regulation, so that the training efficiency and safety are limited, the requirements of athletes on personalized and intelligent training cannot be met, and the training effect of underwater sports cannot be effectively improved.
Disclosure of Invention
The embodiment of the invention provides an intelligent speed regulating system and method for an underwater running platform based on deep learning, which can dynamically regulate the water flow speed according to the real-time physiological state and the athletic performance of athletes, and effectively improve the training effect of underwater sports.
The embodiment of the invention provides an intelligent speed regulation system of an underwater running platform based on deep learning, which comprises the following components:
the heart rate detection device is used for detecting the heart rate of a user moving on the underwater running platform;
the camera is used for shooting the image of the user;
The control device is in communication connection with the heart rate detection device and the camera and is used for:
acquiring the heart rate and the image;
Performing motion gesture analysis on the image by using an image processing algorithm to obtain the motion gesture of the user;
Taking the heart rate and the motion gesture as input data, and inputting the input data into a trained multi-layer perceptron neural network model with an attention mechanism to obtain the motion state of the user on an underwater running platform;
Inputting the motion state into a trained underwater running platform speed regulation model to obtain the optimal training water flow speed of the user on the underwater running platform;
and controlling the current water flow speed of the underwater running platform to be adjusted to the optimal training water flow speed.
As an improvement of the above solution, the inputting the heart rate and the motion gesture as input data into a trained multi-layer perceptron neural network model with an attention mechanism to obtain a motion state of the user on an underwater running platform includes:
extracting the characteristics of the heart rate and the motion gesture to obtain heart rate characteristics and motion gesture characteristics;
performing attention weight calculation and feature stitching on the heart rate features and the motion gesture features to obtain attention weighted fusion features;
and inputting the fusion characteristics into a built multi-layer perceptron neural network model with an attention mechanism, and outputting the motion state of the user on the underwater running platform through multi-layer nonlinear transformation.
As an improvement of the above solution, the calculating the attention weight of the heart rate feature and the motion gesture feature and performing feature stitching to obtain an attention weighted fusion feature includes:
converting heart rate feature vector O HR and motion gesture feature vector O STA into key vectors and value vectors using linear transformations;
The attention weight w i of the heart rate feature vector O HR and the motion-posture feature vector O STA is calculated by the following formula:
Wherein Q k,Ki,Vi represents a query vector, a key vector, and a value vector, respectively, from linear transformations of heart rate feature vector O HR and motion gesture feature vector O STA, respectively, d k is the dimension of the key vector for scaling the dot product result to avoid gradient vanishing;
feature stitching is performed through the following formula to obtain a weighted attention feature vector v:
n is the number of linearly transformed value vectors V i of the heart rate feature vector O HR and the motion gesture feature vector O STA.
As an improvement of the above scheme, the multi-layer perceptron neural network model with the attention mechanism is composed of a plurality of hidden layers, and the calculation formula of each hidden layer is as follows:
hl＝f(Wl·v+bl),l=1,...,L
Wherein h l-1 is the output of the previous layer, W l and b l are the weight matrix and bias vector of the first layer, f is a nonlinear activation function, the output of the last layer L is used as the input of the final layer h L, and the calculation formula of the final layer h L is:
S=g (W S·hL+bS) where W S and b S are the weight matrix and bias vector of the final layer h L, g
And S is the motion state of the user on the underwater running platform.
As an improvement of the scheme, the model formula of the underwater running platform speed regulation model is as follows:
Wherein V opt is the output optimal training water flow speed of the user on the underwater running platform at present; The method is a deep learning model, the parameters are Θ, the parameters comprise a weight matrix W h、Wf and a bias vector b h、bf, sigma (·) is an activation function of an output layer, and tan h (·) is an activation function of a hidden layer, and the method is used for introducing nonlinear transformation and enhancing the expression capability of the model.
Another embodiment of the present invention correspondingly provides an intelligent speed regulation method for an underwater running platform based on deep learning, which is applied to the intelligent speed regulation system for an underwater running platform based on deep learning as described above, and includes:
acquiring the heart rate and the image;
Performing motion gesture analysis on the image by using an image processing algorithm to obtain the motion gesture of the user;
Taking the heart rate and the motion gesture as input data, and inputting the input data into a trained multi-layer perceptron neural network model with an attention mechanism to obtain the motion state of the user on an underwater running platform;
Inputting the motion state into a trained underwater running platform speed regulation model to obtain the optimal training water flow speed of the user on the underwater running platform;
and controlling the current water flow speed of the underwater running platform to be adjusted to the optimal training water flow speed.
As an improvement of the above solution, the inputting the heart rate and the motion gesture as input data into a trained multi-layer perceptron neural network model with an attention mechanism to obtain a motion state of the user on an underwater running platform includes:
