Digital human live broadcast interaction method and device in financial scene

Abstract

The invention discloses a digital human live broadcast interaction method and device in a financial scene, and relates to the field of financial live broadcast interaction, and the method comprises the steps: obtaining a first audience question of a current live broadcast room, and rewriting the first audience question according to a financial live broadcast large model, and obtaining a first digital human question; wherein the financial live broadcast large model is obtained based on training of collected financial scene data; according to the first digital person problem, combining a preset digital person to generate a digital person live broadcast stream; according to the digital person live broadcast stream, controlling the digital person to ask a question for an anchor in a current live broadcast room, and obtaining an anchor live broadcast stream generated by the anchor when the anchor answers the first digital person question; and combining the anchor live broadcast stream and the digital human live broadcast stream to obtain a live broadcast room live broadcast stream, and pushing the live broadcast room live broadcast stream to the current live broadcast room. According to the application, the digital human live broadcast interactivity in a financial scene can be improved.

Description

Technical Field

[0001] The present application relates to the field of financial live broadcast interaction, and in particular to a digital human live broadcast interaction method and device in a financial scenario. Background Art

[0002] There are a variety of digital human live broadcast technologies in the existing technology, including 2D and 3D digital human live broadcast technologies. However, most digital human live broadcast technologies rely on pre-configured text to match the question and answer library, and it is difficult to interact with the audience's comments in real time during the live broadcast. Secondly, some digital humans need to use motion capture technology or real-person voice input during the live broadcast, which leads to interactive delays, consumes unnecessary manpower and reduces the interactivity with the audience. In the existing technology, the live broadcast of digital humans in financial scenarios also has the above technical difficulties. Therefore, how to improve the interactivity of digital human live broadcast interactions in financial scenarios is still a difficult problem that needs to be solved urgently in the existing technology. Summary of the invention

[0003] The present application provides a method and device for digital human live broadcast interaction in a financial scenario to solve the technical problem of lack of interactivity in existing digital human live broadcast interaction in financial scenarios.

[0004] According to a first aspect of the implementation of the present application, a digital human live broadcast interaction method in a financial scenario is provided, comprising:

[0005] Obtain the first audience question in the current live broadcast room, and rewrite the first audience question according to the financial live broadcast model to obtain the first digital human question; wherein the financial live broadcast model is trained based on the collected financial scenario data;

[0006] According to the first digital human problem, combined with the preset digital human, a digital human live stream is generated;

[0007] According to the digital human live stream, controlling the digital human to ask questions to the anchor in the current live broadcast room, and obtaining the anchor live stream generated by the anchor answering the first digital human question;

[0008] The anchor live stream and the digital human live stream are combined to obtain the live stream of the live broadcast room, and the live stream of the live broadcast room is pushed to the current live broadcast room.

[0009] This application first rewrites the first audience question in the current live broadcast room through the financial live broadcast model to obtain the first digital human question, which can rewrite the audience's original question that is not suitable for digital human questions into a question suitable for digital human questions. While obtaining questions through audience questions, it improves the professionalism of the questions and reduces the interaction cost; regenerates the digital human live stream, and controls the digital human to ask questions to the anchor according to the digital human live stream, and obtains the anchor live stream generated by the anchor's answer, and then merges the anchor live stream and the digital human live stream to obtain the live broadcast room live stream and push it to the current live broadcast room, which can present the anchor live stream and the digital human live stream to the audience at the same time, improving the audience's willingness to interact. Compared with the existing digital human live broadcast that relies on the matching of the preset question and answer library, this application can obtain the audience questions in the current live broadcast room and rewrite them, and then the digital human asks questions and the anchor answers, and integrates this question and answer process to present to the audience, which can realize real-time interaction with audience comments, and improve the interactivity of digital human live interaction in financial scenarios.

[0010] In certain implementations of the present application, obtaining the first viewer question in the current live broadcast room specifically includes:

[0011] Get multiple initial audience comments in the current live broadcast room;

[0012] Based on the financial live broadcast model, the plurality of initial audience comments are screened to obtain a plurality of initial audience questions;

[0013] The multiple initial audience questions are arranged according to the order of timestamps to obtain a first question queue, and the first initial audience question at the head of the first question queue is taken as the first audience question.

