Artificial intelligence assisted evaluation method applied to cosmetic medicine

Abstract

The present application relates to an artificial intelligence assisted evaluation method and system applied to cosmetic medical treatment, which is applied to an artificial intelligence recognition analysis module. Wherein, a real-time facial expression evaluation result of a subject is generated by executing an artificial intelligence facial expression evaluation module, and the real-time facial expression evaluation result is input to the artificial intelligence recognition analysis module, and at least one of a medical knowledge rule module and a cosmetic medical treatment assisted evaluation result historical database is selected to execute an artificial intelligence recognition analysis program. Then, the artificial intelligence recognition analysis module generates and outputs a real-time cosmetic medical treatment assisted evaluation result. In this way, a cosmetic medical treatment behavior can be performed according to the real-time cosmetic medical treatment assisted evaluation result, and the treatment effect of personalized beauty can be achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to an artificial intelligence (AI) assisted evaluation method and an AI assisted evaluation system, in particular, an AI assisted evaluation method for cosmetic medical treatment based on real-time facial expression evaluation results obtained by AI technology and an AI assisted evaluation system using the method. BACKGROUND

[0002] The current trend of cosmetic medical treatment, especially facial microcosmetic medical treatment, has been widely accepted by people of all ages.

[0003] The treatment method for facial microcosmetic medical treatment mainly relies on the medical professional skills and knowledge of the doctor to process according to general or normal standard procedures, or with some individual judgment of the doctor's own professional experience.

[0004] However, in actual operation, due to the lack of deep customization, there is always a gap between the actual cosmetic treatment results and the expected treatment effect. In particular, the micro-expression of the face is a muscle movement accumulated over a long period of time, and the subtle changes in facial muscles are not easy to observe with the naked eye in a short period of time, resulting in a significant difference between the actual cosmetic treatment results and the expected treatment effect.

[0005] Even, some actual cosmetic medical treatment results are caused by the doctor's own individual judgment, resulting in poor treatment effect after surgery, causing more medical treatment disputes and medical treatment defects.

[0006] Therefore, how to provide more and better assistance methods and tools for personalized cosmetic medical treatment needs is a technical problem to be solved in the current cosmetic medical industry. Therefore, the applicant has proposed a technical solution of an AI assisted evaluation method and system for cosmetic medical treatment in Chinese patent application No. CN202010388355.8 to achieve the preliminary purpose of solving the above technical problems.

[0007] In order to further optimize the related technical content, the applicant further improves the process of obtaining real-time facial expression evaluation results in the above technical solution and proposes an optimized solution as described in the embodiments of the present application. SUMMARY

[0008] The main purpose of the present application is to provide an AI-assisted evaluation method for cosmetic medical treatment based on real-time facial expression evaluation results obtained by artificial intelligence technology, and an evaluation system using the method, to further optimize the prior art.

[0009] The implementation concept of the present application is mainly to use an AI facial expression evaluation module to first obtain real-time facial expression evaluation results for individuals, and then perform AI recognition analysis procedures based on the aforementioned real-time facial expression evaluation results for individuals, and combine the functional medical anatomy rules and dynamic medical anatomy rules in the medical knowledge rule module with the cosmetic medical treatment assistance evaluation result historical database, to provide real-time cosmetic medical treatment assistance evaluation results, and achieve the provision of exclusive and real-time personalized cosmetic medical treatment assistance suggestions.

[0010] To achieve the above purpose, an embodiment of the present application provides an AI-assisted evaluation method for cosmetic medical treatment, at least comprising the following steps: (a) providing a real-time facial expression evaluation result of a subject; and (b) based on at least one of a medical knowledge rule module and a cosmetic medical treatment assistance evaluation result historical database, performing an AI cosmetic medical treatment recognition analysis procedure for the real-time facial expression evaluation result, and generating a real-time cosmetic medical treatment assistance evaluation result; wherein the step (a) at least comprises the following steps: (a1) performing an AI image detection procedure to obtain a real-time facial image of the subject; (a2) based on the real-time facial image, performing an AI image calibration and feature extraction procedure to obtain facial surface and geometric feature information; (a3) based on the facial surface and the geometric feature information, performing an AI facial motion coding procedure to obtain multiple facial motion coding information; and (a4) based on the multiple facial motion coding information, performing an AI facial emotion recognition procedure to obtain the proportional distribution and combination information of multiple emotion indicators corresponding to the real-time facial image, and form the real-time facial expression evaluation result.

[0011] Preferably, the step (b) further comprises the following step: (c) feeding back and storing the real-time cosmetic medical treatment assistance evaluation result to at least one of the medical knowledge rule module and the cosmetic medical treatment assistance evaluation result historical database.

[0012] Preferably, the real-time cosmetic medical treatment assistance evaluation result at least includes the combination and preferred order of an evaluation treatment site result of the subject, or the combination and preferred order of the evaluation treatment site result and an injection filler type and dose.

[0013] Preferably, the medical knowledge rule module further comprises a functional medical anatomy rule and a dynamic medical anatomy rule.

[0014] Preferably, the artificial intelligence machine learning method applied to the artificial intelligence image detection program comprises a Haar feature combined with adaptive boosting machine learning method belonging to a boundary detection algorithm type, or a histogram of oriented gradients combined with support vector machine machine learning method.

[0015] Preferably, the artificial intelligence image calibration and feature extraction program in the step (a2) at least comprises the following steps: (a21) based on the real-time face image, an artificial intelligence face key point identification program is executed to obtain a face key point identification information; (a22) based on the face key point identification information, a face image calibration program is executed to obtain a normalized face image information; and (a23) based on the face key point identification information and the normalized face image information, a face image feature extraction program is executed to obtain the face image surface and the geometric feature information.

[0016] Preferably, before the artificial intelligence face key point identification program is executed, a certain amount of training data set is used to perform artificial intelligence training of a face key point identification model in the artificial intelligence face key point identification program in a machine learning manner.

[0017] Preferably, the face image calibration program at least comprises an affine transformation technique to eliminate errors caused by different postures in the face key point identification information, and to unify the size and presentation of the face image, thereby obtaining the normalized face image information.

[0018] Preferably, the face image feature extraction program at least comprises a face image surface feature extraction program and a face image geometric feature extraction program; wherein the face image surface feature extraction program comprises executing a histogram of oriented gradients to obtain multi-dimensional vector data, and combining a principal component analysis to reduce the amount of vector data and retain a face image surface feature information, and the face image geometric feature extraction program comprises obtaining a face image geometric feature information based on the face key point identification information.

[0019] Preferably, before the artificial intelligence face action coding program is executed, a certain amount of training data set and a face action coding system are used to perform artificial intelligence training of a face action coding model in the artificial intelligence face action coding program in a machine learning manner.

[0020] Preferably, the artificial intelligence training of the facial action coding model includes training of the facial action coding model under different contexts of a static expression and / or a dynamic expression to obtain a static facial action coding information and / or a dynamic facial action coding information, respectively.

[0021] Preferably, before performing the artificial intelligence facial emotion recognition procedure, the artificial intelligence training of a facial emotion recognition model in the artificial intelligence facial emotion recognition procedure is performed by combining the facial action coding system with at least emotion valence, emotion arousal, and emotion indicator parameters, and by another machine learning method.

[0022] Preferably, the cosmetic medical auxiliary evaluation result historical database includes a plurality of historical cosmetic medical auxiliary evaluation results, wherein each historical cosmetic medical auxiliary evaluation result includes at least a subject name and basic data, a historical facial expression evaluation result, a facial feature, a functional medical anatomy rule and a dynamic medical anatomy rule in the medical knowledge rule module, a combination and preferred order of evaluation treatment site results, and a type and dose of injection filler.

[0023] Preferably, the facial feature includes a habitual expression static wrinkle feature, a static contour line feature, or a skin texture feature.

