A data acquisition, monitoring, analysis and control system for use with a micro-hyperbaric chamber

Abstract

This invention relates to the field of micro hyperbaric oxygen chamber technology, and discloses a data acquisition, monitoring, analysis, and control system for micro hyperbaric oxygen chambers, including: a multi-dimensional sensing module, a digital twin computing module, an adaptive pressure control module, and an active safety defense module. The system acquires multimodal data in real time through physical biosensors and visual devices, constructs a digital twin model using deep learning algorithms, and calculates physiological tolerance. Based on the physiological tolerance, it calls a nonlinear pressure boosting control model, automatically executing adaptive callbacks and dynamic pressure stabilization during pressure boosting to construct a physiological adaptation window. Simultaneously, by calculating a comprehensive risk index, it uses dual logic to determine whether the user is in a sleep or abnormal state, thereby executing wake-up and emergency pressure relief strategies. This invention solves the problem of eardrum damage caused by traditional linear pressure boosting and effectively avoids false alarms and missed alarms in safety monitoring, improving the comfort and safety of equipment operation.

Description

Technical Field

[0001] This invention relates to the field of micro hyperbaric oxygen chamber technology, specifically to a data acquisition, monitoring, analysis and control system for micro hyperbaric oxygen chambers. Background Technology

[0002] Hyperbaric oxygen chambers, as devices that provide an environment slightly above one atmosphere of pressure for the human body to inhale high concentrations of oxygen, are a new type of non-invasive health management device widely used in community health and wellness institutions, rehabilitation centers, and home settings. Their core working principle involves creating a closed, high-pressure environment slightly above one atmosphere (typically 1.3-1.5 ATA), utilizing the principle of physical dissolution to increase the dissolved oxygen content in the blood, thereby improving tissue hypoxia, relieving fatigue, and promoting metabolic circulation. During the operation of a hyperbaric oxygen chamber, the data acquisition, monitoring, and control system is crucial for ensuring equipment efficiency and user experience, responsible for real-time adjustment of the chamber's environmental parameters and feedback on the user's physiological state.

[0003] Most existing micro-hyperbaric oxygen chamber control systems are based on traditional industrial automation logic, typically employing programmable logic controllers (PLCs) to deliver gas and regulate pressure according to preset programs. In terms of operation, the system controls the opening and closing of the intake pump and solenoid valves, driving the chamber pressure to rise unidirectionally from ambient pressure to the set working pressure. Simultaneously, to meet basic safety regulations, some equipment is equipped with pressure sensors, oxygen concentration sensors, and simple video monitoring devices to display current operating data and cut off power or perform physical depressurization when parameters exceed limits, thus maintaining the equipment within basic safety limits.

[0004] However, as users' demands for comfort and safety increase, traditional hyperbaric oxygen chamber control systems have revealed some shortcomings in practical applications: First, in terms of pressure control, traditional systems generally adopt a preset fixed-rate open-loop linear pressurization logic, that is, controlling the gas source to continuously inject air into the chamber at a constant rate. However, the opening of the Eustachian tube and the adjustment of the inner and outer ear pressure balance are physiological processes with lag and intermittency, which cannot be synchronized with the mechanical linear pressurization in real time. When the pressurization rate exceeds the adjustment limit of the user's Eustachian tube, the traditional hyperbaric oxygen chamber control system cannot sense and make adaptive adjustments, such as pausing or reversing, but continues to pressurize. This can easily lead to a rapid accumulation of pressure difference between the two sides of the eardrum in a short period of time, thereby inducing severe ear swelling and pain, congestion, or even barotrauma to the middle ear. Secondly, in terms of safety monitoring, existing technologies mostly rely on threshold alarm mechanisms based on single parameters, such as monitoring only the upper limit of heart rate or relying solely on video footage. They lack the ability to fuse and analyze multi-dimensional data, making it difficult for the system to accurately distinguish between a user's normal sleep state and abnormal states such as sudden pathological coma. This leads to false alarms (disturbing the user's rest) or missed alarms (delaying emergency treatment), posing potential safety hazards to the use of the equipment. Therefore, there is an urgent need for a data acquisition, monitoring, analysis, and control system for micro-hyperbaric oxygen chambers to solve these problems. Summary of the Invention

[0005] To address the problems in related technologies, this invention provides a data acquisition, monitoring, analysis and control system for use in micro hyperbaric oxygen chambers, thereby overcoming the aforementioned technical problems in existing related technologies.

[0006] To solve the aforementioned technical problem, the present invention is achieved through the following technical solution: In a first aspect, embodiments of the present invention provide a data acquisition, monitoring, analysis, and control system for use in a micro hyperbaric oxygen chamber, specifically comprising: a multi-dimensional sensing module for real-time acquisition of users' physiological signs and behavioral posture data through a physical biosensor array and in-chamber vision equipment; a digital twin computing module for constructing a human health digital twin model and calculating the user's real-time physiological tolerance index using a multimodal deep learning algorithm; an adaptive pressure control module for acquiring target pressure increments and mapping them to a specific pressure control range, calling a nonlinear pressure boosting control model to obtain real-time pressure commands, and generating negative pressure commands to execute pressure callbacks; and an active safety defense module for calculating a comprehensive risk index through multimodal data fusion, classifying abnormal states, and executing corresponding strategies.

[0007] As a preferred embodiment of the data acquisition, monitoring, analysis, and control system used in the micro-hyperbaric oxygen chamber of the present invention, the formula for calculating the physiological tolerance index is as follows: ; In the formula, Provides a real-time physiological tolerance index for users. This is the sensitivity coefficient. Basic tolerance threshold, To address inappropriate energy levels.

[0008] As a preferred embodiment of the data acquisition, monitoring, analysis, and control system used in the micro-hyperbaric oxygen chamber described in this invention, the integrated adverse energy... The calculation formula is: ; In the formula, This is the physiological single-modal weight vector. For a single-modal weight vector, Physiological trend characteristics For behavioral feature vectors, This is a cross-fusion matrix.

[0009] As a preferred embodiment of the data acquisition, monitoring, analysis, and control system used in the micro hyperbaric oxygen chamber of the present invention, the calculation formula for the real-time pressure command is as follows: ; In the formula, For real-time pressure commands, As for the current basic pressure, For the target increment, The boost time constant is For the callback stable time constant, As a factor for relieving eardrum pressure, This is the callback coefficient; The callback coefficient Based on real-time physiological tolerance index calculate: ,in This represents the maximum design reduction amount corresponding to the current step range.

[0010] As a preferred embodiment of the data acquisition, monitoring, analysis, and control system used in the micro-hyperbaric oxygen chamber of the present invention, the physiological trend characteristics... It is obtained through prediction using an LSTM model, which uses a forgetting gate. Input gate and output gate Controlling information flow and calculating hidden states And obtained through mapping of fully connected layers : ; In the formula, It is the ReLU activation function. The feature mapping weight matrix, This is the bias vector.

