Intelligent training guiding device and method

Abstract

The invention discloses an intelligent training guiding device and method, and the device comprises a pressure signal collection module which is used for collecting a mechanical / displacement / angle change signal caused by the fluctuation of the chest or abdomen of a user in the breathing process; the signal synchronization module is used for acquiring various basic physiological signals of the user; the signal decoding module is used for calculating physiological signal derivative parameters of the user according to the mechanical / displacement / angle change signals and the basic physiological signals of the breathing process of the user; the physiological path simulation module is used for fusing the basic information related to the personal respiration of the user and the physiological signal derivative parameters into respiration physiological data, inputting the respiration physiological data into a digital twin neural network, and predicting physiological change paths under all possible respiration guide modes; the intelligent optimization and decision module is used for evaluating a plurality of physiological change paths and selecting an optimal physiological change path as an optimal training parameter; and the respiration guiding and adjusting module is used for guiding the user to adapt to the optimal training parameters. According to the invention, self-adaptive real-time guidance can be carried out on the user according to the real-time physiological condition of the user.

Description

Technical Field

[0001] The present invention relates to a breathing training technology, and in particular to an intelligent training guiding device and method. Background Art

[0002] As an important means of regulating physiological functions, relieving psychological stress and improving cardiopulmonary function, breathing training has a direct impact on training effectiveness and safety due to its scientific and standardized nature. With the deepening of medical research, breathing training has developed from simple rhythm control to a complex technical system involving multi-dimensional parameters such as breathing pattern, rate, and strength. Each breathing pattern (such as chest breathing, abdominal breathing, mixed breathing, etc.) is adapted to different physiques and scenarios, and has strict technical requirements for indicators such as core muscle coordination and breathing depth. Studies have shown that breathing training under non-professional guidance may lead to adverse reactions such as hyperventilation syndrome and strain of auxiliary respiratory muscles, and even aggravate the condition of patients with cardiopulmonary diseases. This places higher demands on real-time monitoring of the training process, efficient trial and error, intelligent decision-making, and dynamic guidance.

[0003] At the device development level, the system described in publication number CN202111296808 uses a split airflow monitoring and electrocardiogram (ECG) acquisition device. While this device can accurately capture respiratory rate and heart rate parameters, the facial design significantly impacts user comfort and freedom of daily activities. Another abdominal breathing training device, publication number CN202111044129, uses an abdominal strap to achieve standardized measurement of respiratory parameters, but its reliance on manually adjusted straps is cumbersome to operate.

[0004] At the methodological level, the cluster analysis algorithm proposed in publication number CN202410962308 can establish a personalized breathing parameter recommendation model. However, it quantifies the differences in breathing training patterns among users based on their historical breathing data and the differences in their changing physiological health indicators. Through cluster analysis, it further divides the corresponding breathing training patterns for different users. This algorithm aims to implement different breathing training patterns for different users, but it cannot provide adaptive, real-time training guidance tailored to the specific needs of individual users. Summary of the Invention

[0005] In view of the problems existing in the prior art, the purpose of the present invention is to provide an intelligent training guidance device and method that can adaptively guide a user in real time according to the user's real-time physiological condition.

[0006] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0007] An intelligent training guidance device, comprising:

[0008] The pressure signal acquisition module is used to collect the mechanical / displacement / angle change signals caused by the rise and fall of the chest or abdomen during the user's breathing;

[0009] Signal synchronization module, used to synchronize with smart wearable devices or medical devices to obtain multiple basic physiological signals of users;

[0010] A signal decoding module is used to calculate the user's physiological signal-derived parameters based on the mechanical / displacement / angle change signals and multiple basic physiological signals during the user's breathing process;

[0011] A physiological pathway simulation module is used to fuse the user's personal breathing-related basic information and physiological signal-derived parameters into respiratory physiological data, input it into a pre-built digital twin neural network, and predict the physiological change paths under all possible breathing guidance methods. The digital twin neural network decouples the spatial and temporal characteristics of the respiratory physiological data under each breathing guidance method, obtains the physiological changes in time and space, and forms a physiological change path;

[0012] The intelligent optimization and decision-making module is used to evaluate multiple physiological change paths predicted by the physiological path simulation module based on the medical knowledge database using a reinforcement learning algorithm, and select the optimal physiological change path as the optimal training parameter;

[0013] The breathing guidance adjustment module is used to guide the user to adapt to the determined optimal training parameters through various methods.

[0014] Furthermore, the signal decoding module specifically includes:

[0015] The attention index calculation unit is used to calculate the user's attention index according to the following formula based on the EEG signal fused with the physiological signal:

[0016]

[0017] Where q represents the qth signal segment after the fused physiological signal is divided into time segments, P EEG (f) is the power spectral density of the EEG signal at frequency f, AttentionIndex(q) is the attention index of the qth signal segment, which is a physiological signal-derived parameter and is calculated by the corresponding signal of the qth signal segment.

[0018] Furthermore, the signal decoding module further includes:

[0019] The sympathetic nerve activity evaluation unit is used to search for the peak points of the electrocardiogram signal in the fused physiological signal, form a heart rate sequence by chronologically separating the time intervals between all two adjacent peak points, and calculate the user's sympathetic nerve activity according to the following formula:

[0020]

[0021] SympatheticIndex(q)=w1·HRV LF +w2·ΔBP+w3·Resp-BP Coupling

[0022] Where, HRV LF is the low frequency component of heart rate, P HRV (f) is the power spectral density of the heart rate sequence at frequency f, ΔBP is the blood pressure variability index, BP n is a discrete blood pressure sequence, N is the total number of discrete points in the qth signal segment, is the average value of the blood pressure series, Resp-BP Coupling is the respiratory-blood pressure coupling strength, P(t) is the mechanical / displacement / angle change signal, BP(t+τ) is the blood pressure time series after the time delay τ, t q , t q-1 are the starting time of the q-1th and qth signal segments, respectively. q represents the qth signal segment after the basic physiological signal is divided into time segments. SympatheticIndex(q) is the sympathetic nerve activity of the qth signal segment, which is a physiological signal-derived parameter. w1, w2, and w2 are weight coefficients.

[0023] The vagus nerve activity evaluation unit is used to calculate the user's vagus nerve activity according to the following formula based on the heart rate sequence:

[0024]

[0025] VagalIndex(q)=m1·HRV HF +m2·Resp-HR Coherence

[0026] Where, HRV HF is the high-frequency component of heart rate, P(f) is the auto-power spectrum density of P(t) at frequency f, P HR (f) is the autopower spectral density of the heart rate time series at frequency f, P Resp,HR (f) is the cross-power spectral density between P(t) and the heart rate sequence at frequency f, Coherence(f) is the coherence calculation at frequency f, Resp-HR Coherence is the degree of coherence between P(t) and the heart rate time series, VagalIndex(q) is the vagus nerve activity of the qth signal segment, which is a physiological signal-derived parameter, and m1 and m2 are weight coefficients.

[0027] Furthermore, the signal decoding module further includes:

[0028] The breathing pattern judgment unit is used to obtain the ECG-derived breathing EDR signal by extracting the envelope of the ECG signal from the fused physiological signal, and judge the user's breathing pattern according to the following formula based on the EDR signal and the mechanical / displacement / angle change signals during the user's breathing process:

[0029]

[0030] Δφ(n)=|mod(φ EDR (n)-φ P (n)+π,2π)-π|∈[0,π]

[0031]

[0032] Where, is the imaginary part sequence of the EDR signal, s EDR (n) is the real part sequence of the EDR signal, φ EDR (n) is the phase sequence of the EDR signal, is the imaginary part sequence of the force / displacement / angle change signal, s P (n) is the real part sequence of the discrete mechanics / displacement / angle change signal P(n), φ P (n) is the phase sequence of the discrete mechanical / displacement / angle change signal P(n), P(n) is the discrete sequence after discrete sampling of the mechanical / displacement / angle change signal P(t), Δφ(n) is the phase difference sequence, is the mean value of Δφ(n), mod is the modulo operation, Resp Mode(q) is the breathing mode of the qth signal segment, which is a physiological signal derived parameter;

[0033] The breathing rate calculation unit is used to obtain the user's inhalation and exhalation rates based on the force / displacement / angle change signal P(t):

[0034]

[0035] Where, v insp (q) is the inhalation rate of the qth signal segment, v exp (q) is the exhalation rate of the qth signal segment, which is a physiological signal-derived parameter;

[0036] The respiratory rate calculation unit is used to search for all peak points of the mechanical / displacement / angle change signal P(t) and calculate the time intervals between all adjacent peak points to form the pressure time series T resp (k), get the user's average breathing frequency:

[0037]

