Personalized digital human generation method, system and terminal

Abstract

The invention relates to a personalized digital human generation method, system and terminal, and relates to the field of digital human technology, and the method comprises the steps: controlling a preset selection device to obtain selection information; judging whether the digital person is used by an individual or not according to the selection information; when the digital person is used by a person, acquiring personnel information; determining personnel occupations and personnel genders from a preset personnel information analysis model according to the personnel information; determining personnel wearing according to personnel occupations and personnel genders; matching personal wearing parameters of the digital person from a preset wearing matching model according to the person wearing, and obtaining a person image; determining a personnel appearance according to the personnel image; matching personal five-sense-organ parameters of the digital person from a preset five-sense-organ recognition model according to the appearance of the person; and matching a reference digital person from a preset digital person generation model according to the gender of the person, the personal wearing parameters and the personal five-sense-organ parameters. The method and the device have the effects of improving the use convenience of the digital human and simplifying the generation steps of the digital human.

Description

Technical Field

[0001] The present invention relates to the field of digital human technology, and in particular to a personalized digital human generation method, system and terminal. Background Art

[0002] Digital human refers to a comprehensive product that exists in the non-physical world, is created and used by computer means, and has multiple human characteristics (appearance characteristics, human performance ability, interaction ability, etc.).

[0003] In the existing technology, digital human technology is often used in the live broadcast industry. Digital human technology overcomes problems such as physical fatigue and emotional fluctuations of real anchors, maintains a stable performance state, reduces labor costs, and creates more realistic and amazing visual effects.

[0004] The steps for generating digital humans are rather complicated. When non-professionals need to generate digital humans, they may not be able to obtain the required digital human image, which makes it impossible for non-professionals to use digital human technology. Summary of the invention

[0005] In order to improve the convenience of using digital humans and simplify the steps of generating digital humans, the present invention provides a method, system and terminal for generating personalized digital humans.

[0006] In a first aspect, the present invention provides a method for generating a personalized digital human, which adopts the following technical solution: A method for generating a personalized digital human, comprising: Controlling a preset selection device to obtain selection information; Determine whether the digital person is for personal use based on the selected information; When a digital person is used for personal purposes, obtain personal information; Determine the occupation and gender of the personnel from the preset personnel information analysis model according to the personnel information; Determine what personnel should wear based on their occupation and gender; According to the wear of the person, the personal wear parameters of the digital person are matched from the preset wear matching model, and the image of the person is obtained; Determine the appearance of a person based on his / her image; According to the person's appearance, the digital person's personal facial features parameters are matched from the preset facial features recognition model; A reference digital human is matched from the preset digital human generation model according to the person's gender, personal wearing parameters and personal facial features parameters.

[0007] By adopting the above technical solution, when a user creates a personal digital human, the facial features of the digital human are determined from the user's appearance obtained from the image, and the digital human's clothing is determined from the user's occupation, thereby obtaining a digital human that conforms to the user's actual situation, simplifying the digital human generation steps, and improving the convenience of using the digital human.

[0008] Optionally, a group digital human generation method is also included, and the group digital human generation method includes: When the digital person is not used by an individual, determine whether the digital person is used by a group based on the selected information; When the digital person is used by a group, obtain group information; Determine the group name from a preset group name analysis model according to the group information; Determine the group domain based on the group name; Determine group attire based on group area and gender of participants; According to the group wear, the group wear parameters of the digital human are matched from the preset wear matching model; A benchmark digital human is matched from the preset digital human generation model based on the gender of the person, group wearing parameters and individual facial features parameters.

[0009] By adopting the above technical solution, when a user creates a digital person representing a group, the digital person's clothing is determined based on the group's business direction, so that the digital person's image is close to the group's customers, thereby increasing the group's existing and potential customers' intimacy with the digital person.

[0010] Optionally, the group digital human generation method further includes: When the digital person is used for a group, the male-female ratio is matched from the preset group composition analysis model based on the group information; When the male-to-female ratio is greater than the preset male selection value, male is defined as the group gender; When the male-female ratio is not greater than the preset male selection value, female is defined as the group gender; Determine group attire based on group domain and group gender, and determine group logo from a preset group logo recognition model based on group information; According to the group wear and the group logo, group wear parameters are matched from a preset individual group wear matching model; When the gender of the individual is consistent with the gender of the group, a reference digital human is matched from the preset digital human generation model according to the group gender, group clothing parameters and individual facial features parameters; When the gender of the individual is not consistent with the gender of the group, the baseline facial features parameters are determined according to the gender of the group; A benchmark digital human is matched from the preset digital human generation model according to the group gender, group clothing parameters and benchmark facial features parameters.

[0011] By adopting the above technical solution, when a user creates a digital person representing a group, the gender of the digital person is determined based on the male-female ratio of the group members, so that the image of the digital person reflects the composition of the group members.

