Method for modeling the structure and dynamic expressions of a 4d dynamic skin

Abstract

The application relates to the field of artificial intelligence and discloses a VEM-4D dynamic skin structure and a dynamic expression model construction method, and a 4D dynamic skin structure is provided, under the control of a 4D expression model and a driving module, the dynamic expression of the skin is simulated, wherein, under the expression instruction, an expression function vector is generated in the 4D expression model, excitation currents are provided through the driving module, a parallel electromagnet matrix arranged in an elastic soft magnetic silica gel composite material is driven, two-dimensional surface stretching deformation of the skin is realized, concave-convex electromagnets are driven, three-dimensional concave-convex deformation of the skin is realized, a time-saving current beat is adopted to realize beat per minute (BPM) peristalsis of the skin, and a mapping set corresponding to the expression function vector and the dynamic expression and an expression library are obtained through supervised learning and reinforcement learning; the application has the advantages of simple structure, low cost, convenient secondary development, application in the field of embodied robots, realization of the dynamic expression of facial skin, and solution to the problem of expression personification.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence, specifically to the skin structure and control of animate robots and robotic animals, particularly to the subfield of skin with dynamic expressions for animate robots, with the aim of achieving rhythmic peristalsis of the skin. Background Technology

[0002] With the development of the android industry, the limbs of androids have become widely used due to innovations in various motion components. However, progress in the area of ​​android skin remains unsatisfactory, especially in the facial skin, a core component for the emotional expression of androids using artificial intelligence. To date, no suitable technology or manufacturing solution has been developed, and this predicament actually hinders the development and application of anthropomorphic androids.

[0003] According to the research conducted by the invention team, the current skin-related technical solutions fall into the following categories:

[0004] 1. Materials-driven

[0005] Pneumatic / hydraulic driven soft skin: This type of skin uses elastic materials such as silicone to embed a miniature pneumatic network (pneumatic artificial muscle, PAM). By controlling the air pressure in different cavities, the skin can locally expand, contract, and bend, creating expressions such as smiling and frowning. Advantages include gentle movements and high biomimicry; disadvantages include a relatively slow response speed, the need for a complex pneumatic control system, and the risk of air leakage.

[0006] Shape memory materials:

[0007] Shape memory alloy (SMA) wires / springs: embedded in the skin substrate, they contract when heated by electricity, pulling the skin to produce deformation. Advantages include high strength and compact structure; disadvantages include high power consumption, difficult thermal management, limited cycle life, and slow response speed.

[0008] Shape memory polymers (SMPs) can undergo pre-programmed macroscopic deformations through stimuli such as heat and light. They are mostly used for overall contour changes, but pose a greater challenge for localized, fine, and rapid facial expression simulation.

[0009] Electroactive polymers (EAPs), such as dielectric elastomers (DEs), can generate significant strain under an electric field, similar to "artificial muscles." They hold immense theoretical potential, but practical applications require extremely high driving voltages (kilovolt levels), and material durability and encapsulation technology remain challenges.

[0010] Tendon-rope actuation: Biomimetic tendons (such as high-strength thin ropes or hoses) are placed under flexible skin and pulled by a distal motor to achieve displacement of specific points on the skin. This is a classic biomimetic method with relatively direct control, but the system is complex and prone to mechanical wear and noise.

[0011] 2. Mechanical structure type

[0012] Multi-degree-of-freedom micro-motor / linear actuator array: Micro-servo mechanisms are installed below key facial points (such as eyebrows and corners of the mouth) to directly push and pull the skin. This is currently the mainstream solution for many humanoid robots (such as Sophia and Ameca). The advantages are precise control and fast response; the disadvantages are that the face may appear mechanical, noisy, and the deformation of the skin in non-drive areas may be unnatural.

[0013] Linkage mechanism: A precisely designed mechanical linkage driven by one or a few motors, achieving coordinated movement of multiple facial expression points through mechanism switching. It can simplify the control system, but requires extremely high precision in mechanical design, and the fixed expression patterns limit flexibility.

[0014] 3. Hybrid Drive and Integrated System

[0015] Hybrid actuation: The core areas (such as mouth opening and closing) use motors, while subtle expressions (such as cheek puffing and wrinkles) are actuated by pneumatic or EAP (effortless actuation) software. The aim is to balance power, speed, and naturalness.

[0016] Multi-layered composite skin:

[0017] Functional layering: The outermost layer is a biomimetic textured and colored silicone layer; the middle layer is a strain layer with embedded drive elements (such as SMA wires and fluid channels); the innermost layer may be a base layer connected to the robot's facial skeleton.

[0018] Sensor integration: Integrating flexible pressure, strain, and even temperature sensors into the skin to form a "sensing-actuating" closed loop, enabling robots to sense touch and make more agile responses.

[0019] 4. VEM-Token Model Technology

[0020] This invention team has for the first time proposed a set of novel and innovative VEM (Vocal-Emotion-Multimodal) technologies, comprising a patent pool consisting of the following five already granted invention patents:

[0021] 4.1 VEM-Token Vocal Emotion Multimodal Model

[0022] This refers to the model in "VEM-Token Vocal Emotion Multimodal Tokenization Deep Learning Method for Singing and Accompaniment, CN120126506B", which specifically includes the following steps and methods 1-4:

[0023] (1) Use one or more modalities to record emotions, label vocal emotion multimodalities as VEM, construct VEM classification, VEM coordinate system, VEM function and VEM library. Vocal emotions include one or a combination of joy, sadness, anger, fear, disgust, surprise, calmness, expectation, trust, love, hate, affection and enmity. Multimodalities include one or a combination of lyrics, singing, accompaniment, vocal style, music, emotional basis, accompanying instruments, video and image. The VEM coordinate system includes a coordinate axis system established based on independent emotions, opposite emotion pairs and related opposite emotion groups.

[0024] (2) Collect vocal samples according to VEM classification, and have human vocal experts evaluate the vocal samples in terms of emotion in singing and accompaniment. Use supervised learning and deep learning to train the VEM function to obtain VEM parameters and add them to the VEM library.

[0025] (3) The VEM processor is used to calibrate the rhythm of the vocal file and separate the vocal stream and the accompaniment stream. The vocal file is split into VEM-Tokens according to the rhythm, the vocal stream is converted into a VEM-Token1 sequence, the accompaniment stream is converted into a VEM-Token2 sequence, and added to the preprocessing library.

[0026] (4) Using deep learning, the dialogue expression chart, lyrics chart, VEM-Token song chart, VEM-Token accompaniment chart, and VEM-Token music score are generated respectively.

[0027] 4.2 Model for VEM-Token Beat Capture and Beat Alignment

[0028] This refers to the model in "Method for Constructing VEM-Token Beat Capture and Alignment Model, CN120748450B", which specifically includes the following steps and methods (5-7):

[0029] (5) For vocal files, based on the VEM-Token vocal emotion multimodal model, a beat model including beat capture and beat alignment is set to capture the beat of the vocal file. Based on the beat, the vocal file is divided into VEM-Token sequences, and the starting point and ending point of the beat in each VEM-Token are marked.

[0030] (6) Set the starting point alignment model, including:

[0031] The vocal files, including the sample files and the user files generated by the user imitating the sample files, are divided into VEM-Token1 sequences and VEM-Token2 sequences, respectively. Based on the starting point of each VEM-Token1, a starting point fine-tuning step is used to adjust the starting point of the corresponding VEM-Token2 one by one, so that it is aligned with the starting point of the corresponding VEM-Token1.

[0032] For the dialogue segments and rhythms included in the voice dialogue files in social interactions, taking the first segment as a reference, starting from the second segment, a starting point fine-tuning step is adopted to adjust the starting point of each VEM-Token in each segment one by one, so as to align with the starting point of the VEM-Token at the corresponding position in the first segment, until all loop segments are completed.

[0033] (7) Set the endpoint alignment model, including:

[0034] Based on the endpoint of each VEM-Token1, an endpoint fine-tuning step is used to adjust the endpoint of the corresponding VEM-Token2 one by one, so that it is aligned with the endpoint of the corresponding VEM-Token1.

[0035] For each segment of the dialogue, taking the first segment as a reference, starting from the second segment, a fine-tuning step is used to adjust the endpoint of each VEM-Token in each segment one by one, so that it is aligned with the endpoint of the VEM-Token at the corresponding position in the first segment, until all loop segments are completed.

[0036] 4.3 VEM-Token Vocal Emotion Multimodal Modified Model

[0037] This refers to the model in "Construction Method of VEM-Token Vocal Emotion Multimodal Modification Model, CN120853611B", which specifically includes the following 8 and 9 steps and methods:

[0038] (8) Collect sample files and user files. Based on the VEM-Token model, use beat capture and VEM-Token segmentation to obtain the VEM-Token1 sequence of the sample files and the VEM-Token2 sequence of the user files respectively. Based on the VEM-Token1 sequence, use beat alignment for all VEM-Token2 sequences to generate the VEM-Token2 sequence.