[0014] This application first obtains multiple initial audience comments in the current live broadcast room, then screens them based on the financial live broadcast big model to obtain multiple initial audience questions, and arranges them in order of timestamps to obtain a first question queue. The financial live broadcast big model can be used to screen out non-question type and potential non-question type comments, thereby improving the overall information entropy of the first question queue, thereby reducing the difficulty of subsequent processing and improving processing efficiency when obtaining the first audience question according to the first question queue for subsequent processing.

[0015] In certain embodiments of the present application, rewriting the first audience question according to the financial live broadcast model to obtain the first digital human question specifically includes:

[0016] According to the financial live broadcast model, rewrite the first audience question to obtain the second audience question;

[0017] According to the first audience name corresponding to the first audience question, combined with multiple preset question prefixes and multiple preset question suffixes, the second audience question is updated to obtain the first digital human question; wherein, the first digital human question includes a question prefix, the first audience name, a question suffix and the second audience question.

[0018] This application first rewrites the first audience's question according to the financial live broadcast model to obtain the second audience's question, and can rewrite the audience's original question that is not suitable for digital people to ask into a question that is suitable for digital people to ask, thereby improving the professionalism of the question. Then, according to the first audience name corresponding to the first audience's question, combined with a preset question prefix and a preset question suffix, the second audience's question is updated to obtain the first digital person's question, and the second audience's question can be processed into a first digital person question that is more suitable for the digital person's question format, thereby reducing the interaction cost when the digital person communicates with the anchor.

[0019] In certain implementations of the present application, generating a digital human live stream based on the first digital human problem and in combination with a preset digital human specifically includes:

[0020] The digital human live stream includes a digital human audio live stream and a digital human image live stream;

[0021] Based on the streaming speech synthesis technology, the first digital human question is converted into audio to obtain a digital human audio live stream;

[0022] According to the digital human audio live stream, the facial movements of the digital human are controlled to generate a digital human screen live stream.

[0023] This application first converts the first digital human question into audio based on streaming speech synthesis technology to obtain a digital human audio live stream, and then controls the digital human's facial movements according to the digital human audio live stream to generate a digital human picture live stream, which can respectively obtain the audio and picture of the digital human who is asking questions suitable for the current live broadcast room, thereby attracting audience interaction through the digital human's audio stream and picture stream, and increasing the audience's willingness to interact, thereby improving the interactivity of the digital human live interaction in financial scenarios.

[0024] In certain implementations of the present application, the merging of the anchor live stream and the digital human live stream to obtain the live stream of the live broadcast room specifically includes:

[0025] Extract the audio output and the picture output of the anchor live stream respectively to obtain the anchor audio live stream and the anchor picture live stream;

[0026] According to the preset sound quality parameters, the host audio live stream and the digital human audio live stream are superimposed to obtain the live studio audio live stream;

[0027] Based on the preset screen layout, the digital human screen live stream is superimposed on the anchor screen live stream in the form of a mask to obtain the live studio screen live stream;

[0028] The live broadcast room audio live stream and the live broadcast room picture live stream are integrated to obtain the live broadcast room live stream.

[0029] This application first extracts the audio output and picture output of the host's live stream respectively to obtain the host's audio stream and picture stream, and then superimposes the host's audio stream and the digital human's audio stream according to preset sound quality parameters to obtain the live room audio stream, and superimposes the digital human's picture stream on the host's picture stream in the form of a mask based on the preset picture layout to obtain the live room picture stream. The audio stream and the picture stream can be processed separately to avoid errors caused by simultaneous processing and reduce the error probability, and then superimpose the audio stream and the picture stream of the live room to obtain the live stream of the live room. The host's live stream and the digital human's live stream can be presented to the audience at the same time, thereby increasing the audience's willingness to interact, further realizing the live interaction between digital humans, hosts and audiences, and improving the interactivity of the live interaction of digital humans in financial scenarios.

