Intelligent user synchronization method for animation character based on voiceprint

Through the synchronization processing of voiceprints and image data, the convolutional neural network and Transformer encoder are used to solve the problem of not being able to recognize and imitate user voices in the existing technology, and more personalized animation characters interact with users, improving users A sense of participation and communication experience.

CN119048650AInactive Publication Date: 2024-11-29BEIJING YOUQU TIME CULTURE TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411076672.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-07
Publication Date
2024-11-29
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing intelligent synchronization user method of animation characters cannot effectively recognize and imitate the user's voice, and cannot realize direct interaction of user language characteristics, resulting in poor user participation and communication experience.

Method used

Using a smart synchronized user method based on voiceprints, collect user image data through high-definition cameras, build a convolutional neural network to extract facial features, and combine a microphone to collect voiceprint data, and use Transformer encoder and self-attention mechanism to perform user virtual avatars. Synchronous processing of harmony and voiceprints generates more personalized animated voice.

Benefits of technology

It improves the clarity and contrast of user image data, extracts advanced features of user faces, enhances the stability and accuracy of sound signals, and provides a more personalized and immersive interactive experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119048650A_ABST
    Figure CN119048650A_ABST
Patent Text Reader

Abstract

The invention relates to a voiceprint-based intelligent user synchronization method for animation characters, in particular to the field of multi-modal data processing, which comprises the following steps of: acquiring user image data by using a high-definition camera, zooming the user image data, matching and filling pixel point positions, enhancing the image contrast through histogram equalization, and acquiring a multi-modal image; according to the method, Gaussian filtering weighting is utilized to remove noise existing in user image data, so that the user image data is kept clear and rich in details under a fixed size, a convolutional neural network is constructed to output a final user virtual image based on a dominant hue and a role image of animation, and a microphone is utilized to collect user voiceprint data. The method comprises the following steps: performing frequency spectrum representation on a sound signal of each window through Fourier transform, constructing a Transform encoder, adding a self-attention mechanism and a feedforward neural network, synchronizing and automatically replacing a user virtual image and animation voice converted by a user voiceprint, and providing more personalized and immersive interactive experience for a user.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of multimodal data processing, and more specifically, to a method for intelligently synchronizing users of animated characters based on voiceprints. Background Art

[0002] With the development of human-computer interaction technology, more and more applications are beginning to use virtual characters or animated characters as part of the user interface. This trend not only brings users a more vivid and interesting interface experience, but also brings new possibilities and challenges to human-computer interaction. The use of virtual animated characters gives applications a more vivid appearance and a more personalized interaction form, providing a more intuitive and friendly experience.

[0003] The current method of intelligently synchronizing users with animated characters lacks personalized recognition of user voices, cannot imitate user language characteristics, cannot interact more directly with animated characters through user language, and cannot improve user participation and communication experience. Summary of the invention

[0004] The present invention aims at the technical problems existing in the prior art and provides an intelligent user synchronization method for an animated character based on voiceprint to solve the problems raised in the above background technology.

[0005] The technical solution of the present invention to solve the above technical problem is as follows: a method for intelligently synchronizing users of an animated character based on voiceprints, comprising the following steps:

[0006] S101. Collect user image data using a high-definition camera, scale the user image data, match and fill pixel positions, enhance image contrast through histogram equalization, and remove noise from the user image data using weighted Gaussian filtering;

[0007] S102. Construct a convolutional neural network and receive user image data through the input layer, output the feature mapping of the key points of the user image's face to different user image categories, and output the final user virtual image based on the main color tone and character image of the animation;

[0008] S103. Collect the user voiceprint data using a microphone and divide it into multiple short-term windows, use a Hamming window to perform windowing on the sound signal of each frame of the user voiceprint data, and perform spectrum representation on the sound signal of each window through Fourier transform;

[0009] S104. Take the user's virtual image and the animated speech converted from the user's voiceprint as the input sequence, build a Transformer encoder and add a self-attention mechanism and a feedforward neural network to synchronize and automatically replace the user's virtual image and the animated speech converted from the user's voiceprint.

[0010] In a preferred embodiment, in S101, a high-definition camera is used to collect user image data, the size of the scaling target of the user image data is fixed to 224x224, the position of each pixel in the scaling target is traversed, and the nearest pixel position in the user image data corresponding to each pixel in the scaling target is found and assigned and filled, and the traversal, search and assignment and filling steps are repeated until the complete scaling target is traversed, and the pixel grayscale values ​​of the scaling target of the user image data are redistributed by using histogram equalization to mark the position coordinates of the key points on the user image's face.