[0011] As a preferred embodiment of the data acquisition, monitoring, analysis, and control system used in the micro hyperbaric oxygen chamber described in this invention, the behavioral feature vector... Extracted using a Transformer visual model, which includes a multi-head self-attention mechanism and a feedforward network, outputting sequential features. And obtained through time-dimensional aggregation mapping : ; In the formula, Indicates the first Feature vectors at each time step To output the projection matrix, This is a bias term.

[0012] As a preferred embodiment of the data acquisition, monitoring, analysis, and control system used in the micro hyperbaric oxygen chamber of the present invention, the formula for calculating the comprehensive risk index is as follows: ; In the formula, Score for visual abnormalities, For scoring abnormal biological signs, Score environmental anomalies. These are the corresponding weighting coefficients.

[0013] As a preferred embodiment of the data acquisition, monitoring, analysis, and control system used in the micro hyperbaric oxygen chamber of the present invention, the abnormal state determination includes: setting a visual threshold. Vital signs threshold Risk threshold and wake-up threshold ;when And visual abnormality score Biological sign abnormality score When the patient is in a sleep state, awakening is only performed before the end of the treatment; when... ,and and If the situation is deemed abnormal, the oxygen supply should be immediately cut off and the pressure relief valve should be fully opened.

[0014] As a preferred embodiment of the data acquisition, monitoring, analysis, and control system used in the micro hyperbaric oxygen chamber of the present invention, the adaptive pressure control module includes a preset strategy library, which adjusts the target pressure increment accordingly. Automatic matching control strategy: When When the pressure is in the 5-10 kPa range, the first control strategy is implemented: the exhaust valve is slightly adjusted down by 1-2 kPa and then stabilized, followed by staged pressure increase and stabilization; when When the pressure is in the 10-15 kPa range, the second control strategy is implemented: the control pressure is lowered by 2-3 kPa and then stabilized, followed by phased pressure increase and stabilization; when When the pressure is in the 15-20 kPa range, the third control strategy is implemented: the control pressure is lowered by 3-4 kPa and then stabilized, followed by phased pressure increase and stabilization; when When the pressure is in the 20-35 kPa range, the fourth control strategy is implemented: the control pressure is lowered by 4-5 kPa and then stabilized, the stabilization time interval is shortened, and pressure increase actions are performed in stages with an amplitude of 10-15 kPa or 15-20 kPa; when When the pressure is in the range of 35-50 kPa, the fifth control strategy is implemented: the control pressure is reduced by 5-6 kPa to force the Eustachian tube pressure to be balanced, the pressure stabilization time interval is further shortened, and the pressure is increased in stages.

[0015] Secondly, embodiments of the present invention provide a micro hyperbaric oxygen chamber device, specifically including: a chamber body, an intelligent control cabinet, a gas source system, and a data acquisition, monitoring, analysis, and control system for use in the micro hyperbaric oxygen chamber. The gas source system includes an air compressor, a gas storage tank, a precision filter, and an intelligent valve island assembly. The intelligent control cabinet integrates an edge computing unit, a PLC controller, and a human-machine interface terminal. The edge computing unit is connected to a physical biosensor array, an in-chamber vision device, and the intelligent valve island assembly. The intelligent valve island assembly includes a proportional intake solenoid valve, a micro exhaust solenoid valve, and a safety emergency stop valve. The proportional intake solenoid valve is used to adjust the opening of the intake airflow, the micro exhaust solenoid valve is used to perform adaptive callback actions, and the safety emergency stop valve is used to perform rapid depressurization in abnormal conditions.

[0016] The present invention has the following beneficial effects: 1. This invention uses a digital twin computing module to calculate the user's physiological tolerance in real time and utilizes an adaptive pressure control module to invoke a nonlinear pressure boosting control model. This allows the system to automatically execute an adaptive callback mechanism based on the user's real-time physiological feedback during the pressure boosting process. Specifically, it proactively fine-tunes and stabilizes the chamber pressure when discomfort is detected, creating a physiological adaptation window for the user. This dynamic pressure regulation method achieves real-time synchronization between mechanical pressure boosting and human physiological adaptation, effectively alleviating eardrum pressure and helping to eliminate the stinging sensation and barotraumatic middle ear damage risk associated with traditional linear pressure boosting.

[0017] 2. This invention employs a multi-dimensional sensing module and an active safety defense module, integrating physiological data such as heart rate and blood oxygenation with behavioral and postural data such as eye-closing duration and limb movements to calculate a comprehensive risk index. Through dual logical judgment, it distinguishes between a stable but motionless state of deep sleep and a pathological abnormal state with abnormal signs and no movement or convulsions. This not only avoids disturbing the user's rest due to false alarms, but also ensures that in the event of real dangers such as apparent death, coma, or violent agitation, the system can immediately perform emergency stop depressurization and cut off oxygen supply, thus achieving active safety defense.

[0018] 3. This invention includes a pre-defined strategy library containing various control strategies, enabling the system to automatically match the corresponding boost, pullback, and stabilization rhythms based on the current target pressure increment, such as slight, low-medium, medium-high, and high boost scenarios. For example, in high boost scenarios, the pullback amplitude is automatically increased and the stabilization interval is shortened. This hierarchical control strategy ensures both rapid response under low pressure differentials and sufficient adaptation under high pressure differentials, achieving a balance between boost efficiency and user experience.

[0019] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For those skilled in the art, the drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This invention provides a schematic diagram of a data acquisition, monitoring, analysis and control system for use in a micro hyperbaric oxygen chamber.

[0022] Figure 2 The present invention provides a flowchart of a data acquisition, monitoring, analysis and control method for use in a micro hyperbaric oxygen chamber.

[0023] Figure 3 This is a schematic diagram illustrating the abnormal state determination of the active security defense module provided by the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example 1

[0025] Traditional hyperbaric oxygen chamber control systems often employ an open-loop linear pressurization method, neglecting the nonlinear adaptation process of the human body (especially the eardrum) to pressure changes, which can easily cause ear pain or barotrauma to users. Furthermore, existing safety monitoring relies heavily on single threshold alarms, which cannot accurately distinguish between complex states such as deep sleep and pathological coma, leading to false alarms or missed alarms.

[0026] To solve the above technical problems, such as Figure 1 As shown, Embodiment 1 of the present invention provides a data acquisition, monitoring, analysis, and control system for use in a micro hyperbaric oxygen chamber, specifically including: a multi-dimensional sensing module, a digital twin computing module, an adaptive pressure control module, and an active safety defense module; the multi-dimensional sensing module is used to collect the user's physiological signs and behavioral posture data in real time through a physical biosensor array and in-chamber vision equipment; the digital twin computing module is used to construct a human health digital twin model and calculate the user's real-time physiological tolerance using a multimodal deep learning algorithm; the adaptive pressure control module is used to obtain the target pressure increment and map it to a specific pressure control range, and call a nonlinear pressure boosting control model including adaptive callback and dynamic pressure stabilization to obtain dynamic pressure commands; the active safety defense module is used to calculate a comprehensive risk index through multimodal data fusion, classify abnormal states, and execute corresponding emergency stop or wake-up strategies.