[0030] According to a second aspect of the implementation of the present application, a digital human live interactive device in a financial scenario is provided, comprising a digital human question acquisition module, a digital human live stream generation module, a digital human questioning module and a live room live stream generation module;

[0031] The digital human question acquisition module is used to acquire the first viewer's question in the current live broadcast room, and rewrite the first viewer's question according to the financial live broadcast model to obtain the first digital human question; wherein the financial live broadcast model is trained based on the collected financial scenario data;

[0032] The digital human live stream generation module is used to generate a digital human live stream according to the first digital human question and in combination with a preset digital human;

[0033] The digital human questioning module is used to control the digital human to ask questions to the anchor in the current live broadcast room according to the digital human live broadcast stream, and obtain the anchor live broadcast stream generated by the anchor answering the first digital human question;

[0034] The live broadcast room live stream generation module is used to merge the anchor live stream and the digital human live stream to obtain the live broadcast room live stream, and push the live broadcast room live stream to the current live broadcast room.

[0035] In certain embodiments of the present application, the digital human question acquisition module includes an audience comment acquisition unit, an audience comment screening unit, and an audience question acquisition unit;

[0036] The audience comment acquisition unit is used to acquire a plurality of initial audience comments in the current live broadcast room;

[0037] The audience comment screening unit is used to screen the multiple initial audience comments based on the financial live broadcast model to obtain multiple initial audience questions;

[0038] The audience question acquisition unit is used to arrange the multiple initial audience questions according to the time stamp sequence to obtain a first question queue, and take the first initial audience question at the head of the first question queue as the first audience question.

[0039] In certain embodiments of the present application, the digital human question acquisition module includes an audience question rewriting unit and a digital human question acquisition unit;

[0040] The audience question rewriting unit is used to rewrite the first audience question according to the financial live broadcast model to obtain a second audience question;

[0041] The digital human question acquisition unit is used to update the second audience question according to the first audience name corresponding to the first audience question, combined with multiple preset question prefixes and multiple preset question suffixes, to obtain the first digital human question; wherein the first digital human question includes a question prefix, the first audience name, a question suffix and the second audience question.

[0042] In certain embodiments of the present application, the digital human live stream includes a digital human audio live stream and a digital human image live stream, and the digital human live stream generation module includes a digital human audio stream generation unit and a digital human image stream generation unit;

[0043] The digital human audio stream generating unit is used to convert the first digital human question into audio based on the streaming speech synthesis technology to obtain a digital human audio live stream;

[0044] The digital human picture stream generating unit is used to control the facial movements of the digital human according to the digital human audio live stream to generate a digital human picture live stream.

[0045] In certain embodiments of the present application, the live broadcast room live stream generation module includes a live stream extraction unit, an audio stream superposition unit, a picture stream superposition unit and a live stream integration unit;

[0046] The live stream extraction unit is used to extract the audio output and the picture output of the anchor live stream respectively to obtain the anchor audio live stream and the anchor picture live stream;

[0047] The audio stream superposition unit is used to superimpose the host audio live stream and the digital human audio live stream according to preset sound quality parameters to obtain the live audio stream of the live broadcast room;

[0048] The picture stream superimposing unit is used to superimpose the digital human picture live stream onto the anchor picture live stream in a mask form based on a preset picture layout to obtain a live studio picture live stream;

[0049] The live stream integration unit is used to integrate the live audio stream of the live broadcast room and the live picture stream of the live broadcast room to obtain the live broadcast stream of the live broadcast room.

[0050] This application first rewrites the first audience question in the current live broadcast room through the financial live broadcast model to obtain the first digital human question, which can rewrite the audience's original question that is not suitable for digital human questions into a question suitable for digital human questions. While obtaining questions through audience questions, it improves the professionalism of the questions and reduces the interaction cost; regenerates the digital human live stream, and controls the digital human to ask questions to the anchor according to the digital human live stream, and obtains the anchor live stream generated by the anchor's answer, and then merges the anchor live stream and the digital human live stream to obtain the live broadcast room live stream and push it to the current live broadcast room, which can present the anchor live stream and the digital human live stream to the audience at the same time, improving the audience's willingness to interact. Compared with the existing digital human live broadcast that relies on the matching of the preset question and answer library, this application can obtain the audience questions in the current live broadcast room and rewrite them, and then the digital human asks questions and the anchor answers, and integrates this question and answer process to present to the audience, which can realize real-time interaction with audience comments, and improve the interactivity of digital human live interaction in financial scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 : A flowchart of a digital human live broadcast interaction method in a financial scenario shown in certain embodiments of the present application;

[0052] Figure 2 : A module structure diagram of a digital human live interactive device in a financial scenario shown in certain embodiments of the present application. DETAILED DESCRIPTION

[0053] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions from beginning to end. The embodiments described below in conjunction with the accompanying drawings are exemplary and are only used to explain some embodiments of the present application, and should not be construed as limitations on the embodiments of the present application. Based on the embodiments shown in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0054] In the description of this application, it should be understood that 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. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, unless otherwise clearly and specifically defined, "multiple" and "several" mean two or more.