[0011] Furthermore, the noise in the user image data is removed by using Gaussian filtering weighting, and the specific formula is:

[0012]

[0013] Among them, f represents the denoising result after Gaussian filtering weighting, O i represents the neighborhood point value of the i-th pixel in the user image data, d i represents the distance from the neighborhood point of the i-th pixel in the user image data to the current point, and σ represents the standard deviation of the Gaussian distribution.

[0014] In a preferred embodiment, in S102, a convolutional neural network is constructed and the user image data is received through the input layer, a convolutional layer is added and a batch activation function layer is activated, ReLU is used as the activation function layer to introduce nonlinearity and a convolution kernel is used to perform a convolution operation on the user image data, and the specific formula of the convolution operation is:

[0015]

[0016] Wherein S(i,j) represents an element of the feature of the key points of the facial image of the output user, I(m,n) represents the pixel value of the input user image data at the position coordinates of the key points of the facial image of the user, K(im,jn) represents the weight value of the convolution kernel at the position coordinates of the key points of the facial image of the user, m and n represent the indexes of the rows and columns inside the convolution kernel respectively, i and j represent the indexes of the rows and columns of the features of the key points of the facial image of the output user respectively, the features of the key points of the facial image of the output user output by the convolution layer are downsampled through the pooling layer, the first convolution and pooling operations are repeated, the features of the key points of the facial image of the user in the pooling layer are flattened into a one-dimensional vector through the fully connected layer, and are input into the fully connected layer, the fully connected layer is used to map the features of the key points of the facial image of the output user to different user image categories through the weight matrix, and the final user virtual image is output based on the main color tone and character image of the animation.

[0017] In a preferred embodiment, in S103, a microphone is used to collect user voiceprint data, the user voiceprint data is divided into a plurality of short-term windows, a Hamming window is used to perform windowing processing on the sound signal of each frame of the user voiceprint data, and a Fourier transform is used to perform spectrum representation on the sound signal of each window, and the Fourier transform formula is:

[0018]

[0019] Wherein, Q[k] represents the spectral component with frequency k in Fourier transform, q[M] represents the sampling value with time M in the window, N represents the length of the sound signal of the user voiceprint data, M represents the sampling point index of the sound signal of the user voiceprint data, and u represents the imaginary unit. The user voiceprint is converted into animated speech by minimizing the difference between the spectral representations of the sound signal of the user voiceprint data.

[0020] Furthermore, a generator network is created to map random noise to sound features that match the user's voiceprint features, the generator network structure is designed using deconvolution layers, the feature parameters are adjusted through learning through a multi-layer neural network structure, a discriminator network is created and the target sound features are input to distinguish between the sound features generated by the generator and the real user's voiceprint features, the pit loss is used to drive the generator to generate realistic sound features, and the discriminator is used to help the generator learn.

[0021] In a preferred embodiment, in S104, the user's virtual image and the animated speech converted from the user's voiceprint are used as input sequences and a Transformer encoder is constructed. A self-attention mechanism and a feedforward neural network are added to each encoding layer in the Transformer encoder structure. The input sequence is passed through three fully connected layers to obtain vector representations of the query, key, and value, and is divided into multiple heads through a weight matrix for calculating attention using different subspaces. The multi-head self-attention results are weighted to obtain the output of the self-attention mechanism and residually connected. The nonlinear transformation of the feedforward neural network is used to output the correlation between the user's virtual image and the animated speech converted from the user's voiceprint, and the user's virtual image and the animated speech converted from the user's voiceprint are synchronized and automatically replaced.

