Adaptive overload AI intelligent adjustment posture method and system
By collecting autonomous physiological and environmental signals from the motion theater seats in real time, disability detection and graded adjustment are performed, solving the problem that existing technologies cannot identify the disability status of occupants, and achieving improvements in safety and intelligence.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANDONG TENGWEI TECH CO LTD
- Filing Date
- 2026-05-19
- Publication Date
- 2026-07-21
AI Technical Summary
Existing motion theater seat posture control methods cannot detect the occupant's disability status, especially fear-induced rigidity and fainting, and lack a graded safety intervention mechanism, posing a safety risk.
By collecting users' autonomous physiological signals and environmental constraint signals in real time, disability detection is performed, disability confidence is generated, and through multi-confidence fusion and graded adjustment strategies, the type of disability of the occupant is identified and corresponding safety interventions are carried out.
It achieves accurate identification of occupant disability status, provides progressive safety degradation control from minor intervention to emergency shutdown, reduces secondary injuries caused by abrupt stops, and improves the intelligence level and safety of motion theater seats.
Smart Images
Figure CN122431379A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of automatic control technology, and in particular to an adaptive heavy-load AI intelligent attitude adjustment method and system. Background Technology
[0002] Motion-controlled cinema seats typically possess multiple degrees of freedom of movement, capable of generating posture and movement sequences based on film visuals and sound effects to enhance audience immersion. Existing posture control methods mainly fall into three categories: manually preset motion codes, game-plot-based triggered movements, and music-rhythm-based automatic generation. However, all of these methods are open-loop controls, lacking effective perception and feedback of the physiological and behavioral states of occupants. In practical use, the following challenging problems arise: some occupants, due to excessive fear, sudden fainting, or physical restraint during film viewing, may need to stop their posture but are unable to issue a stop command via buttons or voice. Existing posture control methods cannot perceive that continuing to perform strenuous postures in this state of disability could worsen the injury; fear-induced rigidity, an evolutionarily preserved defense response, manifests as heart rate deceleration, muscle hypertension, and complete body rigidity, but existing posture control methods are completely unable to recognize this state and take targeted measures; when the system is unsure whether an occupant needs to stop, directly and abruptly stopping or completely ignoring the situation poses ethical and safety risks, and existing technologies lack tiered intervention and progressive safety degradation control logic. Summary of the Invention
[0003] The purpose of this invention is to provide an adaptive heavy-load AI intelligent posture adjustment method and system to overcome the shortcomings of existing dynamic cinema seat posture control methods, which cannot sense the occupant's disability state, cannot recognize the special physiological reaction of fear-induced rigidity, and lack a graded safety intervention mechanism.
[0004] The technical implementation of this invention is: an adaptive heavy-load AI intelligent attitude adjustment method, comprising the following steps: S1: Real-time acquisition of the user's autonomous physiological signals and environmental constraint signals during the movie-watching process; S2: Based on the autonomous physiological signals and environmental constraint signals, disability detection is performed, and corresponding disability confidence is generated; S3: Obtain the comprehensive disability confidence score and disability type label based on the generated disability confidence score; S4: Implement a tiered adjustment strategy based on the comprehensive disability confidence level and disability type label.
[0005] Preferably, the real-time acquisition of the user's autonomous physiological signals and environmental constraint signals during the movie-watching process includes: the autonomous physiological signals include heart rate signals and surface electromyography signals; the environmental constraint signals include seat belt tension signals, distributed pressure matrix, seat triaxial acceleration, rotary encoder angle, total seat load, and controller desired motion commands.
[0006] Preferably, the disability detection based on the autonomous physiological signals and environmental constraint signals, and the generation of corresponding disability confidence scores, includes: running fear-induced rigidity detection, constraint-restricted disability detection, and syncope-panic attack detection in parallel based on the collected autonomous physiological signals and environmental constraint signals; the fear-induced rigidity detection uses heart rate signals and surface electromyography signals to obtain rigidity disability confidence scores; the constraint-restricted disability detection uses seat belt tension signals, distributed pressure matrix, and seat triaxial acceleration to obtain restricted disability confidence scores; the syncope-panic attack detection uses heart rate signals, rotary encoder angle, total seat load, and distributed pressure matrix to obtain syncope-panic disability confidence scores; the disability confidence scores include rigidity disability confidence scores, restricted disability confidence scores, and syncope-panic disability confidence scores.
