A terahertz physiotherapy parameter adaptive regulation method and intelligent physiotherapy cabin

Abstract

The application discloses a terahertz physical therapy parameter self-adaptive regulation method and an intelligent physical therapy cabin, and relates to the technical field of intelligent physical therapy equipment; a main controller of the application synchronously collects three paths of near-infrared, visible light and depth face images, extracts five types of features after pretreatment and face registration alignment, filters out transient interference through LSTM time sequence coding, outputs fatigue index, relaxation degree, pain grade and face subarea temperature relying on four parallel inference submodels, divides four levels of user states after checking data, combines multi-index interpolation to calculate global terahertz power, improves subarea physical therapy power according to the low-temperature area of the face, issues a module after safety clamping, linear smoothing transition, and synchronously and gradually changes and regulates sound and light parameters, immediately reduces power and alarms in an abnormal state, detects in a closed loop cycle in the whole physical therapy process, and automatically archives whole-process data at the end of a treatment course; the application realizes global + subarea self-adaptive physical therapy, has multiple safety checks, and improves the intelligentization, safety and physical therapy adaptation effect of the physical therapy cabin.

Description

Technical Field

[0001] This invention relates to the field of intelligent physiotherapy equipment technology, and in particular to a method for adaptive control of terahertz physiotherapy parameters and an intelligent physiotherapy chamber. Background Technology

[0002] At present, terahertz physiotherapy equipment has achieved the integration of multiple physical factors at the hardware level. For example, there are already terahertz light wave health care cabins with multiple radiation sources such as quantum emitters, far-infrared emitters, terahertz emitters and millimeter-wave emitters installed on the inner cabin walls, which can perform multi-band physiotherapy on the human body. Such equipment usually includes basic structures such as outer shell, inner cabin, cabin door, and physiotherapy bed. Some equipment is also equipped with a screw drive device to drive the bed to move, so as to achieve physiotherapy coverage of different parts of the human body.

[0003] CN215584661U discloses a terahertz light wave health care cabin, including an outer shell, an inner cabin, a door, and a treatment bed. The inner cabin is located inside the outer shell and is tubular. Several transmitter mounting slots are provided on the walls of the inner cabin. A telescopic frame parallel to the axial direction of the inner cabin is provided within the inner cabin. The fixed section of the telescopic frame is fixedly connected to the inner cabin. The treatment bed is slidably connected to the inner cabin via a sliding section of the telescopic frame. The door is located at the end of the treatment bed furthest from the inner cabin. However, the following technical shortcomings still exist: On the one hand, commonly... The existing control scheme does not incorporate a time-series filtering mechanism, making it susceptible to sudden interference such as blinking and instantaneous facial movements, which can cause frequent fluctuations in physiotherapy power. The power and audio-visual parameters switch abruptly, and the sudden changes in intensity can easily cause discomfort to users. At the same time, the existing equipment lacks the ability to adjust physiotherapy in different areas of the face, and cannot enhance the local physiotherapy output for areas with weak blood circulation. On the other hand, most products' physiotherapy modules and audio-visual psychological intervention systems are independent of each other and cannot be linked and coordinated. They also lack multi-level user status graded warning and multiple security verification mechanisms, and lack reliable protection measures for communication failures and parameter exceeding limits.

[0004] To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing intelligent physiotherapy cabins, such as the inability to perceive the user's physical and mental state in real time, rigid adjustment of physiotherapy parameters, single monitoring dimensions, lack of zoned physiotherapy, and inadequate safety protection. This invention provides a terahertz physiotherapy parameter adaptive control method and an intelligent physiotherapy cabin. It relies on near-infrared, visible light, and depth three-channel synchronous imaging to extract facial physiological features from multiple dimensions, combines a time-series model to eliminate instantaneous interference, automatically assesses fatigue, relaxation, pain, and facial microcirculation status, adaptively matches the terahertz power of the whole domain and zones, and links sound and light intervention for synchronous and smooth adjustment. It also adds multi-level status warnings, multiple parameter verifications, and communication fault protection to achieve closed-loop intelligent control of the entire physiotherapy process.

[0006] The objective of this invention can be achieved through the following technical solution: an intelligent physiotherapy chamber, comprising a chamber shell, an electrically operated sliding door slidably connected to one side of the chamber shell, a tempered glass observation window provided inside the electrically operated sliding door, a sliding device fixedly installed on the chamber shell, a physiotherapy massage mattress provided at the upper end of the sliding device, an adjustable backrest hinged to the side of the physiotherapy massage mattress near the electrically operated sliding door, an electrically adjustable leg rest hinged to the side of the physiotherapy massage mattress away from the adjustable backrest, an electrically operated push rod lifting mechanism fixedly connected to the lower surface of the physiotherapy massage mattress, a cooling fan fixedly connected inside the chamber shell, and a partition installed inside the chamber shell above the adjustable backrest.