[0038] Where, f resp(q) is the average respiratory frequency of the qth signal segment, which is a physiological signal-derived parameter, and K is the total number of points in the pressure time series;

[0039] The breathing depth calculation unit is used to analyze the fluctuation of the mechanical / displacement / angle change signal P(t) and calculate the user's breathing depth:

[0040]

[0041] Where Depth(q) is the respiratory depth of the qth signal segment, which is a physiological signal derived parameter, and P baseline (t) is the baseline of the force / displacement / angle change signal P(t), t q , t q-1 is the starting time of the qth and q-1th signal segments;

[0042] The breathing intensity calculation unit is used to analyze the amplitude changes of the discrete mechanics / displacement / angle change signal P(n) and calculate the user's breathing intensity:

[0043]

[0044] Where, P baseline (n) is the baseline of the discrete mechanical / displacement / angle change signal P(n), N is the total number of discrete points in the qth signal segment, and S(q) is the respiratory intensity of the qth signal segment, which is a physiological signal-derived parameter;

[0045] The ventilation efficiency calculation unit is used to calculate the peak-to-peak value of the EDR signal and the cuff pressure signal to obtain the user's ventilation efficiency:

[0046]

[0047] Where PK is the number of waveforms of the EDR signal and the force / displacement / angle change signal P(t) that are time-aligned in the qth signal segment, and EDR amp (k) is the peak-to-peak value of the kth corresponding EDR signal, P amp (k) is the peak-to-peak value of the k-th corresponding mechanical / displacement / angle change signal P(t), and Ventilation Efficiency (q) is the ventilation efficiency of the q-th signal segment, which is a physiological signal-derived parameter.

[0048] Furthermore, the signal decoding module further includes:

[0049] The blood flow velocity calculation unit is used to obtain the user's blood flow velocity by calculating the time interval between the electrocardiogram signal in the fused physiological signal and the pulse wave signal in the fused physiological signal:

[0050]

[0051] Where Δt ECG-to-PPG (p) is the time interval between the pth ECG signal waveform and the pulse wave signal waveform, P is the number of ECG signal and pulse wave signal waveforms that are time-aligned in the qth signal segment, v blood (q) is the blood flow velocity of the qth signal segment, which is a physiological signal derived parameter;

[0052] The vascular elasticity calculation unit is used to calculate the user's vascular elasticity by identifying the systolic and diastolic peaks of the pulse wave signal:

[0053]

[0054] Where, P Systolic Peak (p) is the pth contraction peak identified in the pulse wave signal within the qth signal segment, P Dicrotic Peak (p) is the pth diastolic peak identified in the pulse wave signal within the signal segment, ElasticityIndex(q) is the vascular elasticity of the qth signal segment, which is a physiological signal-derived parameter;

[0055] The oxygenation capacity evaluation unit is used to calculate the user's oxygenation capacity by fusing the blood oxygen in the physiological signal and the mechanical / displacement / angle change signal P(t) in the physiological signal:

[0056]

[0057] Where ΔSpO2 is the fluctuation level of blood oxygen, Oxygenation Efficiency (q) is the oxygenation capacity of the qth signal segment, which is a physiological signal-derived parameter, and t q , t q-1 is the starting time of the qth and q-1th signal segments, f resp is the average respiratory frequency of the qth signal segment.

[0058] Furthermore, the physiological pathway simulation module specifically includes:

[0059] a data range setting unit, configured to set a value range of a breathing guidance mode, wherein the breathing guidance mode includes a guided breathing pattern, a guided breathing rate, a guided breathing frequency, a guided inhalation-exhalation ratio, and a guided breathing intensity;

[0060] The action path construction unit is used to traverse the value range and construct the action path of the Q+1th signal segment:

[0061] E m (Q+1)=[κ m (Q+1),ν m (Q+1),f m(Q+1),ρ m (Q+1),∈ m (Q+1)],m=1,…,M

[0062] Where Q is the total number of current signal segments, E m (Q+1) is the m-th action path of the Q+1-th signal segment, i.e., the m-th guidance method, κ m (Q+1),ν m (Q+1),f m (Q+1),ρ m (Q+1),∈ m (Q+1) represents the guided breathing pattern, guided breathing rate, guided breathing frequency, guided inspiration-expiration ratio, and guided breathing intensity in the m-th action path of the Q+1-th signal segment, respectively, which are obtained by traversing the corresponding value range. M is the total number of action paths.

[0063] The spatiotemporal decoupling generator unit is used to take respiratory physiological data formed by various physiological signal-derived parameters and basic information related to the user's breathing as input based on a pre-built digital twin neural network, decouple the temporal and spatial features of the input data, and predict the physiological change path of the user under different motion paths. The digital twin neural network is specifically:

[0064]

[0065] Y m (Q+1)=Sigmoid(W out ·F m (Q+1)),m=1,…,M

[0066] Where q represents the qth signal segment, Y m (Q+1) is the predicted physiological change path caused by the m-th action path of the Q+1-th signal segment, Sigmoid() represents the sigmoid constraint function, tanh() is the tanh constraint function, and W out is the output weight parameter matrix, F m (Q+1) is the set of physiological change paths caused by the predicted m-th action path of the Q+1-th signal segment, W s is the spatial convolution kernel weight matrix, W p is the spatial parameter weight matrix, W t is the temporal convolution kernel weight matrix, W f is the time parameter weight matrix, is the dilated convolution operation, * is the convolution operation, ⊙ is the Hadamard product operation, φ() is the Fourier feature map, and X(Q) is the user's respiratory physiological data including the Q-th signal segment.

[0067] Furthermore, the intelligent optimization and decision-making module specifically includes:

[0068] The physiological reinforcement reward unit is used to calculate the physiological reinforcement reward score for the physiological change path caused by each action path through a solver constructed by physiological reinforcement constraints, wherein the solver constructed by physiological reinforcement constraints is specifically:

[0069]

[0070] Where PI i is the importance ratio of the ith physiological signal derived parameter under the physiological reinforcement constraint, IN is the number of physiological signal derived parameters, is the reinforced rule discriminant space of the ith physiological signal derived parameter constructed based on the medical knowledge base, is the physiological change path Y caused by the predicted m-th action path of the Q+1-th signal segment m (Q+1) is the physiological reinforcement reward score, M is the total number of action paths;

[0071] The safety supervision unit is used to calculate the safety reinforcement reward score for the physiological change path caused by each action path through a solver constructed by safety constraints:

[0072]

[0073] Where, SI i is the importance ratio of the i-th physiological signal derived parameter under safety constraints, is the safety rule discriminant space of the ith physiological signal derived parameter constructed based on the medical knowledge base, Y m (Q+1) security enhancement bonus score;

[0074] The comfort constraint unit is used to calculate the comfort enhancement reward score for the physiological change path caused by each action path through a solver constructed by safety constraints:

[0075]

[0076] Where, CI i is the importance ratio of the i-th physiological signal derived parameter under comfort constraint, M Comforti is the comfort rule discriminant space of the ith physiological signal derived parameter constructed based on the medical knowledge base, Y m (Q+1) comfort enhancement bonus score;

[0077] The comprehensive decision-making unit is used to integrate the physiological reinforcement reward score, safety reinforcement reward score, and comfort reinforcement reward score, and determine the optimal training parameters based on the following policy function constructed through reinforcement learning:

[0078]

[0079] In the formula, γ1, γ2, and γ3 are the adjustment coefficients of each constraint condition, λ m (Q+1) is E m The comprehensive score of (Q+1), which is a learnable strategy function, m * is the index number of the optimal action path solved, The mth segment of the Q+1th signal * action path, Opt(Q+1) is the optimal training parameter for the predicted Q+1th signal segment.

[0080] Furthermore, the breathing guidance adjustment module is specifically an inflatable strap, and the inflatable strap is provided with a communication unit, a speaker unit, a vibration motor unit, an airbag and an air pump unit, and a pressure detection unit. The communication unit is used to notify a smart device with a display function to play a breathing action guidance video that guides the user to train with optimal training parameters. The speaker unit is used to play a breathing action guidance audio that guides the user to train with optimal training parameters. The vibration motor unit is used to guide the user's breathing rhythm to approach the optimal training parameters. The airbag and air pump unit are used to temporarily store gas, inflate and deflate. The pressure detection unit is used to determine whether the pressure generated on the user at the current inflation level meets the optimal training parameter requirements.