[0027] Specific embodiment 1 is as follows: A micro hyperbaric oxygen chamber (designed maximum pressure 1.5 ATA) in a community health and wellness center is equipped with a smart bracelet integrating heart rate, blood oxygen, and blood pressure monitoring, as well as a high-definition wide-angle camera inside the chamber. Before use, the system automatically connects to the cloud to obtain the user's historical health profile, providing a good data foundation for embodiment 1 of the present invention.

[0028] In the specific implementation of the above-mentioned Embodiment 1, the multi-dimensional sensing module is first activated to collect various vital signs data, facial expressions, and limb movements of the user in real time, construct a digital twin model of human health, and calculate the user's real-time physiological tolerance. Secondly, in the pressure boosting stage, the adaptive pressure control module calculates the adaptive callback coefficient based on the nonlinear pressure boosting control model, controls the pressure to actively decrease first to build a physiological adaptation window, and then performs exponential smooth pressure boosting. This method alleviates the eardrum pressure difference through physiological feedback closed loop and avoids the stinging sensation caused by traditional linear pressure boosting. At the same time, during the entire operation, the active safety defense module calculates the comprehensive risk index in real time. If the user is determined to be asleep, the module will only wake the user before the end. If an abnormality is determined, the module will immediately stop and release pressure. This method solves the problem of false alarms or missed alarms caused by existing technologies that rely solely on a single threshold alarm, ensuring the comfort and continuity of the treatment.

[0029] Furthermore, to better illustrate the technical solution of Embodiment 1 of the present invention, such as... Figure 2As shown, this paper describes in detail the data acquisition, monitoring, analysis, and control method for a micro hyperbaric oxygen chamber, including the following: S1. Collect multimodal data through physical biosensor arrays and in-cabin vision devices to construct a digital twin model of human health and calculate the user's real-time physiological tolerance. This includes the following sub-steps: S11. Collect user's physiological data (Including heart rate, blood oxygen saturation, blood pressure, respiratory rate, HRV) and behavioral posture data (Including eye closure, head posture, and frequency of limb twitching); physiological signs data After standardization and integration with the user's historical health records, the data is input into a pre-trained LSTM (Long Short-Term Memory) network model to predict the user's physiological trend characteristics under current stress. Simultaneously, behavioral pose data Input a Transformer visual model and extract behavioral feature vectors. .

[0030] S12, Physiological trend characteristics With behavioral feature vector By integrating these technologies, a digital twin model of human health is constructed, and the user's real-time physiological tolerance index is calculated and output. : ; In the formula, The value range is [0,1], and the lower the value, the worse the user's adaptability to the current pressure; This is the sensitivity coefficient, used to adjust the response speed of the tolerance to pressure changes; The baseline tolerance threshold represents the benchmark stress level of an average person under normal conditions. To comprehensively account for inappropriate energy, the following functional formula can be used: ; In the formula, Physiological single-modal weight vector transpose, For single-mode weight vectors transpose, Physiological trend characteristics Transpose This is a cross-fusion matrix; For example: Physiological single-modal weight vector The data was obtained through offline training, using the following method: First, a large amount of physiological monitoring data was collected from volunteers during the pressurization process in the hyperbaric oxygen chamber; for each moment... Dimensional physiological trend feature vector The label is marked by a professional hyperbaric oxygen therapy doctor, who provides a baseline physiological pressure value. Then, in the network architecture of the human health digital twin model, a physiological unimodal branch is established; this branch contains a fully connected layer, whose weight parameters are defined as having a dimension of [missing information]. vector Furthermore, this fully connected layer does not contain a non-linear activation function and directly performs dot product operations. Next, we set the loss function. The preprocessed dataset is input into the network, and the loss function is calculated relative to the loss function using the backpropagation algorithm. The gradient is calculated and updated using the Adam optimizer. Each element in the dataset. Once training converges, the vector is extracted. Finally, the system performs physical verification on the weight values ​​to ensure that the weight components of the corresponding blood oxygen saturation feature dimension are positive (assuming the feature has been preprocessed as blood oxygen fluctuation), and that the weights of the corresponding heart rate variability (HRV) feature dimension conform to medical logic. After verification, Solidify and store in the control system.

[0031] Behavioral single-modal weight vector The data was obtained through offline training, using the following method: First, video stream data from the in-cabin camera was collected, and then extracted using the Transformer visual model. 3D behavioral feature vector For each video, its behavioral risk level is manually labeled. (For example: stillness = 0, slight agitation = 0.3, violent pounding on the hatch = 0.9, limb twitching = 1.0). Then, in the multimodal network architecture, a behavioral unimodal branch is established; this branch is a linear mapping layer, and its learnable parameters are defined as having a dimension of... vector To avoid the model misjudging actions with irrelevant backgrounds (such as adjusting clothes), during training... When L1 regularization is introduced, it causes the vector to The weights corresponding to invalid action features tend to 0, while only high weights corresponding to dangerous action features (such as painful facial expressions and stiffness) are retained, achieving the dual purpose of feature selection and risk quantification. After training, the weights will be... Solidify and store in the control system.

[0032] Cross-fusion matrix To quantify the correlation gain between physiological and behavioral characteristics, during deep learning model training, the network automatically seeks high-risk physiological-behavioral combinations and accordingly increases the values ​​at the corresponding positions in the matrix. The acquisition process is as follows: First, synchronous physiological data (heart rate, blood oxygen, etc.) and video data from multiple test subjects under different stress conditions are collected, and the true comfort level label for each moment is annotated (e.g., comfortable, mild ear pain, severe pain, coma, etc.). A bilinear pooling layer is constructed in the neural network, with its core weights being of size [size missing]. matrix The initial values ​​are randomly orthogonally initialized. Then, the collected dataset is input into the network, and the Adam optimizer is used to iteratively train the network, optimizing the cross-entropy loss function between the predicted and true labels. Once the loss function converges, the trained matrix is ​​extracted. Physiological single-modal weight vector and behavioral single-modal weight vector The data is burned into the edge computing unit of the micro hyperbaric oxygen chamber. In practical use, the system only needs to perform real-time calculations using sensors and cameras. and The current overall inadequacy energy can be calculated using matrix multiplication. .