[0024] Preferably, before performing the artificial intelligence cosmetic medical recognition analysis procedure, the artificial intelligence training of the artificial intelligence cosmetic medical recognition analysis procedure is performed by using at least one of an artificial neural network algorithm and a deep learning algorithm with the plurality of historical cosmetic medical auxiliary evaluation results.

[0025] Another embodiment of the present application provides an artificial intelligence auxiliary evaluation system for cosmetic medical application, comprising at least an artificial intelligence facial expression evaluation module for providing a real-time facial expression evaluation result of a subject; an artificial intelligence cosmetic medical recognition analysis module connected to the artificial intelligence facial expression evaluation module; and an input / output module connected to the artificial intelligence cosmetic medical recognition analysis module for inputting a basic data and / or a facial feature of the subject and outputting to the artificial intelligence cosmetic medical recognition analysis module; wherein the artificial intelligence cosmetic medical recognition analysis module is used to receive at least one of the basic data and / or the facial feature of the subject and the real-time facial expression evaluation result, and perform an artificial intelligence cosmetic medical recognition analysis procedure according to at least one of a connected medical knowledge rule module and a cosmetic medical auxiliary evaluation result historical database, and generate and output a real-time cosmetic medical auxiliary evaluation result to the input / output module.

[0026] Preferably, the artificial intelligence cosmetic medical recognition analysis module feeds back and stores the real-time cosmetic auxiliary evaluation result to at least one of the medical knowledge rule module and the cosmetic medical auxiliary evaluation result historical database.

[0027] Preferably, the real-time cosmetic medical auxiliary evaluation result at least includes a combination and preferred order of an evaluation treatment site result of the subject, or the combination and preferred order of the evaluation treatment site result and an injection filler type and dose.

[0028] Preferably, the medical knowledge rule module further includes a functional medical anatomy rule and a dynamic medical anatomy rule.

[0029] Preferably, the artificial intelligence facial expression evaluation module includes: an artificial intelligence image detection unit for executing an artificial intelligence image detection program to obtain a real-time facial image of the subject; an artificial intelligence image calibration and feature extraction combined unit connected to the artificial intelligence image detection unit, the artificial intelligence image calibration and feature extraction combined unit is used to execute an artificial intelligence image calibration and feature extraction program based on the real-time facial image, and obtain a facial surface and geometric feature information; an artificial intelligence facial action coding unit connected to the artificial intelligence image calibration and feature extraction combined unit, the artificial intelligence facial action coding unit is used to execute an artificial intelligence facial action coding program based on the facial surface and the geometric feature information, and obtain a plurality of facial action coding information; and an artificial intelligence facial emotion recognition unit connected to the artificial intelligence facial action coding unit, the artificial intelligence facial emotion recognition unit is used to execute an artificial intelligence facial emotion recognition program based on the plurality of facial action coding information, and obtain a proportional distribution and combination information of a plurality of emotion indicators corresponding to the real-time facial image, and form the real-time facial expression evaluation result.

[0030] Preferably, the artificial intelligence machine learning method applied to the artificial intelligence image detection program includes a Haar feature combined with adaptive boosting machine learning method belonging to the boundary detection algorithm type, or a histogram of oriented gradients combined with support vector machine machine learning method.

[0031] Preferably, before executing the artificial intelligence facial action coding program, the artificial intelligence training of a facial action coding model in the artificial intelligence facial action coding program is performed by a certain amount of training data set and a facial action coding system in a machine learning manner.

[0032] Preferably, the artificial intelligence training of the facial action coding model includes training of the facial action coding model under different situations of a static expression and / or a dynamic expression to obtain a static facial action coding information and / or a dynamic facial action coding information, respectively.

[0033] Preferably, before performing the artificial intelligence facial emotion recognition procedure, an artificial intelligence training of a facial emotion recognition model in the artificial intelligence facial emotion recognition procedure is performed by using a number of training data sets and by using a machine learning manner, and the facial emotion recognition procedure is performed by using the facial action coding system and by using another machine learning manner.

[0034] Preferably, the artificial intelligence image calibration and feature extraction combination unit comprises: a facial key point identification unit connected to the artificial intelligence image detection unit, the facial key point identification unit being configured to perform an artificial intelligence facial key point identification procedure and obtain facial key point identification information; a facial calibration and masking unit connected to the facial key point identification unit, the facial calibration and masking unit being configured to perform a facial image calibration procedure based on the facial key point identification information and obtain normalized facial image information; and a facial feature extraction unit connected to the facial calibration and masking unit, the facial feature extraction unit being configured to perform a facial image feature extraction procedure based on the facial key point identification information and the normalized facial image information and obtain facial image surface and geometric feature information.

[0035] Preferably, before performing the artificial intelligence facial key point identification procedure, an artificial intelligence training of a facial key point identification model in the artificial intelligence facial key point identification procedure is performed by using a number of training data sets and by using a machine learning manner.

[0036] Preferably, the facial image feature extraction procedure comprises at least a facial image surface feature extraction procedure and a facial image geometric feature extraction procedure; wherein the facial image surface feature extraction procedure comprises performing a histogram of oriented gradients to obtain multi-dimensional vector data and performing a principal component analysis to reduce the amount of vector data and retain facial image surface feature information, and the facial image geometric feature extraction procedure comprises obtaining facial image geometric feature information based on the facial key point identification information.

[0037] Preferably, the cosmetic medical assistance evaluation result historical database comprises a plurality of historical cosmetic medical assistance evaluation results; wherein each historical cosmetic medical assistance evaluation result comprises at least: a subject name and basic data, a historical facial expression evaluation result, a facial feature, a functional medical anatomy rule and a dynamic medical anatomy rule in the medical knowledge rule module, a combination and preferred order of evaluation treatment site results, and a type and dose of injection filler.

[0038] Preferably, the facial feature comprises a habitual expression static wrinkle feature, a static contour line feature, or a skin quality feature.

[0039] Preferably, before performing the artificial intelligence cosmetic medical recognition analysis procedure, the plurality of historical cosmetic medical auxiliary assessment results are used to perform artificial intelligence training of the artificial intelligence cosmetic medical recognition analysis procedure using at least one of an artificial neural network algorithm and a deep learning algorithm.

[0040] Preferably, based on the artificial intelligence facial expression assessment module, the artificial intelligence cosmetic medical recognition analysis module, and the output input module, an electronic device is assembled; wherein the electronic device is a handheld smart mobile device, a personal computer, or a standalone smart device.

[0041] Preferably, the electronic device is connected to at least one of the cosmetic medical auxiliary assessment result history database and the medical knowledge rule module using at least one of a wireless transmission method and a wired transmission method. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0043] Figure 1 The conceptual diagram of the artificial intelligence auxiliary assessment system provided by an embodiment of the present application.

[0044] Figure 2A The conceptual diagram of the artificial intelligence image detection unit shown in Figure 1

[0045] The embodiment schematic diagram of the artificial intelligence facial action coding unit shown in Figure 2B Figure 1 The embodiment schematic diagram of the artificial intelligence facial action coding unit shown in

[0046] Figure 2C Figure 1 The embodiment schematic diagram of the artificial intelligence facial action coding unit shown in

[0047] Figure 2D The embodiment schematic diagram of the artificial intelligence facial action coding unit shown in Figure 1

[0048] Figure 2E The embodiment schematic diagram of the artificial intelligence facial action coding unit shown in Figure 1 ​​​An embodiment schematic diagram of the facial emotion recognition unit of the artificial intelligence is used to quantitatively analyze the action intensity of the facial muscle group corresponding to each facial action unit in different emotional expressions according to the emotion theory.

[0049] Figure 3 An embodiment schematic diagram of the facial emotion recognition unit of the artificial intelligence is used to quantitatively analyze the action intensity of the facial muscle group corresponding to each facial action unit in different emotional expressions according to the emotion theory. Figure 1 An embodiment schematic diagram of the facial emotion recognition unit of the artificial intelligence is used to quantitatively analyze the action intensity of the facial muscle group corresponding to each facial action unit in different emotional expressions according to the emotion theory.