[0081] Furthermore, the breathing guidance adjustment module is specifically a retractable strap, and the retractable strap is provided with a communication unit, a speaker unit, a vibration motor unit, a mechanical retractable motor unit, and a pressure detection unit. The communication unit notifies the smart device with a display function to play a breathing action guidance video that guides the user to train with the optimal training parameters. The speaker unit is used to play the guidance audio of the breathing action that guides the user to train with the optimal training parameters. The vibration motor unit is used to guide the user's breathing rhythm to approach the optimal training parameters. The mechanical retractable motor unit is used to adjust the length of the strap, and the pressure detection unit is used to determine whether the pressure generated on the user at the current inflation level meets the optimal training parameter requirements.

[0082] An intelligent training guidance method, comprising:

[0083] Collect the mechanical / displacement / angle change signals caused by the rise and fall of the chest or abdomen during the user's breathing;

[0084] Synchronize with smart wearable devices or medical devices to obtain multiple basic physiological signals of users;

[0085] Calculate the user's physiological signal-derived parameters based on the user's breathing process's mechanical / displacement / angle change signals and multiple basic physiological signals;

[0086] The user's personal breathing-related basic information and physiological signal-derived parameters are integrated into respiratory physiological data, which is input into a pre-built digital twin neural network to predict the physiological change paths under all possible breathing guidance methods. The digital twin neural network decouples the spatial and temporal features of the respiratory physiological data under all possible breathing guidance methods, obtains the physiological changes in time and space, and forms the physiological change path;

[0087] Based on the medical knowledge database, multiple physiological change paths predicted by the physiological path simulation module are evaluated, and the optimal physiological change path is selected as the optimal training parameter;

[0088] Guide users to adapt to the determined optimal training parameters through various means.

[0089] Compared with the prior art, the present invention has the following beneficial effects:

[0090] (1) On the technical level, the present invention realizes the calculation of multiple derived physiological parameters in breathing training by detecting the pressure changes caused by the breathing fluctuations of the chest and abdomen, and synchronizing with other signal acquisition equipment, so as to make the real-time dynamic monitoring of individual physiological changes more accurate; it uses digital twin technology to quickly construct multiple pre-trained physiological change paths, and efficiently tests the training plan, reducing the user's trial and error time and energy cost; it combines the medical knowledge database and reinforcement learning technology to enable the generated breathing training plan to achieve accurate, safe and comfortable optimization; based on the innovation of the mechanical structure, a dynamic guidance method is proposed, which can combine vision, hearing, vibration, adaptive voltage transformation and other methods to guide users to adapt personalized breathing training parameters and realize effective adjustment of breathing pattern, rate, frequency, inhalation-exhalation ratio and strength; finally, it realizes the adaptive generation of optimal training parameters for current data based on the real-time collected data, and guides users to perform breathing training;

[0091] (2) From an economic perspective, the present invention has low requirements on hardware and sensor types and is suitable for low-cost, high-efficiency breathing training guidance. Due to the built-in synchronization module, it can also be connected to other smart wearable devices or medical devices to maximize resource utilization, effectively reducing the cost of individuals achieving efficient breathing training and avoiding the waste of health resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] Figure 1 1 is a schematic diagram of a module of an intelligent training guidance device provided by an embodiment of the present invention;

[0093] Figure 22 are two structural schematic diagrams of the breathing guidance adjustment module provided in embodiments of the present invention;

[0094] Figure 3 It is a flowchart of the intelligent training guidance method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0095] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0096] Example 1

[0097] The embodiment of the present invention provides an intelligent training guidance device, such as Figure 1 Shown, including:

[0098] The pressure signal acquisition module is used to collect the mechanical / displacement / angle change signals caused by the rise and fall of the chest or abdomen during the user's breathing;

[0099] Signal synchronization module, used to synchronize with smart wearable devices or medical devices to obtain multiple basic physiological signals of users;

[0100] A signal decoding module is used to calculate the user's physiological signal-derived parameters based on the mechanical / displacement / angle change signals and multiple basic physiological signals during the user's breathing process;

[0101] A physiological pathway simulation module is used to fuse the user's personal breathing-related basic information and physiological signal-derived parameters into respiratory physiological data, input it into a pre-built digital twin neural network, and predict the physiological change paths under all possible breathing guidance methods. The digital twin neural network decouples the spatial and temporal characteristics of the respiratory physiological data under each breathing guidance method, obtains the physiological changes in time and space, and forms a physiological change path;

[0102] The intelligent optimization and decision-making module, based on the medical knowledge database, uses a reinforcement learning algorithm to evaluate multiple physiological change pathways predicted by the physiological pathway simulation module and selects the optimal physiological change pathway as the optimal training parameter;

[0103] The breathing guidance and adjustment module guides users to adapt to the optimal training parameters through various methods, including vision, hearing, vibration, and adaptive voltage variation.

[0104] In specific implementation, Figure 2As shown, the signal acquisition module is specifically a smart bandage, which is equipped with a pressure sensor or an acceleration sensor. The pressure sensor or acceleration sensor is used to detect the degree of fit between the bandage and the body, as well as the mechanical / displacement / angle changes caused by the rise and fall of the chest or abdomen during breathing. The smart bandage material can be plastic and elastic materials, metal materials, fiber materials, or any other new or non-new materials that can be used on the human body and have the ability to carry flexible electrodes and pressure sensors / accelerometers. The pressure sensor / accelerometer can be a device that converts pressure signals / acceleration signals into electrical signals based on any working principle or any material.

[0105] The signal synchronization module is compatible with wireless communication protocols such as Bluetooth, Wi-Fi, and Zigbee, as well as wired or dedicated communication protocols such as USB, UART, and ANT+. It can transmit data through protocols specified by other wearable devices or medical devices to obtain basic physiological signals such as the user's electrocardiogram, electroencephalogram, pulse wave, blood pressure, blood oxygen, and body temperature. The wireless communication protocol compatible with the signal synchronization module can be any new or existing wireless data transmission protocol, and the wired communication protocol can be any new or existing wired data transmission protocol. The signal synchronization module can transmit data with any wearable device or medical device. The acquired ECG can be single-lead or multi-lead, invasive or non-invasive, and obtained by new or non-new measurement methods; the acquired EEG can be single-lead or multi-lead, invasive or non-invasive, and obtained by new or non-new measurement methods; the pulse wave can be single-channel or multi-channel, invasive or non-invasive, and its principle can be piezoelectricity, photoplethysmography, ultrasonic Doppler or other new or non-new measurement methods; blood pressure can be invasive or non-invasive, directly measured or indirectly measured, and new or non-new measurement methods; blood oxygen can be a signal obtained by photoplethysmography, near-infrared spectroscopy, invasive blood oxygen measurement or any other new or non-new measurement methods; any other physiological signal or any required signal collected by any other wearable device or medical device can be synchronized through this module.

[0106] The signal decoding module is specifically used to calculate 12 physiological signal-derived parameters, including attention index, sympathetic nerve activity, vagus nerve activity, breathing pattern, respiratory rate, respiratory frequency, respiratory depth, respiratory force, ventilation efficiency, blood flow velocity, vascular elasticity, and oxygenation capacity, and includes a calculation unit for each physiological signal-derived parameter.

[0107] The attention index calculation unit is used to calculate the user's attention index according to the following formula based on the EEG signal fused with the physiological signal:

[0108]

[0109] Where q represents the qth signal segment after the fused physiological signal is divided into time segments, P EEG (f) is the power spectral density of the EEG signal at frequency f, AttentionIndex(q) is the attention index of the qth signal segment, which is a physiological signal-derived parameter and is calculated by the corresponding signal of the qth signal segment.

[0110] The sympathetic nerve activity evaluation unit is used to search for the peak points of the electrocardiogram signal in the fused physiological signal, form a heart rate sequence by chronologically separating the time intervals between all two adjacent peak points, and calculate the user's sympathetic nerve activity according to the following formula:

[0111]

[0112] SympatheticIndex(q)=w1·HRV LF +w2·ΔBP+w3·Resp-BP Coupling

[0113] Where, HRV LF is the low frequency component of heart rate, P HRV (f) is the power spectral density of the heart rate sequence at frequency f, ΔBP is the blood pressure variability index, BP n is a discrete blood pressure sequence, N is the total number of discrete points in the qth signal segment, is the average value of the blood pressure series, Resp-BP Coupling is the respiratory-blood pressure coupling strength, P(t) is the mechanical / displacement / angle change signal, BP(t+τ) is the blood pressure time series after the time delay τ, t q , t q-1 are the starting time of the q-1th and qth signal segments, respectively. q represents the qth signal segment after the basic physiological signal is divided into time segments. SympatheticIndex(q) is the sympathetic nerve activity of the qth signal segment, which is a physiological signal-derived parameter. w1, w2, and w2 are weight coefficients.