[0033] Basic tolerance threshold The standard hyperbaric oxygen chamber test was determined using a large-sample statistical calibration method. The specific process involved selecting multiple volunteers of different ages, genders, and health conditions to participate. Volunteers were asked to mark the point of physiological discomfort they subjectively perceived as a slight eardrum distension or an immediate need to swallow. The system simultaneously recorded the physiological trend characteristics of each volunteer at that marked moment. and behavioral feature vectors And utilize the pre-trained multimodal weights , and Then, substitute the values ​​into the formula to calculate the comprehensive unsuitability energy value at that moment. Finally, statistical analysis was performed on the collected set of critical energy values, and the mean or median was taken as the basic tolerance threshold. This will allow us to build a digital baseline for measuring the transition of ordinary users from a comfortable state to a slightly uncomfortable state.

[0034] Sensitivity coefficient The gradient fitting method based on the safety response boundary determines the following process: First, define an inappropriate energy value representing the human body's intolerance or impending barotrauma. (For example, the energy level corresponding to an extremely distressed expression or heart rate alarm), and forcibly set the energy level to be appropriate when that inappropriate energy value is reached. At that time, the system's physiological tolerance index It must fall below a very small safety margin to ensure the maximum stress pullback mechanism is triggered. Then, utilize inappropriate energy values. Compared with the above-calibrated baseline tolerance threshold The difference between them is used to deduce the tolerance index by combining the decay characteristics of the nonlinear activation function. Sensitivity coefficient drops rapidly in the discomfort range The sensitivity coefficient Ultimately, this determines the system's sensitivity to changes in physiological discomfort, ensuring that the system exhibits sufficient control rigidity when risk signs are detected.

[0035] In this embodiment 1, a comprehensive discomfort energy function incorporating physiological trend features and behavioral feature vectors is constructed. A cross-fusion matrix is ​​used to quantify the correlation gain of multimodal data, and the physiological tolerance index is calculated in real time based on the Logistic function. This method transforms abstract human comfort into a calculable digital indicator. Deep fusion of multi-dimensional data eliminates noise interference and misjudgments from single sensors, accurately reflecting the user's true physiological adaptation state during the pressurization process. This provides precise decision-making basis for the subsequent adaptive pressure control module, solving the problem that traditional equipment cannot detect latent user discomfort in real time. Specifically, for example: User Zhang San (50 years old, with a history of hypertension) enters the oxygen chamber. Sensors collect his initial resting heart rate of 75 bpm and blood pressure of 130 / 85 mmHg. The system combines his historical data to initialize a digital twin model. In the initial stage of pressurization, if Zhang San appears tense (decreased HRV, limb stiffness), the model calculates his physiological tolerance. (Lower) The system will automatically prompt "Please relax, comfort boost is about to begin".

[0036] S2. Obtain the target pressure increment, invoke the nonlinear boost control model, and calculate the real-time pressure command including an adaptive callback mechanism. This drives the air circuit system to perform refined pressurization. Specifically, it includes the following sub-steps: S21. Based on the set total target pressure (e.g., 50 kPa), query the preset strategy library of the boost control system and divide it into individual... Stepped section. When entering the... One boost phase (starting at...) When calculating the target increment, And extract the baseline callback magnitude corresponding to this period from the preset strategy library. ; The value is the maximum design reduction amount corresponding to that step range in the preset strategy library. For example, in the preset strategy library of the boost control system, when the step range is 5-10 kPa, the pressure is reduced by 1-2 kPa. The value is then set to 2 kPa. Next, based on real-time physiological tolerance... Correcting the callback coefficient The calculation formula is: If the preset strategy library specifies that voltage stabilization is required for this period, then set the start time of the next period. ,in The voltage stabilization duration specified for the strategy.

[0037] S22. Invoke the nonlinear pressure boosting control model to generate real-time pressure commands. The model is defined as follows: ; in, As for the current basic pressure, For the target increment, The boost time constant is For the callback stable time constant, It is an eardrum pressure relief factor used to create a small negative pressure before the pressure-boosting action begins.

[0038] S23, the calculated result The signal is sent to the underlying PLC controller, which controls the opening of the intake solenoid valve and the micro-movement of the exhaust valve through the pressure boosting and balance control system of the micro-hyperbaric oxygen chamber, thereby achieving precise pressure tracking.

[0039] For example: boost time constant The inherent inflation response inertia of the oxygen chamber's gas supply system is determined by the chamber volume, the power of the intake pump, and the flow resistance of the pipeline. It is obtained as follows: First, open the intake solenoid valve to its maximum opening (100%) and record the pressure inside the chamber from... Rise to Real-time pressure curve The processor constructs the standard response equation for a first-order inertial element. Then, the least squares method is used to iteratively solve the problem with the objective function of minimizing the sum of squared residuals: Finally, the result obtained after iterative convergence The value is used as the standard boost time constant for the current system and stored in the control algorithm library.

[0040] Callback steady time constant The pressure decay rate during adaptive callbacks depends on the flow capacity (Cv value) of the pressure relief valve and the instantaneous pressure inside the chamber. It is obtained as follows: First, under stabilizing pressure, the control system sends a standard unit pulse signal to the exhaust solenoid valve, with a duration of... (For example, 500ms); then, high-frequency acquisition of data during the pressure drop phase; according to the gas law, the first derivative of the pressure is proportional to the current pressure, i.e. Take the natural logarithm of the collected pressure drop curve. A linear regression analysis was performed on it. The reciprocal of the absolute value of the slope of the regression line is the callback settling time constant. .

[0041] Ear pressure relief factor The nonlinear flow characteristics of the pressure relief valve are described using the following method: During the factory calibration phase, the exhaust solenoid valve is measured under different opening commands. The flow coefficient below The flow coefficient curve is normalized to obtain the valve's inherent characteristic function. ,in, For opening degree The flow coefficient at that time The maximum flow coefficient when fully open; calculation ,in, It is the sequence of exhaust valve actions issued by the control system (e.g., a trapezoidal opening waveform).

[0042] In this embodiment 1, a nonlinear boost control model is used, and the user's tolerance index is monitored in real time. Dynamically adjust the callback coefficient This drives real-time pressure commands. An adaptive adjustment is performed during the initial pressurization phase. This method utilizes nonlinear control logic to actively construct the physiological adaptation window of the human Eustachian tube, eliminating the lag between mechanical pressurization and physiological regulation. This ensures pressurization efficiency while avoiding stinging sensations or barotrauma induced by accumulated eardrum pressure differences, thus improving user comfort and safety during hyperbaric oxygen therapy. Specifically, for example: the current pressure in the oxygen chamber is 10 kPa, and the plan is to pressurize to 15 kPa (… According to the preset strategy library (see Example 5), this increment belongs to the "slight pressure increase scenario," and the system will adopt the first control strategy. If the user experiences tinnitus symptoms, The callback coefficient is calculated. Then in At the start of the phase, the system automatically slightly opens the pressure relief valve, causing the pressure to decrease from 10 kPa to approximately 1.2 kPa, down to 8.8 kPa, and then enters a 5-minute stabilization period to help the user's eardrums reset and adapt. After the stabilization period, the system invokes a nonlinear pressure boosting control model to drive the pressure to gradually rise to the target value of 15 kPa over the following minutes.