[0007] Preferably, the fear-induced rigidity detection uses heart rate signals and surface electromyography (EMG) signals to obtain a rigidity disability confidence level, including: defining a reference signal vector based on the seat's triaxial acceleration and the controller's desired motion command; performing multi-reference adaptive filtering on the heart rate signals and EMG signals to obtain filtered heart rate signals and EMG signals; and then obtaining the instantaneous phase of the heart rate signal through Hilbert transform. and the instantaneous phase of surface electromyography signals Within a preset sliding time window, the RR interval sequence is obtained based on the original heart rate signal. And calculate the standard deviation of heart rate variability in real time. Based on the instantaneous phase of the heart rate signal, the instantaneous phase of the surface electromyography signal, and the standard deviation of heart rate variability, through... Obtaining rigidity confidence factors ,in The standard deviation of heart rate variability serves as the reference baseline; the confidence level for rigidity disability is obtained based on the rigidity confidence factor. .
[0008] Preferably, the constraint-limited disability detection uses seatbelt tension signal, distributed pressure matrix, and seat triaxial acceleration to obtain the confidence level of constraint-limited disability, including: obtaining the inertial tension component based on the seat triaxial acceleration. Based on the inertial tension component and the original traction force extracted from the seatbelt tension signal; Obtain net tensile force For each sensor point in the distributed pressure matrix, the absolute value of the difference between the current pressure value and the pressure value at the previous moment is calculated sequentially. Sensing points whose absolute values are greater than a preset change sensitivity threshold are recorded as valid change points, and the flux function of the constrained region is obtained based on the valid change points. The constraint confidence index is obtained based on the net tensile force and the flux function of the constrained region. Finally, the restricted disability confidence level is obtained based on the aforementioned constraint confidence index. .
[0009] Preferably, the syncope and panic attack detection uses heart rate signals, a distributed pressure matrix, rotary encoder angles, and total seat load to obtain the confidence level of syncope and panic disability, including: based on the distributed pressure matrix... and seat triaxial acceleration Constructing a joint observation space ,in The distributed pressure matrix is vectorized; the joint observation space is projected onto the separate space to reconstruct the respiratory wave signal. Extract the respiratory amplitude sequence from the respiratory wave signal. and respiratory rate sequence Within a preset sliding time window, the RR interval sequence is obtained based on the heart rate signal. A state space is constructed based on the respiratory amplitude sequence and the RR interval sequence. And define a collapse index based on the state space. ; Calculate the respiratory irregularity index based on the respiratory rate sequence. The confidence level for syncope, panic, and disability was obtained based on the collapse index and the irregular breathing index. .
[0010] Preferably, obtaining the comprehensive disability confidence score and disability type label based on the generated disability confidence score includes: based on the rigidity disability confidence score... Limited disability confidence level and confidence level of fainting, panic, disability Obtain the disability type label and obtain the comprehensive disability confidence score using the comprehensive disability confidence score function. The comprehensive disability confidence function is: ; ; ; ; In the formula As the dominant item; The competition coefficient; For co-amplification terms; This is the magnification factor; This is a dispersion suppression term; These are the discrete coefficients; The mean confidence level for disability; The standard deviation of the confidence level for disability; It is a very small positive number; This is the sensitivity coefficient; This is the inhibition coefficient.
[0011] Preferably, the step of basing the stiffness disability confidence level Limited disability confidence level and confidence level of fainting, panic, disability Obtain disability type labels, including: if the confidence level of rigid disability is higher than the rigidity confidence threshold and the confidence level of restricted disability is lower than the restricted confidence threshold, then the disability type label is marked as rigid; if the confidence level of restricted disability is higher than the restricted confidence threshold and the confidence level of rigid disability is lower than the rigidity confidence threshold, then the disability type label is marked as constrained; if the confidence level of syncope-panic disability is higher than the syncope-panic confidence threshold, then the disability type label is forcibly marked as syncope-panic attack; otherwise, the disability type label is marked as mixed uncertainty.