[0050] Figure 4 An embodiment schematic diagram of the facial emotion recognition unit of the artificial intelligence is used to quantitatively analyze the action intensity of the facial muscle group corresponding to each facial action unit in different emotional expressions according to the emotion theory.

[0051] Figures 5A to 5F An embodiment schematic diagram of the facial emotion recognition unit of the artificial intelligence is used to quantitatively analyze the action intensity of the facial muscle group corresponding to each facial action unit in different emotional expressions according to the emotion theory.

[0052] Figure 6 An embodiment schematic diagram of the facial emotion recognition unit of the artificial intelligence is used to quantitatively analyze the action intensity of the facial muscle group corresponding to each facial action unit in different emotional expressions according to the emotion theory.

[0053] Figure 7 An embodiment schematic diagram of the facial emotion recognition unit of the artificial intelligence is used to quantitatively analyze the action intensity of the facial muscle group corresponding to each facial action unit in different emotional expressions according to the emotion theory.

[0054] Main component symbol explanation:

[0055] 100 is an artificial intelligence assisted assessment system; 110 is an artificial intelligence facial expression assessment module; 111 is an artificial intelligence image detection unit; 112 is an artificial intelligence image calibration and feature extraction combined unit; 1121 is a facial key point identification unit; 1122 is a facial calibration and masking unit; 1123 is a facial feature extraction unit; 113 is an artificial intelligence facial motion coding unit; 114 is an artificial intelligence facial emotion recognition unit; 120 is an artificial intelligence cosmetic medical recognition analysis module; 121 is an artificial intelligence cosmetic medical recognition analysis program; 130 is a medical knowledge rule module; 131 is a functional medical anatomy rule; 132 is a dynamic medical anatomy rule; 140 is a cosmetic medical assisted assessment result historical database; 141 is a historical cosmetic medical assisted assessment result 1-N; 150 is an output input module; 160 is an electronic device; 20 is a rectangular (or square) box; 21 is facial key point identification information; A is a real-time facial expression assessment result; A' is a historical facial expression assessment result; A31-A33 is an emotion index combination; AU1-AUn is static facial motion coding information; A1 is a combination of multiple static facial motion coding information; 22, AU1'-AUn' is dynamic facial motion coding information; A2 is a combination of multiple dynamic facial motion coding information; B is basic data; B1 is gender; B2 is age; C1, C2 is a combination of assessment treatment site results and a preferred combination of preferred order; P is personal facial features; P1, P1' is a static texture feature of habitual expression; P2, P2' is a static contour feature; P3, P3' is a skin texture feature; D is the type of injectable filler; U is the dose of injectable filler; R1, R11-R14 are multiple medical rules of functional medical anatomy rules; R2, R21-R24 are multiple medical rules of dynamic medical anatomy rules; S1-S8 are process steps of an artificial intelligence assisted assessment method. DETAILED DESCRIPTION

[0056] The following detailed description of the proposed embodiments is provided only as an example and does not limit the scope of the invention. In addition, unnecessary components in the embodiments are omitted or components that can be completed with common technology are omitted to clearly show the technical features of the invention. The following further describes the preferred embodiments of the invention with reference to the drawings.

[0057] Please refer to Figure 1 which is a preferred embodiment schematic diagram of the artificial intelligence assisted assessment system in the implementation of the invention.

[0058] As Figure 1As shown, the embodiment of the present application is based on facial expression and applied to the artificial intelligence assisted evaluation system 100 for cosmetic medical treatment, which comprises an artificial intelligence facial expression evaluation module 110, an artificial intelligence cosmetic medical recognition analysis module 120, a medical knowledge rule module 130, a cosmetic medical auxiliary evaluation result historical database 140, and an output input module 150.

[0059] The artificial intelligence facial expression evaluation module 110 at least comprises an artificial intelligence image detection unit 111, an artificial intelligence image calibration and feature extraction combination unit 112, an artificial intelligence facial action coding unit 113, and an artificial intelligence facial emotion recognition unit 114; preferably, the artificial intelligence image calibration and feature extraction combination unit 112 further comprises a facial key point identification unit 1121, a facial calibration and masking unit 1122, and a facial feature extraction unit 1123.

[0060] In addition, the medical knowledge rule module 130 at least comprises a functional medical anatomy rule 131 and a dynamic medical anatomy rule 132; wherein the functional medical anatomy rule 131 and the dynamic medical anatomy rule 132 respectively comprise a plurality of different medical rules R1, R2, which will be described in detail later. Figure 3

[0061] As for the cosmetic medical auxiliary evaluation result historical database 140, it at least comprises a plurality of cosmetic medical auxiliary evaluation results 1-N (141), i.e., the cosmetic medical auxiliary evaluation results 1-N (141) with N historical records in the cosmetic medical auxiliary evaluation result historical database 140.

[0062] Furthermore, the output input module 150 is used to receive input or output various information, for example, receiving input of the basic data B1-B2 and / or personal facial features P1-P3 of the testee and inputting them into the artificial intelligence cosmetic medical recognition analysis module 120; or, outputting a real-time cosmetic medical auxiliary evaluation result received from the artificial intelligence cosmetic medical recognition analysis module 120, which at least includes the preferred combination C1-C2 of the combination and preferred order of the evaluation treatment site result, and / or the type D of the injection filler and the dose U of the injection filler.

[0063] The basic data B1-B2 can respectively refer to the gender and age of the testee and other parameter information; the personal facial features P1-P3 can be used to respectively refer to the static line feature P1, the static contour feature P2, or the skin texture feature P3 of the habitual expression of the testee and other parameter information; in addition, the personal facial features P1'-P3' are another facial feature parameter information of the testee provided directly by the artificial intelligence facial expression evaluation module 110.

[0064] ​Of course, the aforementioned basic data B1-B2 and / or personal facial features P1-P3 (or P1'-P3') of the subject to be tested, or the preferred combination C1-C2 and / or the type D of the injectable filler and the dose U of the injectable filler, or the number or type of each item of medical rule R1, R2, etc. and the parameter information related to the cosmetic medical auxiliary assessment result, are only for convenient description of the subsequent cooperation Figure 3 The preferred embodiment of the cosmetic medical auxiliary assessment result history database is shown in the schematic diagram of the example arranged. The embodiments of the present application are not limited thereto.

[0065] Furthermore, the artificial intelligence facial expression evaluation module 110, the artificial intelligence cosmetic medical identification analysis module 120, and the input and output module 150 can be assembled into an electronic device 160. The electronic device 160 can be a handheld smart mobile device, a personal computer (PC), or a stand-alone smart device. For example, the electronic device 160 can be a tablet computer, a smart mobile device, a notebook computer, a desktop computer, a stand-alone smart device, or a stand-alone smart module, wherein the smart device or the smart module can be assembled or separated from a medical device (not shown in the figure).

[0066] Moreover, the electronic device 160 is connected to the cosmetic medical auxiliary assessment result history database 140 and / or the medical knowledge rule module 130 in a wireless transmission manner and / or a wired transmission manner. For example, the cosmetic medical auxiliary assessment result history database 140 and / or the medical knowledge rule module 130 can be stored in a cloud storage platform, and the electronic device 160 can be connected to the cloud storage platform via various regional / wide area networks (not shown in the figure).