[0043] S3. During the voltage boost and stabilization process, the active safety defense module operates synchronously to calculate the comprehensive risk index. Differentiating between sleep and abnormal states and implementing appropriate strategies includes the following sub-steps: S31, Real-time calculation of visual anomaly scores (Based on duration of eye closure and range of limb movements), abnormal biometric score and environmental anomaly scoring .

[0044] Biological sign abnormality score The calculation formula is: ; In the formula, for Real-time heart rate value at any given moment; The user's resting baseline heart rate; The historical standard deviation of heart rate data; for Real-time blood oxygen saturation at any given moment; The target reference blood oxygen level for the user; This represents the standard deviation of the tolerance fluctuation for blood oxygen data. , , where are weighting coefficients, representing the proportions of regulated heart rate and blood oxygen in the total risk score, respectively.

[0045] For example, in this embodiment 1, the weighting coefficient , The risk level was obtained through a combination of offline sample training and clinical expert verification: First, a physiological monitoring dataset containing a large number of users under different stress conditions was constructed, and professional doctors quantified the physiological risk level at each moment (e.g., 0 represents normal, 1 represents near-shock). Next, a linear regression algorithm was used to analyze the contribution weights of the standardized heart rate deviation and blood oxygen deviation terms to the risk labeling values. Finally, considering that the lethality of hypoxemia under micro-hypertension is higher than that of simple heart rate fluctuations, a safety penalty factor was introduced to correct the regression results, forcibly setting... and The final determined value is then permanently stored in the system's non-volatile memory.

[0046] Visual abnormality score The calculation formula is: ; For silent abnormal components based on eye-closing duration, a translational sigmoid activation function can be used to eliminate interference from normal blinking and short periods of eye-closing rest. ; In the formula, The threshold for the closed-eye alarm; for The duration of continuous eye closure at any given moment; This is the sensitivity coefficient.

[0047] To capture the agitated abnormal component based on the amplitude of limb movements, a sliding window energy integral model is used to capture continuous abnormal movements (such as convulsions or thrashing for help) while filtering out occasional rolling movements. ; In the formula, Base noise level for basic motion; For time windows; is the gain coefficient, used to map the integrated energy value to the interval [0,1].

[0048] For example, in this embodiment 1, the sensitivity coefficient The value is determined by fitting a large amount of statistical distribution data on blink and eye-closing duration. Specifically, it is achieved by analyzing the boundary between normal blinking (milliseconds) and eye-closing during sleep (seconds), and adjusting the coefficient to make the Sigmoid function close to the threshold. The gain coefficient has a steep gradient to ensure rapid differentiation between momentary blinks and sustained eye closure. The calibration was performed through an abnormal behavior simulation experiment. Specifically, the test subject simulated normal rolling over and violent convulsions. The coefficient was adjusted so that the cumulative energy generated by the convulsions reached the saturation region (close to 1.0) after being mapped by the hyperbolic tangent function, while the energy mapping value of normal rolling over remained in the low amplitude region, thereby achieving precise amplification of dangerous movements.

[0049] Environmental anomaly scoring This indicates the degree of threat posed by the physical environment inside the cabin to human life safety: ; In the formula, for Real-time oxygen concentration readings inside the cabin at all times; The safe melting threshold is determined by oxygen concentration. This represents the environmental sensitivity coefficient.

[0050] For example, in this embodiment 1, the environmental sensitivity coefficient The calibration is performed by combining the measurement fluctuation range of the oxygen concentration sensor with the response speed requirements of the safety fuse. Specifically, the sensor is set to activate when the oxygen concentration exceeds the fuse threshold. When dealing with a specific tiny quantity, the function output needs to quickly reach a high-risk value. This can be achieved by substituting the boundary conditions and solving the problem inversely. The value is set to ensure that the system can trigger a rapid increase in the environmental anomaly score in the early stages of exceeding the limit, while avoiding misjudgment caused by the sensor's background noise.

[0051] S32. Calculate the comprehensive risk index The formula is: ; In the formula, The weighting coefficients for the visual anomaly scoring items; The weighting coefficients for the scoring items related to abnormal biological signs. This represents the weighting coefficient for the environmental anomaly scoring item.

[0052] For example, in this embodiment 1, the weighting coefficient , , Using the analytic hierarchy process (AHP) combined with multidimensional risk simulation experiments, it was determined that: firstly, an expert group constructed a risk assessment matrix, establishing that environmental safety and physiological signs had a higher priority than visual behavior (i.e., The eigenvectors are calculated to obtain the basic weight ratios; subsequently, extreme conditions such as hypoxia, apparent death, and violent agitation are reproduced using a simulation chamber, with the requirement that "any single-dimensional extreme danger score (close to 1.0) must be able to independently trigger system emergency stop (i.e., Using “)” as the boundary constraint, the weights are numerically scaled and verified, and finally a set of coefficients that can balance sensitivity and false alarm rate are solidified.

[0053] S33, Setting Vital Sign Thresholds , and if A normal value indicates abnormal physical signs (such as a sudden drop in heart rate or abnormal blood oxygenation); conversely, a normal value indicates stable physical signs.

[0054] Set visual threshold , and if , which means to close your eyes and not move.

[0055] Set risk threshold and wake-up threshold The decision logic is as follows: when At that time, if and The patient is identified as being in a sleep state. The system starts a countdown and emits a gentle wake-up sound through the speaker 10 minutes before the end of the treatment. when At that time, if and The system was identified as being in an abnormal state. It immediately cut off the oxygen supply, fully opened the pressure relief valve, and sent an alarm SMS to the associated mobile phone. when When this occurs, it is considered a normal state.

[0056] For example, in this embodiment 1, the vital sign threshold The visual threshold was determined based on the statistical distribution of physiological baseline data from healthy individuals under micro-high pressure. Specifically, the upper bound of the 95% confidence interval of heart rate and blood oxygen fluctuation data from a large number of healthy subjects inside the chamber was selected as the judgment boundary to filter out normal physiological stress fluctuations. By analyzing the correlation between eye-closing duration and EEG sleep stages, a threshold was established. Specifically, visual characteristic scores corresponding to the onset of light sleep (N1 stage) (such as eye-closing lasting more than 3 minutes) were set as the cutoff value to distinguish between momentary blinking and continuous sleep. Risk threshold. and wake-up threshold The classification performance of the multimodal fusion model was optimized using the receiver operating characteristic (ROC) curve, whereby... The Youden Index point (equilibrium point) that maximizes the identification of sleep states is taken, while Select a high-sensitivity operating point (safety point) that ensures zero missed reports of abnormal conditions.

[0057] Specifically, for example: Scenario 1: User Li Si falls asleep in the cabin, keeping his eyes closed and remaining motionless for 15 minutes. However, her heart rate was 65 bpm and her blood oxygen saturation was 98%. ). Calculated The system does not issue an alarm; it only plays a prompt tone to remind passengers to disembark before the countdown ends.