[0012] Preferably, the tiered adjustment strategy based on the comprehensive disability confidence level and disability type label includes: if the comprehensive disability confidence level is less than a first comprehensive disability threshold, it is marked as a first disability level; if the comprehensive disability confidence level is greater than the first comprehensive disability threshold and less than a second comprehensive disability threshold, it is marked as a second disability level; if the comprehensive disability confidence level is greater than the second comprehensive disability threshold and less than a third comprehensive disability threshold, it is marked as a third disability level; if the comprehensive disability confidence level is greater than the third comprehensive disability threshold, it is marked as a fourth disability level; and different strategies are implemented for different disability levels based on the disability type label.
[0013] Preferably, the adaptive heavy-load AI intelligent attitude adjustment system includes: The data signal acquisition module collects the user's autonomous physiological signals and environmental constraint signals in real time during the movie viewing process; A dedicated disability detection module performs disability detection based on the autonomous physiological signals and environmental constraint signals, and generates a corresponding disability confidence level. The integrated disability fusion module obtains the integrated disability confidence score and disability type label based on the generated disability confidence score; The tiered strategy adjustment module performs a tiered adjustment strategy based on the comprehensive disability confidence level and disability type label.
[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention accurately identifies various disability states of occupants during movie viewing by real-time acquisition of occupants' autonomous physiological signals and environmental constraint signals, and by running three detection mechanisms in parallel: fear-induced rigidity, constraint-related disability, and syncope / panic attacks. Based on this, and using a multi-confidence fusion and graded adjustment strategy, it achieves progressive safety degradation control from mild intervention to emergency shutdown, effectively reducing secondary injuries caused by abrupt stops or blind operation. In particular, a dedicated recognition and response link has been established for the special physiological phenomenon of fear-induced rigidity, filling the perception blind spot of existing open-loop posture control systems and significantly improving the intelligence level and safety of motion theater seats. Attached Figure Description
[0015] Figure 1 A flowchart of an adaptive overloaded AI intelligent posture adjustment method; Figure 2 This is a schematic diagram of an adaptive heavy-load AI intelligent attitude adjustment system. Detailed Implementation
[0016] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.
[0017] Example 1 like Figure 1 As shown, an adaptive reload AI intelligent posture adjustment method includes the following steps: S1: Real-time acquisition of the user's autonomous physiological signals and environmental constraint signals during the movie-watching process; Autonomous physiological signals include heart rate signals and surface electromyography signals; environmental constraint signals include seat belt tension signals, distributed pressure matrix, seat triaxial acceleration, rotary encoder angle, total seat load, and controller desired motion commands.
[0018] It should also be noted that the heart rate signal is acquired by a non-contact photoplethysmography (PPG) sensor integrated into the seat headrest or seat belt shoulder strap at a sampling rate of no less than 250Hz; the surface electromyography (EMG) signal is acquired by high input impedance dry electrodes attached to the armrests on both sides at a sampling rate of 1000Hz for the forearm and shoulder girdle muscles; the seat belt tension signal is acquired by a tension / compression sensor connected in series between the buckle and the webbing; the distributed pressure matrix is composed of 128×128 flexible thin film pressure sensing units in the seat cushion and backrest, which output a two-dimensional pressure distribution at a scanning rate of 50Hz; the seat's triaxial acceleration is measured by a MEMS accelerometer installed at the rigid connection of the chassis; the rotary encoder angle is installed at each motion joint to measure the actual posture angle; the total seat load is obtained by summing the four weighing sensors below; the controller's desired motion command is read from the central controller via the CAN bus, including the desired displacement, velocity, and acceleration trajectory.
[0019] S2: Disability detection is performed based on autonomous physiological signals and environmental constraint signals, and corresponding disability confidence scores are generated; Based on the collected autonomous physiological signals and environmental constraint signals, the system performs parallel detection of fear-induced rigidity, constraint-related disability, and syncope-panic attack. Fear-induced rigidity detection uses heart rate signals and surface electromyography signals to obtain the confidence level of rigidity-related disability. Constraint-related disability detection uses seat belt tension signals, distributed pressure matrix, and seat triaxial acceleration to obtain the confidence level of constraint-related disability. Syncope-panic attack detection uses heart rate signals, rotary encoder angle, total seat load, and distributed pressure matrix to obtain the confidence level of syncope-panic disability. Disability confidence includes rigidity-related disability confidence, constraint-related disability confidence, and syncope-panic disability confidence.