[0067] Please also refer to the artificial intelligence facial expression evaluation module 110 in Figure 1 which is the focus of the embodiments of the present application, and please cooperate with Figure 2A Figure 2E and Figure 3As shown. The AI ​​image detection unit 111 executes an AI image detection program and, after performing an image capture action (e.g., taking a picture with a camera device (not shown), obtains a real-time facial image of the subject. The AI ​​image calibration and feature extraction combination unit 112 is connected to the AI ​​image detection unit 111. Based on the real-time facial image, the AI ​​image calibration and feature extraction combination unit 112 executes an AI image calibration and feature extraction program to obtain facial surface and geometric feature information. The AI ​​facial motion encoding unit 113 is connected to the AI ​​image calibration and feature extraction combination unit 112. Based on the facial surface and geometric feature information, the AI ​​facial motion encoding unit 113 executes an AI facial motion encoding program to obtain corresponding multiple facial motion encoding information. The AI ​​facial emotion recognition unit 114 is connected to the AI ​​facial motion encoding unit 113. The AI ​​facial emotion recognition unit 114 is used to obtain multiple facial emotion encoding information (e.g., such as...) based on the facial motion encoding information. Figure 3 The static facial motion encoding information AU1-AUn and the dynamic facial motion encoding information AU1'-AUn' are used to execute an artificial intelligence facial emotion recognition program, thereby obtaining multiple emotion indicators corresponding to the real-time facial image (e.g., such as...). Figure 3 The proportion and combination information of the emotional indicators (A31-A33) are used to form the real-time facial expression evaluation result A.

[0068] The artificial intelligence machine learning method applied to this artificial intelligence image detection program may include, for example, a machine learning method that combines Haar-like features belonging to the boundary detection algorithm type with adaptive enhancement, or a machine learning method that combines a histogram of oriented gradients (HOG) with a support vector machine (SVM), but the embodiments of the present invention are not limited thereto.

[0069] Please refer to the following: Figure 2A This is in accordance with the embodiments of the present invention. Figure 1 A preferred embodiment of the AI ​​image detection unit 111 is illustrated in the diagram; wherein a rectangular (or square) box 20 is used to mark the detected facial image so that it can be used by the subsequent AI image calibration and feature extraction combination unit 112.

[0070] In addition, before executing the artificial intelligence facial action coding program, the artificial intelligence facial action coding model in the artificial intelligence facial action coding program is trained in advance by a certain amount of training data set and a facial action coding system (FACS) in a machine learning manner; wherein the training of the facial action coding model includes training the facial action coding model in different situations of a static expression and / or a dynamic expression to respectively obtain a static facial action coding information (for example, static facial action coding information AU1-AUn as shown in Figure 3 and / or a dynamic facial action coding information (for example, dynamic facial action coding information AU1'-AUn' as shown in Figure 3 ).

[0071] Another preferred method, in the actual execution of the artificial intelligence facial action coding unit 113 in the artificial intelligence facial expression evaluation module 110, when facing a new subject, it can be required to test in different situations of a static expression and / or a dynamic expression to synchronously obtain the static facial action coding information AU1-AUn and the dynamic facial action coding information AU1'-AUn' of the new subject for subsequent artificial intelligence cosmetic medical recognition analysis module 120 to perform more accurate recognition analysis, and can be stored in the cosmetic medical auxiliary evaluation result historical database 140.

[0072] Of course, please refer to Figure 2B , which is a preferred embodiment of the artificial intelligence facial action coding unit 113 in the present application Figure 1 , and the facial action coding information in this figure is a dynamic facial action coding information 21 showing a smiling face. In addition, Figure 2C is a preferred embodiment of the artificial intelligence facial action coding unit 113 in the present application Figure 1 from the perspective of human anatomy.

[0073] As shown in Figure 2C , the corrugator supercilii muscle at the brow can be defined as the facial action coding information AU4 (Face Action Unit 4), or the depressor anguli oris muscle and the mentails muscle that can make the corners of the mouth droop can be defined as the facial action coding information AU15 (Face Action Unit 15) and the facial action coding information AU17 (Face Action Unit 17), respectively.

[0074] Please refer to Figure 2D , which is a preferred embodiment of the present application Figure 1 , before the artificial intelligence facial action coding unit 113, in different situations of a static expression and a dynamic expression, the artificial intelligence is trained for the facial action coding model to obtain a static facial action coding information and / or a dynamic facial action coding information.

[0075] Among them, as shown in the left side of Figure 2D , a plurality of static facial action coding information (for example, AU1, AU4, AU15) is recorded when the subject is in a static state or the face is in a relaxed state, which is used to perceive and analyze the presentation of some associated specific facial muscle groups of the subject in a static state, and to observe whether there is an involuntary force or action in the static state. In addition, as shown on the right side of Figure 2D , the dynamic facial action coding information (for example, AU1', AU4', AU15', but not shown in the figure) of the subject can dynamically present different facial expressions according to different emotions (for example, sad emotion or more other angry expressions, laughing expressions and different emotions), and the dynamic changes of the associated specific facial muscle groups are observed.

[0076] In addition, before performing the artificial intelligence facial emotion recognition program, the artificial intelligence training of a facial emotion recognition model in the artificial intelligence facial emotion recognition program is performed by combining the facial action coding system (FACS) with at least one emotional valence, emotional arousal and a plurality of emotional indicators (for example, emotional indicators A31-A33 as shown in Figure 3 ) and another machine learning (Machine Learning) method.

[0077] Please refer to Figure 2E , which is a preferred embodiment of the present application Figure 1A preferred embodiment concept diagram of the artificial intelligence facial emotion recognition unit 114 is used to quantitatively analyze the motion intensity of each facial muscle group corresponding to the facial action coding information in different emotional expressions in combination with emotion theory. For example, the artificial intelligence facial expression evaluation module 110 of the embodiment of the present application can, for example, distinguish the facial emotional expressions into 7 categories of definitions, in addition to the neutral expression, including: happy, sad, angry, surprised, scared, disgusted, etc. 6 categories of expression definitions are formed to form various emotional indicators, and the embodiment of the present application is not limited thereto.

[0078] Among them, the emotion theory combined in the artificial intelligence facial emotion recognition program at least includes the "dimensional theory of emotion" proposed by Lisa Feldman Barrett, or the "discrete theory of emotion" proposed by Paul Ekman, and the embodiment of the present application is not limited thereto.

[0079] Further, the aforementioned artificial intelligence image calibration and feature extraction combination unit 112 of the embodiment of the present application can further include: a facial key point identification unit 1121 connected to the artificial intelligence image detection unit 111, the facial key point identification unit 1121 is used to execute an artificial intelligence facial key point identification program, and the facial key point identification information is obtained; a face calibration and masking unit 1122 connected to the facial key point identification unit 1121, the face calibration and masking unit 1122 is used to execute a face image calibration program based on the facial key point identification information, and the normalized face image information is obtained; and a face feature extraction unit 1123 connected to the face calibration and masking unit, the face feature extraction unit 1123 is used to execute a face image feature extraction program based on the facial key point identification information and the normalized face image information, and the face image surface and geometric feature information is obtained.

[0080] Preferably, before executing the artificial intelligence facial key point identification program, a certain amount of training data set is used to perform artificial intelligence training of a facial key point identification model in the artificial intelligence facial key point identification program in a machine learning manner.

[0081] Another preferred implementation, wherein the face image feature extraction procedure comprises at least a face image surface feature extraction procedure and a face image geometric feature extraction procedure; wherein the face image surface feature extraction procedure comprises performing a Histogram of Oriented Gradients (HOG) to obtain multi-dimensional vector data, and combining a Principal Component Analysis (PCA) to reduce the vector data amount and retain important face image surface feature information, and the face image geometric feature extraction procedure comprises obtaining a face image geometric feature information based on the face key point identification information.

[0082] Referring again to Figure 1 , the artificial intelligence cosmetic medical recognition analysis module 120 performs artificial intelligence training of the artificial intelligence cosmetic medical recognition analysis procedure using at least one of an artificial neural network algorithm and a deep learning algorithm based on the plurality of historical cosmetic medical auxiliary evaluation results 1-N (141), but embodiments of the present application are not limited thereto.