[0055] Most of the existing digital human live broadcast technologies rely on pre-configured text to match the question-and-answer library, and lack real-time interaction with the audience during the live broadcast; some digital human live broadcasts require motion capture technology or real-person voice input, which has a certain action delay, resulting in a decrease in interactivity with the audience. The existing digital human live broadcast in financial scenarios also has the above technical difficulties. Therefore, how to improve the interactivity of digital human live broadcast interaction in financial scenarios is still a difficult problem that needs to be solved urgently by existing technologies.

[0056] Based on the above technical background, please refer to Figure 1 The embodiment of the present application provides a digital human live broadcast interaction method in a financial scenario, including steps S101 to S104, each of which is specifically as follows:

[0057] Step S101: Obtain the first audience question in the current live broadcast room, and rewrite the first audience question according to the financial live broadcast model to obtain the first digital human question; wherein the financial live broadcast model is trained based on the collected financial scenario data.

[0058] In certain embodiments of the present application, the financial live broadcast big model is trained based on the collected financial scenario data. Specifically, the financial live broadcast big model is trained on a preset basic big model based on the collected financial scenario data; wherein, the basic big model includes any one of GPT, iFlytek Spark, Zhipu Qingyan, Wenxin Yiyan, Tongyi Qianwen and the big language model fine-tuned by Lora.

[0059] In certain implementations of the present application, obtaining the first viewer question in the current live broadcast room specifically includes:

[0060] Get multiple initial audience comments in the current live broadcast room;

[0061] Based on the financial live broadcast model, the plurality of initial audience comments are screened to obtain a plurality of initial audience questions;

[0062] The multiple initial audience questions are arranged according to the order of timestamps to obtain a first question queue, and the first initial audience question at the head of the first question queue is taken as the first audience question.

[0063] In certain embodiments of the present application, the method for obtaining multiple initial audience comments in the current live broadcast room is specifically: obtaining the multiple initial audience comments in a polling manner through the live broadcast room background comment interface; wherein the multiple initial audience comments include comment content and audience ID.

[0064] In certain embodiments of the present application, based on the financial live broadcast model, the plurality of initial audience comments are screened to obtain a plurality of initial audience questions, specifically:

[0065] Based on the financial live broadcast model, in combination with preset question samples and preset question screening prompts, the multiple initial audience comments are screened to obtain multiple initial audience questions.

[0066] This application first obtains multiple initial audience comments in the current live broadcast room, then screens them based on the financial live broadcast big model to obtain multiple initial audience questions, and arranges them in order of timestamps to obtain a first question queue. The financial live broadcast big model can be used to screen out non-question type and potential non-question type comments, thereby improving the overall information entropy of the first question queue, thereby reducing the difficulty of subsequent processing and improving processing efficiency when obtaining the first audience question according to the first question queue for subsequent processing.

[0067] In certain embodiments of the present application, rewriting the first audience question according to the financial live broadcast model to obtain the first digital human question specifically includes:

[0068] According to the financial live broadcast model, rewrite the first audience question to obtain the second audience question;

[0069] According to the first audience name corresponding to the first audience question, combined with multiple preset question prefixes and multiple preset question suffixes, the second audience question is updated to obtain the first digital human question; wherein, the first digital human question includes a question prefix, the first audience name, a question suffix and the second audience question.

[0070] In certain embodiments of the present application, the first audience question is rewritten according to the financial live broadcast model to obtain the second audience question, specifically:

[0071] Based on the financial live broadcast model, the prompt words are rewritten according to the preset questions, and the first audience question is rewritten to obtain the second audience question.

[0072] Specifically, the question prefix is ​​used to connect the previous question with the current question, and the question suffix is ​​used to connect the first viewer's name with the current question.