[0114] The vagus nerve activity evaluation unit is used to calculate the user's vagus nerve activity according to the following formula based on the heart rate sequence:

[0115]

[0116] VagalIndex(q)=m1·HRV HF +m2·Resp-HR Coherence

[0117] Where, HRV HF is the high-frequency component of heart rate, P(f) is the auto-power spectrum density of P(t) at frequency f, P HR(f) is the autopower spectral density of the heart rate time series at frequency f, P Resp,HR (f) is the cross-power spectral density between P(t) and the heart rate sequence at frequency f, Coherence(f) is the coherence calculation at frequency f, Resp-HR Coherence is the degree of coherence between P(t) and the heart rate time series, VagalIndex(q) is the vagus nerve activity of the qth signal segment, which is a physiological signal-derived parameter, and m1 and m2 are weight coefficients.

[0118] The breathing pattern judgment unit is used to obtain the ECG-derived breathing EDR signal by extracting the envelope of the ECG signal from the fused physiological signal, and judge the user's breathing pattern according to the following formula based on the EDR signal and the mechanical / displacement / angle change signals during the user's breathing process:

[0119]

[0120] Δφ(n)=|mod(φ EDR (n)-φ P (n)+π,2π)-π|∈[0,π]

[0121]

[0122] Where, is the imaginary part sequence of the EDR signal, s EDR (n) is the real part sequence of the EDR signal, φ EDR (n) is the phase sequence of the EDR signal, is the imaginary part sequence of the force / displacement / angle change signal, s P (n) is the real part sequence of the discrete mechanics / displacement / angle change signal P(n), φ P (n) is the phase sequence of the discrete mechanical / displacement / angle change signal P(n), P(n) is the discrete sequence after discrete sampling of the mechanical / displacement / angle change signal P(t), Δφ(n) is the phase difference sequence, is the mean value of Δφ(n), mod is the modulo operation, Resp Mode(q) is the breathing mode of the qth signal segment, which is a physiological signal derived parameter;

[0123] The breathing rate calculation unit is used to obtain the user's inhalation and exhalation rates based on the force / displacement / angle change signal P(t):

[0124]

[0125] Where, v insp (q) is the inhalation rate of the qth signal segment, v exp (q) is the exhalation rate of the qth signal segment, which is a physiological signal-derived parameter;

[0126] The respiratory rate calculation unit is used to search for all peak points of the mechanical / displacement / angle change signal P(t) and calculate the time intervals between all adjacent peak points to form the pressure time series T resp (k), get the user's average breathing frequency:

[0127]

[0128] Where, f resp (q) is the average respiratory frequency of the qth signal segment, which is a physiological signal-derived parameter, and K is the total number of points in the pressure time series;

[0129] The breathing depth calculation unit is used to analyze the fluctuation of the mechanical / displacement / angle change signal P(t) and calculate the user's breathing depth:

[0130]

[0131] Where Depth(q) is the respiratory depth of the qth signal segment, which is a physiological signal derived parameter, and P baseline (t) is the baseline of the force / displacement / angle change signal P(t), t q , t q-1 is the starting time of the qth and q-1th signal segments;

[0132] The breathing intensity calculation unit is used to analyze the amplitude changes of the discrete mechanics / displacement / angle change signal P(n) and calculate the user's breathing intensity:

[0133]

[0134] Where, P baseline (n) is the baseline of the discrete mechanical / displacement / angle change signal P(n), N is the total number of discrete points in the qth signal segment, and S(q) is the respiratory intensity of the qth signal segment, which is a physiological signal-derived parameter;

[0135] The ventilation efficiency calculation unit is used to calculate the peak-to-peak value of the EDR signal and the cuff pressure signal to obtain the user's ventilation efficiency:

[0136]

[0137] Where PK is the number of waveforms of the EDR signal and the force / displacement / angle change signal P(t) that are time-aligned in the qth signal segment, and EDR amp (k) is the peak-to-peak value of the kth corresponding EDR signal, P amp(k) is the peak-to-peak value of the k-th corresponding mechanical / displacement / angle change signal P(t), and Ventilation Efficiency (q) is the ventilation efficiency of the q-th signal segment, which is a physiological signal-derived parameter.

[0138] The blood flow velocity calculation unit is used to obtain the user's blood flow velocity by calculating the time interval between the electrocardiogram signal in the fused physiological signal and the pulse wave signal in the fused physiological signal:

[0139]

[0140] Where Δt ECG-to-PPG (p) is the time interval between the pth ECG signal waveform and the pulse wave signal waveform, P is the number of ECG signal and pulse wave signal waveforms that are time-aligned in the qth signal segment, v blood (q) is the blood flow velocity of the qth signal segment, which is a physiological signal derived parameter;

[0141] The vascular elasticity calculation unit is used to calculate the user's vascular elasticity by identifying the systolic and diastolic peaks of the pulse wave signal:

[0142]

[0143] Where, P Systolic Peak (p) is the pth contraction peak identified in the pulse wave signal within the qth signal segment, P Dicrotic Peak (p) is the pth diastolic peak identified in the pulse wave signal within the signal segment, ElasticityIndex(q) is the vascular elasticity of the qth signal segment, which is a physiological signal-derived parameter;

[0144] The oxygenation capacity evaluation unit is used to calculate the user's oxygenation capacity by fusing the blood oxygen in the physiological signal and the mechanical / displacement / angle change signal P(t) in the physiological signal:

[0145]

[0146] Where ΔSpO2 is the fluctuation level of blood oxygen, Oxygenation Efficiency (q) is the oxygenation capacity of the qth signal segment, which is a physiological signal-derived parameter, and t q , t q-1 is the starting time of the qth and q-1th signal segments, f resp is the average respiratory frequency of the qth signal segment.

[0147] The physiological pathway simulation module specifically includes:

[0148] a data range setting unit, configured to set a value range of a breathing guidance mode, wherein the breathing guidance mode includes a guided breathing pattern, a guided breathing rate, a guided breathing frequency, a guided inhalation-exhalation ratio, and a guided breathing intensity;

[0149] The action path construction unit is used to traverse the value range and construct the action path of the Q+1th signal segment:

[0150] E m (Q+1)=[κ m (Q+1),ν m (Q+1),f m (Q+1),ρ m (Q+1),∈ m (Q+1)],m=1,…,M

[0151] Where Q is the total number of current signal segments, E m (Q+1) is the m-th action path of the Q+1-th signal segment, i.e., the m-th guidance method, κ m (Q+1),ν m (Q+1),f m (Q+1),ρ m (Q+1),∈ m (Q+1) represents the guided breathing pattern, guided breathing rate, guided breathing frequency, guided inspiration-expiration ratio, and guided breathing intensity in the m-th action path of the Q+1-th signal segment, respectively, which are obtained by traversing the corresponding value range. M is the total number of action paths.

[0152] The spatiotemporal decoupling generator unit is used to take respiratory physiological data formed by various physiological signal-derived parameters and basic information related to the user's breathing as input based on a pre-built digital twin neural network, decouple the temporal and spatial features of the input data, and predict the physiological change path of the user under different motion paths. The digital twin neural network is specifically:

[0153]

[0154] Y m (Q+1)=Sigmoid(W out ·F m (Q+1)),m=1,…,M

[0155] Where q represents the qth signal segment, Y m (Q+1) is the predicted physiological change path caused by the m-th action path of the Q+1-th signal segment, Sigmoid() represents the sigmoid constraint function, tanh() is the tanh constraint function, and W out is the output weight parameter matrix, F m(Q+1) is the set of physiological change paths caused by the predicted m-th action path of the Q+1-th signal segment, W s is the spatial convolution kernel weight matrix, W p is the spatial parameter weight matrix, W t is the temporal convolution kernel weight matrix, W f is the time parameter weight matrix, is the dilated convolution operation, * is the convolution operation, ⊙ is the Hadamard product operation, φ() is the Fourier feature map, and X(Q) is the user's respiratory physiological data including the Q-th signal segment.

[0156]

[0157] Where PersonalProfile is the user's basic breathing information, including height, weight, age, vital capacity, and other basic personal information, and Temp(q) is the user's body temperature.