extracting the characteristics of the heart rate and the motion gesture to obtain heart rate characteristics and motion gesture characteristics;
performing attention weight calculation and feature stitching on the heart rate features and the motion gesture features to obtain attention weighted fusion features;
and inputting the fusion characteristics into a built multi-layer perceptron neural network model with an attention mechanism, and outputting the motion state of the user on the underwater running platform through multi-layer nonlinear transformation.
As an improvement of the above solution, the calculating the attention weight of the heart rate feature and the motion gesture feature and performing feature stitching to obtain an attention weighted fusion feature includes:
converting heart rate feature vector O HR and motion gesture feature vector O STA into key vectors and value vectors using linear transformations;
The attention weight w i of the heart rate feature vector O HR and the motion-posture feature vector O STA is calculated by the following formula:
Wherein Q k,Ki,Vi represents a query vector, a key vector, and a value vector, respectively, from linear transformations of heart rate feature vector O HR and motion gesture feature vector O STA, respectively, d k is the dimension of the key vector for scaling the dot product result to avoid gradient vanishing;
Feature stitching is performed through the following formula to obtain a weighted attention feature vector v:
n is the number of linearly transformed value vectors V i of the heart rate feature vector O HR and the motion gesture feature vector O STA.
As an improvement of the above scheme, the multi-layer perceptron neural network model with the attention mechanism is composed of a plurality of hidden layers, and the calculation formula of each hidden layer is as follows:
hl＝f(Wl·v+bl),l=1,...,L
Wherein h l-1 is the output of the previous layer, W l and b l are the weight matrix and bias vector of the first layer, f is a nonlinear activation function, the output of the last layer L is used as the input of the final layer h L, and the calculation formula of the final layer h L is:
S=g (W S·hL+bS) where W S and b S are the weight matrix and bias vector of the final layer h L, g
And S is the motion state of the user on the underwater running platform.
As an improvement of the scheme, the model formula of the underwater running platform speed regulation model is as follows:
Wherein V opt is the output optimal training water flow speed of the user on the underwater running platform at present; The method is a deep learning model, the parameters are Θ, the parameters comprise a weight matrix W h、Wf and a bias vector b h、bf, sigma (·) is an activation function of an output layer, and tan h (·) is an activation function of a hidden layer, and the method is used for introducing nonlinear transformation and enhancing the expression capability of the model.
Compared with the prior art, the embodiment of the invention has the following beneficial effects:
According to the embodiment of the invention, heart rate variation and motion gesture of the athlete are captured in real time, heart rate data and motion gesture characteristics are fused and are input into a trained multi-layer perceptron neural network model with a attention mechanism, the model can deeply analyze the motion state of the athlete, the obtained motion state information is input into a specially trained underwater running platform speed regulation model, and the model learns and predicts the current optimal training water flow speed by utilizing a deep learning technology so as to realize personalized speed regulation. Finally, according to the current optimal training water flow speed, the water flow speed of the underwater running platform is adjusted in real time, so that the athlete can train in the optimal state, the training efficiency is improved, and the training safety is guaranteed. From the analysis, the embodiment of the invention solves the problems of excessive static speed regulation strategy and lack of individuation in the prior art, can dynamically adjust the water flow speed according to the real-time physiological state and the athletic performance of athletes, and effectively improves the training effect of underwater sports.
Drawings
FIG. 1 is a schematic diagram of an architecture of an intelligent speed regulation method system for an underwater running platform based on deep learning according to an embodiment of the present invention;
fig. 2 is a schematic flow chart of an intelligent speed regulation method of an underwater running platform based on deep learning according to an embodiment of the invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, an architecture diagram of an intelligent speed regulation method system for an underwater running platform based on deep learning according to an embodiment of the present invention is provided. The intelligent speed regulation system of the underwater running platform based on deep learning comprises a heart rate detection device 10, a camera 11 and a control device 12, wherein the heart rate detection device 10 is used for detecting the heart rate of a user moving on the underwater running platform, the heart rate detection device 10 can be an intelligent wearing device with a heart rate detection function, the user can wear on a body when moving on an underwater cannon platform, for example, the user can wear an intelligent watch with the heart rate detection function on the hand of the user, the camera 11 is used for shooting images of the user moving on the underwater running platform, the camera 11 can be provided with a waterproof function and arranged in water and aligned with the underwater running platform, the control device 12 (for example, a desktop computer or a control main board used for controlling the underwater running platform) is in communication connection (such as wired connection or wireless communication connection) with the heart rate detection device and the camera, and is used for:
acquiring the heart rate and the image;
Performing motion gesture analysis on the image by using an image processing algorithm to obtain the motion gesture of the user;
Taking the heart rate and the motion gesture as input data, and inputting the input data into a trained multi-layer perceptron neural network model with an attention mechanism to obtain the motion state of the user on an underwater running platform;