[0012] Optionally, the group digital human generation method further includes: When the digital person is used by a group, the hot-selling products are matched from the preset sales database according to the group name; Match the applicable population from the preset product database based on the hot-selling products; Determine the applicable body type and age group based on the applicable population; Matching voice adjustment parameters from a preset voice parameter database according to the gender and applicable age group of the group; Matching posture adjustment parameters from a preset posture adjustment matching model according to applicable age groups and applicable body types; An adjusted digital human is matched from a preset digital human adjustment model according to the voice adjustment parameters, the body adjustment parameters and the reference digital human.

[0013] By adopting the above technical solution, when a user creates a digital person representing a group, the age and body shape of the digital person are determined based on the age group and body shape of the target customers of the group's main products, so that the image of the digital person is closer to the group's customers, thereby increasing the intimacy of the digital person among the group's existing and potential customers in the audience.

[0014] Optionally, a sales display adjustment method is also included, and the sales display adjustment method includes: Obtain product images of products to be sold; Match the main color from the preset main color recognition model according to the product image; Match the main contrasting color from a preset contrasting color matching model according to the main color; Matching hand wear from a preset hand wear matching model according to the main contrasting color and the wear of the person; Matching hand adjustment parameters from a preset hand adjustment model according to the hand wear; Matching and adjusting wearing parameters from a preset wearing parameter adjustment model according to the main contrasting colors and the personal wearing parameters; According to the adjustment of wearing parameters, hand adjustment parameters and the reference digital human, a display digital human is matched from a preset digital human adjustment model.

[0015] By adopting the above technical solution, the color of the digital human's clothing is adjusted according to the color of the product, thereby improving the color contrast between the product and the digital human, making the digital human more obvious when displaying the product, and reducing the situation where the digital human's appearance color is too close to the product's color, resulting in unclear display of product details.

[0016] Optionally, the sales display adjustment method further includes: Match the main proportion from the preset color proportion matching model based on the main color; When the primary proportion is lower than a preset representative threshold, a secondary color is matched from a preset secondary color recognition model according to the product image; Matching a secondary contrasting color from a preset contrasting color matching model according to the secondary color; Hand wear is matched from a preset hand wear matching model according to the secondary contrasting color, the primary contrasting color and the person's wear.

[0017] By adopting the above technical solution, when the product has many colors, the color variety of the gloves worn by the digital person is increased according to the secondary color of the product, so that the colors of the gloves worn by the digital person are distributed at intervals to improve the contrast between the gloves and the product.

[0018] Optionally, a sound adjustment method is also included, and the sound adjustment method includes: When the digital person is used by an individual, the person's voice parameters are obtained; extracting acoustic parameters based on speech parameters; Matching personal voice parameters from a preset voice parameter matching model according to the acoustic parameters; The base digital human voice is adjusted according to the individual voice parameters.

[0019] By adopting the above technical solution, when making a digital human, the user's audio is obtained as a sample, and the user's acoustic features are extracted from the sample, so as to adjust the digital human's voice, thereby making the digital human's voice close to the user's own voice.

[0020] Optionally, a personalized adjustment method is also included, and the personalized adjustment method includes: Based on the situation of generating the reference digital human, judging whether the user needs to adjust the image of the digital human according to the selected information; When the user needs to adjust the image of the digital human, the microphone is controlled to obtain adjustment information; Determine whether the wearing needs to be adjusted according to the adjustment information; When the wearing needs to be adjusted, the wearing requirements are determined according to the adjustment information; According to the wearing requirements and the wearer's wear, the final wearing parameters are matched from the preset wearing replacement model, and whether the facial features need to be adjusted is determined based on the adjustment information; When the facial features need to be adjusted, the requirements for the facial features are determined based on the adjustment information; According to the facial features requirements and the person's appearance, the final facial features parameters are matched from the preset facial features replacement model; The final digital human is matched from the preset digital human generation model according to the person's gender, final wearing parameters and final facial features parameters.

[0021] By adopting the above technical solution, after the digital human is initially generated, semantic recognition is performed based on the user's voice, so that the appearance of the digital human is further modified according to the user's requirements, so that the appearance of the digital human further meets the user's requirements.

[0022] In the second aspect, the present application provides a personalized digital human generation system, which adopts the following technical solutions: A personalized digital human generation system, comprising: An acquisition module, used to acquire selection information, personnel information, personnel images, group information, product images, voice parameters and adjustment information; A memory, used to store a program of any of the above-mentioned personalized digital human generation methods; The program in the processor and the memory can be loaded and executed by the processor to implement any of the above-mentioned personalized digital human generation methods.

[0023] In a third aspect, the present application provides a smart terminal, which adopts the following technical solution: An intelligent terminal comprises a memory and a processor, wherein the memory stores a computer program which can be loaded by the processor and execute any one of the above-mentioned personalized digital human generation methods.