[0058] Scenario 2: User Wang Wu suddenly falls into a coma due to hypoglycemia, with his eyes closed and no movement. ), and the heart rate rose from 80 to 120, accompanied by cold sweats (tested). ). Calculated The system detects an anomaly within 3 seconds, immediately stops oxygen supply, automatically depressurizes, and notifies the administrator.

[0059] In this embodiment 1, multimodal data fusion technology is used to calculate the comprehensive risk index. The system employs a dual-logic approach to assess user status. Specifically, it jointly analyzes visual abnormality scores (reflecting behavioral states, such as whether eyes are closed or the user is still) and biometric abnormality scores (reflecting physiological functions, such as heart rate and blood oxygen stability) to construct a discriminant model that distinguishes between normal sleep and pathological abnormalities. This method addresses the false alarms or missed alarms caused by existing technologies relying solely on single threshold alarms, ensuring the comfort and continuity of the treatment process. Furthermore, it can respond to situations such as apparent death, coma, or severe agitation (abnormal signs without movement or convulsions), immediately executing oxygen supply cut-off and pressure relief strategies to proactively defend and intelligently protect the user's life.

[0060] S4. After the treatment, a multi-dimensional health analysis report is generated and pushed to the user's terminal via the cloud. Specifically, for example, the system summarizes the data throughout the process, compares blood pressure changes before and after entering the chamber (e.g., from 135 / 90 to 128 / 82), blood oxygen changes, and matches abnormal patterns using a knowledge graph. If the report finds that the user triggers high-amplitude callbacks multiple times during the pressurization process, it will suggest that the user choose a gentler pressurization mode next time or have their ear, nose, and throat checked. Example 2

[0061] As a second embodiment of the present invention, based on Embodiment 1, a micro hyperbaric oxygen chamber device is also disclosed, specifically including: a chamber body, an intelligent control cabinet, a gas source system, and the data acquisition, monitoring, analysis, and control system described in Embodiment 1. The specific structure and connection relationship of each component are as follows: The cabin serves as the high-pressure load-bearing structure, and its interior is equipped with interfaces for a physical biosensor array and mounting positions for in-cabin vision devices. The physical biosensor array is specifically a set of medical-grade wearable components (such as a smart bracelet or chest patch) used to collect the user's physiological data. The data is transmitted wirelessly to the intelligent control cabinet via Bluetooth or Wi-Fi. Specifically, the in-cabin vision device is a high-definition infrared night vision camera installed on the cabin ceiling, with a field of view covering the user's head and torso area, used to capture behavioral and posture data in real time. And transmit video streams.

[0062] The air supply system is responsible for executing pressure regulation commands and specifically includes an air compressor, an air tank, a precision filter, and an intelligent valve island assembly. The intelligent valve island assembly is the core actuator for executing the nonlinear boost control model, and it includes a proportional intake solenoid valve, a micro-exhaust solenoid valve, and a safety emergency stop valve. The proportional intake solenoid valve receives analog signals from the control system and can... The opening degree continuously adjusts the intake air flow; the micro-motion exhaust solenoid valve is used to perform the "adaptive callback" action described in Example 1, and its flow coefficient... After specific calibration, to achieve The level of minor pressure relief; the safety emergency stop valve is a normally closed large-diameter solenoid valve, controlled by the "active safety defense module" in Example 1, when the comprehensive risk index... At that time, the valve is physically triggered to achieve rapid pressure relief within milliseconds.

[0063] The intelligent control cabinet integrates an edge computing unit, a PLC controller, and a human-machine interface terminal. The edge computing unit is pre-installed with an LSTM model and a Transformer vision model, responsible for receiving data from sensors and cameras and calculating the user's physiological tolerance in real time. and comprehensive risk index The calculation results are then converted into control parameters; the PLC controller communicates with the edge computing unit to receive dynamic pressure commands that include an adaptive callback mechanism. Furthermore, the PLC internally runs a PID closed-loop control algorithm, which will... Real-time pressure feedback from cabin pressure sensors The system performs comparisons and dynamically adjusts the opening degree of the proportional intake solenoid valve and the on / off state of the micro-exhaust solenoid valve in the gas source system; the human-machine interface terminal is installed on the surface of the control cabinet and is used to display the health analysis report and real-time pressure curve generated by the digital twin model.

[0064] In the specific implementation of Implementation 2 above, after the user enters the cabin and puts on the sensors, they click the start button on the human-machine interface terminal. The edge computing unit in the intelligent control cabinet first reads the user's historical data to construct an initial digital twin model. Subsequently, the gas supply system starts, and the edge computing unit generates an initial pressurization command according to a preset strategy library. During the pressurization process, if the physical and biological sensor array detects the user... (Heart rate variability) Abnormal fluctuations, the edge computing unit immediately calculates the tolerance. The pressure decreases, and the micro-motion exhaust solenoid valve in the intelligent valve island assembly, controlled by the PLC, opens to perform an adaptive callback operation, temporarily reducing and stabilizing the cabin pressure. If the cabin vision equipment detects severe convulsions in the user ( If the air intake is raised (and the active safety defense module determines that the status is abnormal), the PLC controller will directly cut off the air intake and open the emergency stop valve to achieve comprehensive protection for the user. Example 3

[0065] As a third embodiment of the present invention, based on embodiment 1, an LSTM (Long Short-Term Memory) network model is disclosed to predict the physiological trend characteristics of users under current stress. This LSTM model aims to address the time dependence and noise interference issues in physiological data during long-term hyperbaric oxygen chamber operations, accurately capturing the user's delayed physiological response to pressure changes through a gating mechanism. Specifically, it includes the following steps: A1. Constructing the temporal feature input vector: This involves processing the standardized physiological characteristic data... By time step Perform sliding window sampling to construct the input sequence Assume the current time is Select the past The input vector is represented as follows: (Data from 1 time step) (in The dimension of the raw physiological data, such as the number of channels for heart rate, blood oxygen, etc.

[0066] A2. LSTM unit internal computation: The input vector... The information is input into an LSTM unit, and the flow of information is controlled through forget gates, input gates, and output gates to compute the hidden state. The specific calculation logic is as follows: First, calculate the forget gate. Used to determine the cell state from the previous time step. Which information should be discarded (e.g., removing instantaneous heart rate fluctuation noise caused by body position adjustment)? ; In the formula, The Sigmoid activation function maps the output to... interval; Here is the weight matrix for the forget gate; For bias terms; This represents the concatenated vector of the hidden state from the previous time step and the current input.

[0067] Secondly, compute the input gate. and candidate cell state This is used to determine which new physiological changes (such as the slow decline in blood oxygen saturation with increasing blood pressure) will be updated in the cell state: ; ; In the formula, The hyperbolic tangent activation function is used to normalize the values ​​to... ; , These are the corresponding weight matrices; This is a bias term.