[0158] The intelligent optimization and decision-making module specifically includes:

[0159] The physiological reinforcement reward unit is used to calculate the physiological reinforcement reward score for the physiological change path caused by each action path through a solver constructed by physiological reinforcement constraints, wherein the solver constructed by physiological reinforcement constraints is specifically:

[0160]

[0161] Where PI i is the importance ratio of the i-th physiological signal-derived parameter under the physiological reinforcement constraint, which is calculated by the above-mentioned signal decoding module, including: attention index, sympathetic nerve activity, vagus nerve activity, breathing pattern, breathing rate, breathing frequency, breathing depth, breathing force, ventilation efficiency, blood flow velocity, vascular elasticity, and oxygenation capacity, a total of 12 types. The reinforcement rule discriminant space of the i-th physiological signal derived parameter is constructed based on the medical knowledge base. It is obtained by extracting explicit threshold rules related to physiological reinforcement in clinical guidelines through pre-trained language models such as BioBERT or ClinicalBERT. is the physiological change path Y caused by the predicted m-th action path of the Q+1-th signal segment m (Q+1) is the physiological reinforcement reward score, M is the total number of action paths;

[0162] The safety supervision unit is used to calculate the safety reinforcement reward score for the physiological change path caused by each action path through a solver constructed by safety constraints:

[0163]

[0164] Where, SI i is the importance ratio of the i-th physiological signal derived parameter under safety constraints, The safety rule discriminant space of the i-th physiological signal-derived parameter is constructed based on the medical knowledge base. It is obtained by extracting explicit threshold rules related to safety in clinical guidelines through pre-trained language models such as BioBERT or ClinicalBERT. Y m (Q+1) security enhancement bonus score;

[0165] The comfort constraint unit is used to calculate the comfort enhancement reward score for the physiological change path caused by each action path through a solver constructed by safety constraints:

[0166]

[0167] Where, CI i is the importance ratio of the i-th physiological signal derived parameter under comfort constraint, The comfort rule discriminant space of the i-th physiological signal-derived parameter is constructed based on the medical knowledge base. It is obtained by extracting explicit threshold rules related to comfort in clinical guidelines through pre-trained language models such as BioBERT or ClinicalBERT. Y m (Q+1) comfort enhancement bonus score;

[0168] The comprehensive decision-making unit is used to integrate the physiological reinforcement reward score, safety reinforcement reward score, and comfort reinforcement reward score, and determine the optimal training parameters based on the following policy function constructed through reinforcement learning:

[0169]

[0170] In the formula, γ1, γ2, and γ3 are the adjustment coefficients of each constraint condition, λ m (Q+1) is E m The comprehensive score of (Q+1), which is a learnable strategy function, m * is the index number of the optimal action path solved, The mth segment of the Q+1th signal * action path, Opt(Q+1) is the optimal training parameter for the predicted Q+1th signal segment.

[0171] The breathing guidance adjustment module can be specifically an inflatable strap, such as Figure 2As shown, the inflatable strap is provided with a communication unit, a speaker unit, a vibration motor unit, an air bag and an air pump unit, and a pressure detection unit. The communication unit is used to notify a smart device with a display function to play a breathing action guidance video that guides the user to train with the optimal training parameters. The speaker unit is used to play a breathing action guidance audio that guides the user to train with the optimal training parameters. The vibration motor unit is used to guide the user's breathing rhythm to approach the optimal training parameters. The air bag and the air pump unit are used to temporarily store gas, inflate and deflate. The pressure detection unit is used to determine whether the pressure generated on the user at the current inflation level meets the requirements of the optimal training parameters.

[0172] In other embodiments, the breathing guide adjustment module can also be a retractable strap, such as Figure 2 As shown, the retractable strap is provided with a communication unit, a speaker unit, a vibration motor unit, a mechanical retractable motor unit, and a pressure detection unit. The communication unit notifies a smart device with a display function to play a breathing action guidance video that guides the user to train with optimal training parameters. The speaker unit is used to play a breathing action guidance audio that guides the user to train with optimal training parameters. The vibration motor unit is used to guide the user's breathing rhythm to approach the optimal training parameters. The mechanical retractable motor unit is used to adjust the length of the strap. The pressure detection unit is used to determine whether the pressure generated on the user at the current inflation level meets the optimal training parameter requirements.

[0173] Those skilled in the art will appreciate that, in other embodiments, the basic physiological signal may be any one or more of electrocardiogram, electroencephalogram, pulse wave, blood pressure, blood oxygen, body temperature, etc., and the physiological signal-derived parameter may also be any one or more of attention index, sympathetic nerve activity, vagus nerve activity, breathing pattern, respiratory rate, respiratory frequency, respiratory depth, respiratory force, ventilation efficiency, blood flow velocity, vascular elasticity, and oxygenation capacity. When there are fewer physiological signals and fewer physiological signal-derived parameters, the generated training parameters may be offset. However, because the mechanical / displacement / angle change signals during the breathing process are included, the basic training function can still be achieved. However, the optimal choice is to include all parameters.

[0174] In other embodiments, the digital twin neural network may also be other types of neural networks, such as deep neural networks. The intelligent optimization and decision-making module may also select the optimal training parameters in other ways, such as calculating rewards based on any one or more of the physiological enhancement reward score, safety enhancement reward score, and comfort enhancement reward score, selecting the optimal action path, and thus obtaining the optimal training parameters.

[0175] It is worth noting that in the embodiment of the above-mentioned device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0176] The embodiments described above are merely illustrative, wherein the modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, i.e., they may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Those skilled in the art will readily appreciate that each embodiment may be implemented using software plus a necessary general-purpose hardware platform, or may be implemented solely through hardware, as long as the functionality or effect can be achieved.

[0177] Example 2

[0178] The embodiment of the present invention provides an intelligent training guidance method, such as Figure 3 Shown, including:

[0179] S1. Collecting mechanical / displacement / angle change signals caused by the rise and fall of the chest or abdomen during the user's breathing process. This can be achieved based on piezoelectric sensors, accelerometers, gyroscopes, strain gauges, electromagnetic wave radars, capacitive flexible sensors, and optical sensors.

[0180] S2, synchronize with smart wearable devices or medical devices to obtain multiple basic physiological signals of users;

[0181] S3. Calculating the user's physiological signal-derived parameters based on the user's breathing process mechanical / displacement / angle change signals and multiple basic physiological signals;

[0182] S4. Fusion of the user's personal breathing-related basic information and physiological signal-derived parameters into respiratory physiological data, which is input into a pre-built digital twin neural network to predict physiological change paths under all possible breathing guidance methods. The digital twin neural network decouples the spatial and temporal features of the respiratory physiological data under each breathing guidance method, obtains physiological changes in time and space, and forms a physiological change path.

[0183] S5. Based on the medical knowledge database, a reinforcement learning algorithm is used to evaluate multiple physiological change paths predicted by the physiological path simulation module, and the optimal physiological change path is selected as the optimal training parameter;

[0184] S6. Use various methods to guide users to train according to the optimal training parameters.

[0185] Wherein, step S3 specifically includes:

[0186] S301. Calculate the user's attention index according to the following formula based on the EEG signal fused with the physiological signal:

[0187]

[0188] Where q represents the qth signal segment after the fused physiological signal is divided into time segments, P EEG (f) is the power spectral density of the EEG signal at frequency f, AttentionIndex(q) is the attention index of the qth signal segment, which is a physiological signal-derived parameter and is calculated by the corresponding signal of the qth signal segment.

[0189] S302: Search for the peak points of the ECG signal in the fused physiological signal, form a heart rate sequence by chronologically separating the time intervals between all two adjacent peak points, and calculate the user's sympathetic nerve activity according to the following formula:

[0190]

[0191] SympatheticIndex(q)=w1·HRV LF +w2·ΔBP+w3·Resp-BP Coupling

[0192] Where, HRV LF is the low frequency component of heart rate, P HRV (f) is the power spectral density of the heart rate sequence at frequency f, ΔBP is the blood pressure variability index, BP n is a discrete blood pressure sequence, N is the total number of discrete points in the qth signal segment, is the average value of the blood pressure series, Resp-BP Coupling is the respiratory-blood pressure coupling strength, P(t) is the mechanical / displacement / angle change signal, BP(t+τ) is the blood pressure time series after the time delay τ, t q , t q-1 are the starting time of the q-1th and qth signal segments, respectively. q represents the qth signal segment after the basic physiological signal is divided into time segments. SympatheticIndex(q) is the sympathetic nerve activity of the qth signal segment, which is a physiological signal-derived parameter. w1, w2, and w2 are weight coefficients.

[0193] S303. Calculate the user's vagus nerve activity according to the following formula based on the heart rate sequence:

[0194]

[0195] VagalIndex(q)=m1·HRVHF +m2·Resp-HR Coherence

[0196] Where, HRV HF is the high-frequency component of heart rate, P(f) is the auto-power spectrum density of P(t) at frequency f, P HR (f) is the autopower spectral density of the heart rate time series at frequency f, P Resp,HR (f) is the cross-power spectral density between P(t) and the heart rate sequence at frequency f, Coherence(f) is the coherence calculation at frequency f, Resp-HR Coherence is the degree of coherence between P(t) and the heart rate time series, VagalIndex(q) is the vagus nerve activity of the qth signal segment, which is a physiological signal-derived parameter, and m1 and m2 are weight coefficients.