Inputting the motion state into a trained underwater running platform speed regulation model to obtain the optimal training water flow speed of the user on the underwater running platform;
and controlling the current water flow speed of the underwater running platform to be adjusted to the optimal training water flow speed.
According to the embodiment of the invention, heart rate variation and motion gesture of the athlete are captured in real time, heart rate data and motion gesture characteristics are fused and are input into a trained multi-layer perceptron neural network model with a attention mechanism, the model can deeply analyze the motion state of the athlete, the obtained motion state information is input into a specially trained underwater running platform speed regulation model, and the model learns and predicts the current optimal training water flow speed by utilizing a deep learning technology so as to realize personalized speed regulation. Finally, according to the current optimal training water flow speed, the water flow speed of the underwater running platform is adjusted in real time, so that the athlete can train in the optimal state, the training efficiency is improved, and the training safety is guaranteed. From the analysis, the embodiment of the invention solves the problems of excessive static speed regulation strategy and lack of individuation in the prior art, can dynamically adjust the water flow speed according to the real-time physiological state and the athletic performance of athletes, and effectively improves the training effect of underwater sports.
It should be noted that, the image processing algorithm used for the motion gesture analysis of the image may be a prior art, for example OpenPose, deepLab, hourglass Network, etc.
It will be appreciated that an underwater treadmill is typically equipped with a water return system that can adjust the speed and direction of the water flow, thereby affecting the resistance and strength of the runner's exercise. The basic principle of adjusting the water flow speed is that a water return system of the underwater running platform is designed to circulate water and generate water flow. The system may include a water pump, piping, nozzles, valves, and filtration devices. The water pump pumps the water out of the bottom of the pool, filters the water to remove impurities, and then re-injects the water into the pool through the pipeline and the nozzle to form water flow. The core of controlling the water flow rate is to adjust the power of the water pump and/or change the path of the water flow. This can be achieved by:
The rotation speed of the water pump motor is adjusted so as to change the flow and pressure of water.
A variable cross-section nozzle or valve is used to vary the cross-sectional area of the water flow, thereby affecting the flow rate.
The direction of water flow is changed by the guide plate or the baffle plate, so that the density and the speed of the water flow in front of the runner are influenced.
In one embodiment, the inputting the heart rate and the motion gesture as input data into a trained multi-layer perceptron neural network model with an attention mechanism to obtain a motion state of the user on an underwater running platform includes:
extracting the characteristics of the heart rate and the motion gesture to obtain heart rate characteristics and motion gesture characteristics;
performing attention weight calculation and feature stitching on the heart rate features and the motion gesture features to obtain attention weighted fusion features;
and inputting the fusion characteristics into a built multi-layer perceptron neural network model with an attention mechanism, and outputting the motion state of the user on the underwater running platform through multi-layer nonlinear transformation.
In this embodiment, through a deep learning technology, in particular, a multi-layer perceptron neural network model with a attention mechanism, the heart rate data and the motion gesture data acquired in real time are subjected to deep analysis, so as to acquire the accurate motion state of the athlete on the underwater running platform. Specifically, first, raw heart rate data and exercise posture data are preprocessed, and key features capable of reflecting physiological states and exercise performances of the athlete are extracted. This stage aims to convert the complex data into a form that is easy to process by the machine learning model, ensuring the accuracy of the subsequent analysis. Next, the heart rate features and the motion gesture features are weighted using an attention mechanism to emphasize those feature information that are most critical to the current motion state. Through the calculation of the attention weight, the model can autonomously decide which features are more important under the current situation, so that intelligent screening and fusion of the features are performed. The weighted feature stitching not only enhances the understanding capability of the model to complex motion scenes, but also improves the distinguishing degree of the features, and lays a solid foundation for the subsequent motion state prediction. And finally, inputting the weighted fusion characteristics to a pre-trained multi-layer perceptron neural network model with an attention mechanism. The model can capture complex correlations among features through multi-layer nonlinear transformation, so that the motion state of an athlete on an underwater running platform is output, wherein the motion state comprises but is not limited to state indexes of multiple dimensions such as fatigue degree, motion efficiency and the like.
As an improvement of the above solution, the calculating the attention weight of the heart rate feature and the motion gesture feature and performing feature stitching to obtain an attention weighted fusion feature includes:
converting heart rate feature vector O HR and motion gesture feature vector O STA into key vectors and value vectors using linear transformations;
The attention weight w i of the heart rate feature vector O HR and the motion-posture feature vector O STA is calculated by the following formula:
Wherein Q k,Ki,Vi represents a query vector, a key vector, and a value vector, respectively, from linear transformations of heart rate feature vector O HR and motion gesture feature vector O STA, respectively, d k is the dimension of the key vector for scaling the dot product result to avoid gradient vanishing;