[0068] Next, update the cell state at the current moment. This step, through mathematical addition, effectively avoids the gradient vanishing problem, ensuring that long-term physiological tolerance characteristics are preserved. ; In the formula, This represents the Hadamard product, which is the element-wise dot product.

[0069] Finally, calculate the output gate. and the hidden state at the current moment : ; ; In the formula, This is a deep feature representation of physiological states that includes temporal context information.

[0070] A3. Mapping to generate physiological trend features: The hidden states output by the LSTM are... A fully connected layer is used for dimension mapping to obtain the final physiological trend feature vector used in the digital twin model. : ; In the formula, ; The feature mapping weight matrix; It is the bias vector; It is a ReLU activation function used to increase nonlinear expressive power and eliminate negative features (ensuring that physiological energy is non-negative).

[0071] In the specific implementation of the above embodiment 3: the system presets a time window. (Referencing physiological data from the past 30 seconds). If, during the blood pressure increase process, the user experiences a slow but steady increase in respiratory rate while blood oxygen levels remain stable, the LSTM model's forgetting gate... The weights that automatically suppress high-frequency fluctuations in respiratory rate, and the input gate It will enhance the memory of blood oxygen stability. The final output This will be represented as a feature vector characterizing highly compensatory adaptation. This vector is then input into the formula. During the calculation, although There will be some increase, but it will not immediately trigger a drastic reduction command. Instead, it will achieve a smooth transition by fine-tuning the boost rate, thus demonstrating the algorithm's ability to intelligently judge physiological trends. Example 4

[0072] As a fourth embodiment of the present invention, a Transformer visual model is disclosed based on Embodiment 1 for extracting behavioral feature vectors. Specifically, it includes: B1. Constructing Visual Temporal Embedding Vectors: This involves embedding preprocessed behavioral pose data... (Including eye closure, head pose Euler angles, and spatial coordinate change rates of key limb nodes) are mapped to a high-dimensional feature space. Let the input sequence be... ,in The time window length, This represents the original data dimension. To preserve time-series information, location encoding needs to be overlaid. : ; In the formula, Embedded as the initial input for the model; This is a linear projection matrix used for dimension alignment; This represents the hidden layer dimension of the Transformer model.

[0073] B2. Multi-head self-attention mechanism calculation: By calculating the correlation of actions at different times in the sequence, the model can focus on key dangerous action frames. First, the input... (No. The layer output is linearly mapped to the query matrix. Key matrix Sum matrix : ; In the formula, This is a learnable weight matrix.

[0074] Next, a scaling factor is introduced. Calculate the scaled dot product attention score: ; In the formula, The function operates on the last dimension, and the resulting attention map reflects the dependency weights between behavioral features at different times.

[0075] B3. Feedforward Network and Residual Connections: To enhance the model's nonlinear expressiveness and prevent deep network degradation, each sublayer is followed by normalization and residual connections, and includes a feedforward neural network. ; ; In the formula, This indicates a bullish self-attention strategy; It typically involves two linear transformations and a ReLU activation function: .

[0076] B4. Global Feature Aggregation and Mapping: After... After processing by the Transformer encoder, sequence features containing rich contextual information are obtained. To obtain the final behavioral feature vector The time dimension is weighted and aggregated, and then mapped to the target dimension. : ; In the formula, Indicates the first Feature vectors at each time step; To output the projection matrix; For bias terms; The activation function limits the output to Between these, ensure that the calculated dimensions match those of the cross-fusion matrix in Example 1.

[0077] In the specific implementation of the above embodiment 4: the system sets a time window. (That is, analyzing the motion flow over the past minute). For example, user Li Si experiences "eye closure" and "irregular micro-tremors of the head" during the boost process. The self-attention mechanism will detect the persistence and high frequency of the micro-tremors on the timeline, giving them a higher attention weight, while filtering out occasional posture adjustments. The final output... In the vector, the feature dimension corresponding to nervous system abnormalities will increase significantly (approaching 1). When this vector is input into the formula... During the calculation, the overall unsuitable energy will be rapidly increased. This prompts the active safety defense module to determine the risk level. If the risk level exceeds the threshold, the system will automatically perform an emergency stop and pressure relief operation, thereby achieving accurate identification and protection against atypical dangerous actions. Example 5

[0078] This is the fifth embodiment of the present invention, which discloses a preset strategy library for a pressure boosting and balancing control system for a micro hyperbaric oxygen chamber based on embodiment 1. This embodiment specifically defines the "preset strategy library" mentioned in step S2 of embodiment 1. In actual operation, the adaptive pressure control module decomposes the total target pressure into several execution sequences and determines the execution sequence based on the current target pressure increment. Depending on the scope, the following control strategies will be automatically applied: First control strategy (suitable for slightly pressurized scenarios): When an increase in target pressure inside the cabin is detected... Within the pressure range of 5-10 kPa: The system controls the exhaust valve to slightly adjust, automatically lowering the cabin pressure by 1-2 kPa before stabilizing and establishing an initial adaptation window; after stabilizing for 5 minutes, the intake pump is started, controlling the pressure to increase by 5-8 kPa, then the intake is stopped and the pressure is stabilized; after stabilizing for another 5 minutes, the pressure is further increased by 8-10 kPa, then stabilized; if the target is not reached, the pressurization action is performed at 10-13 kPa and 13-18 kPa every 5 minutes until the set pressure value is accurately reached and the final pressure stabilization state is entered.

[0079] The second control strategy (suitable for low-to-medium pressurization scenarios) is triggered when an increase in target pressure inside the cabin is detected. When the pressure is within the range of 10-15 kPa: the system automatically lowers the pressure inside the chamber by 2-3 kPa and then stabilizes it; after stabilizing the pressure for 5 minutes, the control pressure is increased by 8-10 kPa and stabilized; after stabilizing the pressure for another 5 minutes, the control pressure is increased by 10-13 kPa and stabilized; if further pressurization is required, the pressure is increased by 13-15 kPa after stabilizing for another 5 minutes until the set value is reached.

[0080] The third control strategy (suitable for medium pressurization scenarios) is triggered when an increase in target pressure inside the cabin is detected. When the pressure is within the range of 15-20 kPa: the system automatically lowers the pressure inside the chamber by 3-4 kPa and then stabilizes it; after stabilizing the pressure for 5 minutes, it increases the pressure by 10-15 kPa and stabilizes it; then the interval is adjusted to 4 minutes, and the pressure is increased by 13-15 kPa and stabilized; after another 5 minutes, the pressure is increased by 15-20 kPa; finally, after an interval of 4 minutes, the pressure is increased until the set value is reached.