[0039] (9) Based on the VEM parameters included in the VEM-Token model, identify the VEM parameters of the VEM-Token1 sequence, determine the modification scheme by the user, process the VEM parameters of the VEM-Token2 sequence, and modify and generate the VEM-Token2 sequence that matches the style of the VEM-Token1 sequence.

[0040] 4.4 A Layered Fusion Method for VEM-Token Sentiment Synchronization Functions

[0041] This refers to the model in "A Method for Layered Fusion of VEM-Token Sentiment Synchronization Function, CN120913602B", which specifically includes the following steps and methods: 10, 11, and 12.

[0042] (10) The VEM-Token vocal emotion multimodal model is adopted. The song file is divided into VEM-Token sequences in units of beats. The emotion synchronization function is set as VEM-sync. VEM-sync includes a sequence of VEM-sync vectors that correspond to and are aligned with the VEM-Token sequence. The VEM-sync vector includes synchronization content and a synchronization pointer pointing to the corresponding VEM-Token beat.

[0043] (11) Based on the VEM-Token sequence, calculate and obtain the synchronization content of the VEM-sync vector sequence. The synchronization content includes beat attributes and emotion attributes, where the emotion attributes include more than one emotion name, emotion value, and emotion weight.

[0044] (12) The emotion weight function includes: using hierarchical processing of song files to obtain the corresponding VEM-sync vector sequence, and using one or a combination of multi-layer weighted scanning, recurrent neural network, long short-term memory network, self-attention mechanism and retrieval enhancement to generate forward propagation, backward propagation and omnidirectional propagation to obtain the emotion synchronization function with the song file.

[0045] 4.5 VEM-Token World Model Robot Expression Functions

[0046] This refers to the "Method for Constructing Expression Functions of VEM-Token World Model Robots, CN120951102A", which specifically includes the following steps 13 and 14:

[0047] (13) Constructing the VEM-fe expression function model, including: synthesizing one or more expression vectors representing emotional input into the output synthetic expression, aligning the expression vectors in the current beat, eliminating expression conflicts between two or more expression vectors, and transitioning from the synthetic expression of the current beat to the synthetic expression of the next beat, so that the robot can understand emotions and learn expressions.

[0048] (14) The robot uses a physical expression generator or an animated expression generator to form an expression display. The expression driving model is connected by the VEM-fe expression function model to transmit more than one synthetic expression and display facial expressions.

[0049] Insufficiency of existing technical methods

[0050] (1) No product has been found that can be applied to existing solutions for skin expression generation, whether material-driven, mechanical structure-driven, hybrid-driven or integrated system.

[0051] (2) The existing technical solutions are not only difficult to commercialize, but also have high estimated costs. Summary of the Invention

[0052] Based on the shortcomings of existing technologies, the invention team proposes a novel "VEM-4D dynamic skin structure and dynamic expression model construction method". This solution is simple in structure, easy to productize, low in cost, and easy to integrate with current artificial intelligence.

[0053] The purpose and intent of this invention are achieved by the following structure and method:

[0054] 1. Overall structure of VEM-4D dynamic skin

[0055] The present invention, as a structure for VEM-4D dynamic skin, includes, but is not limited to, the following structures:

[0056] VEM-4D dynamic skin includes a 4D dynamic skin and driving module and a 4D expression model, wherein:

[0057] The 4D dynamic skin includes, but is not limited to, two or more electromagnets. The electromagnets include, but are not limited to, parallel electromagnets composed of soft magnetic materials and excitation coils. The parallel electromagnets are arranged side by side at a interval of A. The front of the 4D dynamic skin includes, but is not limited to, the skin that is adhered to the magnetic poles on the front of the parallel electromagnets. The skin is composed of an elastic and stretchable membrane.

[0058] The drive module applies an excitation current to the excitation coil, which generates an attractive or repulsive force between the two magnetic poles of the electromagnet, thus driving the skin to expand and contract.

[0059] VEM-4D dynamic skin also includes, but is not limited to, a 4D facial expression model. The 4D facial expression model provides static and dynamic facial expression commands to the driving module, driving the skin to stretch and deform.

[0060] 2. Transverse partition structure:

[0061] Based on the aforementioned basic solution, the present invention includes, but is not limited to, one or more combinations of the following structures:

[0062] The 4D dynamic skin includes, but is not limited to, a transverse partition with a through-hole buckle, which holds the back magnetic pole of the parallel electromagnet, with a gap B between the through-hole buckle and the back magnetic pole.

[0063] The transverse diaphragm is made of a thin sheet of elastic material, which undergoes vertical elastic deformation when subjected to force; and / or,

[0064] Electromagnets also include, but are not limited to, concave-convex electromagnets with a gap C between the back magnetic pole and the outer side of the transverse partition. The concave-convex electromagnets include, but are not limited to, soft magnetic materials and excitation coils, and are installed on the back plate of the 4D dynamic skin. Using facial expression commands, the excitation coils of the concave-convex electromagnets are driven by the drive module, so that the magnetic poles of the concave-convex electromagnets generate attraction or repulsion on one or more back magnetic poles, driving the skin to produce concave-convex deformation.

[0065] The arrangement of electromagnets and the shape of the 4D dynamic skin are produced according to the actual application. The spacers and backplate also include, but are not limited to, spacer supports and skin positioning posts.

[0066] 3. Subcutaneous tissue structure:

[0067] Based on the foregoing solutions, the present invention includes, but is not limited to, one or more combinations of the following:

[0068] 4D dynamic skin also includes, but is not limited to, subcutaneous tissue filled with elastic material around parallel electromagnets. The epidermis adheres to the surface of the subcutaneous tissue on the front of the 4D dynamic skin. At this time, due to the attraction or repulsion of the front magnetic poles and the action of the concave and convex electromagnets, the subcutaneous tissue causes the epidermis to undergo stretching and concave deformation.

[0069] The subcutaneous tissue also contains, but is not limited to, powders of uniformly mixed soft magnetic materials.

[0070] The subcutaneous tissue also includes, but is not limited to, interconnected cavities filled with heat-dissipating fluid and temperature sensors arranged therein to control the temperature of the epidermis within a set range.

[0071] Subcutaneous tissue encapsulates and seals to form a complete 4D dynamic skin, and provides inlets and outlets for heat dissipation fluid.

[0072] 4. Structure of the driving module and 4D facial expression model:

[0073] Based on the aforementioned solution, the present invention includes, but is not limited to, one or more of the following combinations in the driving module and VEM-4D facial expression model:

[0074] The driving module includes, but is not limited to, an input terminal, a codec, a current source, and an output terminal. The output terminal is connected to the two ends of the excitation coil of each electromagnet. The current source provides the excitation current. The codec provides decoding based on the encoding of the electromagnet. The input terminal is connected to the 4D facial expression model and the driving power supply. The 4D facial expression model drives the electromagnets through the driving module to generate stretching and concave facial expressions on the skin.

[0075] Current sources include, but are not limited to, constant current sources, pulse current sources, and modulated pulse power supplies.

[0076] The driving module also includes, but is not limited to, connecting one or more temperature sensors, which are placed at designated locations in the subcutaneous tissue and transmit temperature signals back to the 4D facial expression model to monitor the temperature of the subcutaneous tissue.

[0077] The drive module also includes, but is not limited to, the connection terminal of the circulation pump that drives the heat dissipation fluid, and completes the temperature control and circulation of the heat dissipation fluid under the control of the 4D facial expression model.

[0078] The driving module also includes, but is not limited to, a filtering circuit to reduce or eliminate the excessive oscillation of the magnetic field in the arrangement of electromagnets caused by the high-frequency pulses of the 4D facial expression model.

[0079] The driving module also includes, but is not limited to, providing a demagnetizing current to the electromagnet under the control of the 4D facial expression model.

[0080] The driving module also includes, but is not limited to, a wireless communication submodule, which communicates with the 4D dynamic skin under the control of the 4D expression model.

[0081] 5. Overall Construction Method of VEM-4D Dynamic Expression Model

[0082] This invention, as a method for constructing VEM-4D dynamic facial expression models, is characterized by including but not limited to one or more combinations of the following:

[0083] S1000: The VEM-4D dynamic expression model includes a 4D expression model, a 4D dynamic skin, and a driving module. Expression commands are loaded at the input end of the 4D expression model, and the expression function vector generated at the output end provides excitation current to the excitation coil of the electromagnet in the 4D dynamic skin through the driving module and controls it. The attraction or repulsion between the magnetic poles of two or more electromagnets drives the epidermis on the front of the 4D dynamic skin to produce rhythmic undulation to simulate dynamic expressions.

[0084] S2000: Expression commands include, but are not limited to, stretch commands, which act on parallel electromagnets arranged side by side in the electromagnets to simulate stretch expressions in the epidermis.

[0085] S3000: Expression commands include, but are not limited to, convex / concave commands, which are applied to electromagnets including, but not limited to, convex / concave electromagnets located on the back of the 4D dynamic skin. Parallel electromagnets are driven by elastic diaphragms to the magnetic poles on the back of the 4D dynamic skin to simulate dynamic expressions on the epidermis, including, but not limited to, convex / concave expressions.

