Respiratory function exercise system based on artificial intelligence

Abstract

The invention relates to the technical field of medical rehabilitation training, in particular to a respiratory function training system based on artificial intelligence, which comprises a multi-dimensional data acquisition module used for acquiring basic information, health condition information and respiratory function information of a user; the data preprocessing module is used for carrying out vectorization processing on the basic information, the health condition information and the respiratory function information to obtain a multi-dimensional vector; the data analysis and feature extraction module performs key feature extraction on the multi-dimensional vector by using a convolutional neural network, constructs a user personalized respiratory function model and outputs the multi-dimensional vector; the scheme customization module is used for searching and optimizing the basic template and generating a personalized exercise scheme; and the real-time monitoring feedback module is used for dynamically adjusting the personalized exercise scheme and feeding back and guiding the user to train in real time. According to the method, the multi-dimensional data of the user is deeply analyzed, an exclusive breathing exercise scheme is customized for the user, the pertinence and effectiveness of training are ensured, and the individual requirements of the user are met.

Description

Technical Field

[0001] The present invention relates to the technical field of medical rehabilitation training, and in particular to a respiratory function training system based on artificial intelligence. Background Art

[0002] Respiratory function training is a training method that uses specific methods and movements to regulate breathing rhythm, enhance respiratory muscle strength, improve lung ventilation function, and enhance respiratory efficiency and health level.

[0003] Currently, most respiratory function training programs utilize fixed, templated training models that lack personalized adaptability. Existing programs are typically formulated based on only a few indicators, such as basic age and gender, and fail to fully consider multiple factors, including the user's health status and respiratory history. This results in inconsistent training results. During exercise, most programs rely on user perception and feedback, failing to implement real-time monitoring and dynamic adjustments. This poses safety risks and can easily lead to a mismatch between training intensity and the user's actual abilities. Furthermore, traditional programs lack intelligent analysis and optimization mechanisms, making it difficult to accurately evaluate training results and continuously improve them, making it difficult to meet user demands for scientific and effective respiratory function training. Summary of the Invention

[0004] The purpose of the present invention is to address the problems existing in the background technology and to propose a respiratory function training system based on artificial intelligence.

[0005] The technical solution of the present invention is: a respiratory function training system based on artificial intelligence, comprising a multi-dimensional data acquisition module, a data preprocessing module, a data analysis and feature extraction module, a program customization module and a real-time monitoring feedback module;

[0006] Multi-dimensional data collection module, used to collect user's basic information, health status information and respiratory function information;

[0007] The data preprocessing module is used to vectorize the basic information, health status information and respiratory function information to obtain the corresponding basic vector, health status vector and respiratory function vector, normalize the vector components respectively, and splice the normalized vectors to obtain a multidimensional vector;

[0008] The data analysis and feature extraction module first uses a convolutional neural network to extract key features from the multidimensional vector. Based on the extracted key features, it uses the support vector machine algorithm to build a user-customized respiratory function model, evaluate the user's current respiratory function status, and update the output multidimensional vector.

[0009] The program customization module is used to integrate the classified breathing exercise program library, match the user's breathing characteristics and health data through reinforcement learning algorithms, search and optimize basic templates, and generate personalized exercise programs;

[0010] The real-time monitoring and feedback module is used to collect breathing rate, depth and rhythm data in real time. After analysis by the sliding window algorithm, it dynamically adjusts the personalized exercise plan and provides real-time feedback to guide user training.

[0011] Preferably, the basic information includes at least age, gender, height and weight; the health status information includes at least respiratory diseases; and the respiratory function information includes at least the current respiratory rate value, respiratory rate fluctuation amplitude, tidal volume, deep inspiratory volume, vital capacity, forced expiratory volume in the first second and maximum ventilation volume.

[0012] Preferably, the solution customization module is connected with a solution library construction unit, a solution screening unit, a template selection unit and a solution generation unit;

[0013] The program library construction unit is used to collect and organize various breathing exercise programs, including basic exercise program templates for different respiratory function problems, different populations, and different exercise scenarios;

[0014] The solution screening unit is used to classify and label each solution in the solution library to facilitate subsequent rapid retrieval and matching based on user characteristics;

[0015] A template selection unit, configured to search and filter a set of candidate exercise program templates suitable for the user's current condition in a program library according to the user's multi-dimensional vector;

[0016] The program generation unit is used to further screen the selected candidate exercise program template set based on the user's multi-dimensional vector using a reinforcement learning algorithm to obtain the optimal exercise program template.