[0073] This application first rewrites the first audience's question according to the financial live broadcast model to obtain the second audience's question, and can rewrite the audience's original question that is not suitable for digital people to ask into a question that is suitable for digital people to ask, thereby improving the professionalism of the question. Then, according to the first audience name corresponding to the first audience's question, combined with a preset question prefix and a preset question suffix, the second audience's question is updated to obtain the first digital person's question, and the second audience's question can be processed into a first digital person question that is more suitable for the digital person's question format, thereby reducing the interaction cost when the digital person communicates with the anchor.

[0074] Step S102: Generate a digital human live stream based on the first digital human problem and in combination with a preset digital human.

[0075] In certain embodiments of the present application, the digital human embodiment includes any one of a 2D digital human and a 3D digital human, and the preferred embodiment is a 2D digital human.

[0076] In certain implementations of the present application, generating a digital human live stream based on the first digital human problem and in combination with a preset digital human specifically includes:

[0077] The digital human live stream includes a digital human audio live stream and a digital human image live stream;

[0078] Based on the streaming speech synthesis technology, the first digital human question is converted into audio to obtain a digital human audio live stream;

[0079] According to the digital human audio live stream, the facial movements of the digital human are controlled to generate a digital human screen live stream.

[0080] In certain embodiments of the present application, the streaming speech synthesis technology is based on a streaming speech synthesis model, and a preferred implementation of the streaming speech synthesis model is a Cosy Voice model.

[0081] In certain embodiments of the present application, the digital human's facial movements are controlled according to the digital human audio live stream to generate a digital human screen live stream, specifically:

[0082] According to the digital human audio live stream, based on a preset audio-driven action model, the digital human's facial movements are controlled, and combined with preset cyclic body movements, a digital human screen live stream is generated.

[0083] In certain embodiments of the present application, embodiments of the audio-driven action model include but are not limited to a Video ReTail king model and an EMAGE model, and a preferred embodiment is the EMAGE model.

[0084] This application first converts the first digital human question into audio based on streaming speech synthesis technology to obtain a digital human audio live stream, and then controls the digital human's facial movements according to the digital human audio live stream to generate a digital human picture live stream, which can respectively obtain the audio and picture of the digital human who is asking questions suitable for the current live broadcast room, thereby attracting audience interaction through the digital human's audio stream and picture stream, and increasing the audience's willingness to interact, thereby improving the interactivity of the digital human live interaction in financial scenarios.

[0085] Step S103: According to the digital human live stream, the digital human is controlled to ask questions to the anchor in the current live broadcast room, and the anchor live stream generated by the anchor answering the first digital human question is obtained.

[0086] Step S104: merging the anchor live stream and the digital human live stream to obtain the live stream of the live broadcast room, and pushing the live stream of the live broadcast room to the current live broadcast room.

[0087] In certain implementations of the present application, the merging of the anchor live stream and the digital human live stream to obtain the live stream of the live broadcast room specifically includes:

[0088] Extract the audio output and the picture output of the anchor live stream respectively to obtain the anchor audio live stream and the anchor picture live stream;

[0089] According to the preset sound quality parameters, the host audio live stream and the digital human audio live stream are superimposed to obtain the live studio audio live stream;

[0090] Based on the preset screen layout, the digital human screen live stream is superimposed on the anchor screen live stream in the form of a mask to obtain the live studio screen live stream;

[0091] The live broadcast room audio live stream and the live broadcast room picture live stream are integrated to obtain the live broadcast room live stream.

[0092] This application first extracts the audio output and picture output of the host's live stream respectively to obtain the host's audio stream and picture stream, and then superimposes the host's audio stream and the digital human's audio stream according to preset sound quality parameters to obtain the live room audio stream, and superimposes the digital human's picture stream on the host's picture stream in the form of a mask based on the preset picture layout to obtain the live room picture stream. The audio stream and the picture stream can be processed separately to avoid errors caused by simultaneous processing and reduce the error probability, and then superimpose the audio stream and the picture stream of the live room to obtain the live stream of the live room. The host's live stream and the digital human's live stream can be presented to the audience at the same time, thereby increasing the audience's willingness to interact, further realizing the live interaction between digital humans, hosts and audiences, and improving the interactivity of the live interaction of digital humans in financial scenarios.