[0024] By adopting the above technical solution, when a user creates a personal digital human, the facial features of the digital human are determined from the user's appearance obtained from the image, and the digital human's clothing is determined from the user's occupation, thereby obtaining a digital human that conforms to the user's actual situation, simplifying the digital human generation steps, and improving the convenience of using the digital human.

[0025] In summary, the present application includes at least one of the following beneficial technical effects: When a user creates a digital person of himself, the facial features of the digital person are determined from the user's appearance obtained from the image, and the digital person's clothing is determined based on the user's occupation, so as to obtain a digital person that conforms to the user's actual situation, simplify the steps of generating the digital person, and improve the convenience of using the digital person; when a user creates a digital person representing a group, the digital person's clothing is determined based on the group's business direction, so that the image of the digital person is close to the group's customers, and the group's existing customers and potential customers in the audience feel more intimate with the digital person; when a user creates a digital person representing a group, the age and body shape of the digital person are determined based on the age group and body shape of the target customers of the group's main products, so that the image of the digital person is further close to the group's customers, and the group's existing customers and potential customers in the audience feel more intimate with the digital person. Figure 1 is a flow chart of a method for generating a personalized digital human; Figure 2 This is the process of the group digital human generation method Figure 1 ; Figure 3 This is the process of the group digital human generation method Figure 2 ; Figure 4 This is the process of the group digital human generation method Figure 3 ; Figure 5 Is the process of sales display adjustment method Figure 1 ; Figure 6 Is the process of sales display adjustment method Figure 2 ; Figure 7 is a flow chart of the sound adjustment method; Figure 8 It is a flow chart of the personalized adjustment method. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0027] The embodiment of the present application discloses a method for generating a personalized digital human. The method is used to integrate information such as the user's appearance, occupation, and voice to automatically generate a digital human that meets the user's situation and is similar in appearance to the user, thereby improving the convenience of using the digital human.

[0028] Reference Figure 1 , a personalized digital human generation method, comprising: Step 100: Control a preset selection device to obtain selection information.

[0029] The selection device refers to a device for the user to select functions. The selection device generally uses a touch screen. Functions such as personal use, group use, and image adjustment can be selected on the selection device. The selection information is a set of information sent after the user selects a function on the selection device. The selection device and the method of obtaining the selection information are selected by the staff based on actual conditions and will not be elaborated here.

[0030] Step 101: Determine whether the digital human is for personal use based on the selected information.

[0031] The personal use analysis model can be used to determine whether the selected information contains data generated after selecting the personal use function, and to determine whether the digital person is used by individuals when the selected information contains data generated after selecting the personal use function. The personal use analysis model refers to a neural network model obtained in advance through sample training.

[0032] Step 102: When the digital person is used by an individual, obtain personal information.

[0033] Personnel information refers to an information set that includes a person's name, gender, occupation and other personal information. Personnel information can be obtained by querying a personnel information database. A personnel information database refers to a database that records the personal information of all users in advance. The method for obtaining personnel information is selected by the staff based on actual conditions and will not be elaborated here.

[0034] Step 103: Determine the occupation and gender of the person from a preset personnel information analysis model based on the personnel information.

[0035] The personnel information analysis model refers to a neural network model obtained through sample training in advance. The personnel information analysis model can extract the user's occupation information and gender information from the personnel information and use them as the personnel occupation and personnel gender.

[0036] Step 104: Determine what the person is wearing according to his / her occupation and gender.

[0037] Personnel wear refers to clothing that matches the user's occupation and gender. Personnel wear can be obtained by querying a wear database. The wear database refers to a database that records in advance the clothing worn by men and women of various occupations.

[0038] Step 105: According to the wear of the person, the personal wear parameters of the digital person are matched from the preset wear matching model, and the image of the person is obtained.

[0039] Personal wearing parameters refer to the data set used to adjust the digital human's wearing. Personal wearing parameters can be obtained through a wearing matching model. The wearing matching model refers to a neural network model obtained in advance through sample training.

[0040] The person image refers to the frontal face picture of the user, which can be obtained through the camera on the terminal, where the terminal refers to a device running a personalized digital human generation method. The acquisition image of the person image is selected by the staff according to the actual situation and will not be elaborated here.

[0041] Step 106: Determine the person's appearance based on the person's image.

[0042] Person appearance refers to facial features parameters that are consistent with the facial features of the person in the person image. Person appearance can be obtained through a facial recognition model, and the facial recognition model refers to a neural network model obtained in advance through sample training.

[0043] Step 107: Match the personal facial features parameters of the digital person from the preset facial features recognition model according to the person's appearance.

[0044] Personal facial features parameters refer to the data set used to adjust the facial features of digital humans. Personal facial features parameters can be obtained through a facial features matching model. The facial features matching model refers to a neural network model obtained in advance through sample training.

[0045] Step 108: Match a reference digital human from a preset digital human generation model according to the person's gender, personal clothing parameters, and personal facial features parameters.