[0017] Preferably, each program template details the exercise movements, movement essentials, exercise intensity, exercise duration, and breathing requirements; each program template contains a feature vector consistent with the dimension of the multidimensional vector, and the value of each dimension is obtained by calculating the big data mean.

[0018] Preferably, the cosine similarity between the user's multidimensional vector and the feature vector in each solution template is calculated using the following formula:

[0019] Where A is the user multidimensional vector; B is the feature vector of the solution template;

[0020] Set a matching threshold, compare the calculated cosine similarity with the threshold, and only retain the solution templates with a similarity greater than the threshold to obtain a preliminary screening template set and enter the subsequent screening process.

[0021] Preferably, the reinforcement learning algorithm uses a Q-learning algorithm to update the strategy of the selection scheme template.

[0022] Preferably, the real-time monitoring feedback module is connected to the real-time monitoring unit, the real-time processing and analysis unit and the solution adjustment unit;

[0023] The real-time monitoring unit is used to collect the breathing rate, breathing depth and breathing rhythm of the user at a fixed frequency during the user's exercise, and integrate them into real-time breathing data and upload them to the system in real time;

[0024] A real-time processing and analysis unit is used to analyze the real-time respiratory parameters using a sliding window algorithm and compare the data with the respiratory requirement information to obtain a determination result of the respiratory parameters;

[0025] The program adjustment unit makes decisions on adjusting the training program based on the judgment results and displays them to the user.

[0026] Preferably, the training program is adjusted and decided based on the determination result, and the method includes: classifying the abnormality into abnormal breathing rate, abnormal breathing depth and disordered breathing rhythm according to the determination result;

[0027] If there are abnormal breathing parameters, the training plan will be adjusted using an adaptive algorithm based on the type of abnormality, and the adjusted training plan will be updated and displayed to the user in a timely manner; if the parameters are normal, no operation will be performed.

[0028] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects:

[0029] (1) The present invention uses advanced artificial intelligence algorithms to deeply analyze multi-dimensional data such as age, gender, health status, respiratory history, etc. input by the user, and combines it with real-time respiratory monitoring information to tailor a unique breathing exercise program for the user, ensuring the pertinence and effectiveness of the training, breaking the limitations of traditional fixed programs, and meeting the user's personalized needs.

[0030] (2) During the exercise, the system continuously monitors the user's breathing frequency, depth, rhythm and other parameters. Once an abnormality or deviation from the predetermined plan is detected, feedback will be given immediately through voice prompts, screen guidance, etc., and the training difficulty and rhythm will be automatically adjusted to ensure the safety and effectiveness of the exercise. It can ensure the safety of the exercise and improve the effect in a timely manner, and has significant advantages in the intelligent and scientific aspects of respiratory function training. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a module structure diagram of the present invention. DETAILED DESCRIPTION

[0032] like Figure 1 As shown, the present invention proposes an artificial intelligence-based respiratory function training system, comprising:

[0033] Multi-dimensional data acquisition module, data preprocessing module, data analysis and feature extraction module, solution customization module and real-time monitoring feedback module;

[0034] A multi-dimensional data collection module is used to collect basic information, health information, and respiratory function information of the user; wherein the basic information includes at least age, gender, height, and weight; the health information includes at least respiratory diseases; and the respiratory function information includes at least the current respiratory rate value, respiratory rate fluctuation amplitude, tidal volume, deep inspiratory volume, vital capacity, forced expiratory volume in one second, and maximum ventilation volume;

[0035] The data preprocessing module is used to vectorize the basic information, health status information and respiratory function information to obtain the corresponding basic vector, health status vector and respiratory function vector, normalize the vector components respectively, and splice the normalized vectors to obtain a multidimensional vector;