[0093] Compared with the prior art, this application first rewrites the first audience question in the current live broadcast room through the financial live broadcast model to obtain the first digital human question, which can rewrite the audience's original question that is not suitable for digital human questions into a question suitable for digital human questions. While obtaining questions through audience questions, the professionalism of the questions is improved and the interaction cost is reduced; then the digital human live broadcast stream is generated, and the digital human is controlled to ask questions to the anchor according to the digital human live broadcast stream, and the anchor live broadcast stream generated by the anchor's answer is obtained, and then the anchor live broadcast stream and the digital human live broadcast stream are merged to obtain the live broadcast room live stream and pushed to the current live broadcast room, and the anchor live stream and the digital human live stream can be presented to the audience at the same time, improving the audience's willingness to interact. Compared with the existing digital human live broadcast's reliance on the matching of the preset question and answer library, this application can obtain the audience questions in the current live broadcast room and rewrite them, and then the digital human asks questions and the anchor answers, and the question and answer process is integrated and presented to the audience, which can realize real-time interaction with audience comments and improve the interactivity of digital human live broadcast interaction in financial scenarios.

[0094] Corresponding to the previous method, see Figure 2 , the embodiment of the present application provides a digital human live interactive device in a financial scenario, including a digital human question acquisition module 210, a digital human live stream generation module 220, a digital human questioning module 230 and a live room live stream generation module 240;

[0095] The digital human question acquisition module 210 is used to acquire the first viewer's question in the current live broadcast room, and rewrite the first viewer's question according to the financial live broadcast model to obtain the first digital human question; wherein the financial live broadcast model is trained based on the collected financial scenario data;

[0096] The digital human live stream generation module 220 is used to generate a digital human live stream according to the first digital human problem and in combination with a preset digital human;

[0097] The digital human questioning module 230 is used to control the digital human to ask questions to the anchor in the current live broadcast room according to the digital human live broadcast stream, and obtain the anchor live broadcast stream generated by the anchor answering the first digital human question;

[0098] The live broadcast room live stream generation module 240 is used to merge the anchor live stream and the digital human live stream to obtain the live broadcast room live stream, and push the live broadcast room live stream to the current live broadcast room.

[0099] In certain embodiments of the present application, the digital human question acquisition module 210 includes an audience comment acquisition unit, an audience comment screening unit, and an audience question acquisition unit;

[0100] The audience comment acquisition unit is used to acquire a plurality of initial audience comments in the current live broadcast room;

[0101] The audience comment screening unit is used to screen the multiple initial audience comments based on the financial live broadcast model to obtain multiple initial audience questions;

[0102] The audience question acquisition unit is used to arrange the multiple initial audience questions according to the time stamp sequence to obtain a first question queue, and take the first initial audience question at the head of the first question queue as the first audience question.

[0103] In certain embodiments of the present application, the digital human question acquisition module 210 includes an audience question rewriting unit and a digital human question acquisition unit;

[0104] The audience question rewriting unit is used to rewrite the first audience question according to the financial live broadcast model to obtain a second audience question;

[0105] The digital human question acquisition unit is used to update the second audience question according to the first audience name corresponding to the first audience question, combined with multiple preset question prefixes and multiple preset question suffixes, to obtain the first digital human question; wherein the first digital human question includes a question prefix, the first audience name, a question suffix and the second audience question.

[0106] In some embodiments of the present application, the digital human live stream includes a digital human audio live stream and a digital human image live stream, and the digital human live stream generation module 220 includes a digital human audio stream generation unit and a digital human image stream generation unit;

[0107] The digital human audio stream generating unit is used to convert the first digital human question into audio based on the streaming speech synthesis technology to obtain a digital human audio live stream;

[0108] The digital human picture stream generating unit is used to control the facial movements of the digital human according to the digital human audio live stream to generate a digital human picture live stream.

[0109] In some embodiments of the present application, the live broadcast room live stream generation module 240 includes a live stream extraction unit, an audio stream superposition unit, a picture stream superposition unit and a live stream integration unit;

[0110] The live stream extraction unit is used to extract the audio output and the picture output of the anchor live stream respectively to obtain the anchor audio live stream and the anchor picture live stream;

[0111] The audio stream superposition unit is used to superimpose the host audio live stream and the digital human audio live stream according to preset sound quality parameters to obtain the live audio stream of the live broadcast room;

[0112] The picture stream superimposing unit is used to superimpose the digital human picture live stream onto the anchor picture live stream in a mask form based on a preset picture layout to obtain a live studio picture live stream;

[0113] The live stream integration unit is used to integrate the live audio stream of the live broadcast room and the live picture stream of the live broadcast room to obtain the live broadcast stream of the live broadcast room.