[0020] Based on the triaxial acceleration of the seat and the desired motion command from the controller, a reference signal vector is defined. Multi-reference adaptive filtering is performed on the heart rate signal and surface electromyography (EMG) signal to obtain the filtered heart rate and EMG signals. Then, the instantaneous phase of the heart rate signal is obtained through Hilbert transform. and the instantaneous phase of surface electromyography signals Within a preset sliding time window, the RR interval sequence is obtained based on the original heart rate signal. And calculate the standard deviation of heart rate variability in real time. Based on the instantaneous phase of the heart rate signal, the instantaneous phase of the surface electromyography signal, and the standard deviation of heart rate variability, Obtaining rigidity confidence factors ,in The standard deviation of heart rate variability serves as the baseline; the confidence level for rigidity disability is obtained based on the rigidity confidence factor. .
[0021] The inertial pull component is obtained from the three-axis acceleration of the seat. The original traction force is extracted based on the inertial tension component and the tension signal through the seatbelt. Obtain net tensile force For each sensor point in the distributed pressure matrix, calculate the absolute value of the difference between the current pressure value and the pressure value at the previous moment. Sensing points whose absolute values are greater than a preset change sensitivity threshold are recorded as valid change points, and the flux function of the constrained region is obtained based on the valid change points. Constraint confidence indices are obtained based on the net tensile force and the flux function of the constrained region. Finally, the restricted disability confidence level is obtained based on the constraint confidence index. .
[0022] Based on the distributed pressure matrix and seat triaxial acceleration Constructing a joint observation space ,in Vectorization of the distributed pressure matrix; reconstructing respiratory wave signals by projecting the joint observation space onto the separate space. Extracting respiratory amplitude sequences from respiratory wave signals and respiratory rate sequence Within a preset sliding time window, the RR interval sequence is obtained based on the heart rate signal. A state space is constructed based on the respiratory amplitude sequence and the RR interval sequence. And define collapse indices based on state space. Calculate the respiratory irregularity index based on the respiratory rate sequence. Confidence level of syncope, panic, and disability was obtained based on collapse index and respiratory irregularity index. .
[0023] It should also be noted that when obtaining confidence levels for rigidity disability, the heart rate signal... and surface electromyography signals Multi-reference adaptive filtering is performed, and the filter structure is as follows: ,in The filtered signal includes the filtered heart rate signal. and surface electromyography signals ; The signal before filtering includes heart rate signal. and surface electromyography signals ; The filter order is 64-128; The adaptive filter coefficients are updated in real time using the least mean square algorithm, with the goal of minimizing the error between the output and the desired clean signal. As a reference signal vector, the three-axis acceleration of the seat and the desired motion command from the controller are combined to form the reference signal vector; the Hilbert transform of the filtered signal is calculated to obtain the instantaneous phase of the heart rate signal. and the instantaneous phase of surface electromyography signals The preset sliding time window is defined as follows: ,in The default value is 5 seconds; for the acquired RR interval series ,pass Obtain the standard deviation of heart rate variability ,in This represents the total number of RR interval sequences within a preset sliding time window; For the first RR interval sequence; The mean of the RR interval sequence within the preset sliding time window; the reference baseline for heart rate variability standard deviation is calculated by collecting the user's heart rate signal at rest before starting to watch the movie, and using the standard deviation of heart rate variability of the RR interval sequence as the reference baseline; rigidity disability confidence. ,in For rigidity confidence factors; The threshold for suspicion of rigidity confidence factors is determined by collecting rigidity confidence factors under normal movie-watching conditions and taking the 80th percentile of these factors as the threshold for suspicion. The 99th percentile of the normal distribution is taken as the confidence threshold for the rigid confidence factor.