[0036] The vectorization processing includes converting numerical data into floating-point data, classifying non-numerical data, performing one-hot encoding on the classified non-numerical data to obtain encoded data, analyzing and calculating the corresponding floating-point data to obtain analysis and calculation results, combining the floating-point data, the analysis and calculation results, and the encoded data to determine the basic vector, health status vector, and respiratory function vector corresponding to the basic information and respiratory function information, respectively;

[0037] The respiratory function vector is specifically {current respiratory rate value, deviation between the current respiratory rate value and the normal range, respiratory rate fluctuation amplitude, tidal volume, inspiratory volume, difference between respiratory depth and normal reference value, respiratory time ratio, vital capacity, forced expiratory volume in the first second, maximum ventilation volume, respiratory disease score};

[0038] The current respiratory rate value is obtained by recording the user's real-time respiratory rate in resting state, daily activity state and exercise state;

[0039] The deviation between the current respiratory rate value and the normal range is calculated by comparing the user's real-time respiratory rate with the normal respiratory rate range of people of the same age, gender, and health status. The specific deviation from the normal range is calculated to determine the degree of abnormality of the respiratory rate. For example, the normal resting respiratory rate of an adult female is 12-20 breaths / minute. If the user's resting respiratory rate is 25 breaths / minute, the deviation value is +5 breaths / minute.

[0040] The amplitude of respiratory rate fluctuation is measured by analyzing the respiratory rate data over a period of time and calculating its standard deviation or range to measure the stability of the respiratory rate; a period of time can be 10 minutes;

[0041] Tidal volume refers to the amount of gas inhaled or exhaled each time the user breathes calmly; inspiratory volume refers to the maximum amount of gas the user can inhale at the end of a calm exhalation; the difference between the breathing depth and the normal reference value is obtained by comparing the user's tidal volume, inspiratory volume and other breathing depth indicators with the normal reference values ​​of the same group of people; the breathing time ratio is obtained by recording the ratio of the user's inhalation time to the exhalation time during the breathing process; the forced expiratory volume in the first second refers to the amount of gas that can be exhaled in the first second after the maximum deep inhalation and the fastest exhalation; maximum ventilation: the maximum amount of gas the user can breathe in a unit of time; it should be noted that tidal volume, inspiratory volume, vital capacity, forced expiratory volume in the first second and maximum ventilation can all be measured by existing measurement techniques, and the details will not be repeated here;

[0042] The respiratory disease score was calculated using the following formula:

[0043]

[0044] S2=z1×D+z2×C+z3×E;(Formula 2)

[0045] S Total =β1×S1+β2×S2; (Formula 3)

[0046] In the formula, S1 is the respiratory function score; FVC is the measured value of forced vital capacity; FVC norm The normal reference value of forced vital capacity for people of the same age, gender and height; FEV1 is the measured value of forced expiratory volume in the first second; FEV 1norm Corresponding normal reference value; FEV1 / FVC is the measured value of the ratio of the two; (FEV1 / FVC) norm is the normal reference ratio; FVC norm , FEV 1norm and (FEV1 / FVC) norm All are based on big data testing and calibration; k1, k2 and k3 are weight coefficients of formula 1; S2 is the symptom score; D is the dyspnea score, where D∈[0,10], 0 means no dyspnea, and 10 means extreme dyspnea; C is the cough score, C∈[0,10] points, 0 means no cough, and 10 means continuous and severe cough; E is the cough score, E∈[0,10] points, 0 means no sputum, and 10 means a large amount of sputum that is difficult to cough up; z1, z2 and z3 are the weight coefficients of formula 2; S Total is the respiratory disease score, β1 and β2 are the weight coefficients of formula 3;

[0047] The data analysis and feature extraction module first uses a convolutional neural network to extract key features from multidimensional vectors. This involves sliding different convolution kernels across the multidimensional feature vector data through the CNN convolution layer to automatically extract key features from the multidimensional vector data. The pooling layer then performs dimensionality reduction on the extracted key features, reducing the amount of data while retaining important features.