[0086] 6. The S1000 includes a design methodology for facial expression commands:

[0087] Based on the foregoing solutions, the present invention includes, but is not limited to, one or more combinations of the following:

[0088] S1100: Expression commands include, but are not limited to, the expression function FE (x, y, z, t, i, p, v, τ), where x, y, z, and t in the expression function vector (x, y, z, t, i, p, v, τ) are the coordinates of the electromagnet in 3D space and the clock value coordinates in the 4th dimension, respectively. i, p, v, and τ are the current value, current polarity, type of electromagnet, and step time, respectively, which are the current value, current polarity, type of electromagnet, and step time. The step time is greater than the clock value and is the time length for transition from the previous expression to the next expression in a dynamic expression. The 4D expression model dynamically drives the skin to simulate a sequence of dynamic expressions based on the changes in the vector sequences in the expression commands.

[0089] S1200: Current sources include, but are not limited to, constant current sources, pulse current sources, and modulated pulse power supplies. 4D facial expression models include, but are not limited to, series and parallel programming of two or more excitation coils through a drive module, specifically including, but not limited to, the following S1210 and S1220:

[0090] S1210: Series programming operates using a current source method. When the excitation coils have the same current value, cycle value, and current polarity, they are connected in series through the drive module.

[0091] S1220: Parallel programming uses a constant voltage source mode. When the excitation coils have the same voltage value, cycle value and current polarity, they are connected in parallel through the drive module.

[0092] S1300: The expression commands also include, but are not limited to, extended system temperature to drive the temperature control system included in the 4D dynamic skin, which includes a temperature sensor, heat dissipation fluid, heat dissipation heater and circulation pump. The 4D expression model performs temperature control through the drive module based on the system temperature in the expression commands.

[0093] S1400: Expression commands include, but are not limited to, VEM expression vector commands that support multimodal voice emotions. VEM expression vector commands include, but are not limited to, expression libraries derived from artificial intelligence supervised learning and reinforcement learning.

[0094] 7. The S1000 also includes a design methodology for magnetic dipole models:

[0095] Based on the foregoing solutions, the present invention includes, but is not limited to, one or more combinations of the following:

[0096] S1500: Using a magnetic dipole model, calculate the absolute value F1 of the attraction or repulsion between the magnetic poles of two adjacent parallel electromagnets with an edge spacing of sp and a magnetic dipole radius of γ. Calculate the magnetic moment saturation current I1. Select the elastic force F3 of the skin elastic material, and ensure that F3≤F1. Determine that the excitation current of the excitation coil is ≥I1 to ensure that the attraction or repulsion of the magnetic poles causes the skin to produce rhythmic peristalsis.

[0097] S1600: Based on the magnetic dipole model, calculate the process and result of rhythmic peristalsis of the skin caused by the attraction or repulsion of the magnetic poles when the polarity of the excitation current in the excitation coil of two or more parallel electromagnets changes. The result of the rhythmic peristalsis is that the skin produces point-like expansion and contraction, line-like expansion and contraction, and block-like expansion and contraction to simulate dynamic facial expressions.

[0098] S1700: Based on the magnetic dipole model, a concave-convex electromagnet is used. The current I2 at which the concave-convex electromagnet reaches magnetic moment saturation is calculated. At this time, the magnetic pole of the concave-convex electromagnet generates an attraction or repulsion force F2 on the magnetic pole of the nearby parallel electromagnet located on the back. The tension of the transverse partition is selected to be F4≤F2, and F2≥F1, so that the transverse partition is deformed. Pushing and pulling the parallel electromagnet causes the skin to produce bulges and depressions, resulting in point-like bulges and depressions, line-like bulges and depressions, and block-like bulges and depressions to simulate dynamic facial expressions.

[0099] S1800: The 4D dynamic skin also includes, but is not limited to, subcutaneous tissue containing elastic materials and uniformly mixed soft magnetic material powder, wherein the elasticity of the subcutaneous tissue is F4, and F4≤F1-F3 is selected to ensure that the attraction or repulsion between the magnetic poles is sufficient to counteract the elasticity of the subcutaneous tissue and the epidermis, so that the epidermis produces rhythmic peristalsis.

[0100] 8. Supervised learning methods for S4000

[0101] Based on the foregoing solutions, the present invention also includes, but is not limited to, one or more of the following combinations:

[0102] S4100: Based on the expression command, the parallel electromagnet is driven to produce point-like, line-like, and block-like expansion and contraction on the skin. The concave-convex electromagnet is driven to produce point-like protrusions and depressions, line-like protrusions and depressions, and block-like protrusions and depressions on the skin. The process and result of forming dynamic expressions on the skin are observed by humans. The human judges and establishes a mapping between a set of expression function vectors and a dynamic expression.

[0103] S4200: The training generates a set of mappings consisting of multiple sets of mappings so that when a set of expression function vectors is input into the expression model, a corresponding dynamic expression is generated on the skin.

[0104] 9. Reinforcement Learning Methods for S5000

[0105] Based on the foregoing solutions, the present invention also includes, but is not limited to, one or more of the following combinations:

[0106] S5100: The 4D expression model includes, but is not limited to, mutually inverse and one-to-one corresponding 4D expression inverse models, which obtain a set of expression function vectors based on a segment of the expression model in the mapping set.

[0107] S5200: The 4D facial expression model also includes, but is not limited to, a 4D camera conversion model, which converts a video recording of a physical facial expression taken by a 4D camera into an facial expression model. The 4D camera specifically includes, but is not limited to, front and side shots of 4D dynamic skin, and combines the front and side shots into an facial expression model based on triangulation.

[0108] S5300: The 4D facial expression model also includes, but is not limited to, the facial expression error model, to calculate the error A between two or more facial expression models. The calculation of error A includes, but is not limited to, the block stretching error, line stretching error, and positional error of the 4D dynamic skin. Based on the actual application, the error A is made to ensure that it does not exceed the allowable value.

[0109] S5400: The 4D facial expression model also includes, but is not limited to, a facial expression vector error model to calculate the error B between two or more sets of facial expression function vectors. The calculation of error B includes the 4D errors of x, y, z, t, i, p, v, τ in the facial expression function vector, and the error B is made to ensure that the error B does not exceed the allowable value according to the actual application.

[0110] S5500: Based on the expression error model, the block stretching error, line stretching error and 4D error of the 4D dynamic skin are adjusted through the 4D camera conversion model. When the error is less than the allowable value, the expression model and expression function vector mapping is deemed qualified and the mapping is included in the mapping set. Otherwise, it is deemed unqualified.

[0111] S5600: Based on the expression vector error model, through the 4D camera conversion model, adjust the vectors x, y, z, t and i, p, v, τ respectively, calculate their errors. When the error is less than the allowable value, the expression model and expression function vector mapping at this time are judged to be qualified, and the mapping is included in the mapping set; otherwise, it is judged to be unqualified.

[0112] 10. Other methods:

[0113] Based on the aforementioned solutions, the present invention specifically includes, but is not limited to, one or more combinations of the following S6100 and S6200:

[0114] S6100: The driver module also includes, but is not limited to, the wireless communication submodule. In this case, the 4D expression model also includes, but is not limited to, the management and support of the wireless communication submodule.

[0115] S6200: In addition to the 4D dynamic skin, it includes, but is not limited to, a dedicated APP and QR codes. The dedicated APP communicates with the wireless communication submodule and the 4D expression model by scanning the QR code on the running hardware. The running hardware includes, but is not limited to, computers, mobile phones, remote controls and dedicated hardware. The dedicated APP also includes, but is not limited to, supervised learning and reinforcement learning functions in the 4D expression model, and submits the calculation results to the 4D expression model through the wireless communication submodule to utilize the running hardware to accelerate computing power.

[0116] 11. Purpose and Intent of the Invention

[0117] The purpose and intent of the present invention, which describes the structure and dynamic expression model construction method of VEM-4D dynamic skin, is as follows:

[0118] (1) Invent a 4D dynamic skin that can simulate the dynamic expressions of the skin.

[0119] (2) A method that can generate rhythmic peristalsis and realize dynamic expression module control.

[0120] (3) Support artificial intelligence methods.

[0121] 12. Beneficial effects of the invention

[0122] (1) It has a simple structure, is easy to produce and easy to develop.

[0123] (2) It can be used for the facial skin of artificial intelligence robots and is easy to connect to existing large model systems.

[0124] (3) It can be used for the facial skin of robotic pets and other intelligent creatures, and is easy to produce offline end-to-end products. Attached Figure Description

[0125] List of attached images:

[0126] Figure 1 Schematic diagram of VEM-4D dynamic skin system

[0127] Figure 2 4D Skin Structure Diagram

[0128] Figure 3 Schematic diagram of the driver module structure

[0129] Figure 4 Schematic diagram of parallel electromagnets contracting the skin

[0130] Figure 5 Schematic diagram of parallel electromagnets stretching skin

[0131] Figure 6 4D Skin Expansion and Depression Diagram

[0132] Figure 7 Schematic diagram of static matrix arrangement of 4D electromagnets

[0133] Figure 8 4D diagram of electromagnet block contraction and line contraction

[0134] Figure 9 4D facial expression model illustration

[0135] Detailed description of the attached diagram:

[0136] For detailed descriptions of each figure, please refer to the description of each figure below. It should be emphasized that these figures are merely one illustration of the innovative concept of this invention, and are simply schematic diagrams. Specific dimensions are not given for the structural parts in the figures. Users of this invention can provide detailed design dimensions based on specific applications. Therefore, the figures here are not intended to limit the innovative concept of this invention. Users of this invention can certainly draw other types of figures and interpret and explain them based on generally accepted knowledge in the industry.