[0007] A method for adaptive adjustment of terahertz therapy parameters in an intelligent physiotherapy chamber includes the following steps:

[0008] Step 1: Relying on the main controller to uniformly control the acquisition timing, acquire near-infrared images, visible light images, and depth images;

[0009] Step 2: After preprocessing and face alignment and registration of the three images, the multi-branch network extracts five types of features and fuses them. Ten frames of features are then used to complete the temporal coding through a preset temporal coding model.

[0010] Step 3: The temporal encoded feature vector is fed into the state assessment model inference engine, which contains four parallel inference sub-models to calculate the fatigue index, relaxation value, pain level and temperature values ​​of five facial regions, and integrates the output parameter package.

[0011] Step 4: First, verify the model output results and correct the data. Based on the state level discrimination conditions, divide the state levels, output the current state level according to the corresponding state level discrimination conditions, and construct the state data package.

[0012] Step 5: Calculate the final theoretical power of the terahertz module using the interpolation formula, increase the partition power in the low temperature region, verify the clamping and linear smooth transition, and send it to each module after safety verification and smooth transition. If there is no response, record the abnormal event and issue an alarm.

[0013] Step 6: Based on the current status level, output the module parameter list corresponding to the current status level, verify the terahertz power range, and if it matches, enter the normal gear synchronous switching process. Level D immediately interrupts the gradual change and issues an audible and visual alarm, while monitoring the remaining physiotherapy time.

[0014] Step 7: The physiotherapy session ends, the cabin door opens automatically, and a complete data record of this physiotherapy session is generated.

[0015] Preferably, the analysis process of the timing coding is as follows:

[0016] The acquired near-infrared images, visible light images, and depth images were sequentially subjected to bilateral filtering for noise reduction and CLAHE illumination normalization preprocessing. At the same time, the MTCNN algorithm was used to complete face alignment and registration.

[0017] The preprocessed near-infrared image, visible light image and depth image are input into a pre-defined facial feature extraction neural network with a multi-branch parallel structure. Each branch extracts the following five types of features: micro-expression features, facial color features, eyelid opening and closing features, skin texture features and facial depth deformation features.

[0018] The extracted five types of features are fused to obtain a single-frame standardized feature vector. Ten consecutive frames of standardized feature vectors are extracted and input into a preset temporal coding model to complete temporal coding, and the output temporal coding feature vector is generated.

[0019] Preferably, the parameter package generation and analysis process is as follows:

[0020] The temporally encoded feature vector is input into the state evaluation model inference engine, which contains four parallel inference sub-models: fatigue index inference sub-model, relaxation degree inference sub-model, pain perception level inference sub-model, and facial temperature distribution evaluation sub-model.

[0021] The output fatigue index, relaxation level, pain level, and temperature values ​​of the five facial regions are used to construct an output parameter package.

[0022] Preferably, the generation and analysis process of the status data packet is as follows:

[0023] The output results of the four parallel inference sub-models are validated for legality: if the fatigue index / relaxation value exceeds the range, it is forcibly corrected to the range boundary value; if the pain level is not an integer between 0 and 5, it is assigned a default value of 0; and if the temperature change amplitude of consecutive frames in the region exceeds the preset threshold, it is directly rejected.

[0024] Set the status level and status level discrimination conditions. The status levels include A: good status, B: normal status, C: deteriorating status and D: abnormal status.

[0025] Based on the set status level, status level discrimination conditions, and output parameter packet, the current status level is output, and a status data packet is constructed based on the current status level and the output parameter packet.

[0026] Preferably, the analysis process for the final theoretical power of the terahertz module is as follows:

[0027] Within the established baseline power range, the main controller combines the fatigue index P, relaxation value F, and pain level D, and uses an interpolation formula to calculate the final theoretical power TPz of the terahertz module. It retrieves the facial temperature distribution assessment results, and when the temperature of areas such as the forehead, cheek, and jaw is identified as being lower than the normal threshold, the terahertz emission partition power corresponding to that area is increased by a preset ratio a1 based on the overall theoretical power to obtain the regional theoretical power.

[0028] Preferably, a safety check is performed on the calculated key indicators. If any indicator exceeds the limit, it is forcibly clamped to a preset safety threshold. The final theoretical power and regional theoretical power of the terahertz module after passing the safety check are collectively referred to as the terahertz module power. The terahertz module power is subjected to linear smooth transition processing, the transition time is set, and the real-time power Pal(t) of the transition stage is calculated by formula.

[0029] After the parameters have been verified and smoothly transitioned, each module drive circuit returns a confirmation frame within a preset confirmation time. If no confirmation frame is returned within the preset confirmation time, its power is automatically reduced to 0 and the abnormal event is recorded. At the same time, an alarm is issued through the display screen.

[0030] Preferably, the analysis process for the module parameter list is as follows:

[0031] Based on the status data packet, read the current status level, and based on the preset comprehensive status level-module parameter linkage table, output the module parameter list corresponding to the current status level.

[0032] Based on the module parameter list, two sets of configurations are obtained at once: the terahertz power range corresponding to the current gear, which has been executed in the previous steps and is only used for consistency verification in this step; and the fixed parameters of acousto-optic subdivision corresponding to the current gear.

[0033] Consistency verification: Compare the terahertz power range obtained from the preset integrated status level-module parameter linkage table, calculate and verify the terahertz module power, and determine whether the terahertz module power falls within the corresponding range.

[0034] If the interval matches: proceed with the normal gear synchronous switching process; if the interval does not match: determine that the parameter is abnormal, record the log and continue using the gear of the previous cycle, and the sound and light psychological intervention module remains unchanged in its current state.

[0035] The normal gear (A / B / C level) synchronous switching process includes:

[0036] Determine if the current gear has changed from the previous historical gear: if the gear has not changed, no switching action is performed; if the gear has changed: initiate linear smooth transition processing, and all audio and visual parameters change synchronously and gradually.

[0037] Transition completion determination: When the cumulative transition time reaches the set transition time, the sound and light psychological intervention module is locked with all parameters of the target level, and the current level switch ends;

[0038] Abnormal gear: Once the current status level is detected as D, immediately terminate all ongoing smooth transition processes, without waiting for the transition to complete, and trigger a voice prompt;

[0039] After parameter adjustment is completed, the remaining duration of the treatment is determined, and the next round of facial status collection, analysis and parameter adjustment is performed or the treatment is switched to a deep relaxation mode.

[0040] The beneficial effects of this invention are as follows:

[0041] (1) This invention simultaneously acquires near-infrared, visible light and depth images, integrates micro-expression, facial color, eyelid, skin texture and depth deformation features, and uses LSTM temporal coding to filter transient interference. Based on the inference model calibrated by EEG and EMG, it can accurately quantify fatigue, relaxation, pain and facial microcirculation. The evaluation accuracy is much higher than that of a single visual acquisition scheme. The model output is corrected by four-level state classification and verification to avoid data abnormalities that cause control deviations.

[0042] (2) Based on the interpolation of the user's physical and mental indicators, the terahertz full-domain power is calculated to enhance the local physiotherapy intensity in the low-temperature area of ​​the face, taking into account both overall conditioning and local microcirculation improvement; the power and sound and light parameters adopt linear smooth transition to eliminate the discomfort caused by sudden intensity changes. The system is equipped with dual parameter verification and bus communication fault monitoring. In case of abnormal status, the power is reduced immediately and sound and light reminders are given. The multi-layer safety mechanism ensures reliable use.

[0043] (3) Physiotherapy and sound and light intervention are linked and adaptively adjusted to form a closed-loop control process. The soothing mode is automatically switched at the end of the treatment. The entire process time sequence data and abnormal logs are automatically archived after the physiotherapy is completed, which facilitates health traceability. No manual intervention is required throughout the process. It takes into account the physiotherapy effect, the comfort of use and the safety of equipment operation, and significantly improves the intelligence and humanization level of the physiotherapy cabin. Attached Figure Description

[0044] The invention will now be further described with reference to the accompanying drawings;

[0045] Figure 1 This is a schematic diagram of the main sectional view of the structure of the present invention;

[0046] Figure 2 This is a schematic diagram of the method of the present invention;

[0047] Figure 3 This is a schematic diagram of image preprocessing according to the present invention;

[0048] Legend: 1. Outer shell of the cabin; 2. Electric sliding cabin door; 3. Tempered glass observation window; 4. Adjustable reclining chair back; 5. Electric lifting leg rest; 6. Physiotherapy massage mattress; 7. Sliding device; 8. Electric push rod lifting mechanism; 9. Cooling fan; 10. Partition. Detailed Implementation

[0049] 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.

[0050] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments;

[0051] Example 1:

[0052] Please see Figures 1 to 3 As shown, the present invention is an intelligent physiotherapy cabin, including a cabin shell 1. An electric sliding door 2 is slidably connected to one side of the cabin shell 1. A tempered glass observation window 3 is provided inside the electric sliding door 2. A sliding device 7 is fixedly installed on the cabin shell 1. A physiotherapy massage mattress 6 is provided at the upper end of the sliding device 7. An adjustable backrest 4 is hinged to the side of the physiotherapy massage mattress 6 near the electric sliding door 2. An electric lifting leg rest 5 is hinged to the side of the physiotherapy massage mattress 6 away from the adjustable backrest 4. An electric push rod lifting mechanism 8 is fixedly connected to the lower surface of the physiotherapy massage mattress 6. A cooling fan 9 is fixedly connected inside the cabin shell 1. A partition 10 is installed inside the cabin shell 1 above the adjustable backrest 4.

[0053] That is: the user opens the electric sliding door 2 to enter the cabin, lies flat on the electrically raised physiotherapy massage mattress 6, and tilts the adjustable backrest 4 to 130° to adapt to the human body's stretching posture. Image acquisition units are symmetrically arranged on both sides of the cabin to collect visible light and near-infrared facial images in real time and transmit them to the bottom main controller. Two sets of terahertz emission arrays are arranged on the top layer, with the upper array covering the human torso and the lower array corresponding to the lower limb area, to achieve full-body zone irradiation. The cabin is equipped with a sound and light relaxation module to provide light, audio and vibration to assist relaxation.

[0054] A method for adaptive adjustment of terahertz therapy parameters in an intelligent physiotherapy chamber includes the following steps:

[0055] Step 1: Relying on the main controller to uniformly control the acquisition timing, acquire near-infrared images, visible light images, and depth images;

[0056] It relies on a near-infrared camera, a visible light camera, and a depth sensor camera installed about 400mm above the user's face on the top of the cabin. The three cameras are controlled by the synchronous trigger signal of the main controller to collect images in a unified manner, and one frame of image is collected synchronously every 2 seconds: near-infrared image (850nm wavelength, collected under low light conditions), visible light image (1920×1080 pixels), and depth image (ToF solution, accuracy ±5mm). All three images are accompanied by a unified timestamp.

[0057] Step 2: After preprocessing and face alignment and registration of the three images, the multi-branch network extracts and fuses five types of features. Ten frames of features are then used to complete the temporal coding through a preset temporal coding model to eliminate instantaneous interference and capture the temporal changes in facial state.

[0058] Specifically, the acquired near-infrared, visible light, and depth images are sequentially processed by bilateral filtering for noise reduction and CLAHE illumination normalization preprocessing. The MTCNN algorithm is then used to detect faces and locate 68 key points. Based on the coordinates of the key points, affine transformation alignment is performed, and the images are cropped into standardized facial images of a preset size. The near-infrared and depth images undergo the same transformation according to the calibration and registration relationship with the visible light images to ensure that the same pixel position in the three images corresponds to the same physical position on the face.

[0059] The preprocessed near-infrared image, visible light image and depth image are input into a pre-defined facial feature extraction neural network with a multi-branch parallel structure. Each branch extracts the following five types of features: micro-expression features, facial color features, eyelid opening and closing features, skin texture features and facial depth deformation features.

[0060] Among them, micro-expression motion features: using existing technologies (such as ResNet-18 backbone network) to identify the activation intensity of 24 commonly used facial action units, in order to infer the user's emotional state;

[0061] Facial features: The visible light image is converted to the HSV color space, and the face is divided into five sub-regions: forehead, left cheek, right cheek, nose, and jaw. The mean and standard deviation of each channel are calculated to generate a fixed-dimensional facial feature vector, which is used to assess blood circulation status and fatigue level.

[0062] Eyelid opening and closing characteristics: Six eyelid key points in each eye were located using a key point regression network, and the average eye opening and closing ratio (EAR value) was calculated to assess drowsiness and fatigue.

[0063] Skin texture features: Local binary mode (LBP) algorithm is used to extract facial skin micro-texture features from near-infrared images to assess skin condition and microcirculation changes;

[0064] Facial depth deformation features: Calculate the pixel-by-pixel difference between the current frame depth image and the user's resting reference depth image, and extract deformation features in four regions: forehead, around the eyes, nasolabial folds, and around the mouth to help improve the accuracy of micro-expression recognition;

[0065] The five extracted features are fused to obtain a single-frame standardized feature vector = {micro-expression action features, facial color features, eyelid opening and closing features, skin texture features, facial depth deformation features};

[0066] To capture the temporal evolution trend of facial states and eliminate instantaneous interference, standardized feature vectors of 10 consecutive frames (e.g., a time window of about 20 seconds) are input into a preset temporal coding model (e.g., a single-layer LSTM network) to complete temporal coding, outputting temporal coded feature vectors to eliminate instantaneous facial state interference and capture the dynamic change trend of states.

[0067] Example 2:

[0068] Step 3: The temporally encoded feature vector is fed into the state assessment model inference engine, which contains four parallel inference sub-models to calculate the fatigue index, relaxation level, pain level, and temperature values ​​of five facial regions, respectively. The integrated output parameter package is as follows:

[0069] The temporally encoded feature vector is input to the state evaluation model inference engine, which contains four parallel inference sub-models:

[0070] Fatigue index inference sub-model: Outputs a continuous fatigue index from 0 to 100, which is calculated based on multi-dimensional features such as the temporal trend of eyelid opening and closing, detection of yawning action units, and changes in facial brightness. The training labels are derived from the synchronously collected EEG α / θ wave power ratio.

[0071] Relaxation inference sub-model: Outputs a continuous relaxation value from 0 to 100, calculated based on features such as facial muscle tension and respiratory rhythm. The training labels are derived from facial electromyography signals and subjective scale scores.

[0072] Pain perception level inference sub-model: outputs pain level from 0 to 5, based on the combination of related action units of the Facial Pain Action Scale (PFACS);

[0073] Facial temperature distribution assessment sub-model: Estimates the temperature values ​​of five facial regions based on the grayscale distribution of near-infrared images to assist in assessing blood circulation status;

[0074] Obtain the fatigue index, relaxation level, pain level, and temperature values ​​of the five facial regions, and construct an output parameter package;

[0075] Step 4: First, verify the model output results and correct the data. Based on the state level discrimination criteria, divide the state into levels, output the current state level according to the corresponding state level discrimination criteria, and construct a state data package. Specifically:

[0076] The legality of the output results of the four parallel inference sub-models is checked: if the fatigue index / relaxation value exceeds the range, it is forcibly corrected to the range boundary value; if the pain level is not an integer between 0 and 5, it is assigned a default value of 0 (judged as no pain); and if the temperature change amplitude of the continuous frame in the region exceeds the preset threshold, it is directly rejected.

[0077] Set the status level and status level discrimination conditions. The status levels include A: good status, B: normal status, C: deteriorating status and D: abnormal status.

[0078] Status level determination criteria:

[0079] If the fatigue index ∈ [0, P1], the relaxation value ∈ [F1, 100], and the pain level is 0, then it is judged as Grade A: good condition;

[0080] If the fatigue index ∈ (P1, P2), relaxation value ∈ [F2, F1] and pain level are both 0 / 1, then it is judged as Grade B: normal condition;

[0081] If any one of the following conditions is met: fatigue index ∈ (P2, P3), relaxation value ∈ [F3, F2), and pain level is 2 / 3, then it is judged as Grade C: declining condition;

[0082] If any one of the following conditions is met: fatigue index ∈ (P3, 100], relaxation value ∈ [0, F3) and pain level is 4 / 5, then it is judged as Grade D: abnormal condition;

[0083] Among them, P1, P2 and P3 are all set fatigue index thresholds, 0 < P1 < P2 < P3 < 100, and F1, F2 and F3 are all set relaxation value thresholds, 100 > F1 > F2 > F3 > 0.

[0084] Based on the set status level, status level discrimination conditions, and output parameter packet, the current status level is output, and a status data packet is constructed based on the current status level, output parameter packet, and other information.

[0085] Example 3:

[0086] Step 5: Calculate the final theoretical power of the terahertz module using interpolation formulas. Increase the power in the low-temperature region, verify the clamping, and perform a linear smooth transition. After safety verification and smooth transition, send the power to each module. If there is no response, record the abnormal event and issue an alarm. Specifically:

[0087] Within the established baseline power range, the main controller, combining the fatigue index P, relaxation value F, and pain level D, uses an interpolation formula to calculate the final theoretical power of the terahertz module:

[0088] TPz=TPmin+(TPmax-TPmin)×[α×(100-P) / 100+β×(100-R) / 100-γ×D / 5];

[0089] Where α, β, and γ are weighting coefficients (default values ​​0.4, 0.3, and 0.3), and TPmax and TPmin are the lower and upper limits of the baseline interval corresponding to the current state level. The lower the fatigue index, the higher the physiotherapy power; the higher the pain level, the lower the power.

[0090] Retrieve facial temperature distribution assessment results. When the temperature of areas such as the forehead, cheeks, and jaw is found to be below the normal threshold, it is determined that the blood circulation of the corresponding human body part is weak. The terahertz emission power of the corresponding part is increased by a preset ratio a1 (a1 is between 10% and 20%) based on the theoretical power of the whole region to obtain the theoretical power of the region: Theoretical power of region = final theoretical power + final theoretical power × preset ratio a1;

[0091] Output the final theoretical power and regional theoretical power of the terahertz module;

[0092] Safety checks are performed on key indicators such as the final theoretical power of the terahertz module, the regional theoretical power, and the drive current. If any indicator exceeds the limit, it will be forcibly clamped to the preset safety threshold. For example, the final theoretical power of the terahertz module and the regional theoretical power must not exceed the rated output power of the terahertz module. If they exceed the limit, the output power will be the rated output power.

[0093] The final theoretical power and regional theoretical power of the terahertz module after passing the safety verification are collectively referred to as the terahertz module power. A linear smooth transition process is performed on the terahertz module power, with a set transition time (e.g., 5 seconds). The real-time power calculation formula during the transition phase is as follows:

[0094] Pal(t) = Pold + (Pnew - Pold) × t / T, where Pold is the original power, Pnew is the final theoretical power of the terahertz module / the theoretical power of the region, t is the time, and T is the total transition time. If a new control command is received during the transition period, the transition will restart from the current real-time power.

[0095] After safety verification and smooth transition, parameters (such as real-time power) are synchronously broadcast to each module via the MCB bus in the form of a broadcast frame. Each module's drive circuit returns a confirmation frame within a preset confirmation time. If no confirmation frame is returned within the preset confirmation time, its power is automatically reduced to 0 and the abnormal event is recorded. At the same time, an alarm is issued through the display screen.

[0096] Step Six: Based on the current status level, output the module parameter list corresponding to the current status level, verify the terahertz power range, and if a match is found, enter the normal gear synchronous switching process. Level D immediately interrupts the gradual change and issues an audible and visual alarm, while simultaneously monitoring the remaining therapy time. Specifically:

[0097] Based on the status data packet, read the current status level, and based on the preset comprehensive status level-module parameter linkage table, output the module parameter list corresponding to the current status level.

[0098] Preset comprehensive status level - module parameter linkage table, such as: Level A (good status), terahertz power range: 70%~90%, sound and light working mode: vitality wake-up mode, sound and light subdivision fixed parameters: volume 70%, color temperature 5500K, brightness 80%, body vibration medium level, etc.

[0099] Based on the module parameter list, two sets of configurations are obtained at once: the terahertz power range corresponding to the current gear, which has been executed in the previous steps and is only used for consistency verification in this step; and the acousto-optic subdivision fixed parameters corresponding to the current gear, which are used for consistency verification and gear analysis.

[0100] Consistency verification: Compare the terahertz power range obtained from the preset integrated status level-module parameter linkage table, calculate and verify the terahertz module power, and determine whether the terahertz module power falls within the corresponding range.

[0101] If the range matches: Proceed to the normal gear shifting process;

[0102] If the interval does not match: it is determined to be a parameter abnormality, the log is recorded and the previous cycle gear is used, and the sound and light psychological intervention module remains unchanged in its current state;

[0103] The normal gear (A / B / C level) synchronous switching process includes:

[0104] Determine if the current gear has changed from the previous historical gear: if the gear has not changed, do not perform the switching action;

[0105] When the gear position changes: a linear smooth transition process is initiated, and all audio and visual parameters change synchronously and gradually.

[0106] Smooth transition of acoustic and optical parameters: The acoustic and optical parameters are updated frame by frame using a linear interpolation formula. Taking any one acoustic and optical parameter as an example: Gnow(t) = Gold + (Gnow - Gold) × t / transition duration;

[0107] Gnow represents the target gear parameter in the module parameter list, and Gold represents the current running parameter;

[0108] Transition completion determination: When the cumulative transition time reaches the set transition time, the sound and light psychological intervention module is locked with all parameters of the target level, and the current level switch ends;

[0109] New parameters (gradual process parameters / final steady-state parameters) of the audio-visual psychological intervention module are sent to the audio-visual psychological intervention module frame by frame through the MCB bus (RS485 physical layer). The driving circuit of each audio-visual psychological intervention module returns an acknowledgment frame within the preset response time. If no acknowledgment frame is returned within the preset response time, a communication failure is determined, an abnormal log is recorded, and the audio-visual psychological intervention module maintains its current output and is not forcibly shut down.

[0110] Abnormal Level (Level D): Once the current status level is detected as Level D, immediately terminate all ongoing smooth transition processes, without waiting for the transition to complete, trigger a voice prompt, announce "User status is abnormal, physiotherapy intensity has been reduced to the minimum", and at the same time, display a pop-up reminder on the screen;

[0111] After completing the parameter adjustment, determine the remaining treatment time: if the remaining treatment time is greater than the preset remaining treatment time, return to step two and continue to perform the next round of facial state collection, analysis and parameter adjustment;

[0112] If the remaining duration of physiotherapy is less than or equal to the preset remaining duration of physiotherapy, the main controller controls all terahertz modules in the zone to linearly reduce the power to the preset power and synchronously switch to the deep relaxation mode.

[0113] Step 7: After the physiotherapy ends, the cabin door opens automatically and a complete data record of this physiotherapy is generated: After the physiotherapy duration ends, the main controller shuts down all terahertz transmission modules and other physiotherapy modules, the cabin door opens automatically, and the safety monitoring module stops working;

[0114] Automatically generate a complete data record for this physiotherapy session, including user identity, physiotherapy duration, full-cycle facial status time-series data, terahertz module parameter time-series data, security event logs, etc.

[0115] In summary, this method integrates multi-dimensional facial physiological features through three-channel synchronous imaging and combines a temporal model to filter out transient interference. Compared with single visible light detection, it can accurately capture subtle changes in fatigue, emotions, pain, and local microcirculation. The evaluation results are consistent with the true values ​​of physiological data such as EEG and EMG, significantly improving the reliability of recognition. It adopts a four-level state grading mechanism and performs boundary verification on the model output to avoid parameter runaway problems caused by acquisition noise. The grading logic is clear and controllable, and it is suitable for different users' physical and mental states.

[0116] Based on multi-index interpolation, the full-domain terahertz power is dynamically calculated, and the treatment intensity is adjusted according to the local lifting of the facial low-temperature area, taking into account both overall adjustment and targeted improvement of local blood circulation. The power and sound and light parameters are all output in a linear and smooth manner to avoid the physical stimulation caused by sudden parameter changes and improve the comfort of treatment. The system adds dual parameter consistency verification and bus communication anomaly monitoring. The module will automatically reduce power alarm if it does not respond. In the case of D-level abnormal state, the gradual change will be interrupted and a voice pop-up prompt will be given. Multiple safety protections ensure safe use.

[0117] The system integrates terahertz therapy with sound and light psychological intervention for synchronized regulation, achieving integrated adaptive adjustment of therapy and emotional relief. It features closed-loop monitoring throughout the therapy process, automatically switching to a relief mode near the end, and automatically archiving full-process time-series data and abnormal logs at the end of the treatment for easy health traceability analysis. The entire control process is fully automated, eliminating the need for manual parameter adjustment. It balances the effectiveness of therapy, user comfort, and equipment safety, significantly improving the intelligence and humanization of the intelligent therapy cabin.

[0118] The threshold is set for comparative analysis of results to determine whether they are good or bad. The value of the threshold is determined by a combination of large-scale model analysis of sample data and human experience. It can also be adjusted appropriately based on seasonal or common-sense influencing factors.

[0119] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An intelligent physiotherapy cabin, characterized in that, An adaptive control method for terahertz therapy parameters in an intelligent physiotherapy chamber includes the following steps: Step 1: Relying on the main controller to uniformly control the acquisition timing, acquire near-infrared images, visible light images, and depth images; Step 2: After preprocessing and face alignment and registration of the three images, the multi-branch network extracts five types of features and fuses them. Ten frames of features are then used to complete the temporal coding through a preset temporal coding model. Step 3: The temporal encoded feature vector is fed into the state assessment model inference engine, which contains four parallel inference sub-models to calculate the fatigue index, relaxation value, pain level and temperature values ​​of five facial regions, and integrates the output parameter package. Step 4: First, verify the model output results and correct the data. Based on the state level discrimination conditions, divide the state levels, output the current state level according to the corresponding state level discrimination conditions, and construct the state data package. Step 5: Calculate the final theoretical power of the terahertz module using the interpolation formula, increase the partition power in the low temperature region, verify the clamping and linear smooth transition, and send it to each module after safety verification and smooth transition. If there is no response, record the abnormal event and issue an alarm. Step 6: Based on the current status level, output the module parameter list corresponding to the current status level, verify the terahertz power range, and if it matches, enter the normal gear synchronous switching process. Level D immediately interrupts the gradual change and issues an audible and visual alarm, while monitoring the remaining physiotherapy time. Step 7: The physiotherapy session ends, the cabin door opens automatically, and a complete data record of this physiotherapy session is generated; The analysis process of the timing coding is as follows: The acquired near-infrared images, visible light images, and depth images were sequentially subjected to bilateral filtering for noise reduction and CLAHE illumination normalization preprocessing. The following five types of features were extracted from the preprocessed near-infrared images, visible light images, and depth images: micro-expression features, facial color features, eyelid opening and closing features, skin texture features, and facial depth deformation features. The extracted five types of features are fused to obtain a single-frame standardized feature vector. The standardized feature vectors of 10 consecutive frames are then input into a preset temporal coding model to complete temporal coding and output a temporal coding feature vector. The analysis process for the final theoretical power of the terahertz module is as follows: Within the established baseline power range, the main controller combines the fatigue index P, relaxation value F, and pain level D, and uses an interpolation formula to calculate the final theoretical power TPz of the terahertz module. It retrieves the facial temperature distribution assessment results, and when the temperature of the forehead, cheek, and jaw area is identified as being lower than the normal threshold, the terahertz emission partition power corresponding to that area is increased by a preset ratio a1 based on the overall theoretical power to obtain the regional theoretical power, where a1 is between 10% and 20%.

2. The intelligent physiotherapy cabin according to claim 1, characterized in that, The parameter package generation and analysis process is as follows: The temporal encoded feature vector is input into the state evaluation model inference engine, which contains four parallel inference sub-models: fatigue index inference sub-model, relaxation degree inference sub-model, pain perception level inference sub-model and facial temperature distribution evaluation sub-model. The output fatigue index, relaxation degree value, pain level and facial five-region temperature values ​​are used to construct the output parameter package.

3. The intelligent physiotherapy cabin according to claim 2, characterized in that, The generation and analysis process of the status data packet is as follows: The output results of the four parallel inference sub-models are validated for legality: if the fatigue index / relaxation value exceeds the range, it is forcibly corrected to the range boundary value; if the pain level is not an integer between 0 and 5, it is assigned a default value of 0; and if the temperature change amplitude of consecutive frames in the region exceeds the preset threshold, it is directly rejected. Set the status level and status level discrimination conditions. The status levels include A: good status, B: normal status, C: deteriorating status and D: abnormal status. Output the current status level, and construct a status data packet based on the current status level and the output parameter packet.

4. The intelligent physiotherapy cabin according to claim 3, characterized in that, Safety checks are performed on the calculated key indicators. If any indicator exceeds the limit, it is forcibly clamped to the preset safety threshold. The final theoretical power and regional theoretical power of the terahertz module after passing the safety check are collectively referred to as the terahertz module power. The terahertz module power is subjected to linear smooth transition processing, the transition time is set, and the real-time power Pal(t) of the transition stage is calculated by formula. After the parameters have been verified and smoothly transitioned, each module's drive circuit returns a confirmation frame within a preset confirmation time. If no confirmation frame is returned within the preset confirmation time, its power is automatically reduced to 0 and the abnormal event is recorded. At the same time, an alarm is issued through the display screen.

5. The intelligent physiotherapy cabin according to claim 4, characterized in that, The analysis process for the module parameter list is as follows: Based on the status data packet, read the current status level, and based on the preset comprehensive status level-module parameter linkage table, output the module parameter list corresponding to the current status level. Based on the module parameter list, consistency verification and power level analysis are performed. Consistency verification: The terahertz power range obtained by comparing with the preset comprehensive status level-module parameter linkage table is calculated and verified to determine whether the terahertz module power falls within the corresponding range. If the interval matches: enter the normal gear synchronous switching process, the normal gear includes A, B and C levels; if the interval does not match: determine that the parameter is abnormal, record the log and use the gear of the previous cycle, the sound and light psychological intervention module remains unchanged in the current state. Gear analysis and processing: The normal gear synchronization switching process includes: Determine if the current gear has changed from the previous historical gear: if the gear has not changed, no switching action is performed; if the gear has changed: initiate linear smooth transition processing, and all audio and visual parameters change synchronously and gradually. Transition completion determination: When the cumulative transition time reaches the set transition time, the sound and light psychological intervention module is locked with all parameters of the target level, and the current level switch ends; Abnormal gear: Once the current status level is detected as D, immediately terminate all ongoing smooth transition processes, without waiting for the transition to complete, and trigger a voice prompt; After parameter adjustment is completed, the remaining duration of the treatment is determined, and the next round of facial status collection, analysis and parameter adjustment is performed or the treatment is switched to a deep relaxation mode.

6. The intelligent physiotherapy cabin according to claim 1, characterized in that, The intelligent physiotherapy cabin includes a cabin shell (1), an electric sliding door (2) is slidably connected to one side of the cabin shell (1), a tempered glass observation window (3) is provided inside the electric sliding door (2), a sliding device (7) is fixedly installed on the cabin shell (1), a physiotherapy massage mattress (6) is provided at the upper end of the sliding device (7), an adjustable backrest (4) is hinged to the side of the physiotherapy massage mattress (6) near the electric sliding door (2), an electric lifting leg rest (5) is hinged to the side of the physiotherapy massage mattress (6) away from the adjustable backrest (4), an electric push rod lifting mechanism (8) is fixedly connected to the lower surface of the physiotherapy massage mattress (6), a cooling fan (9) is fixedly connected inside the cabin shell (1), and a partition (10) is installed inside the cabin shell (1) above the adjustable backrest (4).

Images