[0024] When obtaining the confidence level of limited disability, the component of inertial tension generated by seat movement at the tether point is measured. and the original traction force in the seat belt tension signal pass To obtain the net tensile force; at each sampling time, the current pressure value of all sensing points in the distributed pressure matrix is used. Pressure value at the previous moment pass To obtain the absolute value of the difference; preset change sensitivity threshold. By collecting the background noise standard deviation of each sensor point in the distributed pressure matrix when the seat is unloaded and stationary, a threshold is set to 3-5 times the background noise standard deviation. Sensor points whose absolute value of the difference is greater than this preset change sensitivity threshold are recorded as valid change points. Constructing the flux function of the constrained region ,in This represents the total number of all sensor points; The region weight coefficients for each sensor point are assigned by normalized region weight coefficients based on the constraints of different areas of the seat on human posture and prior knowledge of ergonomics. The characteristic function is used to determine the net tensile force. and the flux function of the constrained region Min-Max normalization was performed to obtain the normalized net tensile force. and the normalized constraint region flux function Then through Obtaining Constraint Confidence Indicators ,in During the calibration period The mean; The length of the half-axis where the net tensile force is applied. , During the calibration period Standard deviation; The preset coefficients, and ; During the calibration period The mean; Let be the semi-axis length of the flux function in the constrained region. , During the calibration period Standard deviation; calibration period defined as the first 30 seconds after the user sits down and the seat remains stationary; limited disability confidence level. ,in This is the integration time window, with a default value of 5 seconds. This is a time scale value, and .
[0025] When obtaining the confidence level of syncope, panic, and disability, the real-time observation signal is projected onto the separation space. , ,in For joint observation space; The projection matrix is obtained by collecting joint observation spatial data when the seat is stationary and the occupant is breathing normally, and then performing independent component analysis to maximize the mutual information between a certain component after projection and the respiratory wave signal. A time-frequency mask based on respiratory physiological priors is applied to each projection component to reconstruct the respiratory wave signal. Then, the respiratory amplitude sequence is extracted from the reconstructed respiratory wave signal. and respiratory rate sequence Z-score standardization was performed on the RR interval sequence and respiratory amplitude sequence to obtain dimensionless variables. and The constructed state space ,in They are respectively and The derivative with respect to time; fitting a biquadratic potential function model during the calibration period. ,in During the calibration period, the least squares method was used for fitting; then, the Lagrangian was calculated based on the biquadratic potential function model. , Based on the obtained Lagrange quantities, through Calculate the degree of conservation index ,in Lagrange quantities during the calibration period The mean; ultimately defining the collapse index. Irregular breathing index ,in The mean of the respiratory rate sequence within a preset sliding time window; Interquartile range of respiratory rate during the calibration period; The median statistical function is taken as follows: Median; confidence level of syncope, panic, and disability ,in As a preset constant, The default value is 2.5. The default value is 1.5.
[0026] S3: Obtain the comprehensive disability confidence score and disability type label based on the generated disability confidence score; Based on the confidence level of rigidity disability Limited disability confidence level and confidence level of fainting, panic, and disability Obtain the disability type label and obtain the comprehensive disability confidence score using the comprehensive disability confidence score function. The comprehensive disability confidence function is: ; ; ; ; In the formula As the dominant item; The competition coefficient; For co-amplification terms; This is the magnification factor; This is a dispersion suppression term; These are the discrete coefficients; The mean confidence level for disability; The standard deviation of the confidence level for disability; It is a very small positive number; This is the sensitivity coefficient; This is the inhibition coefficient.
[0027] If the confidence level for rigid disability is higher than the rigidity confidence threshold and the confidence level for restricted disability is lower than the restricted confidence threshold, then the disability type label is marked as rigid; if the confidence level for restricted disability is higher than the restricted confidence threshold and the confidence level for rigid disability is lower than the rigidity confidence threshold, then the disability type label is marked as constrained; if the confidence level for syncope-panic disability is higher than the syncope-panic confidence threshold, then the disability type label is forcibly marked as syncope-panic attack; otherwise, the disability type label is marked as mixed uncertainty.