[0137] Figure 1 :

[0138] This is a schematic diagram of the VEM-4D dynamic skin system. The structure consists of three parts: a 4D dynamic skin, a driving module, and a 4D expression model. These parts are interconnected. Expression commands are parsed by the 4D expression model and transmitted to the driving module. In the 4D expression model, the uplink connects to the communication interface to obtain expression commands, forming an expression function vector, which is then transmitted to the driving module via the downlink. In the driving module, the expression function vector is decomposed into excitation current, which is output and transmitted to the 4D dynamic skin. The output is connected to the excitation coils of parallel electromagnets and concave-convex electromagnets, driving the 4D dynamic skin according to the 4D expression model to simulate the expressions produced by the skin.

[0139] Additionally, the drive power supply is connected to the drive module to provide power to the current source. Furthermore, users of this invention can also use a constant voltage source instead of a current source to power the excitation coil. The 4D dynamic skin includes parallel electromagnets, and may also include concave-convex electromagnets if needed. The parallel electromagnets generate stretching and contraction deformation on the skin, while the concave-convex electromagnets generate stretching and contraction on the skin through the parallel electromagnets.

[0140] Figure 2 :

[0141] This is a structural diagram of the 4D dynamic skin. It includes: a backplate, a sealing sleeve, an epidermis, subcutaneous tissue, parallel electromagnets, an excitation coil, front magnetic poles, back magnetic poles, transverse septa, transverse septa support pillars, epidermal positioning pillars, concave and convex electromagnets, and heat dissipation fluid inlets and outlets. The subcutaneous tissue is an elastic filler, such as silicone. Furthermore, a heat dissipation device is optional. When heat dissipation is not required, the sealing sleeve is simply made of a non-sealing material, and no heat dissipation fluid inlets or outlets are needed. When heat dissipation is required, the sealing sleeve is simply made of a sealing material, and heat dissipation fluid inlets and outlets are required. It is important to note that the subcutaneous tissue uses foamed silicone material with a high foaming ratio, so that most of the bubbles in the 4D dynamic skin become interconnected spaces, in which heat dissipation fluid is filled. A circulation pump is connected to the heat dissipation fluid inlets and outlets to form a circulating heat dissipation system.

[0142] In addition, the coils of the parallel electromagnets and the concave-convex electromagnets are led to the outside of the sealing sleeve and fixed to the 4D dynamic skin through a cluster connector. The skin is made of an elastic film that can be washed and made up with makeup, and it adheres to the front magnetic pole of the parallel electromagnet. When the front electromagnet moves, it causes the skin to stretch, deform and move.

[0143] The function of the diaphragm support is to maintain the fixed spacing between the diaphragm and the back plate. When a concave-convex electromagnet is energized, it generates an attractive or repulsive force on the back poles of a nearby parallel electromagnet, limiting the deformation of the diaphragm within the range of the diaphragm support. The function of the skin positioning post is to position the skin in three dimensions, preventing it from moving due to the energization of the electromagnet.

[0144] Figure 3 :

[0145] This is a schematic diagram of the drive module. It includes: input interface, processor subsystem, codec, current source matrix, switch matrix, output matrix, and excitation coils for external connections to parallel electromagnet matrix and concave-convex electromagnet matrix. Optional accessories include a wireless communication module, filter circuit, temperature sensor, and connection to a circulating pump.

[0146] In addition, the switch matrix includes on / off switches and polarity reversing switches to control the on / off state and polarity of the excitation current in the excitation coil, causing the electromagnet's poles to generate magnetism and change polarity. A filter circuit is used to improve the electromagnet's operating conditions, reducing vibration and noise. A wireless communication module is used for communication with the outside world.

[0147] At the input of the driver module, the facial expression function vector from the 4D facial expression model and the driving power supply are connected. A temperature sensor and a circulation pump are connected to the processor subsystem within the driver module, forming a temperature control system.

[0148] Figure 4 :

[0149] This is a schematic diagram of a parallel electromagnet retracting skin. It includes: parallel electromagnet 1 and parallel electromagnet 2, each with an excitation coil wound around it. A transverse spacer installed at the bottom of the parallel electromagnet secures the back magnetic poles of the electromagnet (e.g., ...) through a through-hole clip. Figure 2 ), and leave a gap B, the purpose of which is to allow the through hole to latch onto the magnetic pole on the back (such as Figure 2 When the nutation occurs, there is still a gap allowing the parallel electromagnets to rotate around the nutation axis (dashed line in the diagram) and the nutation center. The rotation angle is limited to within the nutation angle. At this time, due to the excitation current in the excitation coil, the N and S magnetic poles on the front side attract each other. The attraction of the front magnetic poles, under the elastic balance of the subcutaneous tissue, causes them to nutationally move together, thus causing the epidermis to contract within the nutation displacement area. On the back side, although opposite poles also attract, they cannot move due to the positional limitation of the transverse septum.

[0150] Figure 5 :

[0151] This is a schematic diagram of parallel electromagnets stretching the skin. Here, the polarity of the front magnetic poles changes due to the excitation current, becoming either N or S poles, causing the skin in that area to be stretched, creating a stretched area.

[0152] Figure 6 :

[0153] This is a schematic diagram of 4D skin stretching and contraction. (See diagram below.) Figure 2 The same annotation, such as Figure 4 , Figure 5 The same principle applies to Figure 6 In the upper part, the excitation currents of the four parallel electromagnets are opposite, resulting in the front magnetic poles being S, N, S, N from top to bottom. This mutual attraction causes the skin at the front magnetic poles to contract, while the surrounding skin stretches, simulating facial expressions. Figure 6 In the lower half, because the back pole of the parallel electromagnet and the adjacent pole of the concave-convex electromagnet are oppositely attracted, the magnetic force causes the transverse septum to bend, pulling the parallel electromagnet to create a depression in the skin, simulating the expression of skin depression (such as dimples on a human face).

[0154] Figure 7 :

[0155] This is a schematic diagram of the static matrix arrangement of 4D electromagnets. This is the front view of the 4D dynamic skin, showing the skin's state in a static state when all electromagnets are not under excitation current. It should be noted that, as needed, skin positioning posts can be placed at required locations on the skin to fix its position and prevent it from moving due to the movement of the electromagnets.

[0156] Figure 8 :

[0157] This is a schematic diagram of the shrinkage of the 4D electromagnet block and the shrinkage of the line. (See diagram below.) Figure 7 As shown, this illustrates the blocky and linear contractions of the skin when the electromagnet is in a certain dynamic state. Similarly, we can use this to simulate skin expressions and design the motion scheme of the electromagnet.

[0158] It should be noted that, in this invention application, from a time perspective, the movement of the electromagnet poles is driven by the expression function vector, and this driving is a temporal rhythm. At the same time, from a spatial perspective, the movement of the electromagnet poles driving the movement of the epidermis is a small-scale peristalsis. Therefore, in this patent application, we define this movement of skin expression simulation as rhythmic peristalsis.

[0159] Figure 9 :

[0160] This is a schematic diagram of a 4D facial expression model. The diagram includes facial expression commands, a computational model, beat decomposition, 3D decomposition, other vector decompositions, beat creep synthesis, facial expression function vectors, 3D camera, 4D camera conversion model, 4D facial expression inverse model, and a mapping set. The blue boxes represent the core mandatory options, while the light red boxes represent optional options. After adding optional options, an information flow is added between the computational model and the mapping set, as shown by the blue bidirectional arc.

[0161] In the computational model, based on facial expression instructions, the facial expression function FE(x, y, z, t, i, p, v, τ) utilizes the 4D time coordinates x, y, z, t from the 4D facial expression model. First, a beat decomposition is performed based on the time coordinate t, then a 3D decomposition is performed based on the spatial coordinates x, y, z, and finally, other vector decompositions are performed based on the other vectors i, p, v, τ, generating the facial expression vector function, which is then sent to the driver module. Here, error A is calculated based on the facial expression error model, and error B is calculated based on the facial expression vector error model. The acceptable values ​​in the facial expression error model and the facial expression vector error model determine the pass / fail status of the facial expression function. Furthermore, when using 4D cameras, the mapping between the facial expression model and the facial expression function vector is determined based on block scaling error and line scaling error, and incorporated into the mapping set. Detailed Implementation

[0162] The objectives and intentions of this invention can be achieved through the following specific embodiments. It should be noted that each specific embodiment has its own specific use and industrial applicability. Therefore, the following embodiments do not encompass all features and steps of this invention, and the accompanying drawings are merely one possible approach and do not constitute a limitation of the invention. The description in the claims is the summary of the invention.