[0046] A benchmark digital human refers to a digital human that is generated according to the person's gender, personal clothing parameters and personal facial features, and is consistent with the person's gender, clothing parameters and appearance. The benchmark digital human can be generated through a digital human generation model, which refers to a neural network model obtained in advance through sample training.

[0047] When a user creates a personal digital human, the user's appearance is identified from the personal image to determine the digital human's facial features, and the digital human's clothing is determined based on the user's occupation and gender, thereby automatically generating a digital human that conforms to the user's actual situation and simplifying the steps of generating the digital human.

[0048] Reference Figure 2 , the group digital human generation method includes: Step 200: When the digital human is not used by an individual, determine whether the digital human is used by a group based on the selected information.

[0049] The group use analysis model can be used to determine whether the selected information contains data generated after selecting the group use function, and to determine whether the digital person is using the group when the selected information contains data generated after selecting the group use function. The group use analysis model refers to a neural network model obtained in advance through sample training.

[0050] Step 201: When the digital human is used by a group, obtain group information.

[0051] Group information refers to an information set that includes group information such as the group's name, personnel composition, and logo. Group information can be obtained by querying the group information database. The group information database refers to a database that records the information of all users' groups in advance. The method of obtaining group information is selected by the staff based on actual conditions and will not be elaborated here.

[0052] Step 202: Determine a group name from a preset group name analysis model according to the group information.

[0053] The group name analysis model refers to a neural network model obtained through sample training in advance. The group name analysis model can extract the name of the group from the group information and use it as the group name.

[0054] Step 203: Determine the group domain according to the group name.

[0055] Group domain refers to the types of fields such as education, medical care, and law that the group's customer groups are in. Group domains can be obtained by querying the domain database. The domain database refers to a database that records in advance the different groups that all users belong to and their corresponding group domains.

[0056] Step 204: Determine group attire based on group area and personnel gender.

[0057] Group wear refers to clothing that is consistent with the group's field and the gender of the personnel. Group wear can be obtained by querying the field database. The field database refers to a database that records the male and female wear of each field in advance.

[0058] Step 205: According to the group wear, the group wear parameters of the digital human are matched from the preset wear matching model.

[0059] The group wearing parameters are the data set matched from the wearing matching model and used to adjust the digital human's wearing.

[0060] Step 206: Match a reference digital human from a preset digital human generation model according to the gender of the person, group wearing parameters and individual facial features parameters.

[0061] The benchmark digital human is a digital human generated from the digital human generation model according to the person's gender, group clothing parameters and personal facial features parameters, and is consistent with the person's gender, group clothing and appearance.

[0062] When creating a digital person representing a group, the digital person's attire is determined based on the customer base that the group is targeting, so that the digital person's image is close to the group's customers, increasing the group's existing and potential customers' sense of familiarity with the digital person.

[0063] Reference Figure 3 , the group digital human generation method also includes: Step 207: When the digital human is used by a group, the male-female ratio is matched from a preset group composition analysis model according to the group information.

[0064] The male-to-female ratio refers to the ratio of the number of male members to the number of female members in a group. The group composition information in the group information can be extracted through the group composition analysis model, and the male-to-female ratio can be calculated based on the group composition information. The group composition analysis model refers to a neural network model obtained through sample training in advance.

[0065] Step 208: When the male-to-female ratio is greater than the preset male selection value, male is defined as the group gender.

[0066] The male selection value refers to the pre-set boundary used to measure the main composition of the group's personnel. When the male-to-female ratio is greater than the male selection value, men are considered to be the main members of the group. The male selection value is selected by the staff based on actual conditions and will not be elaborated here.

[0067] Group gender refers to the main gender of the members of the user's group. If the male-to-female ratio is greater than the male selection value, it means that the main gender of the members of the user's group is male. In this case, male is defined as the group gender.

[0068] Step 209: When the male-female ratio is not greater than the preset male selection value, female is defined as the group gender.

[0069] The male-female ratio is not greater than the male selection value, which means that the majority of the members of the user's group are female. In this case, female is defined as the group gender.

[0070] Step 210: Determine group attire based on group area and group gender, and determine group logo from a preset group logo recognition model based on group information.

[0071] Group wear refers to clothing that matches the group domain and group gender obtained from the domain database. Group logo refers to the logo of the user's group, which can be extracted from group information through a group logo recognition model. The group logo recognition model refers to a neural network model obtained through sample training in advance.

[0072] Step 211: Match group wearing parameters from a preset personalized group wearing matching model according to the group wearing and group logo.

[0073] Group wearing parameters refer to a data set used to adjust the digital human's clothing so that the digital human's clothing conforms to the user's field and can reflect the group logo. The group wearing parameters can be obtained through a personalized group wearing matching model. The personalized group wearing matching model can scale the group logo and assign it to a preset part of the group's clothing, and match the group wearing parameters based on the clothing assigned with the team logo. The preset part is selected by the staff according to the actual situation. The personalized group wearing matching model refers to a neural network model obtained in advance through sample training.