[0081] The fourth control strategy (applicable to medium-high boost scenarios) is to detect an increase in the target pressure inside the cabin. When the pressure is within the range of 20-35 kPa (covering three sub-ranges: 20-25 kPa, 25-30 kPa, and 30-35 kPa): the system automatically lowers the chamber pressure by 4-5 kPa and then stabilizes it; after stabilizing the pressure for 4 minutes (the time is shortened to meet high pressure requirements), the control pressure is increased by 10-15 kPa and stabilized; then, every 4 minutes, the system sequentially performs a pressurization action with an amplitude of 15-20 kPa until the pressure sensor feedback value equals the set pressure value.

[0082] Fifth control strategy (applicable to high-pressure scenarios): When an increase in the target pressure inside the cabin is detected... When the pressure is within the range of 35-50 kPa (including the sub-ranges of 35-40 kPa and 40-50 kPa): the system automatically lowers the pressure inside the chamber by 5-6 kPa, using the larger negative pressure fluctuation to forcibly balance the pressure in the Eustachian tube; after stabilizing the pressure for 3 minutes (with the interval further shortened), the control pressure is increased by 10-15 kPa and stabilized; then every 3 minutes thereafter, the pressure is increased by 15-20 kPa (for the 35-40 kPa range) or 10-15 kPa (for the 40-50 kPa range) until the final pressurization is completed.

[0083] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0084] The preferred embodiments of the invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A data acquisition, monitoring, analysis and control system for use in a micro hyperbaric oxygen chamber, characterized in that, include: The multi-dimensional perception module is used to collect users' physiological signs and behavioral posture data in real time through physical biosensor arrays and in-cabin vision devices; The digital twin computing module is used to construct a digital twin model of human health and calculate the user's real-time physiological tolerance index using multimodal deep learning algorithms. The adaptive pressure control module is used to obtain the target pressure increment and map it to a specific pressure control range, call the nonlinear boost control model to dynamically calculate the real-time pressure command, and generate a negative pressure command to execute the pressure callback. The proactive security defense module is used to calculate a comprehensive risk index through multimodal data fusion, classify abnormal states, and execute corresponding strategies.

2. The data acquisition, monitoring, analysis and control system for the micro hyperbaric oxygen chamber according to claim 1, characterized in that, The formula for calculating the physiological tolerance index is as follows: ； In the formula, Provides a real-time physiological tolerance index for users. This is the sensitivity coefficient. Basic tolerance threshold, To address inappropriate energy levels.

3. The data acquisition, monitoring, analysis, and control system for the micro hyperbaric oxygen chamber according to claim 2, characterized in that, The comprehensive inadequacy energy The calculation formula is: ； In the formula, This is the physiological single-modal weight vector. For a single-modal weight vector, Physiological trend characteristics For behavioral feature vectors, This is a cross-fusion matrix.

4. The data acquisition, monitoring, analysis and control system for the micro hyperbaric oxygen chamber according to claim 2, characterized in that, The calculation formula for the real-time pressure command is as follows: ； In the formula, For real-time pressure commands, As for the current basic pressure, For the target increment, The boost time constant is For the callback stable time constant, As a factor for relieving eardrum pressure, This is the callback coefficient; The callback coefficient Based on real-time physiological tolerance index calculate: ,in This represents the maximum design reduction amount corresponding to the current step range.

5. The data acquisition, monitoring, analysis, and control system for the micro hyperbaric oxygen chamber according to claim 3, characterized in that, Physiological trend characteristics It is obtained through prediction using an LSTM model, which uses a forgetting gate. Input gate and output gate Controlling information flow and calculating hidden states And obtained through mapping of fully connected layers : ； In the formula, It is the ReLU activation function. The feature mapping weight matrix, This is the bias vector.

6. The data acquisition, monitoring, analysis and control system for the micro hyperbaric oxygen chamber according to claim 3, characterized in that, The behavioral feature vector Extracted using a Transformer visual model, which includes a multi-head self-attention mechanism and a feedforward network, outputting sequential features. And obtained through time-dimensional aggregation mapping : ； In the formula, Indicates the first Feature vectors at each time step To output the projection matrix, This is a bias term.

7. The data acquisition, monitoring, analysis and control system for the micro hyperbaric oxygen chamber according to claim 1, characterized in that, The formula for calculating the comprehensive risk index is as follows: ； In the formula, Score for visual abnormalities, For scoring abnormal biological signs, Score environmental anomalies. These are the corresponding weighting coefficients.

8. The data acquisition, monitoring, analysis and control system for the micro hyperbaric oxygen chamber according to claim 7, characterized in that, The abnormal state determination includes: setting a visual threshold. Vital signs threshold Risk threshold and wake-up threshold ;when And visual abnormality score Biological sign abnormality score When the patient is in a sleep state, awakening is only performed before the end of the treatment; when... ,and and If the situation is deemed abnormal, the oxygen supply should be immediately cut off and the pressure relief valve should be fully opened.

9. The data acquisition, monitoring, analysis and control system for the micro hyperbaric oxygen chamber according to claim 4, characterized in that, The adaptive pressure control module includes a preset strategy library, which adjusts according to the target pressure increment. Automatic matching control strategy: when When the pressure is in the range of 5-10 kPa, the first control strategy is executed: the exhaust valve is slightly adjusted down by 1-2 kPa and then stabilized, followed by staged pressure increase and stabilization. when When the pressure is in the range of 10-15 kPa, the second control strategy is implemented: the control pressure is reduced by 2-3 kPa and then stabilized, followed by phased pressure increase and stabilization. when When the pressure is in the range of 15-20 kPa, the third control strategy is implemented: the control pressure is reduced by 3-4 kPa and then stabilized, followed by phased pressure increase and stabilization. when When the pressure is in the range of 20-35 kPa, the fourth control strategy is implemented: the control pressure is reduced by 4-5 kPa and then stabilized, the stabilization time interval is shortened, and the pressure increase action is performed in stages with an amplitude of 10-15 kPa or 15-20 kPa. when When the pressure is in the range of 35-50 kPa, the fifth control strategy is implemented: the control pressure is reduced by 5-6 kPa to force the Eustachian tube pressure to be balanced, the pressure stabilization time interval is further shortened, and the pressure is increased in stages.

10. A micro-hyperbaric oxygen chamber device, characterized in that, Specifically, it includes: The chamber, intelligent control cabinet, gas source system, and data acquisition, monitoring, analysis and control system used in any one of claims 1 to 9, wherein the gas source system includes an air compressor, a gas storage tank, a precision filter and an intelligent valve island assembly; The intelligent control cabinet integrates an edge computing unit, a PLC controller, and a human-machine interaction terminal. The edge computing unit is connected to a physical biosensor array, an in-cabin vision device, and an intelligent valve island component. The intelligent valve island assembly includes a proportional intake solenoid valve, a micro-adjustment exhaust solenoid valve, and a safety emergency stop valve; the proportional intake solenoid valve is used to adjust the opening of the intake air volume, the micro-adjustment exhaust solenoid valve is used to perform adaptive callback action, and the safety emergency stop valve is used to perform rapid pressure relief in abnormal conditions.

Images