feature stitching is performed through the following formula to obtain a weighted attention feature vector v:
n is the number of linearly transformed value vectors V i of the heart rate feature vector O HR and the motion gesture feature vector O STA.
In the embodiment, the depth fusion of the heart rate characteristic and the movement posture characteristic is realized by using an attention mechanism so as to enhance the accuracy and individualization of the model for understanding the movement state of the athlete. Specifically, first, the extracted heart rate feature vector and motion gesture feature vector are converted into key vectors (Keys) and value vectors (Values) through linear transformation. This process aims to lay a foundation for the subsequent calculation of the attention mechanism, so that the model can process the characteristic information of different sources in a unified manner. And then, obtaining the attention weights of the heart rate feature vector and the motion gesture feature vector by calculating the dot product between the Query vector (Query) and the key vector (Keys) and performing scaling and Softmax function processing. The query vector can be regarded as the focus of attention of the model on the current task, and the key vector represents the potential meaning of the feature vector. This calculation process enables the model to adaptively adjust the degree of importance to different features, thereby giving more important features higher weight when feature fusion. And finally, carrying out weighted summation on the value vectors according to the calculated attention weight to obtain the feature vectors after attention weighting. The weighted fusion process not only enhances the interaction between the features, but also ensures that the model can make more comprehensive and accurate decisions from the global perspective. In conclusion, the intelligent fusion of the heart rate characteristics and the movement posture characteristics is realized by introducing the attention mechanism, and the performance of the model in the field of analysis of the movement states of athletes is remarkably improved.
As an improvement of the above scheme, the multi-layer perceptron neural network model with the attention mechanism is composed of a plurality of hidden layers, and the calculation formula of each hidden layer is as follows:
hl＝f(Wl·v+bl),l=1,...,L
Wherein h l-1 is the output of the previous layer, W l and b l are the weight matrix and bias vector of the first layer, f is a nonlinear activation function (such as a ReLU or tanh function), the output of the last layer L is used as the input of the final layer h L, and the calculation formula of the final layer h L is:
S=g (W S·hL+bS), where W S and b S are the weight matrix and bias vector of the final layer h L, g is the output activation function (softmax or linear function may be used), S is the output motion state of the user on the underwater treadmill.
In the embodiment, a multi-layer perceptron neural network model with an attention mechanism is constructed and used for deeply processing and fusing heart rate characteristics and motion gesture characteristics so as to accurately predict the motion state of a user on an underwater running platform. The multi-layer perceptron neural network (MLP) model structure consists of a plurality of hidden layers, each hidden layer carries out linear transformation through a weight matrix and a bias vector, and then the expression capacity of the model is increased through a nonlinear activation function so as to learn a complex feature mapping relation. The attention mechanism enables the model to focus on key parts in the input features by calculating the attention weight, so that the sensitivity and learning efficiency of the model to important features are improved. By means of the attention weighting, the model can adaptively adjust the importance of the features, thereby better capturing patterns closely related to motion state prediction in complex feature spaces. Therefore, the embodiment can automatically screen and learn the information most valuable for predicting the motion state from the complex heart rate and motion gesture features by introducing an attention mechanism, and the accuracy of prediction is remarkably improved. Meanwhile, the deep structure and nonlinear transformation capability of the multi-layer perceptron neural network model enable the model to learn deep abstract features of data, so that good prediction performance is still maintained under different individuals and environmental conditions. In addition, the attention mechanism enables the model to adaptively adjust the prediction strategy according to different user characteristics and motion situations, so that personalized analysis of motion states of each user is realized.
As an improvement of the scheme, the model formula of the underwater running platform speed regulation model is as follows:
Wherein V opt is the output optimal training water flow speed of the user on the underwater running platform at present; The method is a deep learning model, the parameters are Θ, the parameters comprise a weight matrix W h、Wf and a bias vector b h、bf, sigma (·) is an activation function of an output layer (Sigmoid, reLU or linear function can be selected), and tan h (·) is an activation function of a hidden layer, and the method is used for introducing nonlinear transformation and enhancing the expression capability of the model.
In this embodiment, a special underwater running platform speed regulation model is constructed by using a deep learning model to predict the current optimal training water flow speed of the user. The deep learning model is designed to handle complex nonlinear relations, and key features can be automatically learned and extracted from the input motion state so as to predict the optimal training water flow speed. The structure of the model comprises a hidden layer and an output layer, wherein the hidden layer introduces the complexity of the model through a nonlinear activation function, and the modeling capability of the model on the nonlinear relation is enhanced. By predicting the water flow speed which is most matched with the current motion state of the user, the system can provide personalized training conditions for each user, and reduce unnecessary physical consumption and damage risk while maximizing the training effect. And the optimal water flow speed is adjusted in real time, so that the water flow speed is always in an optimal state in the training process, and the endurance, speed and skill of athletes are improved, and the overall training efficiency is improved.