[0197] S304: Extract the envelope of the ECG signal from the fused physiological signal to obtain the ECG-derived respiration EDR signal, and determine the user's breathing pattern according to the following formula based on the EDR signal and the mechanical / displacement / angle change signals during the user's breathing process:

[0198]

[0199] Δφ(n)=|mod(φ EDR (n)-φ P (n)+π,2π)-π|∈[0,π]

[0200]

[0201] Where, is the imaginary part sequence of the EDR signal, s EDR (n) is the real part sequence of the EDR signal, φ EDR (n) is the phase sequence of the EDR signal, is the imaginary part sequence of the force / displacement / angle change signal, s P (n) is the real part sequence of the discrete mechanics / displacement / angle change signal P(n), φ P (n) is the phase sequence of the discrete mechanical / displacement / angle change signal P(n), P(n) is the discrete sequence after discrete sampling of the mechanical / displacement / angle change signal P(t), Δφ(n) is the phase difference sequence, is the mean value of Δφ(n), mod is the modulo operation, Resp Mode(q) is the breathing mode of the qth signal segment, which is a physiological signal derived parameter;

[0202] S305. Obtain the user's inhalation and exhalation rates based on the force / displacement / angle change signal P(t):

[0203]

[0204] Where, v insp (q) is the inhalation rate of the qth signal segment, v exp (q) is the exhalation rate of the qth signal segment, which is a physiological signal-derived parameter;

[0205] S306, search for all peak points of the mechanical / displacement / angle change signal P(t), and calculate the time intervals between all adjacent peak points to form a pressure time series T resp (k), get the user's average breathing frequency:

[0206]

[0207] Where, f resp (q) is the average respiratory frequency of the qth signal segment, which is a physiological signal-derived parameter, and K is the total number of points in the pressure time series;

[0208] S307: Analyze the fluctuation of the force / displacement / angle change signal P(t) and calculate the user's breathing depth:

[0209]

[0210] Where Depth(q) is the respiratory depth of the qth signal segment, which is a physiological signal derived parameter, and P baseline (t) is the baseline of the force / displacement / angle change signal P(t), t q , t q-1 is the starting time of the qth and q-1th signal segments;

[0211] S308. Analyze the amplitude change of the discrete mechanics / displacement / angle change signal P(n) and calculate the user's breathing intensity:

[0212]

[0213] Where, P baseline (n) is the baseline of the discrete mechanical / displacement / angle change signal P(n), N is the total number of discrete points in the qth signal segment, and S(q) is the respiratory intensity of the qth signal segment, which is a physiological signal-derived parameter;

[0214] S309. Calculate the peak-to-peak values of the EDR signal and the cuff pressure signal to obtain the user's ventilation efficiency:

[0215]

[0216] Where PK is the number of waveforms of the EDR signal and the force / displacement / angle change signal P(t) that are time-aligned in the qth signal segment, and EDRamp (k) is the peak-to-peak value of the kth corresponding EDR signal, P amp (k) is the peak-to-peak value of the k-th corresponding mechanical / displacement / angle change signal P(t), and Ventilation Efficiency (q) is the ventilation efficiency of the q-th signal segment, which is a physiological signal-derived parameter.

[0217] S310. Obtain the user's blood flow velocity by calculating the time interval between the electrocardiogram signal in the fused physiological signal and the pulse wave signal in the fused physiological signal:

[0218]

[0219] Where Δt ECG-to-PPG (p) is the time interval between the pth ECG signal waveform and the pulse wave signal waveform, P is the number of ECG signal and pulse wave signal waveforms that are time-aligned in the qth signal segment, v blood (q) is the blood flow velocity of the qth signal segment, which is a physiological signal derived parameter;

[0220] S311. Calculate the user's vascular elasticity by identifying the systolic peak and diastolic peak of the pulse wave signal:

[0221]

[0222] Where, P Systolic Peak (p) is the pth contraction peak identified in the pulse wave signal within the qth signal segment, P Dicrotic Peak (p) is the pth diastolic peak identified in the pulse wave signal within the signal segment, ElasticityIndex(q) is the vascular elasticity of the qth signal segment, which is a physiological signal-derived parameter;

[0223] S312. Calculate the user's oxygenation capacity by fusing the blood oxygen in the physiological signal and the mechanical / displacement / angle change signal P(t) in the physiological signal:

[0224]

[0225] Where ΔSpO2 is the fluctuation level of blood oxygen, Oxygenation Efficiency (q) is the oxygenation capacity of the qth signal segment, which is a physiological signal-derived parameter, and t q , t q-1 is the starting time of the qth and q-1th signal segments, f resp is the average respiratory frequency of the qth signal segment.

[0226] Step S4 specifically includes:

[0227] S401, setting a value range of a breathing guidance mode, wherein the breathing guidance mode includes a guided breathing pattern, a guided breathing rate, a guided breathing frequency, a guided inhalation-exhalation ratio, and a guided breathing intensity;

[0228] S402. Traverse the value range and construct the action path of the Q+1th signal segment:

[0229] E m (Q+1)=[κ m (Q+1),ν m (Q+1),f m (Q+1),ρ m (Q+1),∈ m (Q+1)]m=1,…,M

[0230] Where Q is the total number of current signal segments, E m (Q+1) is the m-th action path of the Q+1-th signal segment, i.e., the m-th guidance method, κ m (Q+1),ν m (Q+1),f m (Q+1),ρ m (Q+1),∈ m (Q+1) represents the guided breathing pattern, guided breathing rate, guided breathing frequency, guided inspiration-expiration ratio, and guided breathing intensity in the m-th action path of the Q+1-th signal segment, respectively, which are obtained by traversing the corresponding value range. M is the total number of action paths.

[0231] S403. Using the pre-built digital twin neural network, the respiratory physiological data formed by the derived parameters of each physiological signal and the basic information related to the user's breathing is used as input, the temporal and spatial characteristics of the input data are decoupled, and the physiological change path of the user under different motion paths is predicted. The digital twin neural network is specifically:

[0232]

[0233] Y m (Q+1)=Sigmoid(W out ·F m (Q+1)),m=1,…,M

[0234] Where q represents the qth signal segment, Y m (Q+1) is the predicted physiological change path caused by the m-th action path of the Q+1-th signal segment, Sigmoid() represents the sigmoid constraint function, tanh() is the tanh constraint function, and W out is the output weight parameter matrix, F m(Q+1) is the set of physiological change paths caused by the predicted m-th action path of the Q+1-th signal segment, W s is the spatial convolution kernel weight matrix, W p is the spatial parameter weight matrix, W t is the temporal convolution kernel weight matrix, W f is the time parameter weight matrix, is the dilated convolution operation, * is the convolution operation, ⊙ is the Hadamard product operation, φ() is the Fourier feature map, and X(Q) is the user's respiratory physiological data including the Q-th signal segment.

[0235] Step S5 specifically includes:

[0236] S501. Calculate a physiological enhancement reward score for each physiological change path caused by each action path using a solver constructed based on physiological enhancement constraints. The solver constructed based on physiological enhancement constraints is specifically:

[0237]

[0238] Where PI i is the importance ratio of the ith physiological signal-derived parameter under the physiological reinforcement constraint. The physiological parameters are calculated by the above-mentioned signal decoding module, including: attention index, sympathetic nerve activity, vagus nerve activity, breathing pattern, breathing rate, breathing frequency, breathing depth, breathing force, ventilation efficiency, blood flow velocity, vascular elasticity, and oxygenation capacity, a total of 12 types. is the reinforced rule discriminant space of the ith physiological signal derived parameter constructed based on the medical knowledge base, is the physiological change path Y caused by the predicted m-th action path of the Q+1-th signal segment m (Q+1) is the physiological reinforcement reward score, M is the total number of action paths;

[0239] S502. Calculate the safety enhancement reward score for the physiological change path caused by each action path using a solver constructed based on safety constraints:

[0240]

[0241] Where, SI i is the importance ratio of the i-th physiological signal derived parameter under safety constraints, is the safety rule discriminant space of the ith physiological signal derived parameter constructed based on the medical knowledge base, Y m (Q+1) security enhancement bonus score;

[0242] S503. Calculate the comfort enhancement reward score for the physiological change path caused by each action path using a solver constructed based on safety constraints:

[0243]

[0244] Where, CI i is the importance ratio of the i-th physiological signal derived parameter under comfort constraint, M Comforti is the comfort rule discriminant space of the ith physiological signal derived parameter constructed based on the medical knowledge base, Y m (Q+1) comfort enhancement bonus score;

[0245] S504: Integrate the physiological reinforcement reward score, the safety reinforcement reward score, and the comfort reinforcement reward score, and determine the optimal training parameters based on the following strategy function constructed through reinforcement learning:

[0246]

[0247] In the formula, γ1, γ2, and γ3 are the adjustment coefficients of each constraint condition, λ m (Q+1) is E m The comprehensive score of (Q+1), which is a learnable strategy function, m * is the index number of the optimal action path solved, The mth segment of the Q+1th signal * action path, Opt(Q+1) is the optimal training parameter for the predicted Q+1th signal segment.