[0083] Referring again to Figure 3 , wherein the cosmetic medical auxiliary evaluation result history database 140 comprises a plurality of historical cosmetic medical auxiliary evaluation results 1-N (141); wherein each historical cosmetic medical auxiliary evaluation result comprises at least: a subject name and basic data B, a historical facial expression evaluation result A', a plurality of medical rules R1, R2 in the functional medical anatomy rule 131 and the dynamic medical anatomy rule 132 in the medical knowledge rule module 130, a preferred combination C1-C2 of a combination and preferred order of an evaluation treatment site result, and a kind D and dose U of an injection filler.

[0084] In actual application, the basic data B can include gender B1 and age B2; the historical facial expression evaluation result A' comprises at least: a combination of static facial action unit code information A1, a combination of dynamic facial action unit code information A2, and a plurality of emotion index combinations A31-A33; wherein the combination of static facial action unit code information A1 can be a plurality of static parameter values of the static facial action unit code information AU1-AUn of the subject without any emotion; the combination of dynamic facial action unit code information A2 can be a plurality of dynamic parameter values of the dynamic facial action unit code information AU1'-AUn' generated by the subject according to different emotions, and the emotion index combination A31-A33 can include, for example: fear index A31, anger index A32 and contempt index A33 which belong to negative emotion indexes, and / or can include happy index A31, moved index A32 and satisfied index A33 which belong to positive emotion indexes.

[0085] Further, the personal facial features P can include static wrinkle features P1, static contour features P2, or skin texture features P3 of habitual expressions; wherein the personal facial features P1-P3 can be provided by at least one of the facial expression assessment module 110 and the input / output module 150.

[0086] As for the medical rules R1 of the functional medical anatomy rules 231, for example, include the stretching degree rules, tension degree rules R11-R14 of each facial muscle group based on different emotional expressions; and, as for the medical rules R2 of the dynamic medical anatomy rules 232, for example, include the linkage rules, contraction rules R21-R24 between each facial muscle group based on different emotional expressions.

[0087] Finally, the preferred combinations C1-C2 can be, for example, one or a combination of a plurality of static facial motion encoding information AU1-AUn (or a plurality of dynamic facial motion encoding information AU1'-AUn') of a part of the subject to be treated; and the type D of the injection filler can include: a hydrogel agent W, a botulinum toxin agent X, a hyaluronic acid agent Y, and a collagen agent Z. Among them, the hydrogel agent W, the hyaluronic acid agent Y, and the collagen agent Z, in addition to being able to reduce the static wrinkles of the human face, thereby reducing the negative emotional indicator combination (sadness indicator, anger indicator, etc.), but also can increase the positive emotional indicator combination (happy indicator, satisfaction indicator, etc.).

[0088] Of course, in the present embodiment, the aforementioned historical cosmetic medical auxiliary assessment results 1-N (141) can be adjusted according to the actual cosmetic medical treatment target requirements, and should not be limited by the present example. And, the content of the aforementioned cosmetic medical auxiliary assessment results 1-N (141) can be variously changed or designed by those skilled in the art, and can be adjusted and designed to adapt to the actual needs of the subject's cosmetic medical treatment.

[0089] Referring to Figure 4 , which is a preferred implementation flow step concept diagram of the artificial intelligence auxiliary assessment method in the inventive concept of the present embodiment, which at least includes the following steps, and please refer to Figures 1 to 3 shown:

[0090] Start;

[0091] Step S1: executing an artificial intelligence image detection program to obtain a real-time facial image of a subject;

[0092] Step S2: based on the real-time facial image, executing an artificial intelligence facial key point identification program, and obtaining a facial key point identification information therefrom;

[0093] Step S3: based on the facial key point identification information, a facial image calibration procedure is performed to obtain normalized facial image information;

[0094] Step S4: based on the facial key point identification information and the normalized facial image information, a facial image feature extraction procedure is performed to obtain facial image surface and geometric feature information;

[0095] Step S5: based on the facial surface and the geometric feature information, an artificial intelligence facial action coding procedure is performed to obtain multiple facial action coding information;

[0096] Step S6: based on the multiple facial action coding information, an artificial intelligence facial emotion recognition procedure is performed to obtain the proportional distribution and combination information of multiple emotion indicators corresponding to the real-time facial image, and form a real-time facial expression evaluation result A;

[0097] Step S7: based on at least one of a medical knowledge rule module and a medical cosmetology auxiliary evaluation result historical database, an artificial intelligence medical cosmetology recognition analysis procedure is performed on the real-time facial expression evaluation result A, and a real-time medical cosmetology auxiliary evaluation result is output; and

[0098] Step S8: feedback and store the real-time medical cosmetology auxiliary evaluation result to at least one of the medical knowledge rule module and the medical cosmetology auxiliary evaluation result historical database;

[0099] End.

[0100] In addition, the following multiple preferred embodiments will be described based on the artificial intelligence auxiliary evaluation method and system of the embodiments of the present application, and how to perform medical cosmetology behavior. Among them, the main treatment target requirements of the following multiple preferred embodiments are to improve the negative emotion indicator combination of the facial expression of the subject 1-3, and to reduce or improve the negative emotion indicator combination caused by negative expression emotion, so as to improve personal charm and interpersonal relationship.

[0101] For example, the main treatment purpose of the multiple preferred embodiments is to reduce or improve the facial expression of involuntary frowning or drooping corners of the mouth, to reduce the feeling of being angry or serious, and even to reduce negative micro-expression, and in other preferred embodiments, the part of the positive emotion indicator combination can be further strengthened to achieve better, more accurate and personalized medical cosmetology effect.

[0102] Please refer to Figures 5A to 5F which is the first preferred embodiment concept diagram of applying the artificial intelligence auxiliary evaluation method and auxiliary evaluation system in the embodiment concept of the present application.

[0103] In the above-mentioned preferred embodiment concept, the main treatment purpose is to reduce or improve the facial expression of involuntary frowning or drooping corners of the mouth, to reduce the feeling of being angry or serious, and even to reduce negative micro-expression, and in other preferred embodiments, the part of the positive emotion indicator combination can be further strengthened to achieve better, more accurate and personalized medical cosmetology effect.Figures 5A to 5F the content, and in combination with Figures 1 to 4 As shown by way of example, based on the artificial intelligence facial expression evaluation module 110, the plurality of static facial action coding information AU1-AUn of the subject 1 is detected, and according to the expression change between the detection result of each static facial action coding information AU1-AUn and the detection result of another static facial action coding information AU1-AUn, dynamic facial action coding information AU1'-AUn' (not shown in the figure) is obtained, and a real-time facial expression evaluation result A is formed accordingly; that is, the real-time facial expression evaluation result A is combined to form each emotional indicator according to the proportion of the combination of static facial action coding information A1 and the combination of dynamic facial action coding information A2.

[0104] As shown in Figure 5A , the subject may unconsciously frown, or the corners of the mouth may sag due to aging, and the micro-expression caused by the plurality of static facial action coding information AU1-AUn is recorded and integrated to form the combination of static facial action coding information A1.

[0105] In addition, as shown in Figure 5B , the combination of dynamic facial action coding information A2 is the different facial expressions of the subject 1 according to different emotions, such as angry expression, laughing expression, etc.

[0106] Then, the artificial intelligence facial emotion analysis and recognition unit 114 further quantitatively analyzes the action intensity of each facial muscle group of the facial action coding information in different emotional expressions (including the combination of static facial action coding information A1 and the combination of dynamic facial action coding information A2), and provides more accurate dynamic parameter values as emotional indicator combinations A31-A33, evaluation treatment site result combinations and preferred combinations C1-C2 of preferred order, and treatment references of the type D of injection fillers and the dose U of fillers.

[0107] In addition, as shown in Figure 5C and Figure 5D , the plurality of emotional indicator combinations A31-A33 of the subject 1 are respectively 35.2% for sadness indicators, 14.1% for anger indicators, and 17.7% for fear indicators, and the related information of each facial action coding information AU1-AUn corresponding to the plurality of emotional indicator combinations A31-A33.