[0048] Based on the extracted key features, a support vector machine algorithm is used to build a personalized respiratory function model for the user. This involves finding the optimal hyperplane to classify samples with different features, accurately dividing the different states and levels of the user's respiratory function in a high-dimensional space. The model is then trained and optimized using historical data, and model parameters are adjusted to enable the model to accurately assess the user's current respiratory function status and update the output multidimensional vector.

[0049] The program customization module is used to integrate the classified breathing exercise program library, match the user's breathing characteristics and health data through reinforcement learning algorithms, search and optimize basic templates, and generate personalized exercise programs;

[0050] The solution customization module is connected with the solution library construction unit, solution screening unit, template selection unit and solution generation unit;

[0051] A program library construction unit is used to collect and organize various breathing exercise programs, including basic exercise program templates for different respiratory function problems, different populations, and different exercise scenarios. Each program template details the exercise movements, movement essentials, exercise intensity, exercise duration, and breathing requirements. Each program template contains feature vectors consistent with the dimensions of the multidimensional vector, and the value of each dimension is calculated by big data mean.

[0052] The program screening unit is used to classify and label each program in the program library, facilitating subsequent rapid retrieval and matching based on user characteristics. For example, the program can be divided into ventilation function improvement and respiratory muscle strength enhancement categories according to respiratory function problem types; it can be divided into elderly population programs and children population programs according to population groups; it can be divided into home exercise programs and outdoor exercise programs according to exercise scenarios; at the same time, detailed labels are added to each program, including the disease type to which the program is applicable, such as asthma, chronic obstructive pulmonary disease, and the type of exercise movement, such as abdominal breathing, pursed lip breathing, and breathing exercises;

[0053] A template selection unit, configured to search and filter a set of candidate exercise program templates suitable for the user's current condition in a program library according to the user's multi-dimensional vector;

[0054] The cosine similarity between the user's multidimensional vector and the feature vector in each solution template is calculated using the following formula:

[0055] Where A is the user multidimensional vector; B is the feature vector of the solution template;

[0056] Set a matching threshold, compare the calculated cosine similarity with the threshold, and retain only the solution templates with a similarity greater than the threshold to obtain a preliminary screening template set and enter the subsequent screening process;

[0057] A program generation unit is used to further screen the selected candidate exercise program template set based on the user's multi-dimensional vector using a reinforcement learning algorithm to obtain the optimal exercise program template;

[0058] The reinforcement learning algorithm uses the Q-learning algorithm to update the strategy for selecting the training plan template. Through continuous trial and error and a reward mechanism, the optimal training plan template is obtained. Specifically, the algorithm randomly selects a basic training plan template from a set of candidate training plan templates as the initial strategy, asks the user to execute the plan, collects user feedback data, calculates the corresponding reward value, and records the user's current state, actions performed, rewards obtained, and the user's next state.

[0059] The Q value formula is as follows:

[0060] Q(s t ,a t )←Q(s t ,a t )+α[r t+1 +γ·max a Q(s t+1 ,a t )-Q(s t ,a t )];

[0061] Where s t is the current state; a t is the execution action of the current state; Q(s t ,a t ) is the Q value of executing action a in the current state s; α is the learning rate, which is used to control the step size of each update and has a value range of 0-1; r t+1 is the reward obtained in the next state after executing action a; γ is the discount factor, which is used to measure the importance of future rewards and has a value range of 0-1; max a Q(s t+1 ,a t ) is the maximum expected future payoff for the next state;

[0062] As the number of training increases, based on the updated Q value, after multiple rounds of training and screening, the program template with the highest Q value is finally determined as the personalized training program;

[0063] The real-time monitoring and feedback module collects respiratory rate, depth, and rhythm data in real time. After analysis using a sliding window algorithm, it dynamically adjusts personalized exercise plans and provides real-time feedback to guide user training.