Referring to fig. 2, a schematic flow chart of an intelligent speed regulation method for an underwater running platform based on deep learning according to an embodiment of the invention is provided. The intelligent speed regulation method of the underwater running platform based on the deep learning is applied to the intelligent speed regulation system of the underwater running platform based on the deep learning, and comprises the following steps of S10 to S14:
s10, acquiring the heart rate and the image;
S11, performing motion gesture analysis on the image by using an image processing algorithm to obtain the motion gesture of the user;
S12, taking the heart rate and the motion gesture as input data, and inputting the input data into a trained multi-layer perceptron neural network model with an attention mechanism to obtain a motion state of the user on an underwater running platform;
S13, inputting the motion state into a trained underwater running platform speed regulation model to obtain the optimal training water flow speed of the user on the underwater running platform;
S14, controlling the current water flow speed of the underwater running platform to be adjusted to the optimal training water flow speed.
According to the embodiment of the invention, heart rate variation and motion gesture of the athlete are captured in real time, heart rate data and motion gesture characteristics are fused and are input into a trained multi-layer perceptron neural network model with a attention mechanism, the model can deeply analyze the motion state of the athlete, the obtained motion state information is input into a specially trained underwater running platform speed regulation model, and the model learns and predicts the current optimal training water flow speed by utilizing a deep learning technology so as to realize personalized speed regulation. Finally, according to the current optimal training water flow speed, the water flow speed of the underwater running platform is adjusted in real time, so that the athlete can train in the optimal state, the training efficiency is improved, and the training safety is guaranteed. From the analysis, the embodiment of the invention solves the problems of excessive static speed regulation strategy and lack of individuation in the prior art, can dynamically adjust the water flow speed according to the real-time physiological state and the athletic performance of athletes, and effectively improves the training effect of underwater sports.
As an improvement of the above solution, the inputting the heart rate and the motion gesture as input data into a trained multi-layer perceptron neural network model with an attention mechanism to obtain a motion state of the user on an underwater running platform includes:
extracting the characteristics of the heart rate and the motion gesture to obtain heart rate characteristics and motion gesture characteristics;
performing attention weight calculation and feature stitching on the heart rate features and the motion gesture features to obtain attention weighted fusion features;
and inputting the fusion characteristics into a built multi-layer perceptron neural network model with an attention mechanism, and outputting the motion state of the user on the underwater running platform through multi-layer nonlinear transformation.
As an improvement of the above solution, the calculating the attention weight of the heart rate feature and the motion gesture feature and performing feature stitching to obtain an attention weighted fusion feature includes:
converting heart rate feature vector O HR and motion gesture feature vector O STA into key vectors and value vectors using linear transformations;
The attention weight w i of the heart rate feature vector O HR and the motion-posture feature vector O STA is calculated by the following formula:
Wherein Q k,Ki,Vi represents a query vector, a key vector, and a value vector, respectively, from linear transformations of heart rate feature vector O HR and motion gesture feature vector O STA, respectively, d k is the dimension of the key vector for scaling the dot product result to avoid gradient vanishing;
feature stitching is performed through the following formula to obtain a weighted attention feature vector v:
n is the number of linearly transformed value vectors V i of the heart rate feature vector O HR and the motion gesture feature vector O STA.
As an improvement of the above scheme, the multi-layer perceptron neural network model with the attention mechanism is composed of a plurality of hidden layers, and the calculation formula of each hidden layer is as follows:
hl＝f(Wl·v+bl),l=1,...,L
Wherein h l-1 is the output of the previous layer, W l and b l are the weight matrix and bias vector of the first layer, f is a nonlinear activation function, the output of the last layer L is used as the input of the final layer h L, and the calculation formula of the final layer h L is:
S=g (W S·hL+bS) where W S and b S are the weight matrix and bias vector of the final layer h L, g is the output activation function, and S is the output motion state of the user on the underwater running deck.
As an improvement of the scheme, the model formula of the underwater running platform speed regulation model is as follows:
Wherein V opt is the output optimal training water flow speed of the user on the underwater running platform at present; The method is a deep learning model, the parameters are Θ, the parameters comprise a weight matrix W h、Wf and a bias vector b h、bf, sigma (·) is an activation function of an output layer, and tan h (·) is an activation function of a hidden layer, and the method is used for introducing nonlinear transformation and enhancing the expression capability of the model.
While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that changes and modifications may be made without departing from the principles of the invention, such changes and modifications are also intended to be within the scope of the invention.