[0108] On the other hand, the facial expression evaluation module 110 can also be combined with a facial three-dimensional (3D) simulation unit and a skin quality detection unit to further provide the habitual expression static line feature P1, static contour feature P2, or skin quality feature P3, etc. (not shown in the figure) of the personal facial features P.

[0109] Then, asFigure 5E As shown, the real-time facial expression assessment result A is inputted into the artificial intelligence cosmetic medical recognition analysis module 120, or the artificial intelligence cosmetic medical recognition analysis module 120 actively receives the real-time facial expression assessment result A, and selects whether to combine at least one of the medical knowledge rule module 130 and the cosmetic medical auxiliary assessment result historical database 140 to execute the artificial intelligence cosmetic medical recognition analysis program 121. Subsequently, the real-time cosmetic medical auxiliary assessment result of the subject 1 is generated and outputted, wherein the real-time cosmetic medical auxiliary assessment result at least includes the preferred combination C1-C2 of the combination and preferred order of the assessment treatment site result, and the type D of the injected filler and the dose U of the filler.

[0110] In this embodiment, the cosmetic medical auxiliary assessment result suggestion is that Botulinum Toxin 8 s.U. is injected to the muscle group (Inner Frontalis) related to the facial action coding information AU1, and Botulinum Toxin DAO 4 s.U. and Mentalis 4 s.U. are injected to the muscle group (depressor anguli oris) related to the facial action coding information AU15 and the muscle group (mentails) related to the facial action coding information A17.

[0111] In this way, referring to Figure 5F As shown, the real-time facial expression assessment result of the subject 1 before treatment, one week after treatment and three weeks after treatment is compared, and it can be found from the comparison that the sadness index of the face of the subject 1 directly decreases from 35.2% to 0% after one week of treatment; in addition, for the change of the anger index of the face of the subject 1, the anger index decreases from 14.1% before treatment to 7.8% after one week of treatment, and even after three weeks of botulinum treatment, the anger index of the face can be completely reduced to 0%.

[0112] On the other hand, the cosmetic medical auxiliary assessment result suggestion for the sadness index can also give another treatment reference path through the implementation of the embodiment of the present application: for example, when the sadness index accounts for more than 10% of the total emotional index combination (total expression), and the facial action coding information AU1, the facial action coding information AU4 and the facial action coding information AU15 in the facial action coding unit all appear to be enhanced (i.e. the proportion is increased), it is suggested that Botulinum Toxin A type can be injected at the corresponding muscle group position; of course, the embodiment of the present application can obviously propose a better treatment suggestion scheme by using more case data (cosmetic medical auxiliary assessment result) and performing multiple artificial intelligence deep learning / training programs, which is not limited to the above-mentioned cosmetic medical auxiliary assessment result.

[0113] Further, please refer to Figure 6 which is a second preferred embodiment schematic diagram of the artificial intelligence assisted evaluation method and the artificial intelligence assisted evaluation system according to the embodiment of the present application.

[0114] As Figure 6 shown, based on the artificial intelligence facial expression evaluation module 110, the plurality of static facial action coding information AU1-AUn (and dynamic facial action coding information AU1'-AUn' (not shown in the figure)) of the subject 2 are detected, so as to provide the real-time facial expression evaluation result A, and it is known that the neutral index in the plurality of emotion index combinations of the subject 2 is 26.3%, which is the highest, followed by the sadness index of 13.9%, which is the second highest, and at the same time, the order of the recommended preferred treatment site in the cosmetic medical assisted evaluation result is the facial action coding information AU1 which mainly causes the sadness index, which is classified as the inner brow raiser.

[0115] Accordingly, the aforementioned real-time facial expression evaluation result A is combined with the medical knowledge rule module 130 and the cosmetic medical assisted evaluation result historical database 140, so that it is known that the muscle group highly related and connected to the aforementioned facial action coding information AU1 is located at the position guided by the functional medical anatomy rule and the dynamic medical anatomy rule, and then the real-time cosmetic medical assisted evaluation result for personalized cosmetic medical treatment can be further provided, for example, the treatment reference suggestion of 8 s.U. of botulinum toxin Toxin is applied to the muscle group related to the facial action coding information AU1 (inner frontalis).

[0116] In this way, by comparing the real-time facial expression evaluation results of the facial expression of the subject 2 before treatment, one week after treatment and three weeks after treatment, it can be found that the sadness index of the subject 2 directly decreases from 13.9% to 8.4% after one week of treatment, and even after three weeks of botulinum treatment, the sadness index of the face can be completely reduced to 0%. In short, the cosmetic medical treatment effect obtained by the artificial intelligence assisted evaluation method and the system thereof applied to cosmetic medical treatment according to the embodiment of the present application is indeed very significant.

[0117] Then, please refer to Figure 7 which is a third preferred embodiment schematic diagram of the artificial intelligence assisted evaluation method and the artificial intelligence assisted evaluation system according to the embodiment of the present application.

[0118] As Figure 7As shown, based on the artificial intelligence facial expression evaluation module 110, the facial action coding information AU1-AUn (and dynamic facial action coding information AU1'-AUn' (not shown in the figure)) of the subject 3 are detected to provide real-time facial expression evaluation results A, so that it can be known that the anger expression of any gender or age of human being is generally related to the muscle groups of the aforementioned facial action coding information AU15 and facial action coding information AU17; and by comparing the facial expression evaluation results of the subject 3 before treatment, one week after treatment and three weeks after treatment, it can be found that the anger index of the subject 3 has been significantly improved after three weeks of treatment.

[0119] That is, the cosmetic medical auxiliary evaluation result suggestion for the anger index can give another treatment reference guide path through the implementation of the embodiment of the present application, for example: when the anger index is more than 10% and the facial action coding information AU15 and the facial action coding information AU17 in the facial action coding unit are enhanced (i.e. the proportion is increased), it is suggested to inject botulinum toxin A type at the corresponding muscle group (depressor anguli oris and geniohyoid muscle) position.

[0120] However, compared with the current cosmetic medical treatment suggestion method based on the personal judgment of the doctor, it can be known that the actual operation in the prior art is indeed limited by the personal experience and stereotypes of the doctor, so that it cannot be considered objectively and there is a major defect of missing the difference of the micro-expression of each person.

[0121] In detail, the treatment target is to reduce the anger index, and the doctor often uses botulinum toxin to reduce the action of the corrugator muscle, while ignoring the fact that the muscle action of the anger expression of each person is actually slightly different. Some people may have their mouths down at the same time, or the chin muscle may contract and rise. Some people also have some action of the medial levator muscle, and the movement of these muscles may only be partially visible to the naked eye. Therefore, it is easy to cause the muscle movement to be too subtle and not easy to detect, thereby causing blind spots and misunderstandings in treatment, and thus causing adverse medical effects and unnecessary medical disputes.

[0122] For example, in the subject 1 in Figures 5A to 5F , if the current cosmetic medical treatment suggestion of the doctor is changed, the main treatment site will be concentrated on the facial action coding information AU15 and the facial action coding information AU17, wherein the doctor misses the facial action coding information AU1 of the anger index, and the cosmetic medical treatment effect is poor (not shown in the figure).

[0123] Secondly, in Figure 6Subject 2, based on the physician's current personal judgment regarding cosmetic medical treatment recommendations, typically focuses on facial movement coding information. The judgment is based on the orbicularis oculi muscle causing drooping eyes, which contributes to the sadness index. However, after the cosmetic procedure, comparing the real-time facial expression assessment results before treatment, one week later, and three weeks later, it was found that Subject 2's sadness index decreased from 6.8% to 5.1% one week after treatment. However, three weeks after Botox treatment, the sadness index rebounded to 6.7%. This was attributed to the physician's misjudgment of the treatment area, leading to poor cosmetic medical treatment results and an inability to effectively improve the sadness index (figure not shown).