[0064] A real-time monitoring feedback module is connected to a real-time monitoring unit, a real-time processing and analysis unit, and a solution adjustment unit;

[0065] The real-time monitoring unit is used to collect the breathing rate, breathing depth and breathing rhythm of the user at a fixed frequency during the user's exercise, and integrate them into real-time breathing data and upload them to the system in real time;

[0066] A real-time processing and analysis unit is used to analyze the real-time respiratory parameters using a sliding window algorithm and compare the data with the respiratory requirement information to obtain a determination result of the respiratory parameters. The method includes setting the window size to 5 data points and the sliding step size to 1 data point, performing statistical calculations on the respiratory parameter data within the window, and comparing the calculated statistical characteristics with a pre-set normal range threshold to determine whether the respiratory parameters are abnormal;

[0067] For example, data analysis is performed on the real-time respiratory rate, including calculating the average and standard deviation of the respiratory rate. The normal range for adults is generally 12-20 breaths / minute. If the average respiratory rate within the window is higher than 20 breaths / minute or lower than 12 breaths / minute, and the standard deviation is large, exceeding the normal fluctuation range, it is determined that the respiratory rate parameters within the window are abnormal; data is compared with the respiratory requirement information, including or if there is a large deviation from the respiratory rate required by the predetermined exercise program, it is determined that the respiratory rate is abnormal;

[0068] The program adjustment unit makes a decision on adjusting the training program based on the determination result and displays it to the user, the method comprising: classifying the abnormality into abnormal breathing rate, abnormal breathing depth and disordered breathing rhythm according to the determination result;

[0069] If there are abnormal breathing parameters, the training plan will be adjusted using an adaptive algorithm based on the type of abnormality, and the adjusted training plan will be updated and displayed to the user in a timely manner; if the parameters are normal, no operation will be performed;

[0070] For example, when the breathing rate is determined to be abnormal, the breathing resistance of the current exercise is reduced, such as lowering the resistance level of the breathing trainer; when the breathing depth does not reach the normal level or does not meet the target depth set in the exercise plan, the breathing depth is determined to be abnormal; when the breathing depth is determined to be abnormal, the duration of each exercise is shortened, and the original duration of each breathing exercise is adjusted from 5 seconds to 4 seconds, so that the user has more time to adjust the breathing rhythm; when the breathing rhythm is determined to be disordered, the original complex inhale-hold-exhale rhythm is temporarily changed to a simple inhale-exhale mode, that is, adjusted to 4 seconds inhale-6 seconds exhale, to reduce the difficulty of operating the breathing rhythm;

[0071] The system fully records all the data of each user's training, including the training start time, end time, training duration, the execution of each exercise, the real-time breathing parameter change curve, and the user's subjective feedback during the training process. This data is stored in a structured form in the user's exclusive data file to establish a long-term training data sequence.

[0072] Utilizing online learning algorithms in machine learning, such as the stochastic gradient descent algorithm (SGD), SGD continuously updates model parameters based on the user's training data and results, enabling the algorithm to better adapt to each user's individual differences and respiratory function changes.

[0073] The system regularly analyzes users' training data to evaluate the effects of different exercise programs on improving their respiratory function. Based on the analysis results, it adjusts the strategy for formulating subsequent training programs. If a certain exercise is found to significantly improve a user's respiratory function, the training ratio or difficulty of that exercise will be increased in subsequent programs. If a program causes discomfort or poor training results, the relevant exercises in the program will be reduced or replaced, achieving continuous optimization and personalized evolution of respiratory function training programs.

[0074] This invention uses advanced artificial intelligence algorithms to deeply analyze multi-dimensional data such as age, gender, health status, and respiratory history input by the user. Combined with real-time respiratory monitoring information, it tailors a unique breathing exercise program for the user, ensuring the pertinence and effectiveness of the training, breaking the limitations of traditional fixed programs and meeting the user's personalized needs.

[0075] During the exercise, the system continuously monitors the user's breathing rate, depth, rhythm and other parameters. Once an abnormality or deviation from the predetermined plan is detected, it will immediately give feedback through voice prompts, screen guidance, etc., and automatically adjust the training difficulty and rhythm to ensure that the exercise is safe and efficient. It can ensure the safety of exercise and improve the effect in a timely manner, and has significant advantages in the intelligent and scientific aspects of respiratory function training.