[0028] It should also be noted that the competition coefficient is obtained by collecting sample data offline under normal viewing conditions and various disability states. A grid search is performed with the goal of maximizing the accuracy of disability detection. The typical value range is 1-5, with larger values taken when the signal-to-noise ratio is high to enhance the discriminative power of the dominant term. The amplification coefficient is calibrated offline using the principle of maximizing the area under the ROC curve. The statistical distribution differences of the pairwise products of the three disability confidence levels under normal and disability states are analyzed to ensure that the synergistic amplification term can moderately improve the overall confidence without saturation. The typical value range is 0.5-2. The discrete coefficient... The confidence level is calibrated by simulating the confidence level distribution under normal and mixed disability conditions based on the ratio distribution of the confidence level standard deviation to the mean. The dispersion suppression term tends to 1 when the confidence level is highly dispersed and tends to 0 when it is concentrated, with a typical value range of 2-4. The sensitivity coefficient is determined by balancing the response speed and false trigger rate of disability detection through field tests, with a typical value range of 0.5-2. The suppression coefficient is optimized by analyzing the overall confidence level stability and safety degradation effect under multiple disability scenarios, with a typical value range of 0.2-0.3.
[0029] The rigidity confidence threshold is calculated by collecting rigidity disability confidence scores under normal movie-watching conditions, taking the 95th percentile as the initial threshold, and then fine-tuning it using simulated or measured rigidity attack data to maximize the detection rate and minimize the false alarm rate. The default value is 0.5. The restricted disability confidence threshold is based on the restricted disability confidence distribution under seat vacancy and active large-scale occupant movement, taking the 90th percentile as the threshold. The default value is 0.4. The syncope and panic confidence threshold uses sample data simulating syncope attacks using physiological signals, and determines the optimal cutoff point through ROC curves. The default value is 0.3.
[0030] S4: Implement a tiered adjustment strategy based on the overall disability confidence level and disability type label.
[0031] If the overall disability confidence level is less than the first overall disability threshold, it is marked as the first disability level; if the overall disability confidence level is greater than the first overall disability threshold and less than the second overall disability threshold, it is marked as the second disability level; if the overall disability confidence level is greater than the second overall disability threshold and less than the third overall disability threshold, it is marked as the third disability level; if the overall disability confidence level is greater than the third overall disability threshold, it is marked as the fourth disability level; different strategies are implemented for different disability levels according to the disability type label.
[0032] It should also be noted that the first comprehensive disability threshold is obtained by collecting the comprehensive disability confidence distribution under normal movie-watching conditions, and taking the 70th percentile as the first threshold, with a default value of 0.3; the second comprehensive disability threshold is based on the comprehensive confidence statistics of early rigidity or mild constraint limitation samples, and taking the 85th percentile as the second threshold, with a default value of 0.6; the third comprehensive disability threshold is based on the comprehensive confidence lower bound of samples with clear syncope or severe rigidity, and taking the 95th percentile as the third threshold, with a default value of 0.9.
[0033] For Level 1 disability, except for the disability type label "fainting / panic attack," other disability type labels maintain their original trajectory, accompanied by a voice prompt asking if assistance is needed. For the "fainting / panic attack" disability type label, additional ventilation is activated, and impact-related movements are suspended. For Level 2 disability, the trajectory movement amplitude is reduced to 50%, and a voice prompt is added. For users with the "rigidity" disability type label, their seat's three-axis acceleration is reduced to 1 / 3 of the gravitational acceleration. For users with the "constraint" disability type label, their seatbelt pretension is reduced. For users with the "fainting / panic attack" disability type label, the trajectory movement amplitude is reduced to 40%, slowly returning to center within 5 seconds, and medium-level ventilation is activated. For users with the "mixed uncertainty" disability type label, impact-related movements are suspended. For Level 3 disability, for users with the "rigidity" disability type label, the index gradually returns to center within 15-30 seconds, maintaining low-frequency vibration after centering, and a wristband warning is sent. For users tagged as "Restricted Injury," the system will quickly center and unlock the seatbelt within 2 seconds, with voice guidance to disembark. For users tagged as "Shock or Panic Attack," the system will center within 5 seconds, tilt the backrest 15°, and activate the emergency broadcast. For users tagged as "Mixed Uncertainty," the system will center within 4 seconds and flash a yellow ambient light. For users at the fourth disability level, tagged as "Rigidity," the system will center and lock within 0.5 seconds, prevent the backrest from tilting back, unlock the restraints, and call for medical assistance. For users tagged as "Restricted Injury," the system will center and lock within 0.5 seconds, forcefully unlock the seatbelt, and trigger the theater broadcast. For users tagged as "Shock or Panic Attack," the system will center within 0.5 seconds, tilt the backrest 30°, broadcast throughout the theater, and call 120 (emergency services). For users tagged as "Mixed Uncertainty," the system will center and lock within 0.5 seconds, prevent the backrest from tilting back, and only vibrate the staff wristband.