[0124] Finally, Figure 7 If Subject 3's cosmetic medical treatment recommendation were changed to the current physician's personal judgment, the main treatment site would typically be an injection of 4 units of botulinum toxin A at facial motion code AU17. However, after the cosmetic medical procedure, comparing Subject 3's facial expressions before treatment, one week later, and three weeks later, it was found that Subject 3's facial anger index dropped from 10.9% to 5.9% after one week of treatment, but rose to 13.9% three weeks after the botulinum toxin treatment. This was because the physician not only failed to determine the treatment site (facial motion code AU15) for Subject 3, but also injected insufficient doses of botulinum toxin A at facial motion code AU17, resulting in the opposite effect of increasing the anger index instead of improving the subject's facial expression (figure not shown).

[0125] Compared to the AI-assisted assessment method and system for cosmetic medicine in this embodiment of the invention, this method provides high-quality real-time facial expression assessment results A for the subject based on the AI ​​facial expression assessment module 110. Then, the AI ​​cosmetic medicine identification and analysis module 120 selects and combines multiple medical rules from the medical knowledge rule module with a historical database of cosmetic medicine assisted assessment results to execute the AI ​​identification and analysis program, thereby generating and outputting cosmetic medicine assisted assessment results. These results include at least the combination and preferred order of the subject's assessed treatment site results and / or the type and dosage of injected fillers. In this way, this embodiment of the invention can not only accurately analyze and assess the correct and complete treatment site, but also accurately provide the type and dosage of injected fillers, thereby achieving a personalized aesthetic cosmetic medicine treatment effect.

[0126] In addition, the embodiment of the present application can also perform a cosmetic medical behavior of strengthening the positive emotion index combination, or make a preventive improvement cosmetic treatment suggestion for the treatment site of facial aging, such as the mouth corner sagging after the facial muscle relaxation, which may cause the angry indicator face.

[0127] On the other hand, the method and system of the embodiment of the present application can be applied to various cosmetic medical or aesthetic fields, and can also be used as a basis for judging the treatment effect before and after the cosmetic medical operation, and can also be applied to the medical teaching field, and can be used as a training doctor to improve the blind spot and misunderstanding in the previous treatment.

[0128] The present application has been disclosed in the foregoing with a preferred embodiment, and those skilled in the art should understand that the embodiment is only used to depict the present application, and should not be interpreted as limiting the scope of the present application. It should be noted that any equivalent changes and substitutions of the embodiment should be considered as covered within the scope of the present application. Therefore, the protection scope of the present application is defined by the following patent claim.

Claims

1. An AI-assisted evaluation method for cosmetic medical applications, characterized in that, It should include at least the following steps: (a) Provide a real-time facial expression assessment of a subject; and (b) Based on at least one of a medical knowledge rule module and a historical database of cosmetic medical auxiliary assessment results, an artificial intelligence cosmetic medical identification and analysis program is executed on the real-time facial expression assessment result, and a real-time cosmetic medical auxiliary assessment result is generated and output. Step (a) includes at least the following steps: (a1) Execute an artificial intelligence image detection program to obtain a real-time facial image of the subject; (a2) Based on the real-time facial image, perform an artificial intelligence image calibration and feature extraction procedure to obtain facial surface and geometric feature information; (a3) Based on the facial surface and geometric feature information, an artificial intelligence facial motion coding program is executed to obtain multiple facial motion coding information; and (a4) Based on the multiple facial motion encoding information, an artificial intelligence facial emotion recognition program is executed to obtain the proportional distribution and combination information of multiple emotion indicators corresponding to the real-time facial image, and to form the real-time facial expression evaluation result. Furthermore, the AI ​​image calibration and feature extraction procedure in step (a2) includes at least the following steps: (a21) Based on the real-time facial image, an artificial intelligence facial landmark identification program is executed to obtain facial landmark identification information. (a22) Based on the facial key point marker information, a facial image calibration procedure is performed to obtain a normalized facial image information; wherein the facial image calibration procedure includes at least using an affine transformation technique to eliminate errors in the facial key point marker information caused by different poses, and standardizing the size and presentation of the facial image to obtain the normalized facial image information; and (a23) Based on the facial key point identification information and the normalized facial image information, a facial image feature extraction procedure is performed to obtain the facial image surface and geometric feature information; wherein, the facial image feature extraction procedure includes at least a facial image surface feature extraction procedure and a facial image geometric feature extraction procedure; and, the facial image surface feature extraction procedure includes performing a directional gradient histogram to obtain multi-dimensional vector data, and combining it with a principal component analysis to reduce the amount of vector data and retain facial image surface feature information, and the facial image geometric feature extraction procedure includes obtaining facial image geometric feature information based on the facial key point identification information.

2. The AI-assisted evaluation method for cosmetic medical applications as described in claim 1, characterized in that, Step (b) is followed by the following steps: (c) Feedback and storage of the real-time cosmetic medical auxiliary assessment results to at least one of the medical knowledge rules module and the cosmetic medical auxiliary assessment results historical database.

3. The AI-assisted evaluation method for cosmetic medical applications as described in claim 1, wherein the real-time cosmetic medical assisted evaluation result includes at least: The combination and preferred order of the assessment results of the treatment site of the subject, or the combination and preferred order of the assessment results of the treatment site and a type and dosage of injectable filler.

4. The AI-assisted evaluation method for cosmetic medical applications as described in claim 1, characterized in that, The medical knowledge rules module also includes a functional medical anatomy rule and a dynamic medical anatomy rule.

5. The AI-assisted evaluation method for cosmetic medical applications as described in claim 1, characterized in that, The artificial intelligence machine learning methods applied to this AI image detection program include a machine learning method that combines Haar features with adaptive enhancement, which belongs to the boundary detection algorithm type, or a machine learning method that combines a directional gradient histogram with a support vector machine.

6. The AI-assisted evaluation method for cosmetic medical applications as described in claim 1, characterized in that, Before executing the AI ​​facial landmark identification program, a certain number of training datasets are used to train the facial landmark identification model in the AI ​​facial landmark identification program using a machine learning approach.

7. The AI-assisted evaluation method for cosmetic medical applications as described in claim 1, characterized in that, Before executing the AI ​​facial motion coding program, an AI training method is used to train the facial motion coding model in the AI ​​facial motion coding program using a certain amount of training dataset and a facial motion coding system.

8. The AI-assisted evaluation method for cosmetic medical applications as described in claim 7, characterized in that, The artificial intelligence training of the facial motion coding model includes training the facial motion coding model in different contexts of a static expression and / or a dynamic expression, so as to obtain static facial motion coding information and / or dynamic facial motion coding information respectively.

9. The AI-assisted evaluation method for cosmetic medical applications as described in claim 7, characterized in that, Before executing the AI ​​facial emotion recognition program, an AI training method is used to train a facial emotion recognition model in the AI ​​facial emotion recognition program, which includes at least one emotion positive / negative attribute, one emotion arousal level and multiple emotion index parameters, combined with the facial motion coding system.

10. The AI-assisted evaluation method for cosmetic medical applications as described in claim 1, characterized in that, The historical database of cosmetic medical auxiliary assessment results includes multiple historical cosmetic medical auxiliary assessment results; each historical cosmetic medical auxiliary assessment result includes at least: a subject's name and basic data, a historical facial expression assessment result, a facial feature, a functional medical anatomy rule and a dynamic medical anatomy rule in the medical knowledge rule module, a combination and preferred order of assessment results for treatment sites, and a type and dosage of injectable filler.

11. The AI-assisted evaluation method for cosmetic medical applications as described in claim 10, characterized in that, The individual's facial features include a static texture feature of a habitual expression, a static contour feature, or a skin texture feature.