[0074] Step 212: When the gender of the individual is consistent with the gender of the group, a reference digital human is matched from a preset digital human generation model according to the group gender, group clothing parameters and individual facial features parameters.

[0075] The individual gender and group gender represent the user's gender, which is the main gender of the members of the user's group. At this time, the user's personal facial features parameters are used as the facial features of the generated digital person.

[0076] Step 213: When the individual's gender is inconsistent with the group's gender, the baseline facial features parameters are determined according to the group's gender.

[0077] The gender of an individual is inconsistent with the gender of the group, which means that the gender of the user is not the main gender of the members of the group to which the user belongs. At this time, it is necessary to generate facial feature parameters that conform to the gender of the group. The baseline facial feature parameters are the facial feature parameters generated according to the gender of the group. The baseline facial feature parameters can be obtained by querying the facial feature database. The facial feature database refers to a database that has pre-recorded facial feature parameters of men and women.

[0078] Step 214: Match a reference digital human from a preset digital human generation model according to the group gender, group clothing parameters and reference facial features parameters.

[0079] The benchmark digital human is a digital human that is consistent with the group gender, group clothing and benchmark appearance and is generated from the group gender, group clothing parameters and benchmark facial feature parameters generated from the digital human generation model.

[0080] The gender of the digital person is determined by the ratio of male to female members in the group, so that the personnel composition of the group can be reflected through the image of the digital person.

[0081] Reference Figure 4 , the group digital human generation method also includes: Step 215: When the digital person is used by a group, the hot-selling products are matched from the preset sales database according to the group name.

[0082] The hot-selling product refers to the product with the highest profit among all the products of the group. The hot-selling product can be obtained by querying from the sales database. The sales database refers to the database that records the most profitable products of each team.

[0083] Step 216: Match the applicable population from the preset product database based on the hot-selling products.

[0084] The applicable population refers to a set of information on the occupation, body shape, age group, etc. of the user group that the hot-selling products are targeting. The applicable population can be obtained by querying the product database. The product database refers to a database that records the applicable population of each product.

[0085] Step 217: Determine the applicable body type and applicable age group according to the applicable population.

[0086] The applicable body shape refers to the body shape of the user group that the best-selling product is aimed at, and the applicable age group refers to the age group of the user group that the best-selling product is aimed at. The applicable body shape and the applicable age group can be extracted from the applicable population through the applicable population analysis model. The applicable population analysis model refers to a neural network model obtained in advance through sample training.

[0087] Step 218: Match the voice adjustment parameters from the preset voice parameter database according to the group gender and applicable age group.

[0088] The voice adjustment parameters refer to the parameters used to adjust the voice of the digital human so that the voice of the digital human is close to the applicable age group and group gender of the users of the hot-selling product. The voice adjustment parameters can be obtained by querying the voice parameter database. The voice parameter database refers to a database that has previously recorded the voice adjustment parameters of males and females of different age groups.

[0089] Step 219: Matching posture adjustment parameters from a preset posture adjustment matching model according to applicable age groups and applicable body types.

[0090] The posture adjustment parameters refer to the parameters used to adjust the body shape of the digital human so that the body shape of the digital human is close to the applicable age group and applicable body shape of the users of the hot-selling product. The posture adjustment parameters can be obtained through the posture adjustment matching model. The posture adjustment matching model refers to a neural network model obtained in advance through sample training.

[0091] Step 220: Match an adjusted digital human from a preset digital human adjustment model according to the voice adjustment parameters, the body adjustment parameters and the reference digital human.

[0092] The adjusted digital human is a digital human obtained by adjusting the voice and body shape of a reference digital human according to voice adjustment parameters and body adjustment parameters. The adjusted digital human can be generated through a digital human adjustment model. The digital human adjustment model refers to a neural network model obtained in advance through sample training.

[0093] Reference Figure 5 , sales display adjustment methods include: Step 300: Acquire product images of products to be sold.

[0094] Product images refer to pictures of products to be sold. Product images can be obtained through a camera. The method for obtaining product images is selected by the staff based on actual conditions and will not be elaborated here.

[0095] Step 301: Match the main color from a preset main color recognition model according to the product image.

[0096] The main color refers to the color that accounts for the largest proportion in the product image. The main color can be identified from the product image through a main color recognition model. The main color recognition model refers to a neural network model obtained in advance through sample training.

[0097] Step 302: Match the main contrasting color from a preset contrasting color matching model according to the main color.

[0098] The main contrasting color refers to a color that is 180° away from the main color on the color wheel. The main contrasting color can be obtained through a contrasting color matching model. The contrasting color matching model refers to a neural network model obtained in advance through sample training.

[0099] Step 303: Match the hand wear from a preset hand wear matching model according to the main contrasting colors and the wear of the person.