[0034] Example 2 like Figure 2 As shown, based on Embodiment 1, an adaptive heavy-load AI intelligent attitude adjustment system includes: The data signal acquisition module collects the user's autonomous physiological signals and environmental constraint signals in real time during the movie viewing process; A dedicated disability detection module detects disability based on autonomous physiological signals and environmental constraint signals, and generates corresponding disability confidence levels. The integrated disability fusion module obtains the integrated disability confidence score and disability type label based on the generated disability confidence score; The tiered strategy adjustment module adjusts the strategy based on the comprehensive disability confidence level and disability type label.
[0035] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. An adaptive overload AI intelligent attitude adjustment method, characterized in that, Includes the following steps: S1: Real-time acquisition of the user's autonomous physiological signals and environmental constraint signals during the movie-watching process; S2: Based on the autonomous physiological signals and environmental constraint signals, disability detection is performed, and corresponding disability confidence is generated; S3: Obtain the comprehensive disability confidence score and disability type label based on the generated disability confidence score; S4: Implement a tiered adjustment strategy based on the comprehensive disability confidence level and disability type label.
2. The adaptive heavy-load AI intelligent attitude adjustment method according to claim 1, characterized in that, The real-time acquisition of the user's autonomous physiological signals and environmental constraint signals during the movie-watching process includes: the autonomous physiological signals include heart rate signals and surface electromyography signals; the environmental constraint signals include seat belt tension signals, distributed pressure matrix, seat triaxial acceleration, rotary encoder angle, total seat load, and controller desired motion commands.
3. The adaptive heavy-load AI intelligent attitude adjustment method according to claim 1, characterized in that, The disability detection based on the autonomous physiological signals and environmental constraint signals, generating corresponding disability confidence scores, includes: parallel operation of fear-induced rigidity detection, constraint-restricted disability detection, and syncope-panic attack detection based on the collected autonomous physiological signals and environmental constraint signals; the fear-induced rigidity detection uses heart rate signals and surface electromyography signals to obtain rigidity disability confidence scores; the constraint-restricted disability detection uses seat belt tension signals, distributed pressure matrix, and seat triaxial acceleration to obtain restricted disability confidence scores; the syncope-panic attack detection uses heart rate signals, rotary encoder angle, total seat load, and distributed pressure matrix to obtain syncope-panic disability confidence scores; the disability confidence scores include rigidity disability confidence scores, restricted disability confidence scores, and syncope-panic disability confidence scores.
4. The adaptive heavy-load AI intelligent attitude adjustment method according to claim 2, characterized in that, The fear-induced rigidity detection uses heart rate and surface electromyography (EMG) signals to obtain a rigidity disability confidence level, including: defining a reference signal vector based on the seat's triaxial acceleration and the controller's desired motion command; performing multi-reference adaptive filtering on the heart rate and EMG signals to obtain filtered heart rate and EMG signals; and then obtaining the instantaneous phase of the heart rate signal through Hilbert transform. and the instantaneous phase of surface electromyography signals Within a preset sliding time window, the RR interval sequence is obtained based on the original heart rate signal. And calculate the standard deviation of heart rate variability in real time. Based on the instantaneous phase of the heart rate signal, the instantaneous phase of the surface electromyography signal, and the standard deviation of heart rate variability, through... Obtaining rigidity confidence factors ,in The standard deviation of heart rate variability serves as the reference baseline; the confidence level for rigidity disability is obtained based on the rigidity confidence factor. .
5. The adaptive heavy-load AI intelligent attitude adjustment method according to claim 2, characterized in that, The constrained disability detection uses seatbelt tension signal, distributed pressure matrix, and seat triaxial acceleration to obtain constrained disability confidence, including: obtaining inertial tension component based on the seat triaxial acceleration. Based on the inertial tension component and the original traction force extracted from the seatbelt tension signal; Obtain net tensile force For each sensor point in the distributed pressure matrix, the absolute value of the difference between the current pressure value and the pressure value at the previous moment is calculated sequentially. Sensing points whose absolute values are greater than a preset change sensitivity threshold are recorded as valid change points, and the flux function of the constrained region is obtained based on the valid change points. The constraint confidence index is obtained based on the net tensile force and the flux function of the constrained region. Finally, the restricted disability confidence level is obtained based on the aforementioned constraint confidence index. .