12. The AI-assisted evaluation method for cosmetic medical applications as described in claim 10, characterized in that, Before executing the AI-powered cosmetic medical identification and analysis program, the AI-powered cosmetic medical identification and analysis program is trained using the multiple historical cosmetic medical auxiliary evaluation results and at least one of an artificial neural network algorithm and a deep learning algorithm.

13. An artificial intelligence-assisted evaluation system for cosmetic medical applications, characterized in that, At least including: An AI-powered facial expression assessment module provides a real-time facial expression assessment result for a test subject; An AI-powered cosmetic medical recognition and analysis module is connected to the AI-powered facial expression assessment module; and An input / output module is connected to the AI-powered cosmetic medical identification and analysis module. It is used to input basic data of the subject and / or facial features, and output to the AI-powered cosmetic medical identification and analysis module. The AI-powered cosmetic medical identification and analysis module receives at least one of the subject's basic data and / or personal facial features, as well as the real-time facial expression assessment result. Based on at least one of a connected medical knowledge rule module and a historical database of cosmetic medical auxiliary assessment results, it executes an AI-powered cosmetic medical identification and analysis program, and generates and outputs a real-time cosmetic medical auxiliary assessment result to the input / output module. An electronic device is assembled based on the AI ​​facial expression assessment module, the AI-powered cosmetic medical identification and analysis module, and the input / output module. The AI ​​facial expression assessment module includes: an AI image detection unit for executing an AI image detection program to obtain a real-time facial image of the subject; and an AI image calibration and feature extraction combination unit connected to the AI ​​image detection unit, which is used to execute an AI image calibration and feature extraction program based on the real-time facial image to obtain facial surface and geometric feature information. The AI ​​image calibration and feature extraction program includes at least the sequential execution of an AI facial key point identification program, a facial image calibration program, and a facial image feature extraction program. The AI ​​image calibration and feature extraction combination unit executes the facial image calibration program in response to facial key point identification information generated by the AI ​​facial key point identification program, thereby obtaining normalized facial image information. The AI ​​image calibration and feature extraction combination unit also executes the facial image feature extraction program in response to the facial key point identification information and the normalized facial image information, thereby obtaining the facial image surface and geometric feature information. The facial image calibration procedure includes at least using an affine transformation technique to eliminate errors caused by different poses in the facial key point identification information, and unifying the size and presentation of the facial image to obtain the normalized facial image information. Furthermore, the facial image feature extraction procedure includes at least a facial image surface feature extraction procedure and a facial image geometric feature extraction procedure. The facial image surface feature extraction procedure includes performing a histogram of directional gradients to obtain multi-dimensional vector data, and combining this with principal component analysis to reduce the amount of vector data and retain facial image surface feature information. The facial image geometric feature extraction procedure includes obtaining facial image geometric feature information based on the facial key point identification information.

14. The AI-assisted evaluation system for cosmetic medical applications as described in claim 13, characterized in that, The AI-powered cosmetic medical identification and analysis module feeds back and stores the real-time cosmetic assistance assessment results to at least one of the medical knowledge rules module and the historical database of cosmetic medical assistance assessment results.

15. The AI-assisted evaluation system for cosmetic medical applications as described in claim 13, characterized in that, The real-time cosmetic medical auxiliary assessment results include at least a combination and preferred order of assessment results for the treatment site of the subject, or a combination and preferred order of assessment results for the treatment site and a type and dosage of injectable filler.

16. The AI-assisted evaluation system for cosmetic medical applications as described in claim 13, characterized in that, The medical knowledge rules module also includes a functional medical anatomy rule and a dynamic medical anatomy rule.

17. The AI-assisted evaluation system for cosmetic medical applications as described in claim 13, characterized in that, This AI-powered facial expression assessment module further includes: An AI facial motion encoding unit is connected to the AI ​​image calibration and feature extraction combination unit. This AI facial motion encoding unit executes an AI facial motion encoding program based on the facial surface and geometric feature information, thereby obtaining multiple facial motion encoding information. An AI facial emotion recognition unit is connected to the AI ​​facial motion encoding unit. The AI ​​facial emotion recognition unit is used to execute an AI facial emotion recognition program based on the multiple facial motion encoding information, thereby obtaining the proportional distribution and combination information of multiple emotion indicators corresponding to the real-time facial image, and forming the real-time facial expression evaluation result.

18. The AI-assisted evaluation system for cosmetic medical applications as described in claim 17, characterized in that, The artificial intelligence machine learning methods applied to this AI image detection program include a machine learning method that combines Haar features with adaptive enhancement, which belongs to the boundary detection algorithm type, or a machine learning method that combines a directional gradient histogram with a support vector machine.

19. The AI-assisted evaluation system for cosmetic medical applications as described in claim 17, characterized in that, Before executing the AI ​​facial motion coding program, an AI training method is used to train the facial motion coding model in the AI ​​facial motion coding program using a certain amount of training dataset and a facial motion coding system.

20. The AI-assisted evaluation system for cosmetic medical applications as described in claim 19, characterized in that, The artificial intelligence training of the facial motion coding model includes training the facial motion coding model in different contexts of a static expression and / or a dynamic expression, so as to obtain static facial motion coding information and / or dynamic facial motion coding information respectively.

21. The AI-assisted evaluation system for cosmetic medical applications as described in claim 19, characterized in that, Before executing the AI ​​facial emotion recognition program, an AI training method is used to train a facial emotion recognition model in the AI ​​facial emotion recognition program, which includes at least one emotion positive / negative attribute, one emotion arousal level and multiple emotion index parameters, combined with the facial motion coding system.

22. The AI-assisted evaluation system for cosmetic medical applications as described in claim 17, characterized in that, This AI-powered image calibration and feature extraction unit includes: A facial key point identification unit is connected to the artificial intelligence image detection unit. The facial key point identification unit is used to execute the artificial intelligence facial key point identification program and thereby obtain the facial key point identification information. A face calibration and masking unit is connected to the face key point identification unit. This unit is used to perform a face image calibration procedure based on the face key point identification information, thereby obtaining the normalized face image information; and A facial feature extraction unit is connected to the facial calibration and masking unit. The facial feature extraction unit is used to perform a facial image feature extraction program based on the facial key point identification information and the normalized facial image information, and thereby obtain the surface and geometric feature information of the facial image.

23. The AI-assisted evaluation system for cosmetic medical applications as described in claim 22, characterized in that, Before executing the AI ​​facial landmark identification program, an AI training method is used to train a facial landmark identification model in the AI ​​facial landmark identification program using a certain amount of training dataset.

24. The AI-assisted evaluation system for cosmetic medical applications as described in claim 22, characterized in that, The historical database of cosmetic medical auxiliary assessment results includes multiple historical cosmetic medical auxiliary assessment results; each historical cosmetic medical auxiliary assessment result includes at least: a subject's name and basic data, a historical facial expression assessment result, a facial feature, a functional medical anatomy rule and a dynamic medical anatomy rule in the medical knowledge rule module, a combination and preferred order of assessment results for treatment sites, and a type and dosage of injectable filler.

25. The AI-assisted evaluation system for cosmetic medical applications as described in claim 24, characterized in that, The individual's facial features include a static texture feature of a habitual expression, a static contour feature, or a skin texture feature.

26. The AI-assisted evaluation system for cosmetic medical applications as described in claim 24, characterized in that, Before executing the AI-powered cosmetic medical identification and analysis program, the AI-powered cosmetic medical identification and analysis program is trained using the multiple historical cosmetic medical auxiliary evaluation results and at least one of an artificial neural network algorithm and a deep learning algorithm.

27. The AI-assisted evaluation system for cosmetic medical applications as described in claim 13, characterized in that, The electronic device can be a handheld smart mobile device, a personal computer, or a stand-alone smart device.

28. The AI-assisted evaluation system for cosmetic medical applications as described in claim 27, characterized in that, The electronic device is connected to at least one of the historical database of cosmetic medical auxiliary assessment results and the medical knowledge rules module via at least one of a wireless transmission method and a wired transmission method.

Images