[0163] Specific embodiments of the present invention are illustrated below:

[0164] Innovative method for constructing VEM-4D dynamic skin structure and dynamic expression models

[0165] Diagram Explanation

[0166] This embodiment mainly includes, but is not limited to, the following schematic diagrams, which are described in detail in the accompanying drawings. Figures 1 to 9 .

[0167] Structure and Method Implementation Instructions

[0168] This embodiment mainly includes steps 1 to 10. The first four steps constitute the structure of this application, and the last six steps constitute the method. Each of these ten parts includes several substructures or submethods. This is explained below:

[0169] SXXXX (where XXXX is a number) are identifiers for logical lines in the claims. Unless otherwise specified, the numbers in the identifiers do not indicate the order of logical lines or their relative sizes. "Furthermore" in a logical line indicates that this logical line is a mandatory option following the previous logical line, while "preferably" indicates that the logical line is not mandatory but optional in some applications.

[0170] Unless otherwise specified, these substructures and submethods are not strictly necessary.

[0171] Unless otherwise specified, their order is not required.

[0172] Unless otherwise specified, the selection of materials and design parameters is not mandatory.

[0173] Instead, the patent implementer makes preferred and further selections based on the specific task requirements.

[0174] 1. Overall structure of VEM-4D dynamic skin

[0175] The present invention, as the overall structure of VEM-4D dynamic skin, includes, but is not limited to, the following main structures:

[0176] VEM-4D dynamic skin includes a 4D dynamic skin and driving module and a 4D expression model, wherein:

[0177] The 4D dynamic skin includes, but is not limited to, two or more electromagnets. The electromagnets include, but are not limited to, parallel electromagnets composed of soft magnetic materials and excitation coils. The parallel electromagnets are arranged side by side at a interval of A. The front of the 4D dynamic skin includes, but is not limited to, the skin that is adhered to the magnetic poles on the front of the parallel electromagnets. The skin is composed of an elastic and stretchable membrane.

[0178] The drive module applies an excitation current to the excitation coil, which generates an attractive or repulsive force between the two magnetic poles of the electromagnet, thus driving the skin to expand and contract.

[0179] Preferably, the VEM-4D dynamic skin also includes, but is not limited to, a 4D expression model, which provides static and dynamic expression commands to the driving module to drive the skin's stretching and deformation.

[0180] Reference here Figure 1 .

[0181] Figure 1 This is a schematic diagram of the VEM-4D dynamic skin system. Under normal circumstances, the system structure is divided into three parts: 4D dynamic skin, driving module and 4D expression model, which includes the "preferred" structure.

[0182] It should be noted that, in the simplified mode, the 4D expression model may not be included in the mandatory scheme of this invention. In this case, the function of the 4D expression model is summarized into the upper-level system. That is, when the upper-level system connects to the VEM-4D dynamic skin, it directly uses the expression function vector to communicate with the VEM-4D dynamic skin.

[0183] 2. Transverse partition structure:

[0184] Based on the aforementioned basic solution, the present invention includes, but is not limited to, one or more combinations of the following structures:

[0185] The 4D dynamic skin includes, but is not limited to, a transverse partition with a through-hole buckle, which holds the back magnetic pole of the parallel electromagnet, with a gap B between the through-hole buckle and the back magnetic pole.

[0186] Furthermore, the diaphragm is made of a thin sheet of elastic material, which undergoes vertical elastic deformation when subjected to force.

[0187] Preferably, the electromagnet also includes, but is not limited to, a concave-convex electromagnet located on the back magnetic pole and outside the transverse partition with a gap C. The concave-convex electromagnet includes, but is not limited to, soft magnetic material and excitation coil, and is installed on the back plate of the 4D dynamic skin. Using facial expression commands, the excitation coil of the concave-convex electromagnet is driven by the drive module, so that the magnetic pole of the concave-convex electromagnet generates an attractive or repulsive force on one or more back magnetic poles, driving the skin to produce concave-convex deformation.

[0188] Preferably, the arrangement of the electromagnets and the shape of the 4D dynamic skin are produced according to the actual application, and the spacers and backplate also include, but are not limited to, spacer supports and skin positioning posts.

[0189] Reference here Figure 2 , Figure 4 , Figure 5 , Figure 6 , Figure 7 and Figure 8 It should be noted that there are two optional "preferred" logical lines here.

[0190] The function of the transverse septum is to fix the back magnetic poles of the parallel electromagnet. When attractive or repulsive forces are generated, the surface only undergoes expansion and contraction deformation on the front side, while the back side cannot expand or contract due to the limitation of the transverse septum.

[0191] The function of the spacer A is to separate the parallel electromagnets, so that a gap is maintained between the magnetic poles, and when the magnetic poles generate attraction or repulsion, the front magnetic pole drives the skin to undergo expansion and contraction deformation.

[0192] The purpose of the interval B is to allow the parallel electromagnet to maintain radial nutation when there is attraction or repulsion between the front magnetic poles, rather than jamming the parallel electromagnet.

[0193] The function of the spacer C is to ensure that when the concave-convex electromagnet generates an attractive or repulsive force on the back pole of the parallel electromagnet, it can push and pull the transverse spacer to move laterally, thereby causing the parallel electromagnet to move radially and ultimately achieving the concave-convex deformation of the skin.

[0194] The backplate serves as the central fixation base for 4D dynamic skin, and also forms the basis for the epidermis's stretching, deformation, and unevenness.

[0195] The function of the diaphragm support is to support the gap between the diaphragm and the back plate, so as to isolate the range of attraction or repulsion of the diaphragm by the electromagnet, and limit the bending range of the diaphragm by the electromagnet to the range of the diaphragm support.

[0196] The function of the skin positioning post is to fix the surface position point of the skin. The skin positioning post is fixed between the transverse septum and the skin to prevent the skin at that position point from moving due to the attraction or repulsion of the surrounding magnetic poles.

[0197] Soft magnetic materials can be made of permalloy (iron-nickel alloy), which is a soft magnetic material with high permeability and low coercivity.

[0198] The excitation coil is made of enameled wire, and the wire diameter must be sufficient to withstand the saturation current of the electromagnet.

[0199] The saturation current is the current value in the excitation coil when the soft magnetic material enters magnetic saturation.

[0200] 3. Subcutaneous tissue structure:

[0201] Based on the foregoing solutions, the present invention includes, but is not limited to, one or more combinations of the following:

[0202] 4D dynamic skin also includes, but is not limited to, subcutaneous tissue filled with elastic material around parallel electromagnets. The epidermis adheres to the surface of the subcutaneous tissue on the front of the 4D dynamic skin. At this time, due to the attraction or repulsion of the front magnetic poles and the action of the concave and convex electromagnets, the subcutaneous tissue causes the epidermis to undergo stretching and concave deformation.

[0203] Preferably, the subcutaneous tissue also contains, but is not limited to, powder of a uniformly mixed soft magnetic material.

[0204] Preferably, the subcutaneous tissue also includes, but is not limited to, interconnected cavities filled with heat-dissipating fluid and temperature sensors arranged therein to control the temperature of the epidermis within a set range.

[0205] Preferably, the subcutaneous tissue wraps and seals to form a complete 4D dynamic skin, and provides inlets and outlets for heat dissipation fluid.

[0206] It is important to note that:

[0207] As subcutaneous tissue, there are optional structural combinations that can be used or not. In the structure without subcutaneous tissue, the epidermis is directly adhered to the front magnetic poles of parallel electromagnets. In this case, the epidermis uses an elastic material to balance the attractive and repulsive forces of the two front electromagnets. When using subcutaneous tissue, a heat dissipation fluid can be used for heat dissipation or to ensure a constant epidermal temperature. A circulating heat dissipation fluid option is also available, in which case the inlet and outlet of the heat dissipation fluid need to be set on the 4D dynamic skin.

[0208] Specifically, the subcutaneous tissue uses silicone of mixed soft magnetic composite materials with different volume ratios to enhance the magnetic permeability between parallel electromagnets. The soft magnetic materials include, but are not limited to, carbonyl iron powder or iron-silicon-aluminum powder, and the volume filling rate with silicone is about 10% to 90%.

[0209] like Figure 2 and Figure 6 As shown.

[0210] 4. Structure of the driving module and 4D facial expression model:

[0211] Based on the aforementioned solution, the present invention includes, but is not limited to, one or more of the following combinations in the driving module and VEM-4D facial expression model:

[0212] The driving module includes, but is not limited to, an input terminal, a codec, a current source, and an output terminal. The output terminal is connected to the two ends of the excitation coil of each electromagnet. The current source provides the excitation current. The codec provides decoding based on the encoding of the electromagnet. The input terminal is connected to the 4D facial expression model and the driving power supply. The 4D facial expression model drives the electromagnets through the driving module to generate stretching and concave facial expressions on the skin.

[0213] Preferably, the current source includes, but is not limited to, a constant current source, a pulse current source, and a modulated pulse power supply.

[0214] Preferably, the driving module also includes, but is not limited to, connecting one or more temperature sensors, which are placed at a designated location in the subcutaneous tissue and transmit temperature signals back to the 4D facial expression model to monitor the temperature of the subcutaneous tissue.