[0100] Hand wear refers to a glove of a digital human that fits the needs of a person and is of a primary contrasting color. The hand wear can be obtained through a hand wear matching model, which refers to a neural network model obtained in advance through sample training.

[0101] Step 304: Match hand adjustment parameters from a preset hand adjustment model according to the hand wear.

[0102] The hand adjustment parameters refer to the parameters used to adjust the appearance of the digital human so that the digital human can wear the hand wear. The hand adjustment parameters can be obtained by matching the hand adjustment model. The hand adjustment model refers to the neural network model obtained in advance through sample training.

[0103] Step 305: Match and adjust the wearing parameters from a preset wearing parameter adjustment model according to the main contrasting colors and the personal wearing parameters.

[0104] Adjusting the wearing parameters refers to adjusting the appearance of the digital human so that the digital human's clothing presents parameters of mainly contrasting colors. The adjusting wearing parameters can be obtained by matching from a wearing parameter adjustment model. The wearing parameter adjustment model refers to a neural network model obtained in advance through sample training.

[0105] Step 306: Matching a display digital human from a preset digital human adjustment model according to the adjustment wearing parameters, hand adjustment parameters and the reference digital human.

[0106] The display digital human refers to a digital human after adjusting the wearing parameters and hand adjustment parameters so that the reference digital human wears the hand wear and the wear of the digital human is in a main contrasting color. The display digital human can be generated by the digital human adjustment model. The digital human adjustment model refers to a neural network model obtained in advance through sample training.

[0107] The main color of the product is identified through the product image, so that the digital human's clothing is adjusted to a contrasting color with a significant contrast to the main color, thereby improving the color contrast between the digital human and the product when the digital human displays the product to improve the clarity of the product display.

[0108] Reference Figure 6 , sales display adjustment methods also include: Step 307: Match the main proportion from a preset color proportion matching model based on the main color.

[0109] The main proportion refers to the area proportion of the main color on the surface of the product. The color proportion matching model can be used to identify the area value of the main color and the area value of the entire product from the product image and calculate the quotient of the two to obtain the main proportion. The color proportion matching model refers to a neural network model obtained in advance through sample training.

[0110] Step 308: When the primary proportion is lower than a preset representative threshold, a secondary color is matched from a preset secondary color recognition model according to the product image.

[0111] The representative threshold refers to the minimum ratio of the main color to the main body of the product. The representative threshold is selected by the staff according to the actual situation and will not be elaborated here. If the main proportion is lower than the representative threshold, it means that the main color has a low proportion in the product image, that is, there are many types of colors on the product surface. At this time, it is necessary to increase the types of contrasting colors to improve the contrast between the digital human and the product.

[0112] The secondary color refers to the color that accounts for the largest proportion in the product image except the primary color. The secondary color can be identified from the product image through a secondary color recognition model. The secondary color recognition model refers to a neural network model obtained in advance through sample training.

[0113] Step 309: Match a secondary contrasting color from a preset contrasting color matching model according to the secondary color.

[0114] The secondary contrasting color refers to a color that is 180° away from the secondary color on the color wheel. The secondary contrasting color can be obtained through a contrasting color matching model. The contrasting color matching model refers to a neural network model obtained in advance through sample training.

[0115] Step 310: Match the hand wear from a preset hand wear matching model according to the secondary contrasting color, the primary contrasting color and the person's wear.

[0116] Hand wear refers to gloves matched from the hand wear matching model that are suitable for personnel to wear and have patterns formed by the intervals of secondary contrasting colors and primary contrasting colors. The type of pattern on the hand wear is selected by the staff according to actual conditions.

[0117] Reference Figure 7 , the sound adjustment methods include: Step 400: When the digital human is used by an individual, the voice parameters of the individual are obtained.

[0118] Voice parameters refer to the recording data of a user for a certain period of time. Voice parameters can be obtained through the microphone on the terminal. The method for obtaining voice parameters is selected by the staff according to the actual situation and will not be elaborated here.

[0119] Step 401: Extracting acoustic parameters according to speech parameters.

[0120] Acoustic parameters refer to parameters such as frequency and loudness used to distinguish the voices of different people. Acoustic parameters can be obtained through an acoustic extraction model. The acoustic extraction model refers to a neural network model obtained in advance through sample training.

[0121] Step 402: Match individual voice parameters from a preset voice parameter matching model according to the acoustic parameters.

[0122] Personal voice parameters refer to parameters used to adjust the voice of a digital person so that the voice of the digital person is close to the voice of the user. Personal voice parameters can be obtained through a voice parameter matching model. The voice parameter matching model refers to a neural network model obtained in advance through sample training.

[0123] Step 403: Adjust the voice of the reference digital person according to the personal voice parameters.