Claims (8)

1. An intelligent speed regulation system of running platform under water based on degree of depth study, its characterized in that includes:
the heart rate detection device is used for detecting the heart rate of a user moving on the underwater running platform;
the camera is used for shooting the image of the user;
The control device is in communication connection with the heart rate detection device and the camera and is used for:
acquiring the heart rate and the image;
Performing motion gesture analysis on the image by using an image processing algorithm to obtain the motion gesture of the user;
Taking the heart rate and the motion gesture as input data, and inputting the input data into a trained multi-layer perceptron neural network model with an attention mechanism to obtain the motion state of the user on an underwater running platform;
Inputting the motion state into a trained underwater running platform speed regulation model to obtain the optimal training water flow speed of the user on the underwater running platform, wherein the model formula of the underwater running platform speed regulation model is as follows:
Wherein V opt is the output optimal training water flow speed of the user on the underwater running platform at present; the method is a deep learning model, the parameters are Θ, the parameters comprise a weight matrix W h、Wf and a bias vector b h、bf, sigma (·) is an activation function of an output layer, and tan h (·) is an activation function of a hidden layer, and the parameters are used for introducing nonlinear transformation to enhance the expression capability of the model;
and controlling the current water flow speed of the underwater running platform to be adjusted to the optimal training water flow speed.
2. The intelligent speed regulation system of the underwater running platform based on deep learning according to claim 1, wherein the step of inputting the heart rate and the motion gesture as input data into a trained multi-layer perceptron neural network model with an attention mechanism to obtain the motion state of the user on the underwater running platform comprises the following steps:
extracting the characteristics of the heart rate and the motion gesture to obtain heart rate characteristics and motion gesture characteristics;
performing attention weight calculation and feature stitching on the heart rate features and the motion gesture features to obtain attention weighted fusion features;
and inputting the fusion characteristics into a built multi-layer perceptron neural network model with an attention mechanism, and outputting the motion state of the user on the underwater running platform through multi-layer nonlinear transformation.
3. The intelligent speed regulation system of the underwater running platform based on deep learning as set forth in claim 2, wherein the calculating the attention weight of the heart rate feature and the motion gesture feature and the feature stitching to obtain the attention weighted fusion feature includes:
converting heart rate feature vector O HR and motion gesture feature vector O STA into key vectors and value vectors using linear transformations;
The attention weight w i of the heart rate feature vector O HR and the motion-posture feature vector O STA is calculated by the following formula:
Wherein Q k,Ki,Vi represents a query vector, a key vector, and a value vector, respectively, from linear transformations of heart rate feature vector O HR and motion gesture feature vector O STA, respectively, d k is the dimension of the key vector for scaling the dot product result to avoid gradient vanishing;
feature stitching is performed through the following formula to obtain a weighted attention feature vector v:
n is the number of linearly transformed value vectors V i of the heart rate feature vector O HR and the motion gesture feature vector O STA.
4. The intelligent speed regulation system of an underwater running platform based on deep learning as set forth in claim 3, wherein the neural network model of the multi-layer perceptron with a attention mechanism is composed of a plurality of hidden layers, and the calculation formula of each hidden layer is:
hl＝f(Wl·v+bl),l=1,...,L
Wherein h l-1 is the output of the previous layer, W l and b l are the weight matrix and bias vector of the first layer, f is a nonlinear activation function, the output of the last layer L is used as the input of the final layer h L, and the calculation formula of the final layer h L is:
S=g (W S·hL+bS), where W S and b S are the weight matrix and bias vector of the final layer h L, g is the output activation function, and S is the output motion state of the user on the underwater running deck.