[0215] Preferably, the drive module also includes, but is not limited to, the connection terminal of the circulation pump that drives the heat dissipation fluid, and completes the temperature control and circulation of the heat dissipation fluid under the control of the 4D facial expression model.

[0216] Preferably, the driving module also includes, but is not limited to, a filtering circuit to reduce or eliminate the excessive oscillation of the magnetic field in the arrangement of electromagnets caused by the high-frequency pulses of the 4D facial expression model.

[0217] Preferably, the driving module also includes, but is not limited to, providing a demagnetizing current to the electromagnet under the control of the 4D facial expression model.

[0218] Preferably, the driving module also includes, but is not limited to, a wireless communication submodule, which communicates with the 4D dynamic skin under the control of the 4D expression model.

[0219] Figure 3 This is a schematic diagram of the driver module structure.

[0220] It is important to note that:

[0221] When it is necessary to reduce the excitation current, the current source can use pulse width modulation (PWM) or variable pulse width modulation (VPWM). For example, at the beginning of the current source, the pulse width can be increased to overcome the distance between the magnetic poles at the beginning of driving the magnetic poles. After the distance between the magnetic poles becomes smaller, a smaller excitation current is used to maintain the smaller distance between the magnetic poles.

[0222] The "preferred" logical lines here are all optional structures, and users of this invention can choose to execute them according to their respective applications. In this embodiment, all optional options are selected.

[0223] 5. Overall Construction Method of VEM-4D Dynamic Expression Model

[0224] This invention, as a method for constructing VEM-4D dynamic facial expression models, is characterized by including but not limited to one or more combinations of the following:

[0225] S1000: The VEM-4D dynamic expression model includes a 4D expression model, a 4D dynamic skin, and a driving module. Expression commands are loaded at the input end of the 4D expression model, and the expression function vector generated at the output end provides excitation current to the excitation coil of the electromagnet in the 4D dynamic skin through the driving module and controls it. The attraction or repulsion between the magnetic poles of two or more electromagnets drives the epidermis on the front of the 4D dynamic skin to produce rhythmic undulation to simulate dynamic expressions.

[0226] S2000: Further, the expression commands include, but are not limited to, stretch commands, which act on the parallel electromagnets arranged side by side included in the electromagnet, and include the stretch expressions in the epidermal simulation of dynamic expressions.

[0227] S3000: Preferably, the expression command includes, but is not limited to, the concave-convex command, which acts on the concave-convex electromagnet located on the back of the 4D dynamic skin, including but not limited to, the electromagnet, through the elastic diaphragm, to drive the parallel electromagnet at the magnetic pole located on the back of the 4D dynamic skin, and simulates the concave-convex expression in the epidermis.

[0228] Figure 9 This is a schematic diagram of a 4D facial expression model. It also represents the overall methodological structure of the VEM-4D dynamic facial expression model of this invention.

[0229] It is important to note that:

[0230] The logical row marker S3000 is optional, meaning that in some applications, the convex / concave electromagnet and convex / concave instructions may not be required.

[0231] 6. Design methods for facial expression commands and rhythmic undulations:

[0232] Based on the foregoing solutions, the present invention includes, but is not limited to, one or more combinations of the following:

[0233] S1100: Expression commands include, but are not limited to, the expression function FE (x, y, z, t, i, p, v, τ), where x, y, z, and t in the expression function vector (x, y, z, t, i, p, v, τ) are the coordinates of the electromagnet in 3D space and the clock value coordinates in the 4th dimension, respectively. i, p, v, and τ are the current value, current polarity, type of electromagnet, and step time, respectively, which are the current value, current polarity, type of electromagnet, and step time. The step time is greater than the clock value and is the time length for transition from the previous expression to the next expression in a dynamic expression. The 4D expression model dynamically drives the skin to simulate a sequence of dynamic expressions based on the changes in the vector sequences in the expression commands.

[0234] S1200: Further, the current source includes, but is not limited to, a constant current source, a pulse current source, and a modulated pulse power supply. The 4D expression model includes, but is not limited to, series programming and parallel programming of two or more excitation coils through a drive module, specifically including, but not limited to, the following S1210 and S1220:

[0235] S1210: Series programming operates using a current source method. When the excitation coils have the same current value, cycle value, and current polarity, they are connected in series through the drive module.

[0236] S1220: Preferably, the parallel programming adopts a constant voltage source mode. When they have the same voltage value, cycle value and current polarity, these excitation coils are connected in parallel through the drive module.

[0237] S1300: Preferably, the expression command also includes, but is not limited to, an extended system temperature to drive the temperature control system included in the 4D dynamic skin, including a temperature sensor, heat dissipation fluid, heat dissipation heater and circulation pump. The 4D expression model performs temperature control through the drive module based on the system temperature in the expression command.

[0238] S1400: Preferably, the expression commands include, but are not limited to, VEM expression vector commands that support multimodal speech emotions. The VEM expression vector commands include, but are not limited to, expression libraries derived from artificial intelligence supervised learning and reinforcement learning.

[0239] rhythmic peristalsis

[0240] like Figure 9 In this context, facial expression commands are composed of facial expression functions. Specifically, the facial expression function FE(x, y, z, t, i, p, v, τ) determines the state of any point on the skin using the facial expression function vector (x, y, z, t, i, p, v, τ). It's important to note that the 4D referred to in this invention refers to the 3D coordinates (x, y, z) and 1D time value (t) of this state, totaling a 4D spatial state, i.e., a 4D facial expression state. At the same 3D spatial point on the skin, the state of this point can continuously change over time. Furthermore, this 1D time change is measured according to the shortest time—the beat—therefore, the change at this point on the skin—the creeping—also creeps according to the beat, resulting in a beat-like creeping process. The beat-like creeping method is determined according to (i, p, v, τ).

[0241] It should be noted that the 3D point coordinates (x, y, z) on the epidermis referred to in this invention are supported by electromagnets beneath the epidermis. The expression function vector is generated by the rhythmic peristalsis produced by the attraction or repulsion of the electromagnet at that point.

[0242] Regarding preferred options

[0243] Logic row markers S1100 and S1200 are mandatory, while S1210 and S1220 are either one of two or both can be selected on different electromagnets. Logic rows S1300 and S1400 are options for different applications.

[0244] In this embodiment, all of the above logical lines are used.

[0245] It needs to be emphasized that:

[0246] Regarding temperature control, the following two points should be noted:

[0247] (1) Skin temperature issue. In some applications, such as simulating human skin, in order to make the simulation more similar to human skin, it is necessary to control the skin temperature within the same temperature range as human skin, such as 33°C to 38°C. In this case, it is necessary to select the option of the temperature control system.

[0248] (2) Skin heat dissipation problem. If the distance between the magnetic poles of the electromagnet is large, according to the magnetic dipole model, the excitation current may be large enough to allow the electromagnet to enter the magnetic saturation region. Since the large current will cause heat generation, it may exceed the upper temperature limit of the skin when simulating skin. In this case, it is also necessary to select the option of temperature control system.

[0249] 7. Design method of magnetic dipole model:

[0250] Based on the foregoing solutions, the present invention includes, but is not limited to, one or more combinations of the following:

[0251] S1500: Using a magnetic dipole model, calculate the absolute value F1 of the attraction or repulsion between the magnetic poles of two adjacent parallel electromagnets with an edge spacing of sp and a magnetic dipole radius of γ. Calculate the magnetic moment saturation current I1. Select the elastic force F3 of the skin elastic material, and ensure that F3≤F1. Determine that the excitation current of the excitation coil is ≥I1 to ensure that the attraction or repulsion of the magnetic poles causes the skin to produce rhythmic peristalsis.

[0252] S1600: Further, based on the magnetic dipole model, the process and result of the skin undergoing rhythmic peristalsis caused by the attraction or repulsion of the magnetic poles when the polarity of the excitation current changes in the excitation coils of two or more parallel electromagnets. The result of the rhythmic peristalsis is that the skin undergoes point-like expansion and contraction, line-like expansion and contraction, and block-like expansion and contraction to simulate dynamic facial expressions.

[0253] S1700: Preferably, based on the magnetic dipole model, a concave-convex electromagnet is used, and the current I2 at which the concave-convex electromagnet reaches magnetic moment saturation is calculated. At this time, the magnetic pole of the concave-convex electromagnet generates an absolute value F2 of attraction or repulsion force on the magnetic pole of the nearby parallel electromagnet located on the back. The tension of the transverse partition is selected to be F4≤F2, and F2≥F1, so that the transverse partition is deformed. Pushing and pulling the parallel electromagnet causes the skin to produce bulges and depressions, resulting in point-like bulges and depressions, line-like bulges and depressions, and block-like bulges and depressions to simulate dynamic expressions.

[0254] S1800: Preferably, the 4D dynamic skin also includes, but is not limited to, subcutaneous tissue of elastic material and uniformly mixed soft magnetic material powder, wherein the elasticity of the subcutaneous tissue is F4, and F4≤F1-F3 is selected to ensure that the attraction or repulsion between the magnetic poles is sufficient to counteract the elasticity of the subcutaneous tissue and the epidermis, so that the epidermis produces rhythmic peristalsis.