[0114] This application first rewrites the first audience question in the current live broadcast room through the financial live broadcast model to obtain the first digital human question, which can rewrite the audience's original question that is not suitable for digital human questions into a question suitable for digital human questions. While obtaining questions through audience questions, it improves the professionalism of the questions and reduces the interaction cost; regenerates the digital human live stream, and controls the digital human to ask questions to the anchor according to the digital human live stream, and obtains the anchor live stream generated by the anchor's answer, and then merges the anchor live stream and the digital human live stream to obtain the live broadcast room live stream and push it to the current live broadcast room, which can present the anchor live stream and the digital human live stream to the audience at the same time, improving the audience's willingness to interact. Compared with the existing digital human live broadcast that relies on the matching of the preset question and answer library, this application can obtain the audience questions in the current live broadcast room and rewrite them, and then the digital human asks questions and the anchor answers, and integrates this question and answer process to present to the audience, which can realize real-time interaction with audience comments, and improve the interactivity of digital human live interaction in financial scenarios.

[0115] It should be understood that the device provided in the embodiment of the present application corresponds to the aforementioned method, and the digital human live broadcast interaction device in a financial scenario provided in the embodiment of the present application can implement the digital human live broadcast interaction method in a financial scenario provided in any embodiment of the present application.

[0116] Adaptively, the embodiments of the present application also provide a computer device and a computer-readable storage medium.

[0117] The computer device comprises: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor;

[0118] Among them, when the processor executes the computer program, a digital human live broadcast interaction method in a financial scenario of the present application is implemented.

[0119] The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor to execute a digital human live broadcast interaction method in a financial scenario of the present application.

[0120] The above are some implementation methods of the present application, and the purpose, technical scheme and beneficial effects of the present application are further described in detail. It should be clear that the above-mentioned partial implementation methods of the present application cannot be understood as limiting the present application. It is particularly pointed out that for those skilled in the art, any changes, modifications, equivalent substitutions and variations made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A digital human live interactive method in a financial scenario, characterized in that: include: Obtain the first audience question in the current live broadcast room, and rewrite the first audience question according to the financial live broadcast model to obtain the first digital human question; wherein the financial live broadcast model is trained based on the collected financial scenario data; According to the first digital human problem, combined with the preset digital human, a digital human live stream is generated; According to the digital human live stream, controlling the digital human to ask questions to the anchor in the current live broadcast room, and obtaining the anchor live stream generated by the anchor answering the first digital human question; The anchor live stream and the digital human live stream are combined to obtain the live stream of the live broadcast room, and the live stream of the live broadcast room is pushed to the current live broadcast room.

2. According to the method for digital human live broadcast interaction in a financial scenario according to claim 1, it is characterized in that: The obtaining of the first audience question in the current live broadcast room specifically includes: Get multiple initial audience comments in the current live broadcast room; Based on the financial live broadcast model, the plurality of initial audience comments are screened to obtain a plurality of initial audience questions; The multiple initial audience questions are arranged according to the order of timestamps to obtain a first question queue, and the first initial audience question at the head of the first question queue is taken as the first audience question.

3. The digital human live interactive method in a financial scenario according to claim 1, characterized in that: The first audience question is rewritten according to the financial live broadcast model to obtain the first digital human question, specifically including: According to the financial live broadcast model, rewrite the first audience question to obtain the second audience question; According to the first audience name corresponding to the first audience question, combined with multiple preset question prefixes and multiple preset question suffixes, the second audience question is updated to obtain the first digital human question; wherein, the first digital human question includes a question prefix, the first audience name, a question suffix and the second audience question.

4. The digital human live interactive method in a financial scenario according to claim 1, characterized in that: The step of generating a digital human live stream based on the first digital human problem and in combination with a preset digital human specifically includes: The digital human live stream includes a digital human audio live stream and a digital human image live stream; Based on the streaming speech synthesis technology, the first digital human question is converted into audio to obtain a digital human audio live stream; According to the digital human audio live stream, the facial movements of the digital human are controlled to generate a digital human screen live stream.