5. An intelligent speed regulation method of an underwater running platform based on deep learning, which is applied to the intelligent speed regulation system of the underwater running platform based on deep learning as set forth in any one of claims 1 to 4, and is characterized by comprising the following steps:
acquiring the heart rate and the image;
Performing motion gesture analysis on the image by using an image processing algorithm to obtain the motion gesture of the user;
Taking the heart rate and the motion gesture as input data, and inputting the input data into a trained multi-layer perceptron neural network model with an attention mechanism to obtain the motion state of the user on an underwater running platform;
Inputting the motion state into a trained underwater running platform speed regulation model to obtain the optimal training water flow speed of the user on the underwater running platform;
and controlling the current water flow speed of the underwater running platform to be adjusted to the optimal training water flow speed.
6. The intelligent speed regulation method of the underwater running platform based on deep learning according to claim 5, wherein the step of inputting the heart rate and the motion gesture as input data into a trained multi-layer perceptron neural network model with an attention mechanism to obtain the motion state of the user on the underwater running platform comprises the following steps:
extracting the characteristics of the heart rate and the motion gesture to obtain heart rate characteristics and motion gesture characteristics;
performing attention weight calculation and feature stitching on the heart rate features and the motion gesture features to obtain attention weighted fusion features;
and inputting the fusion characteristics into a built multi-layer perceptron neural network model with an attention mechanism, and outputting the motion state of the user on the underwater running platform through multi-layer nonlinear transformation.
7. The intelligent speed regulation method of the underwater running platform based on deep learning as set forth in claim 6, wherein the calculating the attention weight of the heart rate feature and the motion gesture feature and the feature stitching to obtain the attention weighted fusion feature includes:
converting heart rate feature vector O HR and motion gesture feature vector O STA into key vectors and value vectors using linear transformations;
The attention weight w i of the heart rate feature vector O HR and the motion-posture feature vector O STA is calculated by the following formula:
Wherein Q k,Ki,Vi represents a query vector, a key vector, and a value vector, respectively, from linear transformations of heart rate feature vector O HR and motion gesture feature vector O STA, respectively, d k is the dimension of the key vector for scaling the dot product result to avoid gradient vanishing;
Feature stitching is performed through the following formula, and a feature vector upsilon after attention weighting is obtained:
n is the number of linearly transformed value vectors V i of the heart rate feature vector O HR and the motion gesture feature vector O STA.
8. The intelligent speed regulation method of the underwater running platform based on deep learning as set forth in claim 7, wherein the multi-layer perceptron neural network model with the attention mechanism consists of a plurality of hidden layers, and the calculation formula of each hidden layer is:
hl＝f(Wl·v+bl),l=1,...,L
Wherein h l-1 is the output of the previous layer, W l and b l are the weight matrix and bias vector of the first layer, f is a nonlinear activation function, the output of the last layer L is used as the input of the final layer h L, and the calculation formula of the final layer h L is:
S=g (W S·hL+bS), where W S and b S are the weight matrix and bias vector of the final layer h L, g is the output activation function, and S is the output motion state of the user on the underwater running deck.
CN202410906937.9A 2024-07-08 2024-07-08 Underwater running platform intelligent speed regulation system and method based on deep learning Active CN119015671B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202410906937.9A CN119015671B (en) 2024-07-08 2024-07-08 Underwater running platform intelligent speed regulation system and method based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202410906937.9A CN119015671B (en) 2024-07-08 2024-07-08 Underwater running platform intelligent speed regulation system and method based on deep learning