[0255] Regarding preferred options

[0256] The logical line markers S1500 and S1600 are required. These are the calculation methods for magnetic dipoles in "Electromagnetism" and the methods for parallel electromagnets to drive the skin to produce point-like, line-like, and block-like expansion and contraction.

[0257] The logical row markers S1700 and S1800 are optional. These are methods for creating uneven stretching of the epidermis by using a concave-convex electromagnet, and for generating rhythmic peristalsis by balancing the elasticity of the subcutaneous tissue.

[0258] It should be noted that, in order to increase the magnetic permeability and actual effect of subcutaneous tissue, the elastic material can be a colloid such as silicone, into which more than 20% of soft magnetic material powder, such as carbonyl iron powder or permalloy powder, can be mixed in by volume.

[0259] In this embodiment, all of the above logical lines are used.

[0260] 8. Methods of Supervised Learning

[0261] Based on the foregoing solutions, the present invention also includes, but is not limited to, one or more of the following combinations:

[0262] S4100: Based on the expression command, the parallel electromagnet is driven to produce point-like, line-like, and block-like expansion and contraction on the skin. The concave-convex electromagnet is driven to produce point-like protrusions and depressions, line-like protrusions and depressions, and block-like protrusions and depressions on the skin. The process and result of forming dynamic expressions on the skin are observed by humans. The human judges and establishes a mapping between a set of expression function vectors and a dynamic expression.

[0263] S4200: Further, a mapping set consisting of multiple mappings is generated during training so that when a set of expression function vectors is input into the expression model, a corresponding dynamic expression is generated on the skin.

[0264] Supervised learning in artificial intelligence is employed to train a one-to-one mapping between dynamic facial expressions and facial expression function vectors during human evaluation. Examples include the mapping between blinking and facial expression function vectors; the mapping between the pronunciation of vowels (ah, ee, um, uh, oh) and the facial expression function vectors of mouth shapes; the mapping between the pronunciation of consonants and the facial expression function vectors of mouth shapes; the mapping between emotional pronunciation and the facial expression function vectors of mouth shapes, and so on. The training results are incorporated into a mapping set, becoming the basic elements of the mapping set.

[0265] It needs to be emphasized that:

[0266] This mapping relationship is rhythmic, meaning it produces a rhythmic undulation, a dynamic process, rather than just a static image.

[0267] In the initial stage of supervised learning training, it is recommended that human experts collect typical dynamic expressions, and gradually establish a basic mapping set through manual observation and adjustment of expression parameters (expression function vectors).

[0268] The supervised learning method can be omitted in later 4D dynamic skin products. A wireless communication submodule can be used to connect the 4D dynamic skin product to a network that stores the mapping set, or store it locally in the 4D dynamic skin product.

[0269] 9. Reinforcement learning methods

[0270] Based on the foregoing solutions, the present invention also includes, but is not limited to, one or more of the following combinations:

[0271] S5100: The 4D expression model includes, but is not limited to, mutually inverse and one-to-one corresponding 4D expression inverse models, which obtain a set of expression function vectors based on a segment of the expression model in the mapping set.

[0272] S5200: Further, the 4D expression model also includes, but is not limited to, a 4D camera conversion model, which converts a video of a physical expression captured by a 4D camera into an expression model. Specifically, the 4D camera includes, but is not limited to, front and side shots of 4D dynamic skin, and the front and side shots are combined into an expression model based on triangulation.

[0273] S5300: Further, the 4D expression model also includes, but is not limited to, an expression error model to calculate the error A between two or more expression models, wherein the calculation of error A includes, but is not limited to, the block stretching error, line stretching error and position error of the 4D dynamic skin, and according to the actual application, the error A does not exceed the allowable value.

[0274] S5400: Further, the 4D expression model also includes, but is not limited to, an expression vector error model to calculate the error B between two or more sets of expression function vectors, wherein the calculation of error B includes the 4D errors of x, y, z, t, i, p, v, τ in the expression function vector, and according to the actual application, the error B does not exceed the allowable value.

[0275] S5500: Further, based on the expression error model, the block stretching error, line stretching error and 4D error of the 4D dynamic skin are adjusted respectively through the 4D camera conversion model. When the error is less than the allowable value, the expression model and expression function vector mapping at this time are judged to be qualified and the mapping is included in the mapping set; otherwise, it is judged to be unqualified.

[0276] S5600: Preferably, based on the expression vector error model, the vectors x, y, z, t and i, p, v, τ are adjusted respectively through the 4D camera conversion model, and their errors are calculated. When the error is less than the allowable value, the expression model and expression function vector mapping at this time are judged to be qualified, and the mapping is included in the mapping set; otherwise, it is judged to be unqualified.

[0277] Regarding preferred options

[0278] Logical row markers S5100 to S5500 are mandatory, and logical row S5600 is optional. However, in this embodiment, logical row S5600 is also mandatory.

[0279] We employ reinforcement learning in artificial intelligence, using block scaling error, line scaling error, and 4D error of 4D dynamic skin as trial-and-error criteria for reinforcement learning. We then incorporate qualified expression models and expression function vector mapping results into the mapping set.

[0280] 10. Other methods

[0281] Based on the foregoing solutions, the present invention specifically includes, but is not limited to, one or more of the following combinations:

[0282] S6100: The driver module also includes, but is not limited to, the wireless communication submodule. In this case, the 4D expression model also includes, but is not limited to, the management and support of the wireless communication submodule.

[0283] S6200: Further, in addition to the 4D dynamic skin, there are including but not limited to a dedicated APP and a QR code. The dedicated APP communicates with the wireless communication submodule and the 4D expression model by scanning the QR code on the running hardware. The running hardware includes but is not limited to a computer, mobile phone, remote control and dedicated hardware. The dedicated APP also includes but is not limited to the supervised learning and reinforcement learning functions in the 4D expression model, and submits the calculation results to the 4D expression model through the wireless communication submodule to utilize the running hardware to accelerate computing power.

[0284] In this embodiment, a wireless communication submodule and a specially designed dedicated APP are used. Based on the application type of the 4D dynamic skin, a QR code is set for the 4D dynamic skin. At the same time, a cloud mode is used to store the mapping set, and the wireless communication submodule is connected to the mapping set to work.

Claims

1. The structure of VEM-4D dynamic skin, characterized in that, include: VEM-4D dynamic skin includes a 4D dynamic skin and a driving module, among which, The 4D dynamic skin includes two or more electromagnets. The electromagnets are parallel electromagnets composed of soft magnetic materials and excitation coils. The parallel electromagnets are arranged side by side at a distance A. The front of the 4D dynamic skin includes a skin that is attached to the magnetic poles on the front of the parallel electromagnets. The skin is made of an elastic and stretchable membrane. The drive module applies an excitation current to the excitation coil, which generates an attractive or repulsive force between the two magnetic poles of the electromagnet, thus driving the skin to produce expansion and contraction deformation. VEM-4D dynamic skin also includes a 4D expression model, which provides static and dynamic expression commands to the driving module, driving the skin to stretch and deform.

2. The structure according to claim 1, characterized in that, It also includes the following structures: The 4D dynamic skin includes a transverse septum with through-hole buckles, which holds the back magnetic poles of the parallel electromagnets, with a gap B between the through-hole buckles and the back magnetic poles; The transverse diaphragm is made of a thin sheet of elastic material, which undergoes vertical elastic deformation when subjected to force; and / or, The electromagnet also includes a concave-convex electromagnet located between the back magnetic pole and the outer side of the transverse partition, with a gap C. The concave-convex electromagnet includes a soft magnetic material and an excitation coil and is mounted on the back plate of the 4D dynamic skin. Using facial expression commands, the excitation coil of the concave-convex electromagnet is driven by a drive module, causing the magnetic poles of the concave-convex electromagnet to generate an attractive or repulsive force on one or more back magnetic poles, driving the skin to produce concave-convex deformation; and / or, The arrangement of electromagnets and the shape of the 4D dynamic skin are produced according to actual applications. The spacers and backplate also include spacer supports and skin positioning posts.

3. The structure according to claim 2, characterized in that, It also includes structure: 4D dynamic skin also includes subcutaneous tissue filled with elastic material surrounding parallel electromagnets. The epidermis adheres to the surface of the subcutaneous tissue on the front side of the 4D dynamic skin. Due to the attraction or repulsion of the front magnetic poles and the action of the concave and convex electromagnets, the subcutaneous tissue causes stretching and concave deformation of the epidermis; and / or, The subcutaneous tissue also includes powder of a uniformly mixed soft magnetic material; and / or, The subcutaneous tissue also includes interconnected cavities filled with heat-dissipating fluid and temperature sensors to control the temperature of the epidermis within a set range. Subcutaneous tissue encapsulates and seals to form a complete 4D dynamic skin, and provides inlets and outlets for heat dissipation fluid.