[0022] The beneficial effects of the present invention are as follows: by scaling, matching and filling pixel positions, the user image data is kept clear and detailed at a fixed size; by histogram equalization, the pixel grayscale values ​​can be redistributed, the contrast of the image is improved, and the details are clearer; by using Gaussian filtering weighting to remove the noise existing in the user image data, the image is cleaner and clearer, and interference is reduced; by using ReLU as an activation function to introduce nonlinearity, it is helpful to capture the complex features in the user image data; at the same time, through multiple convolution and pooling operations, the high-level features of the user image data can be gradually extracted, which is helpful to better characterize the user's facial features; by dividing the voiceprint data into multiple short-term windows and applying a Hamming window to the sound signal of each frame for windowing processing, it is helpful to reduce the amplitude jump at both ends of the frame and avoid spectrum leakage, thereby improving the stability and accuracy of the sound signal; by minimizing the difference between the spectrum representations of the sound signal of the user voiceprint data, a more personalized and immersive interactive experience is provided for the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flow chart of the method of the present invention; DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0025] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, "plurality" means two or more, unless otherwise clearly and specifically defined.

[0026] In the description of the present application, the term "for example" is used to mean "used as an example, illustration or description". Any embodiment described as "for example" in the present application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid unnecessary details to obscure the description of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present application.

[0027] This embodiment provides Figure 1 The method for intelligently synchronizing users of animated characters based on voiceprints specifically includes the following steps:

[0028] S101. Collect user image data using a high-definition camera, scale the user image data, match and fill pixel positions, enhance image contrast through histogram equalization, and remove noise from the user image data using weighted Gaussian filtering;

[0029] Furthermore, a high-definition camera is used to collect user image data, and the size of the scaling target of the user image data is fixed to 224x224. The position of each pixel in the scaling target is traversed, and the position of the nearest pixel in the user image data corresponding to each pixel in the scaling target is found and assigned and filled. The traversal, finding and assignment and filling steps are repeated until the complete scaling target is traversed. Histogram equalization is used to redistribute the pixel grayscale values ​​of the scaling target of the user image data to enhance the contrast of the user image data, so that the details of the user image data are cleaner, and the position coordinates of the key points of the user image's face are marked, including the temples, eyebrow corners, eye corners, nose tip, and mouth corners. Among them, the temples, eyebrow corners, eye corners, nose tip and mouth corners are more prominent feature points on the user image's face. The generation of the user image is assisted by marking each key point, so as to achieve more refined user image data analysis and processing, and improve the accuracy and stability of user image generation.

[0030] Furthermore, the noise in the user image data is removed by using Gaussian filtering weighting, and the specific formula is:

[0031]

[0032] Among them, f represents the denoising result after Gaussian filtering weighting, O i represents the neighborhood point value of the i-th pixel in the user image data, d irepresents the distance from the neighborhood point of the i-th pixel in the user image data to the current point, σ represents the standard deviation of the Gaussian distribution, where the Gaussian filter reduces the noise in the user image data by weighted averaging the neighboring pixels around each pixel in the user image data, which can effectively reduce Gaussian noise and mild salt and pepper noise, and can retain the overall details of the image.

[0033] S102. Construct a convolutional neural network and receive user image data through the input layer, output the feature mapping of the key points of the user image's face to different user image categories, and output the final user virtual image based on the main color tone and character image of the animation;

[0034] Further, a convolutional neural network is constructed and the user image data is received through the input layer, a convolutional layer is added and a batch activation function layer is activated, wherein each convolutional layer includes multiple convolution kernels, ReLU is used as the activation function layer to introduce nonlinearity and the convolution kernel is used to perform a convolution operation on the user image data, which is used to output the features of the key points of the user image face, wherein the ReLU activation function is introduced as a nonlinear activation function, which can increase the expression ability of the network, and the specific formula of the convolution operation is:

[0035]

[0036] Where S(i,j) represents an element of the feature of the key points of the user's facial image output, I(m,n) represents the pixel value of the input user's facial image data at the position coordinates of the key points of the user's facial image, K(im,jn) represents the weight value of the convolution kernel at the position coordinates of the key points of the user's facial image, m and n represent the indexes of the row and column inside the convolution kernel, i and j represent the indexes of the row and column of the feature of the key points of the user's facial image output, and the specific formula of the ReLU as the activation function is:

[0037] f(x)=max(0,x)

[0038] When the input value x is greater than or equal to 0, the output of the activation function is f(x)=x, and when the input value x is less than 0, the output of the activation function is f(x)=0. Given an input value x, the activation function returns a larger value to introduce nonlinearity by comparing the size of x with 0. The features of the facial key points of the user image output by the convolution layer are downsampled through the pooling layer to reduce the feature size of the facial key points of the user image and retain important feature information. The first convolution and pooling operations are repeated to extract high-level features of the user image data. The features of the facial key points of the user image of the pooling layer are flattened into a one-dimensional vector through the fully connected layer and input into the fully connected layer. The fully connected layer is used to map the features of the output facial key points of the user image to different user image categories through the weight matrix, and the final user virtual image is output based on the main color tone and character image of the animation. Through the layer-by-layer abstraction and learning ability of the convolutional neural network, the complex features in the user image data are better captured, which helps to realize the deep feature extraction and mapping of the user image data and improve the accuracy and fidelity of the generated virtual image.