[0248] Among them, the adjustment module used for breathing guidance is specifically an inflatable strap, and the inflatable strap is provided with a communication unit, a speaker unit, a vibration motor unit, an airbag and an air pump unit, and a pressure detection unit. The communication unit is used to notify the smart device with a display function to play a breathing action guidance video that guides the user to train with the optimal training parameters. The speaker unit is used to play the breathing action guidance audio that guides the user to train with the optimal training parameters. The vibration motor unit is used to guide the user's breathing rhythm to approach the optimal training parameters. The airbag and air pump unit are used to temporarily store gas, inflate and deflate. The pressure detection unit is used to determine whether the pressure generated on the user at the current inflation level meets the optimal training parameter requirements.

[0249] Among them, the adjustment module used for breathing guidance can also be a retractable strap, which is provided with a communication unit, a speaker unit, a vibration motor unit, a mechanical retractable motor unit, and a pressure detection unit. The communication unit notifies the smart device with a display function to play a breathing action guidance video that guides the user to train with the optimal training parameters. The speaker unit is used to play the breathing action guidance audio that guides the user to train with the optimal training parameters. The vibration motor unit is used to guide the user's breathing rhythm to approach the optimal training parameters. The mechanical retractable motor unit is used to adjust the length of the strap, and the pressure detection unit is used to determine whether the pressure generated on the user at the current inflation level meets the optimal training parameter requirements.

[0250] The method of the embodiment of the present invention corresponds to the system of the embodiment one by one and has the same technical effect. If it is not specified in detail, please refer to the first embodiment and will not be repeated.

[0251] It should be understood that the above embodiments and descriptions only describe the principles, main features and advantages of the present invention. Without departing from the spirit and scope of the present invention, the present invention may be subject to various changes and improvements, and these changes and improvements all fall within the scope of protection of the present invention.

Claims

1. An intelligent training guidance device, characterized in that: include: The pressure signal acquisition module is used to collect the mechanical / displacement / angle change signals caused by the rise and fall of the chest or abdomen during the user's breathing; Signal synchronization module, used to synchronize with smart wearable devices or medical devices to obtain multiple basic physiological signals of users; A signal decoding module is used to calculate the user's physiological signal-derived parameters based on the mechanical / displacement / angle change signals and multiple basic physiological signals during the user's breathing process; A physiological pathway simulation module is used to fuse the user's personal breathing-related basic information and physiological signal-derived parameters into respiratory physiological data, input it into a pre-built digital twin neural network, and predict the physiological change paths under all possible breathing guidance methods. The digital twin neural network decouples the spatial and temporal characteristics of the respiratory physiological data under each breathing guidance method, obtains the physiological changes in time and space, and forms a physiological change path; The intelligent optimization and decision-making module is used to evaluate multiple physiological change paths predicted by the physiological path simulation module based on the medical knowledge database using a reinforcement learning algorithm, and select the optimal physiological change path as the optimal training parameter; The breathing guidance adjustment module is used to guide the user to adapt to the determined optimal training parameters through various methods.

2. The intelligent training guidance device according to claim 1, characterized in that: The signal decoding module specifically includes: The attention index calculation unit is used to calculate the user's attention index according to the following formula based on the EEG signal fused with the physiological signal: Where q represents the qth signal segment after the fused physiological signal is divided into time segments, P EEG (f) is the power spectral density of the EEG signal at frequency f, AttentionIndex(q) is the attention index of the qth signal segment, which is a physiological signal-derived parameter and is calculated by the corresponding signal of the qth signal segment.

3. The intelligent training guidance device according to claim 1, characterized in that: The signal decoding module also includes: The sympathetic nerve activity evaluation unit is used to search for the peak points of the electrocardiogram signal in the fused physiological signal, form a heart rate sequence by chronologically separating the time intervals between all two adjacent peak points, and calculate the user's sympathetic nerve activity according to the following formula: SympatheticIndex(q)=w1·HRV LF +w2·ΔBP+w3·Resp-BP Coupling Where, HRV LF is the low frequency component of heart rate, P HRV (f) is the power spectral density of the heart rate sequence at frequency f, ΔBP is the blood pressure variability index, BP n is a discrete blood pressure sequence, N is the total number of discrete points in the qth signal segment, is the average value of the blood pressure series, Resp-BP Coupling is the respiratory-blood pressure coupling strength, P(t) is the mechanical / displacement / angle change signal, BP(t+τ) is the blood pressure time series after the time delay τ, t q , t q-1 are the starting time of the q-1th and qth signal segments, respectively. q represents the qth signal segment after the basic physiological signal is divided into time segments. SympatheticIndex(q) is the sympathetic nerve activity of the qth signal segment, which is a physiological signal-derived parameter. w1, w2, and w2 are weight coefficients. The vagus nerve activity evaluation unit is used to calculate the user's vagus nerve activity according to the following formula based on the heart rate sequence: VagalIndex(q)=m1·HRV HF +m2·Resp-HR Coherence Where, HRV HF is the high-frequency component of heart rate, P(f) is the auto-power spectrum density of P(t) at frequency f, P HR (f) is the autopower spectral density of the heart rate time series at frequency f, P Resp,HR (f) is the cross-power spectral density between P(t) and the heart rate sequence at frequency f, Coherence(f) is the coherence calculation at frequency f, Resp-HR Coherence is the degree of coherence between P(t) and the heart rate time series, VagalIndex(q) is the vagus nerve activity of the qth signal segment, which is a physiological signal-derived parameter, and m1 and m2 are weight coefficients.

4. The intelligent training guidance device according to claim 1, characterized in that: The signal decoding module also includes: The breathing pattern judgment unit is used to obtain the ECG-derived breathing EDR signal by extracting the envelope of the ECG signal from the fused physiological signal, and judge the user's breathing pattern according to the following formula based on the EDR signal and the mechanical / displacement / angle change signals during the user's breathing process: Δφ(n)=|mod(φ EDR (n)-φ P (n)+π,2π)-π|∈[0,π] Where, is the imaginary part sequence of the EDR signal, s EDR (n) is the real part sequence of the EDR signal, φ EDR (n) is the phase sequence of the EDR signal, is the imaginary part sequence of the force / displacement / angle change signal, s P (n) is the real part sequence of the discrete mechanics / displacement / angle change signal P(n), φ P (n) is the phase sequence of the discrete mechanical / displacement / angle change signal P(n), P(n) is the discrete sequence after discrete sampling of the mechanical / displacement / angle change signal P(t), Δφ(n) is the phase difference sequence, is the mean value of Δφ(n), mod is the modulo operation, Resp Mode(q) is the breathing mode of the qth signal segment, which is a physiological signal derived parameter; The breathing rate calculation unit is used to obtain the user's inhalation and exhalation rates based on the force / displacement / angle change signal P(t): Where, v insp (q) is the inhalation rate of the qth signal segment, v exp (q) is the exhalation rate of the qth signal segment, which is a physiological signal-derived parameter; The respiratory rate calculation unit is used to search for all peak points of the mechanical / displacement / angle change signal P(t) and calculate the time intervals between all adjacent peak points to form the pressure time series T resp (k), get the user's average breathing frequency: Where, f resp (q) is the average respiratory frequency of the qth signal segment, which is a physiological signal-derived parameter, and K is the total number of points in the pressure time series; The breathing depth calculation unit is used to analyze the fluctuation of the mechanical / displacement / angle change signal P(t) and calculate the user's breathing depth: Where Depth(q) is the respiratory depth of the qth signal segment, which is a physiological signal derived parameter, and P baseline (t) is the baseline of the force / displacement / angle change signal P(t), t q , t q-1 is the starting time of the qth and q-1th signal segments; The breathing intensity calculation unit is used to analyze the amplitude changes of the discrete mechanics / displacement / angle change signal P(n) and calculate the user's breathing intensity: Where, P baseline (n) is the baseline of the discrete mechanical / displacement / angle change signal P(n), N is the total number of discrete points in the qth signal segment, and S(q) is the respiratory intensity of the qth signal segment, which is a physiological signal-derived parameter; The ventilation efficiency calculation unit is used to calculate the peak-to-peak value of the EDR signal and the cuff pressure signal to obtain the user's ventilation efficiency: Where PK is the number of waveforms of the EDR signal and the force / displacement / angle change signal P(t) that are time-aligned in the qth signal segment, and EDR amp (k) is the peak-to-peak value of the kth corresponding EDR signal, P amp (k) is the peak-to-peak value of the k-th corresponding mechanical / displacement / angle change signal P(t), and Ventilation Efficiency (q) is the ventilation efficiency of the q-th signal segment, which is a physiological signal-derived parameter.