[0076] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A respiratory function training system based on artificial intelligence, characterized in that: It includes multi-dimensional data acquisition module, data pre-processing module, data analysis and feature extraction module, solution customization module and real-time monitoring feedback module; Multi-dimensional data collection module, used to collect user's basic information, health status information and respiratory function information; The data preprocessing module is used to vectorize the basic information, health status information and respiratory function information to obtain the corresponding basic vector, health status vector and respiratory function vector, normalize the vector components respectively, and splice the normalized vectors to obtain a multidimensional vector; The data analysis and feature extraction module first uses a convolutional neural network to extract key features from the multidimensional vector. Based on the extracted key features, it uses the support vector machine algorithm to build a user-customized respiratory function model, evaluate the user's current respiratory function status, and update the output multidimensional vector. The program customization module is used to integrate the classified breathing exercise program library, match the user's breathing characteristics and health data through reinforcement learning algorithms, search and optimize basic templates, and generate personalized exercise programs; The real-time monitoring and feedback module is used to collect breathing rate, depth and rhythm data in real time. After analysis by the sliding window algorithm, it dynamically adjusts the personalized exercise plan and provides real-time feedback to guide user training.

2. The artificial intelligence-based respiratory function training system according to claim 1, characterized in that: Basic information includes at least age, gender, height and weight; health status information includes at least respiratory diseases; The respiratory function information includes at least the current respiratory rate value, respiratory rate fluctuation amplitude, tidal volume, deep inspiratory volume, vital capacity, forced expiratory volume in one second and maximum ventilation volume.

3. The artificial intelligence-based respiratory function training system according to claim 1, characterized in that: The solution customization module is connected with the solution library construction unit, solution screening unit, template selection unit and solution generation unit; The program library construction unit is used to collect and organize various breathing exercise programs, including basic exercise program templates for different respiratory function problems, different populations, and different exercise scenarios; The solution screening unit is used to classify and label each solution in the solution library to facilitate subsequent rapid retrieval and matching based on user characteristics; A template selection unit, configured to search and filter a set of candidate exercise program templates suitable for the user's current condition in a program library according to the user's multi-dimensional vector; The program generation unit is used to further screen the selected candidate exercise program template set based on the user's multi-dimensional vector using a reinforcement learning algorithm to obtain the optimal exercise program template.

4. The artificial intelligence-based respiratory function training system according to claim 3, characterized in that: Each plan template details the exercise movements, movement essentials, exercise intensity, exercise duration, and breathing requirements; each plan template contains a feature vector consistent with the dimension of the multidimensional vector, and the value of each dimension is obtained by calculating the big data mean.

5. The artificial intelligence-based respiratory function training system according to claim 4, characterized in that: The cosine similarity between the user's multidimensional vector and the feature vector in each solution template is calculated using the following formula: Where A is the user multidimensional vector; B is the feature vector of the solution template; Set a matching threshold, compare the calculated cosine similarity with the threshold, and only retain the solution templates with a similarity greater than the threshold to obtain a preliminary screening template set and enter the subsequent screening process.

6. The artificial intelligence-based respiratory function training system according to claim 5, characterized in that: The reinforcement learning algorithm uses the Q-learning algorithm to update the strategy of selecting the solution template.

7. The artificial intelligence-based respiratory function training system according to claim 6, characterized in that: A real-time monitoring feedback module is connected to a real-time monitoring unit, a real-time processing and analysis unit, and a solution adjustment unit; The real-time monitoring unit is used to collect the breathing rate, breathing depth and breathing rhythm of the user at a fixed frequency during the user's exercise, and integrate them into real-time breathing data and upload them to the system in real time; A real-time processing and analysis unit is used to analyze the real-time respiratory parameters using a sliding window algorithm and compare the data with the respiratory requirement information to obtain a determination result of the respiratory parameters; The program adjustment unit makes decisions on adjusting the training program based on the judgment results and displays them to the user.

8. The artificial intelligence-based respiratory function training system according to claim 7, characterized in that: Adjusting the training program and making decisions based on the determination results include: classifying abnormal conditions into abnormal breathing rate, abnormal breathing depth, and disordered breathing rhythm according to the determination results; If there are abnormal breathing parameters, the training plan will be adjusted using an adaptive algorithm based on the type of abnormality, and the adjusted training plan will be updated and displayed to the user in a timely manner; if the parameters are normal, no operation will be performed.

Images