[0124] The voice of the reference digital human is adjusted according to personal voice parameters so that the voice of the digital human is close to the voice of the user. The voice of the digital human can be adjusted through a sound calibration model, which refers to a neural network model obtained in advance through sample training.

[0125] Reference Figure 8 , personalized adjustment methods include: Step 500: Based on the situation of generating a reference digital human, determine whether the user needs to adjust the image of the digital human according to the selected information.

[0126] The image adjustment analysis model can be used to determine whether the selected information contains data generated after selecting the image adjustment function, and when the selected information contains data generated after selecting the image adjustment function, it can be determined that the image of the digital human needs to be adjusted. The image adjustment analysis model refers to a neural network model obtained through sample training in advance.

[0127] Step 501: When the user needs to adjust the image of the digital human, the microphone is controlled to obtain adjustment information.

[0128] Adjustment information refers to the recorded data containing the user's adjustment requirements for the digital human image. The user can be prompted to start stating the adjustment requirements through the audio voice on the terminal, and the adjustment information can be recorded through the microphone on the terminal. The method for obtaining the adjustment information is selected by the staff based on actual conditions and will not be elaborated here.

[0129] Step 502: Determine whether the wear needs to be adjusted based on the adjustment information.

[0130] The wear adjustment analysis model can be used to determine whether the adjustment information contains keywords related to wear, such as clothes, pants, gloves, etc., and determine that the wear needs to be adjusted when the adjustment information contains keywords related to wear, such as clothes, pants, gloves, etc. The wear adjustment analysis model refers to a neural network model obtained in advance through sample training.

[0131] Step 503: When wearing adjustment is required, the wearing requirement is determined according to the adjustment information.

[0132] Wearing requirements refer to the specific requirements of users for adjusting different aspects of wearing, such as clothes, pants, gloves, etc. The wearing requirements can be obtained through a wearing requirement recognition model. The wearing requirement recognition model refers to a neural network model obtained in advance through sample training.

[0133] Step 504: Match the final wearing parameters from the preset wearing replacement model according to the wearing requirements and the wearer's wearing, and determine whether the facial features need to be adjusted according to the adjustment information.

[0134] The final wearing parameters refer to the parameters used to adjust the digital human's wearing so that the digital human's wearing meets the wearing requirements. The final wearing parameters can be obtained through the wearing replacement model. The wearing replacement model refers to the neural network model obtained in advance through sample training.

[0135] Step 505: When the facial features need to be adjusted, the facial features requirements are determined based on the adjustment information.

[0136] The facial features requirements refer to the specific requirements of users for the adjustment of different facial organs such as eyes, nose, and mouth. The facial features requirements can be obtained through a facial features requirement recognition model. The facial features requirement recognition model refers to a neural network model obtained in advance through sample training.

[0137] Step 506: Match the final facial feature parameters from the preset facial feature replacement model according to the facial feature requirements and the person's appearance.

[0138] The final facial features parameters refer to the parameters used to adjust the facial features of the digital human so that the facial features of the digital human meet the facial features requirements. The final facial features parameters can be obtained through a facial features replacement model. The facial features replacement model refers to a neural network model obtained in advance through sample training.

[0139] Step 507: According to the gender of the person, the final wearing parameters and the final facial parameters, a final digital human is matched from the preset digital human generation model.

[0140] The final digital human is a digital human generated from the digital human generation model according to the gender of the person, the final wearing parameters and the final facial features parameters, and is generated after adjusting the wearing and facial features of the baseline digital human according to the wearing requirements and facial features requirements.

[0141] Based on the same inventive concept, an embodiment of the present invention provides a personalized digital human generation system, including: An acquisition module, used to acquire selection information, personnel information, personnel images, group information, product images, voice parameters and adjustment information; A memory, used to store a program of any of the above-mentioned personalized digital human generation methods; The program in the processor and the memory can be loaded and executed by the processor to implement any of the above-mentioned personalized digital human generation methods.

[0142] Based on the same inventive concept, an embodiment of the present invention provides an intelligent terminal, including a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and execute a method for generating a personalized digital human.

[0143] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device and unit described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0144] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A method for generating a personalized digital human, characterized in that: include: Controlling a preset selection device to obtain selection information; Determine whether the digital person is for personal use based on the selected information; When a digital person is used for personal purposes, obtain personal information; Determine the occupation and gender of the personnel from the preset personnel information analysis model according to the personnel information; Determine what personnel should wear based on their occupation and gender; According to the wear of the person, the personal wear parameters of the digital person are matched from the preset wear matching model, and the image of the person is obtained; Determine the appearance of a person based on his / her image; According to the person's appearance, the digital person's personal facial features parameters are matched from the preset facial features recognition model; A reference digital human is matched from the preset digital human generation model according to the person's gender, personal wearing parameters and personal facial features parameters.