5. The intelligent training guidance device according to claim 1, characterized in that: The signal decoding module also includes: The blood flow velocity calculation unit is used to obtain the user's blood flow velocity by calculating the time interval between the electrocardiogram signal in the fused physiological signal and the pulse wave signal in the fused physiological signal: Where Δt ECG-to-PPG (p) is the time interval between the pth ECG signal waveform and the pulse wave signal waveform, P is the number of ECG signal and pulse wave signal waveforms that are time-aligned in the qth signal segment, v blood (q) is the blood flow velocity of the qth signal segment, which is a physiological signal derived parameter; The vascular elasticity calculation unit is used to calculate the user's vascular elasticity by identifying the systolic and diastolic peaks of the pulse wave signal: Where, P Systolic Peak (p) is the pth contraction peak identified in the pulse wave signal within the qth signal segment, P Dicrotic Peak (p) is the pth diastolic peak identified in the pulse wave signal within the signal segment, ElasticityIndex(q) is the vascular elasticity of the qth signal segment, which is a physiological signal-derived parameter; The oxygenation capacity evaluation unit is used to calculate the user's oxygenation capacity by fusing the blood oxygen in the physiological signal and the mechanical / displacement / angle change signal P(t) in the physiological signal: Where ΔSpO2 is the fluctuation level of blood oxygen, Oxygenation Efficiency (q) is the oxygenation capacity of the qth signal segment, which is a physiological signal-derived parameter, and t q , t q-1 is the starting time of the qth and q-1th signal segments, f resp is the average respiratory frequency of the qth signal segment.

6. The intelligent training guidance device according to claim 1, characterized in that: The physiological pathway simulation module specifically includes: a data range setting unit, configured to set a value range of a breathing guidance mode, wherein the breathing guidance mode includes a guided breathing pattern, a guided breathing rate, a guided breathing frequency, a guided inhalation-exhalation ratio, and a guided breathing intensity; The action path construction unit is used to traverse the value range and construct the action path of the Q+1th signal segment: From m (Q+1)=[κ m (Q+1),ν m (Q+1),f m (Q+1),ρ m (Q+1),∈ m (Q+1)],m=1,…,M Where Q is the total number of current signal segments, E m (Q+1) is the m-th action path of the Q+1-th signal segment, i.e., the m-th guidance method, κ m (Q+1),ν m (Q+1),f m (Q+1),ρ m (Q+1),∈ m (Q+1) represents the guided breathing pattern, guided breathing rate, guided breathing frequency, guided inspiration-expiration ratio, and guided breathing intensity in the m-th action path of the Q+1-th signal segment, respectively, which are obtained by traversing the corresponding value range. M is the total number of action paths. The spatiotemporal decoupling generator unit is used to take respiratory physiological data formed by various physiological signal-derived parameters and basic information related to the user's breathing as input based on a pre-built digital twin neural network, decouple the temporal and spatial features of the input data, and predict the physiological change path of the user under different motion paths. The digital twin neural network is specifically: Y m (Q+1)=Sigmoid(W out ·F m (Q+1)),m=1,…,M Where q represents the qth signal segment, Y m (Q+1) is the predicted physiological change path caused by the m-th action path of the Q+1-th signal segment, Sigmoid() represents the sigmoid constraint function, tanh() is the tanh constraint function, and W out is the output weight parameter matrix, F m (Q+1) is the set of physiological change paths caused by the predicted m-th action path of the Q+1-th signal segment, W s is the spatial convolution kernel weight matrix, W p is the spatial parameter weight matrix, W t is the temporal convolution kernel weight matrix, W f is the time parameter weight matrix, is the dilated convolution operation, * is the convolution operation, ⊙ is the Hadamard product operation, φ() is the Fourier feature map, and X(Q) is the user's respiratory physiological data including the Q-th signal segment.

7. The intelligent training guidance device according to claim 1, characterized in that: The intelligent optimization and decision-making module specifically includes: The physiological reinforcement reward unit is used to calculate the physiological reinforcement reward score for the physiological change path caused by each action path through a solver constructed by physiological reinforcement constraints, wherein the solver constructed by physiological reinforcement constraints is specifically: Where PI i is the importance ratio of the ith physiological signal derived parameter under the physiological reinforcement constraint, IN is the number of physiological signal derived parameters, is the reinforced rule discriminant space of the ith physiological signal derived parameter constructed based on the medical knowledge base, is the physiological change path Y caused by the predicted m-th action path of the Q+1-th signal segment m (Q+1) is the physiological reinforcement reward score, M is the total number of action paths; The safety supervision unit is used to calculate the safety reinforcement reward score for the physiological change path caused by each action path through a solver constructed by safety constraints: Where, SI i is the importance ratio of the i-th physiological signal derived parameter under safety constraints, is the safety rule discriminant space of the ith physiological signal derived parameter constructed based on the medical knowledge base, Y m (Q+1) security enhancement bonus score; The comfort constraint unit is used to calculate the comfort enhancement reward score for the physiological change path caused by each action path through a solver constructed by safety constraints: Where, CI i is the importance ratio of the i-th physiological signal derived parameter under comfort constraint, is the comfort rule discriminant space of the ith physiological signal derived parameter constructed based on the medical knowledge base, Y m (Q+1) comfort enhancement bonus score; The comprehensive decision-making unit is used to integrate the physiological reinforcement reward score, safety reinforcement reward score, and comfort reinforcement reward score, and determine the optimal training parameters based on the following policy function constructed through reinforcement learning: In the formula, γ1, γ2, and γ3 are the adjustment coefficients of each constraint condition, λ m (Q+1) is E m The comprehensive score of (Q+1), which is a learnable strategy function, m * is the index number of the optimal action path solved, The mth segment of the Q+1th signal * action path, Opt(Q+1) is the optimal training parameter for the predicted Q+1th signal segment.

8. The intelligent training guidance device according to claim 1, characterized in that: The breathing guidance adjustment module is specifically an inflatable strap, which is provided with a communication unit, a speaker unit, a vibration motor unit, an airbag and an air pump unit, and a pressure detection unit. The communication unit is used to notify a smart device with a display function to play a breathing action guidance video that guides the user to train with optimal training parameters. The speaker unit is used to play a breathing action guidance audio that guides the user to train with optimal training parameters. The vibration motor unit is used to guide the user's breathing rhythm to approach the optimal training parameters. The airbag and air pump unit are used to temporarily store gas, inflate and deflate. The pressure detection unit is used to determine whether the pressure generated on the user at the current inflation level meets the optimal training parameter requirements.

9. The intelligent training guidance device according to claim 1, characterized in that: The breathing guidance adjustment module is specifically a retractable strap, which is provided with a communication unit, a speaker unit, a vibration motor unit, a mechanical retractable motor unit, and a pressure detection unit. The communication unit notifies a smart device with a display function to play a breathing action guidance video that guides the user to train with optimal training parameters. The speaker unit is used to play a guidance audio of the breathing action that guides the user to train with optimal training parameters. The vibration motor unit is used to guide the user's breathing rhythm to approach the optimal training parameters. The mechanical retractable motor unit is used to adjust the length of the strap. The pressure detection unit is used to determine whether the pressure generated on the user at the current inflation level meets the optimal training parameter requirements.

10. An intelligent training guidance method, characterized in that: include: Collect the mechanical / displacement / angle change signals caused by the rise and fall of the chest or abdomen during the user's breathing; Synchronize with smart wearable devices or medical devices to obtain multiple basic physiological signals of users; Calculate the user's physiological signal-derived parameters based on the user's breathing process's mechanical / displacement / angle change signals and multiple basic physiological signals; The user's personal breathing-related basic information and physiological signal-derived parameters are integrated into respiratory physiological data, which is input into a pre-built digital twin neural network to predict the physiological change paths under all possible breathing guidance methods. The digital twin neural network decouples the spatial and temporal features of the respiratory physiological data under all possible breathing guidance methods, obtains the physiological changes in time and space, and forms the physiological change path; Based on the medical knowledge database, multiple physiological change paths predicted by the physiological path simulation module are evaluated, and the optimal physiological change path is selected as the optimal training parameter; Guide users to adapt to the determined optimal training parameters through various means.

Images