6. The adaptive heavy-load AI intelligent attitude adjustment method according to claim 2, characterized in that, The syncope and panic attack detection uses heart rate signals, a distributed pressure matrix, rotary encoder angles, and total seat load to obtain the confidence level of syncope, panic, and disability, including: based on the distributed pressure matrix... and seat triaxial acceleration Constructing a joint observation space ,in The distributed pressure matrix is vectorized; the joint observation space is projected onto the separate space to reconstruct the respiratory wave signal. Extract the respiratory amplitude sequence from the respiratory wave signal. and respiratory rate sequence Within a preset sliding time window, the RR interval sequence is obtained based on the heart rate signal. A state space is constructed based on the respiratory amplitude sequence and the RR interval sequence. And define a collapse index based on the state space. ; Calculate the respiratory irregularity index based on the respiratory rate sequence. The confidence level for syncope, panic, and disability was obtained based on the collapse index and the irregular breathing index. .
7. The adaptive heavy-load AI intelligent attitude adjustment method according to claim 3, characterized in that, The step of obtaining the comprehensive disability confidence score and disability type label based on the generated disability confidence score includes: based on the rigidity disability confidence score... Limited disability confidence level and confidence level of fainting, panic, and disability Obtain the disability type label and obtain the comprehensive disability confidence score using the comprehensive disability confidence score function. The comprehensive disability confidence function is: ； ； ； ； In the formula As the dominant item; The competition coefficient; For co-amplification terms; This is the magnification factor; This is a dispersion suppression term; These are the discrete coefficients; The mean confidence level for disability; The standard deviation of the confidence level for disability; It is a very small positive number; This is the sensitivity coefficient; This is the inhibition coefficient.
8. The adaptive heavy-load AI intelligent attitude adjustment method according to claim 7, characterized in that, The confidence level based on the rigidity and disability Limited disability confidence level and confidence level of fainting, panic, and disability Obtain disability type labels, including: if the confidence level of rigid disability is higher than the rigidity confidence threshold and the confidence level of restricted disability is lower than the restricted confidence threshold, then the disability type label is marked as rigid; if the confidence level of restricted disability is higher than the restricted confidence threshold and the confidence level of rigid disability is lower than the rigidity confidence threshold, then the disability type label is marked as constrained; if the confidence level of syncope-panic disability is higher than the syncope-panic confidence threshold, then the disability type label is forcibly marked as syncope-panic attack; otherwise, the disability type label is marked as mixed uncertainty.
9. The adaptive heavy-load AI intelligent attitude adjustment method according to claim 1, characterized in that, The tiered adjustment strategy based on the comprehensive disability confidence level and disability type label includes: if the comprehensive disability confidence level is less than a first comprehensive disability threshold, it is marked as a first disability level; if the comprehensive disability confidence level is greater than the first comprehensive disability threshold and less than a second comprehensive disability threshold, it is marked as a second disability level; if the comprehensive disability confidence level is greater than the second comprehensive disability threshold and less than a third comprehensive disability threshold, it is marked as a third disability level; if the comprehensive disability confidence level is greater than the third comprehensive disability threshold, it is marked as a fourth disability level; and different strategies are applied to different disability levels based on the disability type label.
10. An adaptive heavy-load AI intelligent attitude adjustment system, used to implement the adaptive heavy-load AI intelligent attitude adjustment method described in any one of 1-9, characterized in that, include: The data signal acquisition module collects the user's autonomous physiological signals and environmental constraint signals in real time during the movie viewing process; A dedicated disability detection module performs disability detection based on the autonomous physiological signals and environmental constraint signals, and generates a corresponding disability confidence level. The integrated disability fusion module obtains the integrated disability confidence score and disability type label based on the generated disability confidence score; The tiered strategy adjustment module performs a tiered adjustment strategy based on the comprehensive disability confidence level and disability type label.