5. The digital human live broadcast interaction method in a financial scenario according to claim 4, characterized in that: The merging of the anchor live stream and the digital human live stream to obtain the live stream of the live broadcast room specifically includes: Extract the audio output and the picture output of the anchor live stream respectively to obtain the anchor audio live stream and the anchor picture live stream; According to the preset sound quality parameters, the host audio live stream and the digital human audio live stream are superimposed to obtain the live studio audio live stream; Based on the preset screen layout, the digital human screen live stream is superimposed on the anchor screen live stream in the form of a mask to obtain the live studio screen live stream; The live broadcast room audio live stream and the live broadcast room picture live stream are integrated to obtain the live broadcast room live stream.

6. A digital human live interactive device in a financial scenario, characterized in that: It includes a digital human question acquisition module, a digital human live stream generation module, a digital human questioning module and a live stream generation module in the live broadcast room; The digital human question acquisition module is used to acquire the first viewer's question in the current live broadcast room, and rewrite the first viewer's question according to the financial live broadcast model to obtain the first digital human question; wherein the financial live broadcast model is trained based on the collected financial scenario data; The digital human live stream generation module is used to generate a digital human live stream according to the first digital human question and in combination with a preset digital human; The digital human questioning module is used to control the digital human to ask questions to the anchor in the current live broadcast room according to the digital human live broadcast stream, and obtain the anchor live broadcast stream generated by the anchor answering the first digital human question; The live broadcast room live stream generation module is used to merge the anchor live stream and the digital human live stream to obtain the live broadcast room live stream, and push the live broadcast room live stream to the current live broadcast room.

7. The digital human live interactive device in a financial scenario according to claim 6, characterized in that: The digital human question acquisition module includes an audience comment acquisition unit, an audience comment screening unit and an audience question acquisition unit; The audience comment acquisition unit is used to acquire a plurality of initial audience comments in the current live broadcast room; The audience comment screening unit is used to screen the multiple initial audience comments based on the financial live broadcast model to obtain multiple initial audience questions; The audience question acquisition unit is used to arrange the multiple initial audience questions according to the time stamp sequence to obtain a first question queue, and take the first initial audience question at the head of the first question queue as the first audience question.

8. The digital human live interactive device in a financial scenario according to claim 6, characterized in that: The digital human question acquisition module includes an audience question rewriting unit and a digital human question acquisition unit; The audience question rewriting unit is used to rewrite the first audience question according to the financial live broadcast model to obtain a second audience question; The digital human question acquisition unit is used to update the second audience question according to the first audience name corresponding to the first audience question, combined with multiple preset question prefixes and multiple preset question suffixes, to obtain the first digital human question; wherein the first digital human question includes a question prefix, the first audience name, a question suffix and the second audience question.

9. The digital human live interactive device in a financial scenario according to claim 6, characterized in that: The digital human live stream includes a digital human audio live stream and a digital human image live stream, and the digital human live stream generation module includes a digital human audio stream generation unit and a digital human image stream generation unit; The digital human audio stream generating unit is used to convert the first digital human question into audio based on the streaming speech synthesis technology to obtain a digital human audio live stream; The digital human picture stream generating unit is used to control the facial movements of the digital human according to the digital human audio live stream to generate a digital human picture live stream.

10. The digital human live interactive device in a financial scenario according to claim 9, characterized in that: The live broadcast room live stream generation module includes a live stream extraction unit, an audio stream superposition unit, a picture stream superposition unit and a live stream integration unit; The live stream extraction unit is used to extract the audio output and the picture output of the anchor live stream respectively to obtain the anchor audio live stream and the anchor picture live stream; The audio stream superposition unit is used to superimpose the host audio live stream and the digital human audio live stream according to preset sound quality parameters to obtain the live audio stream of the live broadcast room; The picture stream superimposing unit is used to superimpose the digital human picture live stream onto the anchor picture live stream in a mask form based on a preset picture layout to obtain a live studio picture live stream; The live stream integration unit is used to integrate the live audio stream of the live broadcast room and the live picture stream of the live broadcast room to obtain the live broadcast stream of the live broadcast room.

Images