[0039] S103. Collect the user voiceprint data using a microphone and divide it into multiple short-term windows, use a Hamming window to perform windowing on the sound signal of each frame of the user voiceprint data, and perform spectrum representation on the sound signal of each window through Fourier transform;

[0040] Furthermore, a microphone is used to collect user voiceprint data, and the user voiceprint data is divided into multiple short-term windows, where each window becomes a frame, a frame is 20ms, and adjacent frames overlap, to ensure the stability of the sound signal of the user voiceprint data in each time window during the collection process, and a Hamming window is used to perform windowing processing on the sound signal of the user voiceprint data of each frame to reduce the amplitude jump at both ends of the frame and avoid spectrum leakage, and a Fourier transform is used to perform spectrum representation on the sound signal of each window to obtain the spectrum feature information of the sound signal of each frame, so as to realize the effective representation and extraction of the user voiceprint data. The Fourier transform formula is:

[0041]

[0042] Among them, Q[k] represents the spectral component with frequency k in Fourier transform, q[M] represents the sampling value of time M in the window, N represents the length of the sound signal of the user voiceprint data, M represents the sampling point index of the sound signal of the user voiceprint data, and u represents an imaginary unit. The user voiceprint is converted into animated speech by minimizing the difference between the spectral representations of the sound signal of the user voiceprint data, and the speech synthesis process of the animated character is optimized, so that the generated animated speech is closer to the user's voice characteristics.

[0043] Furthermore, a generator network is created to map random noise to sound features that match the user's voiceprint features, and a deconvolution layer is used to design the generator network structure, including the number of layers, the number of nodes in each layer, and the activation function, wherein the random noise is a fixed-length vector used to generate target sound features, and feature parameters are adjusted through learning through a multi-layer neural network structure. A discriminator network is created and the target sound features are input to distinguish between the sound features generated by the generator and the real user voiceprint features, wherein the pit loss is used in the generator network to drive the generator to generate realistic sound features, and the discriminator is used to help the generator learn, which can convert the user's voiceprint into animated speech, thereby realizing the learning and conversion of acoustic features.

[0044] S104. Take the user's virtual image and the animated speech converted from the user's voiceprint as input sequences, construct a Transformer encoder and add a self-attention mechanism and a feedforward neural network to synchronize and automatically replace the user's virtual image and the animated speech converted from the user's voiceprint;

[0045] Furthermore, taking the user's virtual image and the animated speech converted from the user's voiceprint as input sequences and constructing a Transformer encoder can effectively capture the correlation information between the input sequences and realize the synchronous replacement function of the user's virtual image and the animated speech converted from the voiceprint. In the Transformer encoder structure, a self-attention mechanism and a feedforward neural network are added to each encoding layer. The input sequence passes through three fully connected layers to obtain the vector representation of the query, key, and value, and is divided into multiple heads through a weight matrix for calculating attention using different subspaces. The multi-head self-attention results are weighted to obtain the output of the self-attention mechanism and residually connected. The nonlinear transformation of the feedforward neural network is used to output the correlation between the user's virtual image and the animated speech converted from the user's voiceprint, so as to better measure and optimize the correlation between the user's virtual image and the animated speech converted from the voiceprint, further improve the quality of synchronous replacement, and synchronize and automatically replace the user's virtual image and the animated speech converted from the user's voiceprint.

[0046] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0047] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0048] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0049] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0050] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0051] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0052] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An intelligent user synchronization method for animation characters based on voiceprints, characterized in that: The specific steps include: S101. Collect user image data using a high-definition camera, scale the user image data, match and fill pixel positions, enhance image contrast through histogram equalization, and remove noise from the user image data using weighted Gaussian filtering; S102. Construct a convolutional neural network and receive user image data through the input layer, output the feature mapping of the key points of the user image's face to different user image categories, and output the final user virtual image based on the main color tone and character image of the animation; S103. Collect the user voiceprint data using a microphone and divide it into multiple short-term windows, use a Hamming window to perform windowing on the sound signal of each frame of the user voiceprint data, and perform spectrum representation on the sound signal of each window through Fourier transform; S104. Take the user's virtual image and the animated speech converted from the user's voiceprint as the input sequence, build a Transformer encoder and add a self-attention mechanism and a feedforward neural network to synchronize and automatically replace the user's virtual image and the animated speech converted from the user's voiceprint.