Publications (2)

Publication Number Publication Date
CN119015671A CN119015671A (en) 2024-11-26
CN119015671B true CN119015671B (en) 2025-03-14

Family

ID=93523866

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202410906937.9A Active CN119015671B (en) 2024-07-08 2024-07-08 Underwater running platform intelligent speed regulation system and method based on deep learning

Country Status (1)

Country Link
CN (1) CN119015671B (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2024060321A1 (en) * 2022-09-22 2024-03-28 之江实验室 Joint modeling method and apparatus for enhancing local features of pedestrians
CN118161818A (en) * 2024-04-23 2024-06-11 康辉医疗科技(苏州)有限公司 Hydrotherapy rehabilitation training device for correcting exercise posture on line

Family Cites Families (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6430997B1 (en) * 1995-11-06 2002-08-13 Trazer Technologies, Inc. System and method for tracking and assessing movement skills in multidimensional space
JP2005267179A (en) * 2004-03-18 2005-09-29 Seiko Epson Corp Sports facility use system and method of use
KR101606768B1 (en) * 2014-06-24 2016-03-28 (주)이튜 Motion extraction-based ambulation posture correcting apparatus
KR20160064890A (en) * 2014-11-28 2016-06-08 주식회사 삼육오엠씨네트웍스 Treadmill system with running posture correction
CN108434664A (en) * 2018-04-08 2018-08-24 上海应用技术大学 A kind of treadmill intelligent safe protector and guard method
WO2020170001A1 (en) * 2019-02-22 2020-08-27 INSERM (Institut National de la Santé et de la Recherche Médicale) Device and method for patient exercising in aquatic conditions
US11972353B2 (en) * 2020-01-22 2024-04-30 Electronic Arts Inc. Character controllers using motion variational autoencoders (MVAEs)
CN212642266U (en) * 2020-05-27 2021-03-02 深圳金诚卫浴有限公司 Treadmill swimming pool
KR20220064291A (en) * 2020-11-11 2022-05-18 정재훈 System and method based on deep learing for recognizing weight training equipments
CN112774098B (en) * 2021-02-22 2022-03-15 中山大学 Dynamic calibration method for treadmill force platform
CN113730892A (en) * 2021-09-30 2021-12-03 天津工业大学 Athlete training method and system
EP4300435A1 (en) * 2022-07-01 2024-01-03 Toyota Jidosha Kabushiki Kaisha Motion representation calculation method and system, training method, computer program, readable medium and system
CN116226749A (en) * 2022-12-24 2023-06-06 上海海洋大学 A Method for Estimating the State of Submersible Swimming and Swimming Movement
CN116570273A (en) * 2023-05-29 2023-08-11 深圳市界水医疗科技有限公司 Aquatic movement rehabilitation method and system based on movement detection
CN117909922A (en) * 2024-01-29 2024-04-19 南京邮电大学 Depth feature fusion and optimization method and system for multi-mode data

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2024060321A1 (en) * 2022-09-22 2024-03-28 之江实验室 Joint modeling method and apparatus for enhancing local features of pedestrians
CN118161818A (en) * 2024-04-23 2024-06-11 康辉医疗科技(苏州)有限公司 Hydrotherapy rehabilitation training device for correcting exercise posture on line

Also Published As

Publication number Publication date
CN119015671A (en) 2024-11-26

Similar Documents

Publication Publication Date Title
Yang et al. IA-Net $: $ An inception–attention-module-based network for classifying underwater images from others
CN111627052B (en) Action recognition method T-STAM based on double-flow space-time attention mechanism
Cai et al. Deep historical long short-term memory network for action recognition
CN110478883B (en) A kind of fitness action teaching and correction system and method
CN114841319B (en) A multispectral image change detection method based on multi-scale adaptive convolution kernel
CN115223201B (en) Method, system and device for estimating three-dimensional human joint points based on monocular sequence images
CN114186672A (en) Efficient high-precision training algorithm for impulse neural network
CN111598968A (en) Image processing method and device, storage medium and electronic equipment
CN113128424A (en) Attention mechanism-based graph convolution neural network action identification method
Martin et al. 3D attention mechanism for fine-grained classification of table tennis strokes using a Twin Spatio-Temporal Convolutional Neural Networks
Xu et al. Isolated Word Sign Language Recognition Based on Improved SKResNet‐TCN Network
CN119015671B (en) Underwater running platform intelligent speed regulation system and method based on deep learning
Gao et al. Resfnn: Residual structure-based feedforward neural network for action quality assessment in sports consumer electronics
CN113457108B (en) Cognitive characterization-based exercise performance improving method and device
Xu et al. Dynamic Context Removal: A General Training Strategy for Robust Models on Video Action Predictive Tasks.
CN119810909A (en) Dance motion capture and real-time feedback control system
CN114067429A (en) Action recognition processing method, device and equipment
CN118490226B (en) A method, device and system for assessing concentration for amblyopia training
Joung et al. Contrastive learning for action assessment using graph convolutional networks with augmented virtual joints
Zhang et al. LMFNet: human activity recognition using attentive 3-D residual network and multistage fusion strategy
Iosifidis et al. Multi-view regularized extreme learning machine for human action recognition
Pan et al. Generalizing Sensor-based Human Activity Recognition with Domain Specific Ensemble Learning
Cui et al. Design and Implementation of Sports Network Teaching Platform Based on Discrete Similarity Method.
Wang et al. Coordination Training and Testing of Upper and Lower Limbs in Aerobics Under Neural Networks
Rustogi et al. Long-term Spatio-temporal Contrastive Learning framework for Skeleton Action Recognition

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant