Handheld intelligent physiotherapy instrument OTA system

Abstract

A handheld intelligent physiotherapy instrument OTA system comprises a handheld intelligent physiotherapy instrument, a background server can be operated conveniently by introducing a high-definition camera module, custom wrinkle calibration is performed, the severity of wrinkles is intelligently recognized, and the problems of high-precision recognition memory and severity judgment of wrinkle parts are solved. Meanwhile, the function is continuously updated through the OTA technology, the load and the production cost of the intelligent chip are reduced, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to an OTA system, in particular to a handheld intelligent physiotherapy instrument OTA system, belonging to the field of beauty therapy. Background Art

[0002] When it comes to the five major functions of cleansing, nourishing, lifting and firming, eye care, and blue light calming, existing beauty devices often only provide approximate action time and frequency in the instruction manual, as well as recommended action time and frequency for different skin types. The control of these parameters is left to the user to actively choose according to their skin type.

[0003] Faced with the update and iteration of multiple functions, especially the addition of intelligent functions, how to achieve partial updates of the system without the need to re-download the entire system has become an urgent problem to be solved. In the automotive industry, OTA technology is often used in car systems. However, this technology has not been implemented for small physiotherapy products because the general industry believes that the functions of physiotherapy instruments are fixed. No differentiation in customer needs has been found. However, customization has become the goal of personalized design today. For example, how to intelligently select parameters suitable for the user's skin has become the primary task of function development, which requires continuous system updates. Summary of the invention

[0004] In the face of the above four major topics, the present invention considers the following three aspects: first, the parameters of the five major functions are refined, and the personalized parameters required for one of the five major functions selected by the user are intelligently identified through machine learning based on face skin recognition; second, the automatic application scheme of the randomly matched pre-action physiotherapy fluid is considered; third, a mechanical face-lifting mechanism is designed, which is an external mechanical auxiliary function intervened when executing the five major functions, to realize a multiple-repetition physiotherapy scheme of external mechanical action + electric pulse + optical action; fourth, a high-definition shooting function is set on the physiotherapy device, and the computing power of the smart chip is strengthened; fifth, an OTA system with continuously updated functions is realized.

[0005] In view of the above considerations, an object of the present invention is to provide a handheld intelligent physiotherapy instrument OTA system, including a handheld intelligent physiotherapy instrument and a background server, wherein the handheld intelligent physiotherapy instrument communicates with the background server via OTA, so that the handheld intelligent physiotherapy instrument can selectively download updates of original functions or expanded new functions from the background server, wherein the handheld intelligent physiotherapy instrument includes: a handheld shell, a physiotherapy head, a switch / mode button, a charging port, a mode display and a charging display, a pinhole camera for collecting skin images of a human face, and a battery in the handheld shell, wherein the physiotherapy head includes a A massage head for vibrating and / or MFIP and a light therapy care source capable of emitting amber light, red light, crimson light, and blue light, wherein the mode display comprises a display for switching five functions of cleaning, nourishing, lifting and firming, eye care, and blue light calming in sequence by short pressing a switch / mode button, characterized in that the physiotherapy head further comprises a touch sensing part coupled to the massage head and a mechanical skin lifting mechanism arranged on the physiotherapy head outside the massage head, wherein the touch sensing part automatically or manually activates the mechanical skin lifting mechanism according to the facial skin contacting the massage head, so as to achieve a multi-directional extension effect on the skin level simultaneously with the physiotherapy;

[0006] The handheld intelligent physiotherapy instrument also includes a high-definition camera module, which can be used to capture faces, and a sub-shell that is detachably extended from one side of the handheld shell and is provided with a high-definition touch screen with a front camera. A gripping finger groove is provided between the sub-shell and the extended handheld shell on one side to facilitate gripping and operating the high-definition touch screen. When the high-definition camera module is started, a real-time face image appears on the high-definition touch screen. At this time, the holding hand does not move, and the other hand can zoom in on the real-time face image, circle the wrinkle area, and determine to upload it to the background server for storage. When the next facelift is performed, the high-definition camera module recognizes the current real-time face image through the face recognition function in the smart chip, and calls out the saved face image of the circled wrinkle area to indicate the user's desired facelift area. The user can delete the area of ​​action on the high-definition camera module and confirm to upload it to the background server again to save and update the face image, wherein the optimization of the face recognition function is performed by the background server and the user selectively chooses whether to OTA download it to the smart chip for update. The optimization of face recognition can be developed to reduce the amount of calculation, such as analyzing the detailed features of the facial features, collecting the facial contour and comparing it with the previously saved contour, or analyzing the distribution of the detailed features of the facial features in the contour. When the facelift effect is achieved and the user believes that it is not necessary to do it again, the user can choose to delete it. Due to the reduction in the amount of calculation, there may be a certain probability of recognition errors, so it is necessary to consider the user discovering and reporting the errors to the backend server.

[0007] Therefore, preferably, when the user finds that the called face image is incorrect, the user reports it to the background server through the high-definition touch screen, and through the uploaded registered account, the background server sends the correct face image, so that the high-definition touch screen is updated in OTA mode, and the wrong recognition result is automatically replaced with the OTA update result at the next recognition to make a correct recognition. In this way, the wrong result is automatically replaced with the correct OTA result without optimizing the algorithm itself, which is equivalent to optimizing the algorithm. The user will not be aware of this, but only believes that the recognition accuracy has improved.

[0008] A plurality of physiotherapy agent discharge holes are arranged on the side of the physiotherapy head away from the mechanical facelift mechanism. The physiotherapy agent discharge holes are connected through conduits to a plurality of physiotherapy agent receiving chambers inside the handheld shell, which are sealed by a transparent material body and one side of which passes through the surface of the handheld shell to display the piston position therein. Each of the plurality of physiotherapy agent receiving chambers can be pushed by a piston push rod to squeeze out the physiotherapy agent contained in the physiotherapy agent receiving chamber. A push portion arranged at one end of the piston push rod protrudes from the guide groove on the outer surface of the handheld shell. When the push portion is moved along the guide groove, the push portion is pushed out of the guide groove. When the guide groove moves, the current pushed position of the piston can be observed through the transparent material body, and the corresponding physiotherapy agent can be discharged from the physiotherapy agent receiving chamber, discharged from the corresponding physiotherapy agent discharge hole through the conduit and applied to the finger or the face to be treated, and then applied by means of the finger or the side of the physiotherapy head respectively; protective caps are provided on the multiple physiotherapy agent discharge holes, so that the protective caps can be covered when not in use to prevent the physiotherapy agent components from volatilizing and / or deteriorating; the piston push rod can be pulled out by disassembling the handheld housing to add the corresponding physiotherapy agent into the physiotherapy agent receiving chamber;

[0009] The smart chip arranged inside the handheld housing is used to set parameters based on smart settings according to the skin image acquisition and one of the functions of cleaning, nourishing, lifting and firming, eye care, and blue light calming selected by the user, so that after the parameter setting is completed and the physiotherapy agent is applied, the mechanical skin-lifting mechanism is automatically or manually started by the contact between the touch sensing part and the facial skin, and the physiotherapy of the selected function is performed; when the facial skin or the skin of any other part contacts the touch sensing part and then leaves the touch sensing part to present a non-contact state, the mechanical skin-lifting mechanism is reversely moved and the physiotherapy of the selected function is automatically suspended. At this time, the user can forcibly press the switch / mode button for 2-3 seconds to terminate the physiotherapy and enter the mode of reselecting the function, and press the switch / mode button twice in succession to confirm the selected function; or press the switch / mode button for 4-5 seconds to shut down the handheld smart physiotherapy instrument.

[0010] And the age input part, the age input part button or voice input method, for the button input method, that is, cyclic switching among multiple age groups (the current switched age group can be displayed in the manual gear shift panel), when selected, press and hold the age input part button for 1-2 seconds to confirm; for the voice input method, voice input can be performed multiple times and the most recent voice input will be cyclically overwritten until no voice input is received within 2-5 seconds.

[0011] Preferably, the user can establish a treatment record through the high-definition touch screen, and control the front camera to collect images of the push-up position through the high-definition touch screen. The back-end server calculates the recommended push-up position based on the collected skin image to adapt to the user's skin, and downloads the recommended push-up position image to the high-definition touch screen through OTA, and automatically inputs it into the smart chip after the secondary shell is connected to the handheld shell, so as to calculate the recommended position again for each future use, wherein the calculation algorithm for calculating the recommended push-up position based on the collected skin image can be continuously optimized by the back-end server, and the user can choose whether to update it regularly through OTA. That is, when the user thinks that the current algorithm is satisfactory, the OTA update can be selected.

[0012] Optionally, the recalculation method is that, based on the current recommended position of the background server, the amount of physiotherapy agent usage corresponding to the recommended positions in the future shows a decreasing function trend; the OTA update also includes the user being able to choose whether to make the following update: that is, when the deletion operation is performed on the high-definition touch screen, a dialog box will pop up to display a message reminding the user that the facelift needs to be consolidated and persisted, and asking whether it is really necessary to delete it. Therefore, through the notice, the user is prevented from being overconfident and missing the opportunity for consolidation to produce a rebound effect.

[0013] Optionally, the subtraction function is linear, and a scale is provided on the handheld housing passing through part of the edge, which can read the position of the piston on the scale by comparing it with the piston position. After the recalculation is completed, the smart chip can display the recommended scale position indication for multiple future uses on the high-definition touch screen for user reference.

[0014] Preferably, the pause time of the physical therapy that automatically pauses the execution of the selected function is included in the compensation time after the next contact for physical therapy, and the compensation time is not more than 1 minute.

[0015] It is understandable that if the contact again is with a non-therapy area or any other object, the user is generally likely to move away immediately or move away due to a change of therapy area. This is equivalent to pausing again, and the time lost in therapy due to incorrect contact or the time incurred during the change of area is also included in the compensation time, thereby avoiding as much as possible the shortening of therapy time due to incorrect contact.

[0016] Preferably, the handheld shell includes a detachable front part and a rear part based on a partition, and the partition divides the space enclosed by the installed front part and the rear part into a front cavity and a rear cavity that are independently sealed from each other, and the rear cavity accommodates the multiple therapeutic agent accommodating cavities, the corresponding piston push rods, part of the push part, and at least part of the catheter, and the therapeutic agent discharge hole is located on the side of the therapeutic head opposite to the mechanical facelift mechanism; the front cavity is used to accommodate the battery and the smart chip in the handheld shell; the switch / mode button, mode display and charging display are located at the front, and the charging hole is arranged at the bottom of the front cavity at one end belonging to the front.

[0017] Preferably, the plurality of therapeutic agent containing chambers are detachably fixed on the partition.

[0018] Optionally, in one embodiment, the mechanical facelift mechanism includes a back plate fixed to the therapy head and having a center column, a slave gear sleeved on the center column of the back plate, a main gear meshing with the slave gear, a motor accommodated in the internal cavity of the therapy head and connected to the main gear via a reducer or directly connected to the main gear, and a plurality of push heads, wherein the central hole of the main gear is connected to the reducer shaft through a through hole on the back plate, the slave gear has a plurality of rotary grooves the same in number as the push heads, the push head has a push rod, the push rod can be slidably embedded in the slide groove of the back plate, one end of the push rod has a convex rod passing through the rotary groove, and when the motor drives the When the main gear rotates clockwise or counterclockwise, the convex rod is driven by the rotation of the slave gear to slide in the rotation groove, thereby pushing the pusher head rods of the multiple pusher heads radially outward in the slide groove to make the multiple pusher heads open. Thereafter, at any time when the motor drives the main gear to rotate in the opposite direction, the pusher head rod moves radially inward in the slide groove to make the multiple pusher heads converge in the opposite direction, thereby achieving external auxiliary pulling effect on the skin. The motor is communicatively connected to the smart chip, and the back plate has at least one wire hole between the pusher head rods for passing the components and / or wires of the touch sensing part, massage head, and light therapy care source through so as to communicate with the smart chip.

[0019] Optionally, the motor is a stepper motor.

[0020] In another embodiment, the mechanical facelift mechanism includes a facelift scissors, the front end of which has a pair of push arcs arranged on the side where the massage head is located and can be pressed against the skin, and the rear end handle passes through the internal cavity of the therapy head and out of the outer shell of the therapy head so that the hand can hold it and press the handle relative to perform a shearing operation. After the skin is pressed, the handle is released to achieve a skin tightening operation, thereby achieving reverse push-pull and tightening of the skin under the compression state by the front end push arc. The part of the therapy head that is penetrated forms a pair of guide grooves to guide the handle to perform the shearing operation and the skin tightening operation. At this time, when the user chooses to push the push part, the therapeutic agent will be discharged from the therapeutic agent discharge hole and applied to the fingers or other application materials (such as cotton pads, etc.) to apply the therapeutic agent to the treatment area.

[0021] Optionally, the intelligent setting method comprises the following steps:

[0022] S1 set up five test groups: cleaning group, nourishing group, lifting and firming group, eye care group, and blue light calming group. Each test group was divided into six groups: 18-24 years old, 25-35 years old, 35-45 years old, 45-60 years old, 60-75 years old, and over 75 years old. Each group was set up with a training set, a validation set, and a test set, and the ratio of the three was 5-3:1-2:1. All subjects followed the preset work and rest routine for 1-3 months. The work and rest schedule is to get up at 7:30 am for those aged 18-24, 7 am for those aged 25-35, 7 am for those aged 35-45, 6:30 am for those aged 45-60, 6 am for those aged 60-75, and 6 am for those aged 75 and above. All test groups go to bed between 10-11 pm. Three meals a day, breakfast, lunch, and dinner, start within 30 minutes after getting up, between 11:30-12:30, and between 17:30-18:30 respectively.

[0023] The five experimental groups in S2 were trained with the following parameters for the artificial intelligence model:

[0024] Cleaning group, low frequency vibration low, medium and high three gears, amber light intensity low, medium and high three gears, total cleaning time under warm effect 4min, seven parameters,

[0025] Nourishing group, high-frequency vibration low, medium, high three gears, red light intensity low, medium, high three gears, total nourishing time under warming effect 4min, seven parameters,

[0026] Lifting and firming group, low frequency vibration low, medium and high three gears, high frequency vibration low, medium and high three gears, MFIP pulse frequency high, medium and low three gears, pulse intensity high, medium and low three gears, deep red light intensity, amber light intensity, total lifting and firming time under warm effect 4min, fifteen parameters,

[0027] Eye care group, MFIP pulse frequency high, medium and low three gears, pulse intensity high, medium and low three gears, low frequency vibration low, medium and high three gears, high frequency vibration low, medium and high three gears, high and low frequency vibration interval, total eye care time under warming effect 4min, fourteen parameters,

[0028] Blue light calms, blue light intensity, ice compress effect is to cool down the massage head for 4 minutes, one parameter;

[0029] When S3 is turned on, it uses a pinhole camera to collect images of the face to be treated and uploads them to the backend server. The backend server identifies whether the skin is sensitive based on a preset sensitive skin recognition model (i.e., based on facial features such as burn scars, bloodshot eyes, and blush in the captured images). If so, the action time of all groups is halved, and the training sets of each group are subjected to a cross-test program of parameter selection for the five major functional therapies. The top N parameter configurations (N≤5) are obtained through an overall evaluation, and the top three classification scores of the therapy results for each parameter configuration are taken. The corresponding captured images are input into the corresponding artificial intelligence model for training, and the validation set is used to verify the accuracy. The loss function is used to adjust the model The network parameters are adjusted until the accuracy and loss function values ​​are stable, and the training is stopped, thereby obtaining the classification of the top three scores of each parameter configuration of each functional therapy under each group as good, medium, and poor (that is, the three scores are divided into three levels from the highest to the relatively lowest); by randomly selecting the middle parameter configuration of each parameter configuration corresponding to the good or medium prediction result of each parameter configuration as the corresponding functional therapy under the group, the parameter configuration with the highest score among the middle parameter configurations of N parameter configurations is compared as the best parameter configuration of the functional therapy under the group, and if there are multiple highest scores that are the same, one of the parameter configurations is randomly selected as the best parameter configuration of the functional therapy under the group, thereby forming a trained artificial intelligence model;

[0030] The S4 high-definition touch screen downloads the artificial intelligence model trained in step S3 from the background server via OTA, inputs the age of the subject in the test set, uses a pinhole camera to capture images of the subject's face to be treated, and inputs the captured images into the corresponding trained artificial intelligence model to obtain five sets of optimal parameter configurations predicted under each functional treatment, thereby selecting the corresponding optimal parameter configuration for treatment preparation based on the function selected by the subject, and performing treatment based on the subject's face to be treated contacting the touch sensing part.

[0031] Through the cross-test in the early stage of training, obviously inappropriate parameter configurations for each age group were eliminated, so that the parameter configurations of the top N in training were concentrated. Considering the accuracy of training, there was an additional step to randomly select the intermediate parameters between good and medium as a compensation solution with a certain probability of simulation prediction deviation. Because one of the good and medium must be the potential best parameter configuration solution, if the model prediction is wrong, then the medium is the best choice. When the model estimation is correct, even if the medium is finally selected, the probability of producing undesirable results is relatively low. Finally, the highest score among all parameter configurations is used as the best parameter configuration for the corresponding function under this group.

[0032] Optionally, the therapeutic agents corresponding to cleansing, nourishing, lifting and firming, and eye care are makeup removers; the corresponding therapeutic agents are at least one of toner, lotion, face cream, essential oil, and facial mask; the corresponding therapeutic agents are at least one of gel, viscous essence, and firming essence; eye cream.

[0033] Preferably, the artificial intelligence model is a convolutional neural network CNN, or a ResNet based on a residual mechanism.

[0034] Optionally, the handheld intelligent physiotherapy device also includes a gear button arranged on the front, which is used to set the frequency and intensity of electric pulses, and / or the vibration frequency, and / or the intensity of light emitted by the phototherapy care source, and / or the rotation angle of the motor; the mode display also includes a manual gear adjustment panel, which is used to switch the display between the adjustment modes of electric pulses, the intensity of light emitted by the phototherapy care source, and the rotation angle of the motor by pressing the switch / mode button for 1-2 seconds multiple times, and then use the gear button to select the gear, and finally press the switch / mode button again to successfully set it.

[0035] Preferably, the therapy head is also provided with a microphone for receiving voice control sounds, and the front part also includes at least one sound outlet hole for voice prompts to issue age input prompts, respond to user voice commands, user operations, and voice prompts for function execution status.

[0036] Preferably, there are two switch / mode buttons and one of them is a spare button. The user can select one of them for operation at any time when the device is turned on. When one of the buttons fails, the spare button can still be used for operation.

[0037] Preferably, the mode display and charging display can be displayed on the high-definition touch screen, the switch / mode button, the front gear button, the standby button, the age input button, and the OTA button all form touch buttons on the high-definition touch screen, and a face image acquisition button is provided on the high-definition touch screen; the smart chip also calls a pre-trained wrinkle recognition model to perform identification marking on a face image based on a face image captured by the user pressing the face image acquisition button.

[0038] Optionally, the wrinkle recognition model is a convolutional neural network model with a residual mechanism, and the training method is to obtain volunteers of different age groups, and for each age group, take pictures of multiple different parts that the volunteers themselves identify as needing a facelift, and divide the unacceptable degree of wrinkles into three levels: negligible, non-negligible, and deep processing according to the volunteers' own identification, and use the high-definition camera module to collect multiple high-definition images of the three levels of parts for different age groups for each part, and divide the multiple high-definition images of the parts into training sets and verification sets according to different parts, and upload them to the background server through the high-definition touch screen, and the background server forms a sub-convolutional neural network with a residual mechanism for each part , use the training set for training, calculate the loss function to optimize the model parameters, and use the validation set to verify the model accuracy to obtain the final wrinkle recognition model. At this time, the high-definition touch screen downloads the trained wrinkle recognition model to the smart chip through the OTA method; during recognition, the user selects his own age group and clicks on the required face-lifting part on the collected face image, so that the smart chip can take a screenshot and enlarge the image part, or the user imports the saved face image of the circled wrinkle part through the high-definition touch screen operation, clicks on the required face-lifting part on the collected face image, and inputs it into the sub-convolutional neural network with residual mechanism corresponding to the trained corresponding part, so as to finally obtain the classification level of wrinkle recognition in each age group, which is displayed on the high-definition touch screen.

[0039] Through the description of the above algorithms, the present invention uses the OTA method to hand over some model training tasks (including the calculation algorithm of the push part position, the optimization of the face recognition function, the artificial intelligence model in the intelligent setting method, and the wrinkle recognition model) to the background server, thereby reducing the cost of the smart chip.

[0040] Preferably, both sides of the auxiliary housing have interfaces electrically connected to the handheld housing, and when one side is connected to the handheld housing, the high-definition touch screen is turned on, thereby realizing the usage habits of left-handed and right-handed people.

[0041] Preferably, the high-definition touch screen is independently powered by a battery in the sub-shell, and a sub-shell battery charging port is provided on the sub-shell. When the electrical connection is disconnected, the captured facial image can still be enlarged and the front camera can be controlled to capture the image of the push-up position.

[0042] Another object of the present invention is to provide a facial skin physiotherapy method using the handheld intelligent physiotherapy device provided by the present invention, which is characterized by comprising the following steps:

[0043] (1) Turn on the device and input age, then face the pinhole camera to collect images and complete parameter setting based on intelligent settings;

[0044] (2) Select the therapy function by using the switch / mode button;

[0045] (3) discharging the corresponding physiotherapy agent through the physiotherapy agent discharge hole to act on the facial skin to be treated;

[0046] (4) contacting the facial skin to be treated with the therapeutic agent in step (3) with the touch sensitive part to perform the selected therapeutic function; or,

[0047] After powering on, start directly from step (2) by using the switch / mode button to wake up the manual gear adjustment panel; then long press the switch / mode button to select manual adjustment of the electric pulse frequency and intensity, and / or the vibration frequency, and / or the intensity of the light emitted by the phototherapy care source, and / or the rotation angle of the motor, use the gear button to select the gear, and finally long press the switch / mode button again to successfully set;

[0048] Preferably, when image acquisition fails, a voice prompt of image acquisition failure can be given through the sound outlet every 3-5 seconds until the user successfully acquires the image, or the manual adjustment panel is awakened to enter the next step or prompted to enter the manual adjustment parameter mode and then perform steps (3) and (4) in sequence.

[0049] The present invention also includes a wrinkle tightening treatment method based on the above-mentioned handheld intelligent physiotherapy instrument, which specifically includes starting the high-definition touch screen on the sub-shell, facing the high-definition camera module to collect real-time high-definition images of the user, manually amplifying the real-time face image, circling the wrinkle area, and confirming to save. When the face is lifted next time, the high-definition camera module recognizes the current real-time face image through the face recognition function in the smart chip, and prompts on the high-definition touch screen whether to call out the saved face image of the circled wrinkle area. When "yes" is selected, it is asked again whether the recognition model is needed. When "not needed" is selected, the user's desired face lift area is indicated on the called displayed image; when "needed" is selected, the wrinkle recognition model step is entered; when "no" is selected, the wrinkle recognition model step is also entered. Wrinkle recognition model step, when entering the wrinkle recognition model step, press the face image acquisition button to collect a face image, call the pre-trained wrinkle recognition model to perform recognition marking on the face image, during specific recognition, the user first selects his own age group, clicks on the desired face-lifting part on the collected face image, and the smart chip can take a screenshot to enlarge the image part, or the user imports the saved face image of the circled wrinkle part through the high-definition touch screen operation, clicks on the desired face-lifting part on the collected face image, and inputs the sub-convolutional neural network with residual mechanism corresponding to the trained corresponding part, so as to finally obtain the classification level of wrinkle recognition in each age group, and the user determines the type, dosage and tightening time of the physiotherapy agent according to the classification level.

[0050] Beneficial Effects

[0051] (1) By introducing a physiotherapy agent cavity into the housing and using the physiotherapy head as an application and physiotherapy agent outlet, the preparation of the physiotherapy agent is facilitated;

[0052] (2) Using a mechanical face-lift mechanism can produce additional mechanical skin tensioning during physical therapy, making it easier to achieve good physical therapy results;

[0053] (3) Using artificial intelligence algorithms, image acquisition is used to identify skin type and appropriate treatment parameters, and a customized treatment plan is implemented for the user's skin;

[0054] (4) The addition of voice control provides a better user experience for the beauty device, making it easier for users to operate it and obtain prompt information related to the operation and treatment status.

[0055] (5) By collecting the user's real-time facial image and the precise parts around the user's customized calibration, after targeted facial recognition during use, the user's calibration record is called, thereby better and more conveniently solving the problem of forgetting or inaccurate memory of face lifting. At the same time, considering dividing volunteers into different age groups, the labels of the training set are defined according to the different levels of wrinkle treatment by each volunteer, which is more in line with the actual treatment needs of users of different age groups. For example, the younger the volunteers, the more likely they are to define fine wrinkles as a higher severity level in a statistical sense, while the older volunteers are more likely to focus on defining obvious wrinkles as a high severity level, and fine wrinkles can be treated or not. The wrinkle recognition model trained in this way can better perform face lifting and targeted beauty treatments based on the consensus of users of different ages. Finally, by selecting a wrinkle recognition scheme that imports images, it can be applied to some users who prepare for treatment in advance and do not know the severity of their own wrinkles, which is of guiding significance.

[0056] (6) A detachable external interface is used to connect the handheld housing to a high-definition touch screen, all physical therapy function operations and displays are integrated on the touch screen, and the touch screen is used to display the collected real-time facial images.

[0057] (7) By establishing a backend server, many constructed models can be trained and optimized on the backend server, so that functions can be continuously updated through OTA technology, reducing the load and production cost of the smart chip and improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 Schematic diagram of the overall structure of the OTA system of the handheld intelligent physiotherapy device in Example 1 of the present invention; Figure 2a Schematic diagram of the overall structure of the handheld intelligent physiotherapy device according to Embodiment 1 of the present invention; Figure 2b A rear view of the embodiment 1 of the present invention with the rear portion of the handheld housing removed; Figure 2c for Figure 2a A side sectional view of the central axis of the Figure 2d It is a schematic diagram of the high-definition touch screen being powered on and lit up after the secondary housing is electrically connected to the handheld housing from the left side; Figure 2e Flowchart of face recognition error correction; Figure 3 A schematic diagram of the structure of a physiotherapy head according to Embodiment 2 of the present invention, wherein a is a front view, b is a rear view, c is a side sectional view of the central axis in a, d is a first embodiment of a mechanical face-lifting mechanism, e shows a side sectional view of a second embodiment of a mechanical face-lifting mechanism, and f is a schematic diagram of an arc-pushing structure and an opening direction of the second embodiment of a mechanical face-lifting mechanism; Figure 4 Flow chart of the key switching method of the handheld intelligent physiotherapy device in embodiment 3 of the present invention; Figure 5 A simplified flowchart of the intelligent setting method in Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0063] Example 1

[0064] This embodiment will explain the overall structure of the handheld intelligent physiotherapy instrument OTA system. Figure 1 A handheld intelligent physiotherapy instrument OTA system is shown, comprising a handheld intelligent physiotherapy instrument and a background server, wherein the handheld intelligent physiotherapy instrument communicates with the background server via OTA, so that the handheld intelligent physiotherapy instrument can selectively download updates of original functions or expanded new functions from the background server, wherein the handheld intelligent physiotherapy instrument comprises a physiotherapy head, a front portion and a rear portion of a handheld housing detachably mounted based on a partition ( Figure 2c ), the front and rear parts form independently sealed front and rear cavities with the partition respectively, the front part includes a cleaning mode display including five major function displays of cleaning, nourishing, lifting and firming, eye care, and blue light calming, and a manual gear shift panel, a switch / mode button (including spare), a gear button for setting the frequency and intensity of electric pulses, and / or the vibration frequency, and / or the intensity of the light emitted by the phototherapy care source, and / or the rotation angle of the motor, a charging display ( Figure 2a ),as well as( Figure 2a , 2c) A TPYE-C charging port is provided on the front portion of the bottom of the front cavity for charging the battery in the handheld housing in the front cavity, and the front cavity also contains a smart chip fixedly connected to the partition.

[0065] The manual gear adjustment panel is used to switch and display the adjustment modes among the electric pulses, the intensity of light emitted by the phototherapy care source, and the rotation angle of the motor by pressing the switch / mode button for 1-2 seconds multiple times, and then use the gear button to select the gear, and finally long press the switch / mode button again to successfully set it.

[0066] The front part is also provided with two sound holes for voice prompts, which can issue age input prompts, respond to user voice commands, user operations, and function execution status (including the name of the current therapy function or manual gear adjustment when the user switches functions, executing, terminating therapy, shutting down and welcoming again, etc.). A microphone is provided on one side of the therapy head close to the front part for receiving voice control sounds ( Figure 2a The front part is also provided with a pinhole camera for collecting facial images, for building a sensitive skin recognition model (obtained by training a neural network model through image collection and real skin type) and an intelligent setting method.

[0067] like Figure 2a As shown, the handheld intelligent physiotherapy instrument also includes a high-definition camera module provided on the handheld housing, which can be used to capture human faces, and a sub-housing with a high-definition touch screen that can be extended from the left and right sides of the handheld housing. There is a gripping finger groove between the sub-housing and the extended left handheld housing indicated in the figure, so as to facilitate holding and operating the high-definition touch screen after completing the electrical connection in the direction of the arrow in the figure. The sub-housing has a pair of upper and lower interfaces, and a pair on each side, which can be electrically connected to the handheld housing respectively. The high-definition touch screen independently adopts a battery in the sub-housing ( Figure 2a ) is powered, and a battery charging port (not shown) is provided on the auxiliary housing. Figure 2a When the electrical connection is disconnected, the acquired face image can still be enlarged (for example Figure 2d The front camera is controlled to collect images of the push-up position. The upper part of the high-definition touch screen has a front camera.

[0068] Figure 2a As shown in , a pair of interfaces on the left side of the auxiliary housing can also be electrically connected to the right side of the handheld housing, which is also friendly to users who are accustomed to left-handed touch.

[0069] After the electrical connection is completed, click Figure 2d By pressing the middle switch button, the high-definition camera module is started, and a real-time facial image appears on the high-definition touch screen. At this time, the holding hand does not move, while the other hand can zoom in on the real-time facial image, circle the wrinkles, and confirm to upload it to the background server for storage.

[0070] The user can delete the affected part on the high-definition camera module and confirm again to upload it to the backend server to save and update the facial image.

[0071] When the next face lift is performed, the high-definition camera module recognizes the current real-time face image (such as Figure 2d ), and prompts on the high-definition touch screen ( Figure 2d(not shown) whether to call out the saved face image of the circled wrinkle area. When "yes" is selected, it is asked again whether the wrinkle recognition model is needed. When "no" is selected, the user is indicated on the image that is called and displayed where the face lift is needed, such as the forehead wrinkles of the enlarged forehead, the crow's feet of the right eye, and the nasolabial folds of the mouth. When "yes" is selected, the wrinkle recognition model step is entered; when "no" is selected, the wrinkle recognition model step is also entered.

[0072] For example Figure 2d As shown, the mode display and charging display can be displayed on the high-definition touch screen, the switch / mode button, the front gear button, the standby button, the age input button, and the OTA button all form touch buttons on the high-definition touch screen, and a face image acquisition button is provided on the high-definition touch screen; the smart chip also calls a pre-trained wrinkle recognition model to perform recognition marking on a face image based on a face image captured by the user pressing the face image acquisition button in the figure.

[0073] The face recognition function is built and optimized by the backend server. Figure 2d Click the OTA button with the word "OTA" in the pop-up dialog box (not shown in the figure) to select the OTA object to implement the corresponding update. When the user finds that the called face image is incorrect, it will be reported to the background server through the high-definition touch screen, and the background server will send the correct face image through the uploaded registered account, so that the high-definition touch screen will be updated in OTA mode, and the wrong recognition result will be automatically replaced with the OTA update correction result at the next recognition to make a correct recognition. The algorithm logic diagram is shown in the figure Figure 2e When a recognition error occurs, the backend server calls the correctly saved image based on the registered account and replaces the recognition result.

[0074] Therefore, when face recognition is performed again in the future, the recognition result will search whether there is a record of the previous recognition result being replaced due to the user reporting an error. If so, it will be automatically replaced with the correct result, otherwise the recognition result will still be adopted.

[0075] like Figure 2dAs shown, the wrinkle recognition model is a convolutional neural network model with a residual mechanism. The training method is to obtain volunteers of different age groups, and for each age group, take photos of multiple different parts that the volunteers themselves identify as needing a facelift, and divide the unacceptable degree of wrinkles into three levels: negligible, non-negligible, and deep processing according to the volunteers' own identification. The high-definition camera module is used to collect multiple high-definition images of the three levels of different age groups for each part (which can be obtained by automatically taking screenshots and enlarging them by the smart chip), and the multiple high-definition images of the parts are divided into training sets and verification sets according to different parts. By uploading them to the background server, a sub-convolutional neural network with a residual mechanism is formed for each part. The network is trained using the training set, the loss function is calculated to optimize the model parameters, the model accuracy is verified by the verification set, and the final wrinkle recognition model is obtained. At this time, the high-definition touch screen downloads the trained wrinkle recognition model to the smart chip through the OTA method; during recognition, the user selects his own age group and clicks on the desired face-lifting part on the collected face image, so that the smart chip can take a screenshot and enlarge the image part, or the user imports the saved face image with the wrinkle part circled by himself through the high-definition touch screen operation, clicks on the desired face-lifting part on the collected face image, and inputs it into the sub-convolutional neural network with residual mechanism corresponding to the trained corresponding part, so as to finally obtain the classification level of wrinkle recognition in each age group, and display the classification level on the display screen.

[0076] In this way, for the program that chooses to import images, users can obtain the severity level of wrinkles in various parts of their body based on the consensus of their age group. It is especially suitable for users who do not pay much attention to their wrinkles, and occasionally prepare for physical treatment in advance to meet the needs of certain occasions, or for users who are not clear about the severity of wrinkles in various parts of their body.

[0077] Among them, volunteers of different age groups are divided into six groups in this embodiment and embodiment 4: 18-24 years old, 25-35 years old, 35-45 years old, 45-60 years old, 60-75 years old, and over 75 years old.

[0078] like Figure 2b and 2c The rear cavity contains four physiotherapy agent containing chambers which are detachably fixed on the partition, each of which has a piston push rod inside, and the end of the piston push rod away from the piston is provided with a push portion ( Figure 2c ). The four physiotherapy agent receiving chambers are formed by a transparent material body and one side passes through the surface of the handheld shell to form a through portion to show the position of the piston. Figure 2b), a scale is provided on the hand-held housing passing through part of the edge, and the position of the piston can be read by comparing with the piston position. Taking the second accommodating chamber from the left as an example, the arrows in the figure indicate the lowest and highest scales (the middle scale is not shown in the figure).

[0079] The outlets of the four physiotherapy agent containing chambers are respectively connected to catheters, which extend through the rear cavity to the inner cavity of the physiotherapy head and are connected to the four physiotherapy agent discharge holes arranged on the outer shell of the physiotherapy head.

[0080] The four physiotherapy agent discharge holes are located on the side of the outer shell of the physiotherapy head closer to the rear cavity in the front cavity and the rear cavity, and are arranged in sequence at the middle position of the horizontal diameter of the physiotherapy head ( Figure 2b The rear portion is provided with a guide groove (not shown). The guide groove is sealed by transparent organic glass ( Figure 2b ), the hand-held housing is on the side of the guide groove, and when the push part is moved along the guide groove, the corresponding physiotherapy agents in the four physiotherapy agent receiving chambers are discharged from the physiotherapy agent receiving chambers, and are discharged from the physiotherapy agent discharge holes through the catheter. The physiotherapy agent discharge holes are sealed by protective caps ( Figure 2c ) to prevent the therapeutic agent from volatilizing or deteriorating.

[0081] Figure 2b In the method, after the piston push rod is pulled out from the therapeutic agent receiving chamber, or after the therapeutic agent receiving chamber is removed first and then the piston push rod is pulled out, the therapeutic agent can be added to each therapeutic agent receiving chamber under the premise that the protective cap is closed and the catheter is connected.

[0082] From left to right, the four therapeutic agent storage chambers store makeup remover, essential oil, gel, and eye cream for cleaning, nourishing, lifting and firming, eye care, and preliminary application of therapeutic agents.

[0083] The user can create therapy records through the high-definition touch screen and control the front camera ( Figure 2a ) collects images of the push-up position, and the backend server calculates the recommended push-up position based on the collected skin images to adapt to the user's skin, and downloads the recommended push-up position image to the high-definition touch screen through OTA, and automatically inputs it into the smart chip after the secondary shell is connected to the handheld shell, so as to recalculate the recommended position for each future use, wherein the calculation algorithm for calculating the recommended push-up position based on the collected skin image can be continuously optimized by the backend server, and regularly provided to the user through OTA to choose whether to update. That is, when the user thinks that the current algorithm is satisfactory, he or she may not choose OTA update.

[0084] The recommended push-up position calculated based on the collected skin image is determined by the intelligent setting in Example 4. That is, the dosage is gradually increased by three levels: low, medium, and high.

[0085] Example 2

[0086] This embodiment will explain in detail the structure of the physiotherapy head in Embodiment 1. Figure 3 The physiotherapy head shown in the figure includes a massage head, a touch sensing part coupled to the center of the massage head, and a light therapy care source between the center of the massage head and the peripheral massage head that can emit amber light, red light, crimson light, and blue light. A mechanical facelift mechanism ( Figure 3 a). Figure 3 As shown in b, four therapeutic agent discharge holes of Example 1 are arranged on the other side of the therapeutic head opposite to the massage head. The face lifting mechanism can be as follows Figure 3 c and 3d, including a back plate ( Figure 3 c), a slave gear sleeved on the central column of the back plate, a master gear meshed with the slave gear, a motor connected to the master gear through a reducer and accommodated in the inner cavity of the physiotherapy head, and a plurality of push heads ( Figure 3 d).

[0087] The central hole of the main gear is connected to the reducer shaft through the through hole on the back plate, and the slave gear has 6 rotating grooves whose number is the same as the number of the pusher heads. The pusher head has a push rod, and the push rod can be slidably engaged in the slide groove of the back plate ( Figure 3 d not shown), one end of the push rod has a convex rod passing through the rotary groove, when the motor drives the main gear to rotate clockwise, the convex rod is driven by the gear to rotate and slide in the rotary groove, thereby pushing the push head rods of the 6 push heads radially outward in the slide groove to make the multiple push heads open, and thereafter at any time when the motor drives the main gear to rotate counterclockwise, the push head rod moves radially inward in the slide groove to make the multiple push heads converge in the opposite direction, thereby achieving external auxiliary pulling effect on the skin.

[0088] The motor is communicatively connected to the smart chip, and the back plate has a wire hole (5 in total) between each of the push head rods for passing components and / or wires of the touch sensing part, massage head, and light therapy care source through so as to communicate with the smart chip.

[0089] Or the face-lifting mechanism is as follows Figure 3 e and 3f, including a facelift scissors ( Figure 3 e), the front end of which has a pair of push arcs ( Figure 3 f), and the rear end handle passes through the inner cavity of the physiotherapy head and out of the physiotherapy head shell for hand holding as shown in FIG. Figure 3e) to perform a shearing operation by pressing the handles relative to each other, and to tighten the skin when releasing the handles (specifically, the front end can resume the movement trend of opening after releasing by the compression spring arranged across the handles and / or across the front end), so as to achieve the skin being pressed by the front end push arc as shown in FIG. Figure 3 Push and pull in the opposite direction as shown by the arrows in f.

[0090] The part of the treatment head through which it is inserted forms a pair of guide grooves ( Figure 3 e is not shown). At this time, Figure 3 As shown in FIG. 5 , when the user chooses to push the pushing portion, the therapeutic agent will be discharged from the therapeutic agent discharge hole and applied to the finger to apply the therapeutic agent to the treatment area.

[0091] It is understandable that due to the obstruction and space occupied by the handle passing through the shell of the physiotherapy head, there is no suitable angle to use the physiotherapy agent discharge hole on the shell of the physiotherapy head to directly apply the physiotherapy agent to the skin and perform the application operation. Figure 3 As shown in e, we can still use Figure 3 The therapeutic agent discharge hole is set slightly lower than b to discharge the therapeutic agent to the fingers and apply the therapeutic agent.

[0092] Example 3

[0093] This embodiment will explain the key switching method of the handheld intelligent physiotherapy instrument. Figure 4 After long pressing the light / mode button for 4-5 seconds to turn on the beauty instrument, the age is input (via voice) into the beauty instrument. The user can cycle through the functions of the light / mode button by short pressing. If it is judged that the light / mode button is pressed twice in a row, the function is selected (for the sake of simplicity, this step is not shown in the figure) and it is further judged whether it is long pressed for 2-3 seconds. If it is judged that it is not pressed twice in a row, the function can continue to be cycled through short pressing;

[0094] If it is judged that the long press is 2-3 seconds, the selected function is exited (for the sake of simplicity, this step is not shown in the figure), and then it is judged whether it is long pressed for 4-5 seconds. If so, the device is turned off, otherwise the function cycle continues;

[0095] If it is judged that the button is not pressed for 2-3 seconds, the selected function will be treated through contact sensing. If it is judged that the button is pressed for 2-3 seconds, the treatment will be terminated and the function will be switched in a loop. Otherwise, it will be judged every 1-2 seconds whether it is non-contact. If it is, the compensation time will be calculated. Otherwise, the selected function will be treated. If the non-therapy part is touched, the user will generally move away. At this time, if it is judged that the button is not pressed for 2-3 seconds, it will continue to be judged as non-contact, so as to calculate the compensation time, until it is judged that the skin is touched again to perform the selected function treatment and compensate the treatment time. The time spent on adjusting to the correct therapy part due to the mistaken contact is included in the compensation time.

[0096] During the calculation of the compensation time, it is determined every 1-2 seconds whether the skin is touched. If it is still not touched, it is also determined at 1-2 second intervals whether it is pressed for 2-3 seconds. If so, the physical therapy is terminated and the function cycle is continued. Otherwise, it is determined whether it is pressed for 4-5 seconds. If so, the device is shut down. Otherwise, it continues to determine whether the skin is touched.

[0097] Example 4

[0098] This embodiment will illustrate the method of intelligent setting. Figure 5 As shown, the following steps are included:

[0099] S1 set up five test groups: cleansing group, nourishing group, lifting and firming group, eye care group, and blue light calming group. Each test group was divided into six groups: 18-24 years old, 25-35 years old, 35-45 years old, 45-60 years old, 60-75 years old, and over 75 years old. Figure 4 Taking the cleaning function as an example, the training set, validation set and test set are set, and the ratio of the three is 3:1:1; all the subjects follow the preset work and rest rules for 2 months, which are 18-24 years old get up at 7:30 am, 25-35 years old get up at 7 am, 35-45 years old get up at 7 am, 45-60 years old get up at 6:30 am, 60-75 years old get up at 6 am, and those over 75 years old get up at 6 am. All experimental groups go to bed between 10-11 pm, and three meals a day, breakfast, lunch, and dinner, start within 30 minutes after getting up, between 11:30-12:30, and 17:30-18:30 respectively;

[0100] The S2 clean test group trained ResNet according to the following parameters:

[0101] Low frequency vibration low, medium and high three gears, amber light intensity low, medium and high three gears, total cleaning time under warm effect 4min, seven parameters,

[0102] When S3 is turned on, it uses a pinhole camera to collect images of the face to be treated. It identifies whether the skin is sensitive based on the pre-set sensitive skin recognition model (i.e., based on facial features such as burn scars, bloodshot eyes, and blush in the collected images). If so, the action time of all groups is halved. The training sets of each group are subjected to a cross-test scheme for parameter selection of the five major functional treatments, and the top three parameter configurations are obtained through an overall evaluation. The classification scores of the top three treatment results of each parameter configuration are taken. The corresponding collected images are input into the corresponding ResNet model for training, and the verification set is used to verify the accuracy. The network parameters of the model are adjusted using the loss function until the accuracy is achieved. When the accuracy and loss function values ​​are stable, the training is stopped, thereby obtaining the classification of the top three scores of each parameter configuration of each functional therapy under each group into good, medium and poor (i.e., three levels are divided according to the three scores from the highest to the relatively lowest); by randomly selecting the middle parameter configuration of each parameter configuration corresponding to the functional therapy under the group corresponding to the good or medium prediction result of each parameter configuration, the parameter configuration with the highest score among the middle parameter configurations of the three parameter configurations is compared as the best parameter configuration of the functional therapy under the group; if there are multiple highest scores that are the same, one of the parameter configurations is randomly selected as the best parameter configuration of the functional therapy under the group, thereby forming a trained artificial intelligence model;

[0103] The S4 high-definition touch screen downloads the artificial intelligence model trained in step S3 from the background server through OTA, inputs the age of the subject in the test set, uses a pinhole camera to collect images of the subject's face to be treated, and inputs the collected images into the corresponding trained ResNet model to obtain a set of optimal parameter configurations predicted under cleaning, so as to select the corresponding optimal parameter configuration for cleaning preparation according to the cleaning selected by the subject, and perform treatment according to the subject's face to be treated contacting the touch sensing part.

[0104] Example 5

[0105] This embodiment will illustrate two facial skin physiotherapy methods provided by the handheld intelligent physiotherapy apparatus of embodiments 1-4, including the following steps:

[0106] The first is the intelligent physiotherapy program:

[0107] (1) Turn on the device and input age, then face the pinhole camera to collect images and complete parameter setting based on intelligent settings;

[0108] (2) Select the therapy function by using the switch / mode button;

[0109] (3) discharging the corresponding physiotherapy agent through the physiotherapy agent discharge hole to act on the facial skin to be treated;

[0110] (4) The facial skin to be treated by the therapeutic agent in step (3) is brought into contact with the touch-sensitive part to perform the selected therapeutic function.

[0111] The second is manual therapy

[0112] After powering on, start directly from step (2) by using the switch / mode button to wake up the manual gear adjustment panel; then long press the switch / mode button to select manual adjustment of the electric pulse frequency and intensity, and / or the vibration frequency, and / or the intensity of the light emitted by the phototherapy care source, and / or the rotation angle of the motor, use the gear button to select the gear, and finally long press the switch / mode button again to successfully set;

[0113] When image acquisition fails, the sound outlet can provide a voice prompt of image acquisition failure every 3 seconds until the user succeeds in image acquisition, or wake up the manual adjustment panel to enter the next step or prompt to enter the manual adjustment parameter mode and then perform steps (3) and (4) in sequence.

[0114] Table 1 compares the effects of the two physiotherapy programs (with the assistance of tension), where (0) indicates that no definite improvement was measured without the assistance of tension, and (-) indicates that the data without the assistance of tension were relatively poor (between 0.2-0.5%).

[0115]

[0116]

[0117] The results showed that the use of intelligent solution + tension assistance showed better results in all indicators.

Claims

1. A handheld intelligent physiotherapy instrument OTA system, characterized in that: The invention comprises a handheld intelligent physiotherapy instrument and a background server, wherein the handheld intelligent physiotherapy instrument communicates with the background server via OTA, so that the handheld intelligent physiotherapy instrument can selectively download updated or expanded new functions of the original functions from the background server, wherein the handheld intelligent physiotherapy instrument comprises: a handheld shell, a physiotherapy head, a switch / mode button, a charging port, a mode display and a charging display, a pinhole camera for collecting skin images of a human face, and a battery in the handheld shell, wherein the physiotherapy head comprises a button that can perform vibration and / or MFIP action. The massage head and the light therapy care source that can emit amber light, red light, crimson light and blue light, the mode display includes a display of five major functions of cleaning, nourishing, lifting and firming, eye care and blue light calming that can be realized by short pressing the switch / mode button, and the feature is that the therapy head also includes a touch sensing part coupled to the massage head and a mechanical skin lifting mechanism arranged on the therapy head outside the massage head, the touch sensing part automatically or manually starts the mechanical skin lifting mechanism according to the facial skin contacting the massage head, so as to realize the multi-directional extension effect on the skin level at the same time as the therapy; The handheld intelligent physiotherapy instrument also includes a high-definition camera module, which can be used to capture human faces, and a sub-shell that is detachably extended from one side of the handheld shell and is provided with a high-definition touch screen with a front camera. There is a gripping finger groove between the sub-shell and the extended handheld shell on one side to facilitate gripping and operating the high-definition touch screen. When the high-definition camera module is started, a real-time face image appears on the high-definition touch screen. At this time, the holding hand does not move, and the other hand can zoom in on the real-time face image, circle the wrinkle area, and determine to upload it to the backend server for storage. When the face is lifted next time, the high-definition camera module recognizes the current real-time face image through the face recognition function in the smart chip, and calls out the saved face image of the circled wrinkle area to indicate the user's desired face lift area. The user can delete the area of ​​action on the high-definition camera module and confirm to upload it to the backend server again to save and update the face image, wherein the optimization of the face recognition function is performed by the backend server and the user selectively chooses whether to OTA download it to the smart chip for update; A plurality of physiotherapy agent discharge holes are arranged on the side of the physiotherapy head away from the mechanical facelift mechanism. The physiotherapy agent discharge holes are connected through conduits to a plurality of physiotherapy agent receiving chambers inside the handheld shell, which are sealed by a transparent material body and one side of which passes through the surface of the handheld shell to display the piston position therein. Each of the plurality of physiotherapy agent receiving chambers can be pushed by a piston push rod to squeeze out the physiotherapy agent contained in the physiotherapy agent receiving chamber. A push portion arranged at one end of the piston push rod protrudes from the guide groove on the outer surface of the handheld shell. When the push portion is moved along the guide groove, the push portion is pushed out of the guide groove. When the guide groove moves, the current pushed position of the piston can be observed through the transparent material body, and the corresponding physiotherapy agent can be discharged from the physiotherapy agent receiving chamber, discharged from the corresponding physiotherapy agent discharge hole through the conduit and applied to the finger or the face to be treated, and then applied by means of the finger or the side of the physiotherapy head respectively; protective caps are provided on the multiple physiotherapy agent discharge holes, so that the protective caps can be covered when not in use to prevent the physiotherapy agent components from volatilizing and / or deteriorating; the piston push rod can be pulled out by disassembling the handheld housing to add the corresponding physiotherapy agent into the physiotherapy agent receiving chamber; The smart chip arranged inside the handheld housing is used to set parameters based on smart settings according to the skin image acquisition and one of the functions of cleaning, nourishing, lifting and firming, eye care, and blue light calming selected by the user, so that after the parameter setting is completed and the physiotherapy agent is applied, the mechanical skin-lifting mechanism is automatically or manually started by the contact between the touch sensing part and the facial skin, and the physiotherapy of the selected function is performed; when the facial skin or the skin of any other part contacts the touch sensing part and then leaves the touch sensing part to present a non-contact state, the mechanical skin-lifting mechanism is reversely moved and the physiotherapy of the selected function is automatically suspended. At this time, the user can forcibly press the switch / mode button for 2-3 seconds to terminate the physiotherapy and enter the mode of reselecting the function, and press the switch / mode button twice in succession to confirm the selected function; or press the switch / mode button for 4-5 seconds to shut down the handheld smart physiotherapy instrument. And the age input part, the age input part button or voice input method, for the button input method, that is, cyclically switching among multiple age groups, when selected, long press the age input part button for 1-2 seconds to confirm; for the voice input method, voice input can be performed multiple times and the most recent voice input is cyclically overwritten until no voice input is received within 2-5 seconds.

2. The system according to claim 1, characterized in that When the user finds that the called face image is incorrect, the user reports it to the backend server through the high-definition touch screen, and uses the uploaded registered account to let the backend server send the correct face image, so that the high-definition touch screen can be updated OTA, and the wrong recognition result will be automatically replaced with the OTA update result at the next recognition to make a correct recognition; The pause time of the treatment that automatically pauses the selected function is included in the compensation time after the next contact for treatment, and the compensation time shall not exceed 1 minute; The recalculation method is that, based on the current recommended position of the background server, the amount of physiotherapy agent usage corresponding to the recommended positions in the future shows a decreasing function trend; the OTA update also includes the user being able to choose whether to make the following updates: that is, when the deletion operation is performed on the high-definition touch screen, a dialog box will pop up to display a message reminding the user that the face lift needs to be consolidated and persisted, and asking whether it is really necessary to delete; The reduction function is linear, and a scale is provided on the handheld housing passing through part of the edge, which can read the position of the piston on the scale by comparing it with the piston position. After the smart chip completes the recalculation, it can display the recommended scale position indication for multiple future uses on the high-definition touch screen for user reference.

3. The system according to claim 1, characterized in that The handheld housing comprises a detachable front part and a rear part based on a partition, and the partition divides the space surrounded by the installed front part and the rear part into a front cavity and a rear cavity that are independently sealed from each other, and the rear cavity contains the multiple physiotherapy agent accommodating cavities, the corresponding piston push rod, part of the push part, and at least part of the catheter, and the physiotherapy agent discharge hole is located on the side of the physiotherapy head opposite to the mechanical skin lifting mechanism; the front cavity is used to accommodate the battery and the smart chip in the handheld housing; the switch / mode button, mode display and charging display are located in the front part, and the charging hole is set at the bottom of the front cavity and belongs to one end of the front part; the multiple physiotherapy agent accommodating cavities are detachably fixed on the partition; The handheld intelligent physiotherapeutic instrument further comprises a gear button arranged at the front part, which is used to set the frequency and intensity of electric pulses, and / or the vibration frequency, and / or the intensity of light emitted by the phototherapy care source, and / or the rotation angle of the motor; the mode display further comprises a manual gear adjustment panel, which is used to switch and display the adjustment modes among the electric pulses, the intensity of light emitted by the phototherapy care source, and the rotation angle of the motor by pressing the switch / mode button for 1-2 seconds multiple times, and then the gear button is used to select the gear, and finally the switch / mode button is pressed for a long time again to successfully set the gear; The mode display and charging display can be displayed on the high-definition touch screen, the switch / mode button, the front gear button, the standby button, the age input button, and the OTA button all form touch buttons on the high-definition touch screen, and a face image acquisition button is provided on the high-definition touch screen; the smart chip also calls a pre-trained wrinkle recognition model to perform identification marking on a face image based on a face image captured by the user pressing the face image acquisition button.

4. The system according to claim 3, characterized in that The wrinkle recognition model is a convolutional neural network model with a residual mechanism. The training method is to obtain volunteers of different age groups, and for each age group, take photos of multiple different parts that the volunteers themselves identify as needing a facelift, and divide the unacceptable degree of wrinkles into three levels: negligible, non-negligible, and deep processing according to the volunteers' own identification. The high-definition camera module is used to collect multiple high-definition images of each part of the three levels for different age groups, and the multiple high-definition images of the parts are divided into training sets and verification sets according to different parts, and uploaded to the background server by the high-definition touch screen. The background server forms a sub-convolutional neural network with a residual mechanism for each part, which is used to The training set is used for training, the loss function is calculated to optimize the model parameters, and the model accuracy is verified by the validation set to obtain the final wrinkle recognition model. At this time, the high-definition touch screen downloads the trained wrinkle recognition model to the smart chip through the OTA method; during recognition, the user selects his own age group and clicks on the desired face-lifting part on the collected face image, so that the smart chip can take a screenshot and enlarge the image part, or the user imports the saved face image of the wrinkle part circled by the user through the high-definition touch screen operation, clicks on the desired face-lifting part on the collected face image, and inputs it into the sub-convolutional neural network with residual mechanism corresponding to the trained corresponding part, so as to finally obtain the classification level of wrinkle recognition in each age group, which is displayed on the high-definition touch screen; Both sides of the sub-shell have interfaces electrically connected to the handheld shell. When one of the sides is connected to the handheld shell, the high-definition touch screen is turned on; the high-definition touch screen is independently powered by the battery in the sub-shell, and a sub-shell battery charging port is provided on the sub-shell. When the electrical connection is disconnected, the captured facial image can still be enlarged and the front camera can be controlled to capture the image of the push-up position.

5. The system according to claim 1, characterized in that The intelligent setting method comprises the following steps: S1 set up five test groups: cleaning group, nourishing group, lifting and firming group, eye care group, and blue light calming group. Each test group was divided into six groups: 18-24 years old, 25-35 years old, 35-45 years old, 45-60 years old, 60-75 years old, and over 75 years old. Each group was set up with a training set, a validation set, and a test set, and the ratio of the three was 5-3:1-2:

1. All subjects followed the preset work and rest routine for 1-3 months. The work and rest schedule is to get up at 7:30 am for those aged 18-24, 7 am for those aged 25-35, 7 am for those aged 35-45, 6:30 am for those aged 45-60, 6 am for those aged 60-75, and 6 am for those aged 75 and above. All test groups go to bed between 10-11 pm. Three meals a day, breakfast, lunch, and dinner, start within 30 minutes after getting up, between 11:30-12:30, and between 17:30-18:30 respectively. The five experimental groups in S2 were trained with the following parameters for the artificial intelligence model: Cleaning group, low frequency vibration low, medium and high three gears, amber light intensity low, medium and high three gears, total cleaning time under warm effect 4min, seven parameters, Nourishing group, high-frequency vibration low, medium, high three gears, red light intensity low, medium, high three gears, total nourishing time under warming effect 4min, seven parameters, Lifting and firming group, low frequency vibration low, medium and high three gears, high frequency vibration low, medium and high three gears, MFIP pulse frequency high, medium and low three gears, pulse intensity high, medium and low three gears, deep red light intensity, amber light intensity, total lifting and firming time under warm effect 4min, fifteen parameters, Eye care group, MFIP pulse frequency high, medium and low three gears, pulse intensity high, medium and low three gears, low frequency vibration low, medium and high three gears, high frequency vibration low, medium and high three gears, high and low frequency vibration interval, total eye care time under warming effect 4min, fourteen parameters, Blue light calms, blue light intensity, ice compress effect is to cool down the massage head for 4 minutes, one parameter; When S3 is turned on, it uses a pinhole camera to collect images of the face to be treated and uploads them to the backend server. The backend server identifies whether the skin is sensitive based on the preset sensitive skin recognition model. If it is, the action time of all groups is halved. The training sets of each group are subjected to the parameter selection cross-test scheme of the five major functional treatments, and the top N parameter configurations are obtained through overall evaluation, N≤5, and the classification scores of the top three treatment results of each parameter configuration are taken; The corresponding collected images are respectively input into the corresponding artificial intelligence model for training, and the accuracy is verified using the validation set. The network parameters of the model are adjusted using the loss function until the accuracy and the loss function value are stable, and the training is stopped, thereby obtaining the classification of the top three scores of each parameter configuration of each functional physiotherapy under each group as good, medium, and poor; by randomly selecting the middle parameter configuration of each parameter configuration corresponding to the good or medium prediction result of each parameter configuration as the corresponding functional physiotherapy under the group, the parameter configuration with the highest score among the middle parameter configurations of N parameter configurations is compared as the best parameter configuration of the corresponding functional physiotherapy under the group, if there are multiple highest scores with the same score, one of the parameter configurations is randomly selected as the best parameter configuration of the corresponding functional physiotherapy under the group, thereby forming a trained artificial intelligence model; the artificial intelligence model is a convolutional neural network CNN, or a ResNet based on a residual mechanism; The S4 high-definition touch screen downloads the artificial intelligence model trained in step S3 from the background server via OTA, and inputs the collected images into the corresponding trained artificial intelligence model to obtain five sets of optimal parameter configurations predicted under each functional therapy. In this way, the corresponding optimal parameter configuration is selected according to the function selected by the subject for therapy preparation, and therapy is performed according to the subject's facial contact with the touch-sensitive part to be treated.

6. The system according to any one of claims 1 to 5, characterized in that: The mechanical facelift mechanism includes a back plate fixedly connected to the physiotherapy head and having a center column, a slave gear sleeved on the center column of the back plate, a main gear meshing with the slave gear, a motor accommodated in the internal cavity of the physiotherapy head and connected to the main gear through a reducer or directly connected, and a plurality of push heads, wherein the central hole of the main gear is connected to the reducer shaft through a through hole on the back plate, the slave gear has a plurality of rotary grooves whose number is the same as the number of the push heads, the push head has a push rod, the push rod can be slidably embedded in the slide groove of the back plate, one end of the push rod has a convex rod passing through the rotary groove, when the motor drives the main gear to rotate clockwise Or when rotating counterclockwise, the convex rod drives the convex rod to slide in the rotating groove due to the rotation of the slave gear, thereby pushing the push head rods of the multiple push heads radially outward in the slide groove to make the multiple push heads open, and thereafter at any time when the motor drives the main gear to rotate in the opposite direction, the push head rod moves radially inward in the slide groove to make the multiple push heads converge in the opposite direction, thereby achieving external auxiliary pulling effect on the skin, the motor is communicatively connected to the smart chip, and the back plate has at least one wire hole between the push head rods for passing the components and / or wires of the touch sensing part, massage head, and light therapy care source to communicate with the smart chip.

7. The system according to any one of claims 1 to 5, characterized in that: The mechanical facelift mechanism includes a facelift scissors, the front end of which has a pair of push arcs arranged on the side where the massage head is located and can be pressed against the skin, and the rear end handle passes through the internal cavity of the physiotherapy head and passes through the outer shell of the physiotherapy head so that the hand can hold it and press the handle relatively to perform a shearing operation. After the skin is pressed, the handle is released to achieve a skin tightening operation, and the front end push arc pushes and pulls the skin in the reverse state of tightening. The part of the physiotherapy head that is penetrated forms a pair of guide grooves for guiding the handle to perform the shearing operation and the skin tightening operation. At this time, when the user chooses to push the push part, the physiotherapy agent will be discharged from the physiotherapy agent discharge hole and applied to the fingers or other application materials to apply the physiotherapy agent to the treatment area; and / or, The physiotherapy head is also provided with a microphone for receiving voice control sound, and the front portion also includes at least one sound outlet for voice prompts to issue age input prompts, respond to user voice commands, user operations, and voice prompts for function execution status; There are two switch / mode buttons, one of which is a spare button. The user can select one of them for operation at any time when the device is turned on. When one of the buttons fails, the spare button can still be used for operation.

8. A facial skin treatment method using the system according to any one of claims 1 to 7, characterized in that: The steps include: (1) Turn on the device and input age, then face the pinhole camera to collect images and complete parameter setting based on intelligent settings; (2) Select the therapy function by using the switch / mode button; (3) discharging the corresponding physiotherapy agent through the physiotherapy agent discharge hole to act on the facial skin to be treated; (4) contacting the facial skin to be treated with the therapeutic agent in step (3) with the touch sensitive part to perform the selected therapeutic function; or, After powering on, start directly from step (2) by using the switch / mode button to wake up the manual gear adjustment panel; then long press the switch / mode button to select manual adjustment of the electric pulse frequency and intensity, and / or the vibration frequency, and / or the intensity of the light emitted by the phototherapy care source, and / or the rotation angle of the motor, use the gear button to select the gear, and finally long press the switch / mode button again to successfully set; When image acquisition fails, the voice prompt of image acquisition failure can be given through the sound outlet every 3-5 seconds until the user succeeds in image acquisition, or the manual adjustment panel is awakened to enter the next step or prompted to enter the manual adjustment parameter mode and then perform steps (3) and (4) in sequence.

9. The facial skin treatment method according to claim 8, characterized in that: After long pressing the light / mode button for 4-5 seconds to turn on the instrument, enter the age into the beauty instrument. The user can cycle through the functions by short pressing the light / mode button. If it is judged that the light / mode button is pressed twice in a row, the function is selected and it is further judged whether it is long pressed for 2-3 seconds. If it is judged that it is not pressed twice in a row, the function can be cycled through short pressing. If the long press is continued for 2-3 seconds, the selected function will be exited, and then it will be determined whether it is long pressed for 4-5 seconds. If so, the device will be shut down, otherwise, the function cycle will continue; If it is judged that the button is not pressed for 2-3 seconds, the selected function will be treated through contact sensing. If it is judged that the button is pressed for 2-3 seconds, the treatment will be terminated and the function will be switched in a loop. Otherwise, it will be judged whether it is non-contact every 1-2 seconds. If it is, the compensation time will be calculated. Otherwise, the selected function will be treated. If the non-treatment part is touched, the user will generally move away. At this time, if it is judged that the button is not pressed for 2-3 seconds, it will continue to be judged as non-contact, so as to calculate the compensation time, until it is judged that the skin is touched again, the selected function will be treated and the treatment time will be compensated. During the calculation of the compensation time, it is determined every 1-2 seconds whether the skin is touched. If it is still not touched, it is also determined at 1-2 second intervals whether it is pressed for 2-3 seconds. If so, the physical therapy is terminated and the function cycle is continued. Otherwise, it is determined whether it is pressed for 4-5 seconds. If so, the device is shut down. Otherwise, it continues to determine whether the skin is touched.

10. The facial skin treatment method according to claim 9, characterized in that: It also includes a wrinkle tightening treatment method, which specifically includes starting the high-definition touch screen on the sub-shell, facing the high-definition camera module to collect real-time high-definition images of the user, manually zooming in on the real-time face image, circling the wrinkle area, and confirming to save. When doing a face lift next time, the high-definition camera module recognizes the current real-time face image through the face recognition function in the smart chip, and prompts on the high-definition touch screen whether to call out the saved face image of the circled wrinkle area. When "Yes" is selected, it is asked again whether model recognition is required. When "Not required" is selected, the user is indicated on the image that is displayed where the face lift is required; when "Required" is selected, the wrinkle recognition model step is entered; when "No" is selected, the wrinkle recognition model step is also entered, and the wrinkle recognition model step is entered. During the step, a face image is collected by pressing the face image collection button, and a pre-trained wrinkle recognition model is called to perform recognition marking on the face image. During specific recognition, the user first selects his own age group, and clicks on the desired face-lifting part on the collected face image, so that the smart chip can take a screenshot and enlarge the image part, or the user imports a saved face image of the circled wrinkle part through the high-definition touch screen operation, clicks on the desired face-lifting part on the collected face image, and inputs the input into the sub-convolutional neural network with residual mechanism corresponding to the trained corresponding part, so as to finally obtain the classification level of wrinkle recognition in each age group, and display the classification level on the high-definition touch screen. The user determines the type, dosage and tightening time of the physiotherapy agent according to the classification level.

Images