2. A personalized digital human generation method according to claim 1, characterized in that: The invention also includes a group digital human generation method, wherein the group digital human generation method includes: When the digital person is not used by an individual, determine whether the digital person is used by a group based on the selected information; When the digital person is used by a group, obtain group information; Determine the group name from a preset group name analysis model according to the group information; Determine the group domain based on the group name; Determine group attire based on group area and gender of participants; According to the group wear, the group wear parameters of the digital human are matched from the preset wear matching model; A benchmark digital human is matched from the preset digital human generation model based on the gender of the person, group wearing parameters and individual facial features parameters.

3. A personalized digital human generation method according to claim 2, characterized in that: The group digital human generation method further comprises: When the digital person is used for a group, the male-female ratio is matched from the preset group composition analysis model based on the group information; When the male-to-female ratio is greater than the preset male selection value, male is defined as the group gender; When the male-female ratio is not greater than the preset male selection value, female is defined as the group gender; Determine group attire based on group domain and group gender, and determine group logo from a preset group logo recognition model based on group information; According to the group wear and group logo, group wear parameters are matched from a preset individual group wear matching model; When the gender of the individual is consistent with the gender of the group, a reference digital human is matched from the preset digital human generation model according to the group gender, group clothing parameters and individual facial features parameters; When the gender of the individual is not consistent with the gender of the group, the baseline facial features parameters are determined according to the gender of the group; A benchmark digital human is matched from the preset digital human generation model according to the group gender, group clothing parameters and benchmark facial features parameters.

4. A personalized digital human generation method according to claim 3, characterized in that: The group digital human generation method further comprises: When the digital person is used by a group, the hot-selling products are matched from the preset sales database according to the group name; Match the applicable population from the preset product database based on the hot-selling products; Determine the applicable body type and age group based on the applicable population; Matching voice adjustment parameters from a preset voice parameter database according to the gender and applicable age group of the group; Matching posture adjustment parameters from a preset posture adjustment matching model according to applicable age groups and applicable body types; An adjusted digital human is matched from a preset digital human adjustment model according to the voice adjustment parameters, the body adjustment parameters and the reference digital human.

5. The method for generating a personalized digital human according to claim 1, characterized in that: Also included is a sales display adjustment method, the sales display adjustment method comprising: Obtain product images of products to be sold; Match the main color from the preset main color recognition model according to the product image; Match the main contrasting color from a preset contrasting color matching model according to the main color; Matching hand wear from a preset hand wear matching model according to the main contrasting color and the wear of the person; Matching hand adjustment parameters from a preset hand adjustment model according to the hand wear; Matching and adjusting wearing parameters from a preset wearing parameter adjustment model according to the main contrasting colors and the personal wearing parameters; According to the adjustment of wearing parameters, hand adjustment parameters and the reference digital human, a display digital human is matched from a preset digital human adjustment model.

6. A personalized digital human generation method according to claim 5, characterized in that: The sales display adjustment method also includes: Match the main proportion from the preset color proportion matching model based on the main color; When the primary proportion is lower than a preset representative threshold, a secondary color is matched from a preset secondary color recognition model according to the product image; Matching a secondary contrasting color from a preset contrasting color matching model according to the secondary color; Hand wear is matched from a preset hand wear matching model according to the secondary contrasting color, the primary contrasting color and the person's wear.

7. A personalized digital human generation method according to claim 1, characterized in that: Also included is a sound adjustment method, the sound adjustment method comprising: When the digital person is used by an individual, the person's voice parameters are obtained; extracting acoustic parameters based on speech parameters; Matching personal voice parameters from a preset voice parameter matching model according to the acoustic parameters; The base digital human voice is adjusted according to the individual voice parameters.

8. A personalized digital human generation method according to claim 1, characterized in that: Also included is a personalized adjustment method, the personalized adjustment method comprising: Based on the situation of generating the reference digital human, judging whether the user needs to adjust the image of the digital human according to the selected information; When the user needs to adjust the image of the digital human, the microphone is controlled to obtain adjustment information; Determine whether the wearing needs to be adjusted according to the adjustment information; When the wearing needs to be adjusted, the wearing requirements are determined according to the adjustment information; According to the wearing requirements and the wearer's wear, the final wearing parameters are matched from the preset wearing replacement model, and whether the facial features need to be adjusted is determined based on the adjustment information; When the facial features need to be adjusted, the requirements for the facial features are determined based on the adjustment information; According to the facial features requirements and the person's appearance, the final facial features parameters are matched from the preset facial features replacement model; The final digital human is matched from the preset digital human generation model according to the person's gender, final wearing parameters and final facial features parameters.

9. A personalized digital human generation system, characterized in that: include: An acquisition module, used to acquire selection information, personnel information, personnel images, group information, product images, voice parameters and adjustment information; A memory, used to store a program of a personalized digital human generation method according to any one of claims 1 to 8; The program in the memory can be loaded and executed by the processor.

10. An intelligent terminal, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executes a method for generating a personalized digital human as claimed in any one of claims 1 to 8.

Images