2. The method for intelligently synchronizing users of animated characters based on voiceprints according to claim 1, characterized in that: In S101, the position of each pixel in the zoom target is traversed, and the nearest pixel position in the user image data corresponding to each pixel in the zoom target is found and assigned and filled, and the traversal, search and assignment and filling steps are repeated until the entire zoom target is traversed.

3. The method for intelligently synchronizing users of animated characters based on voiceprint according to claim 1, characterized in that: In S101, histogram equalization is used to redistribute the pixel grayscale values ​​of the scaling target of the user image data, the position coordinates of the key points of the user image's face are marked, and the noise in the user image data is removed by weighted Gaussian filtering.

4. The method for intelligently synchronizing users of animated characters based on voiceprints according to claim 3 is characterized in that: The specific formula for Gaussian filter weighting is: Among them, f represents the denoising result after Gaussian filtering weighting, O i represents the neighborhood point value of the i-th pixel in the user image data, d i represents the distance from the neighborhood point of the i-th pixel in the user image data to the current point, and σ represents the standard deviation of the Gaussian distribution.

5. The method for intelligently synchronizing users of animated characters based on voiceprints according to claim 1, characterized in that: In S102, a convolutional neural network is constructed and user image data is received through an input layer, a convolutional layer is added and a function layer is activated in batches, a convolution operation is performed on the user image data using a convolution kernel, a downsampling operation is performed on the features of the user image facial key points output by the convolutional layer through a pooling layer, the first convolution and pooling operation is repeated and iterated, the features of the user image facial key points of the pooling layer are flattened into a one-dimensional vector through a fully connected layer, and the features are input into the fully connected layer, and the features of the output user image facial key points are mapped to different user image categories using a weight matrix using the fully connected layer.

6. The method for intelligently synchronizing users of animated characters based on voiceprints according to claim 5 is characterized in that: The specific formula of the convolution operation is: Where S(i,j) represents an element of the feature of the key points of the user's facial image output, I(m,n) represents the pixel value of the input user image data at the position coordinates of the key points of the user's facial image, K(im,jn) represents the weight value of the convolution kernel at the position coordinates of the key points of the user's facial image, m and n represent the row and column indexes inside the convolution kernel respectively, and i and j represent the row and column indexes of the feature of the key points of the user's facial image output respectively.

7. The method for intelligently synchronizing users of animated characters based on voiceprints according to claim 1, characterized in that: In S103, a microphone is used to collect user voiceprint data, the user voiceprint data is divided into a plurality of short-term windows, a Hamming window is used to perform windowing processing on the sound signal of each frame of the user voiceprint data, a Fourier transform is used to perform spectrum representation on the sound signal of each window, and the user voiceprint is converted into animated speech by minimizing the difference between the spectrum representations of the sound signal of the user voiceprint data.

8. The method for intelligently synchronizing users of animated characters based on voiceprints according to claim 7 is characterized in that: The Fourier transform formula is: Among them, Q[k] represents the spectral component with frequency k in Fourier transform, q[M] represents the sampling value with time M in the window, N represents the length of the sound signal of the user voiceprint data, M represents the sampling point index of the sound signal of the user voiceprint data, and u represents the imaginary unit.

9. The method for intelligently synchronizing users of animated characters based on voiceprint according to claim 1, characterized in that: In the S104, a self-attention mechanism and a feedforward neural network are added to each encoding layer in the Transformer encoder structure, and the input sequence is respectively passed through three fully connected layers to obtain vector representations of the query, key, and value, and is divided into multiple heads through a weight matrix. The multi-head self-attention results are weighted to obtain the output of the self-attention mechanism and residually connected, and the nonlinear transformation of the feedforward neural network is used to output the correlation between the user's virtual image and the animated voice converted from the user's voiceprint, and the user's virtual image and the animated voice converted from the user's voiceprint are synchronized and automatically replaced.