4. The structure according to claim 3, characterized in that, It also includes the following structures: The drive module includes an input terminal, a codec, a current source, and an output terminal. The output terminal connects to the two ends of the excitation coils of each electromagnet. The current source provides the excitation current, and the codec provides decoding based on the electromagnet's encoding. The input terminal connects to the 4D facial expression model and the drive power supply. The 4D facial expression model uses the drive module to drive the electromagnets to generate stretching and concave facial expressions on the skin; and / or, Current sources include constant current sources, pulse current sources, and modulated pulse power supplies; And / or, The driving module also includes connecting one or more temperature sensors, which are placed at designated locations in the subcutaneous tissue to transmit temperature signals back to the 4D facial expression model to monitor the temperature of the subcutaneous tissue. The drive module also includes a connection terminal for a circulation pump that drives the cooling fluid, which, under the control of the 4D facial expression model, performs temperature control and circulation of the cooling fluid; and / or, The drive module also includes a filter circuit to reduce or eliminate the excessive oscillation of the magnetic field in the arrangement of electromagnets caused by the high-frequency pulses of the 4D facial expression model. The drive module also includes a demagnetizing current that provides demagnetization to the electromagnet under the control of the 4D facial expression model; and / or, The driver module also includes a wireless communication submodule, which communicates with the 4D dynamic skin under the control of the 4D expression model.

5. A method for constructing VEM-4D dynamic facial expression models, characterized in that, include: S1000: The VEM-4D dynamic expression model includes a 4D expression model, a 4D dynamic skin, and a driving module. Expression commands are loaded at the input end of the 4D expression model, and the expression function vector generated at the output end provides excitation current to the excitation coil of the electromagnet in the 4D dynamic skin through the driving module and controls it. The attraction or repulsion between the magnetic poles of two or more electromagnets drives the epidermis on the front of the 4D dynamic skin to produce rhythmic undulation to simulate dynamic expressions. S2000: Expression commands include stretching commands, acting on parallel electromagnets arranged side-by-side within the electromagnet system, to simulate stretching expressions on the skin; and / or, S3000: Expression commands include convex / concave commands, which act on the convex / concave electromagnets located on the back of the 4D dynamic skin, including electromagnets. The parallel electromagnets are driven by elastic diaphragms to the magnetic poles on the back of the 4D dynamic skin, simulating the convex / concave expressions included in the dynamic expressions on the epidermis.

6. The method according to claim 5, characterized in that, S1000 specifically includes the following methods: S1100: The expression command includes the expression function FE(x, y, z, t, i, p, v, τ), where x, y, z, and t in the expression function vector (x, y, z, t, i, p, v, τ) are the coordinates of the electromagnet in 3D space and the clock value coordinates in the 4th dimension, respectively. i, p, v, and τ are the current value, current polarity, type of electromagnet, and step time, respectively, which are the current value, current polarity, type of electromagnet, and step time. The step time is greater than the clock value and is the time length for transition from the previous expression to the next expression in the dynamic expression. The 4D expression model dynamically drives the skin to simulate a sequence of dynamic expressions based on the changes of each set of vector sequences in the expression command. S1200: The current source includes a constant current source, a pulse current source, and a modulated pulse power supply. The 4D facial expression model includes series and parallel programming of two or more excitation coils through the drive module, specifically including the following S1210 and S1220: S1210: Series programming operates using a current source method. When excitation coils have the same current value, cycle value, and current polarity, they are connected in series via the drive module; and / or, S1220: Parallel programming operates using a constant voltage source. When excitation coils have the same voltage, cycle time, and current polarity, they are connected in parallel via the drive module; and / or, S1300: The expression commands also include an extended system temperature to drive the temperature control system included in the 4D dynamic skin, which includes a temperature sensor, heat dissipation fluid, heat dissipation heater, and circulation pump. The 4D expression model performs temperature control via the drive module based on the system temperature in the expression commands; and / or, S1400: The facial expression commands include VEM facial expression vector commands that support multimodal voice emotion. The VEM facial expression vector commands are derived from an facial expression library that has undergone artificial intelligence supervised learning and reinforcement learning.

7. The method according to claim 6, characterized in that, S1000 also includes the following methods: S1500: Using a magnetic dipole model, calculate the absolute value F1 of the attraction or repulsion between the magnetic poles of two adjacent parallel electromagnets with an edge spacing of sp and a magnetic dipole radius of γ, calculate the magnetic moment saturation current I1, select the elastic force F3 of the skin elastic material, and ensure that F3≤F1, determine that the excitation current of the excitation coil ≥I1, and ensure that the attraction or repulsion of the magnetic poles causes the skin to produce rhythmic peristalsis; S1600: Based on the magnetic dipole model, calculate the process and result of rhythmic peristalsis caused by the attraction or repulsion of the magnetic poles in two or more parallel electromagnets when the polarity of the excitation current changes in their respective excitation coils. The result of this rhythmic peristalsis is point-like, line-like, and block-like expansion and contraction of the skin, simulating dynamic facial expressions; and / or, S1700: Based on the magnetic dipole model, using a concave-convex electromagnet, calculate the current I2 at which the concave-convex electromagnet reaches magnetic moment saturation. At this point, the magnetic poles of the concave-convex electromagnet generate an attractive or repulsive force F2 on the magnetic poles of the nearby parallel electromagnets located on the back side. The tension of the transverse partition is F4. We choose F4≤F2 and F2≥F1, causing the transverse partition to deform. Pushing and pulling the parallel electromagnets causes bulges and depressions on the surface, resulting in point-like, linear, and block-like bulges and depressions to simulate dynamic facial expressions; and / or, S1800: The 4D dynamic skin also includes subcutaneous tissue of elastic material and / or uniformly mixed soft magnetic material powder, wherein the elastic force of the subcutaneous tissue is less than or equal to the difference between the absolute value of the attraction or repulsion between the magnetic poles of the parallel electromagnets and the elastic force of the epidermal elastic material, so as to ensure that the attraction or repulsion between the magnetic poles is sufficient to counteract the elastic force of the subcutaneous tissue and the epidermis, so that the epidermis generates rhythmic peristalsis.

8. The method according to claim 7, characterized in that, It also includes supervised learning methods for S4000, specifically including: S4100: Drives parallel electromagnets to produce point-like, line-like, and block-like expansion and contraction on the skin according to facial expression commands, and drives concave and convex electromagnets to produce point-like protrusions and depressions, line-like protrusions and depressions, and block-like protrusions and depressions on the skin. Through manual observation of the process and results of forming dynamic facial expressions on the skin, manual evaluation and establishment of a mapping between a set of facial expression function vectors and a dynamic facial expression. S4200: The training generates a set of mappings consisting of multiple sets of mappings so that when a set of expression function vectors is input into the expression model, a corresponding dynamic expression is generated on the skin.

9. The method according to claim 8, characterized in that, It also includes reinforcement learning methods for the S5000, specifically including: S5100: The 4D expression model includes mutually inverse and one-to-one corresponding 4D expression inverse models, and obtains a set of expression function vectors based on a segment of the expression model in the mapping set; S5200: The 4D expression model also includes a 4D camera conversion model, which converts a video of a physical expression captured by a 4D camera into an expression model. Specifically, the 4D camera includes front and side shots of 4D dynamic skin, and the front and side shots are combined into an expression model based on triangulation. S5300: The 4D facial expression model also includes an facial expression error model to calculate the error A between two or more facial expression models. The calculation of error A includes the block stretching error, line stretching error, and 4D error of the 4D dynamic skin. Based on the actual application, the error A is made to ensure that it does not exceed the allowable value. S5400: The 4D expression model also includes an expression vector error model to calculate the error B between two or more sets of expression function vectors. The calculation of error B includes the 4D errors of x, y, z, t in the expression function vector, and the errors of i, p, v, τ, and, depending on the actual application, the error B does not exceed the allowable value. S5500: Based on the expression error model, and through the 4D camera conversion model, the block scaling error, line scaling error, and 4D error of the 4D dynamic skin are adjusted respectively. When the error is less than the allowable value, the expression model and expression function vector mapping are deemed qualified, and the mapping is included in the mapping set; otherwise, it is deemed unqualified; and / or, S5600: Based on the expression vector error model, through the 4D camera conversion model, adjust the vectors x, y, z, t and i, p, v, τ respectively, calculate their errors. When the error is less than the allowable value, the expression model and expression function vector mapping at this time are judged to be qualified, and the mapping is included in the mapping set; otherwise, it is judged to be unqualified.

10. The method according to claim 7, 8, or 9, characterized in that, It also includes the following methods of S6100 and S6200: S6100: The driver module also includes a wireless communication submodule. At this time, the 4D expression model also includes the management and support of the wireless communication submodule. S6200: In addition to the 4D dynamic skin, it includes a dedicated APP and a QR code. The dedicated APP communicates with the wireless communication submodule and the 4D expression model by scanning the QR code on the running hardware. The running hardware includes a computer, mobile phone, remote control and dedicated hardware. The dedicated APP also includes supervised learning and reinforcement learning functions in the 4D expression model, and submits the calculation results to the 4D expression model through the wireless communication submodule to utilize the running hardware to accelerate computing power.

Images