Voice data processing method of AIGC model

Abstract

The invention relates to a voice data processing method of an AIGC model applied to the technical field of artificial intelligence, and the method comprises the steps: recording a plurality of pieces of voice information of a plurality of persons through a plurality of voice recording end modules at the same time, distinguishing voices according to sounding persons through a voice dividing module, and then carrying out the content extraction of the speech of each person through a content extraction module, according to the sound production characteristics of each participating member, the relative position of the sound receiving ball and the directional outer cover is adjusted through telescoping of the telescopic pipe, so that the state of the elastic sound adjusting pipe is adjusted to adapt to the participating members with different sound production volumes, the voice distinguishing effect between different members is effectively improved, and the voice analysis quality is effectively improved; and the content summarization module summarizes the speaking voice content of each person through the content obtained by the content extraction module to obtain a final discussion structure copywriting, so that summarization communication of the voice content of multiple persons is facilitated, the quality of data input into the AIGC model is effectively improved, and the quality of generated works is effectively improved.

Description

Technical Field

[0001] The present invention relates to a voice data processing method, and in particular to a voice data processing method of an AIGC model applied in the field of artificial intelligence technology. Background Art

[0002] The AIGC model is an artificial intelligence content generation model. It represents the use of artificial intelligence technology to generate various types of content, realize the efficient conversion between people and AI, and then produce an artificial intelligence content generation model. This model is applied to text, images, audio, and video. The AIGC model requires a large amount of training data and advanced algorithms. These contents are not only diverse in form, but also require a large amount of data for training and learning the AIGC model.

[0003] The specification of Chinese invention patent CN202410324909.6 discloses "A method and system for processing speech data of AIGC model", which performs dynamic spectrum mapping processing on the original speech data to match the preset AIGC model; divides the processed speech data into multiple speech samples, and extracts the acoustic features of each speech sample; calculates the entropy value of the extracted acoustic features to obtain the entropy voiceprint; classifies the speech samples according to the entropy voiceprint, and generates the target speech content through the preset AIGC model after the classified speech samples. It can analyze and process speech data more comprehensively, improve the multidimensionality and diversity of acoustic feature extraction, and enhance the naturalness and accuracy of speech generated by the AIGC model.

[0004] Since the quality of works generated by the AIGC model depends on the quality of the input data, the quality of data provided by a single person is limited. Therefore, in order to improve the quality of AIGC model AI-generated works, multiple people are needed to participate in data compilation. However, the existing AIGC model voice data processing is not suitable for multi-person discussion, which hinders the promotion of AIGC technology in more application scenarios and hinders the promotion of AIGC technology in group application scenarios. Summary of the invention

[0005] In view of the above-mentioned prior art, the technical problem to be solved by the present invention is that the existing AIGC model voice data processing is not suitable for multi-person discussion, which hinders the promotion of AIGC technology in more application scenarios and hinders the promotion of AIGC technology in group application scenarios.

[0006] To solve the above problems, the present invention provides a voice data processing method of an AIGC model, comprising a system processor for voice data processing, wherein the input end of the system processor is connected to a voice collection end module, the output end of the system processor is connected to a voice analysis module, the voice analysis module internally comprises a voice division module and a content extraction module, the output end of the voice analysis module is connected to a content summary module, and the output end of the content summary module is connected to a data output module;

[0007] The specific processing steps include:

[0008] S1: The voice data is collected through the voice collection module, and the voice content is pre-processed during the input process;

[0009] S2: The speech analysis module analyzes the recorded speech content. First, the speech classification module is used to distinguish the speakers according to their voices. Then, the content extraction module extracts the content of each speaker's speech to obtain the sentence text.

[0010] S3: The content summarization module uses the content obtained by the content extraction module to summarize the speech content of each person and obtain the final discussion structure copy. The data output module then processes the copy into code data that adapts to the AIGC model.

[0011] The voice recording end module is composed of a microphone body, the top of the microphone body is rotatably connected to a sound receiving sphere, the outside of the sound receiving sphere is fixedly covered with a directional outer cover, the open end of the directional outer cover is fixedly connected to an elastic sound tuning tube, the elastic sound tuning tube is sleeved on the sound receiving sphere, and a telescopic tube is fixedly connected between the end of the directional outer cover away from the elastic sound tuning tube and the sound receiving sphere.

[0012] In the voice data processing method of the above-mentioned AIGC model, multiple voice information of multiple people are collected at the same time through the voice collection end module, and the voice content is divided according to the individual pronunciation characteristics, which is convenient for summarizing and communicating the voice content of multiple people, effectively improving the data quality of the input AIGC model, thereby effectively improving the quality of the generated works.

[0013] As a further improvement of the present application, the preprocessing steps of the voice data recording module in the voice recording process also include: eliminating noise and background noise from the voice data, improving the clarity of the voice content, effectively improving the quality of the recorded voice, and facilitating the system's analysis and processing of the voice.

[0014] As a further improvement of the present application, the voice division module includes a volume comparison unit, a timbre comparison unit, a pitch comparison unit and a rhythm comparison unit, which distinguishes the voice information by recording the volume, timbre, pitch and speaking rhythm of the voice, thereby effectively improving the effect of distinguishing the voice information of different people.

[0015] As a further improvement of the present application, the directional outer cover is made of sound insulation material, and the elastic sound-adjusting tube is made of elastic rubber sound insulation material, which effectively improves the sound insulation effect of the directional outer cover in directions other than the collecting direction of the sound-receiving sphere, and effectively reduces the impact of other people's pronunciation on the voice recording effect.

[0016] As another improvement of the present application, the outer surface of the elastic sound-tuning tube is fixedly inlaid with supporting ribs made of elastic plastic, one end of the supporting ribs is fixedly connected to the directional outer cover, and the other end of the supporting ribs is arranged in an arc shape corresponding to the surface of the sound-receiving sphere. The supporting ribs are utilized to effectively improve the elastic deformation recovery ability of the elastic sound-tuning tube, and at the same time, the arc-shaped supporting ribs can be easily stretched by the pressure of the sound-receiving sphere, thereby facilitating the adjustment of the sound-receiving ability of the sound-receiving sphere to adapt to members with different speaking volumes, thereby effectively improving the quality of voice recording.

[0017] As another improved supplement of the present application, the content summary module also includes a real-time display module and a content ambiguity module. The real-time display module is used to display the text of the voice content in real time, and the real-time display module is selected as a projector device and a display screen device. The content ambiguity module is used to mark the ambiguous sentence description information in the real-time copy content. The real-time display module is used to display the results of the discussion in real time, so that the participating members can understand the progress of the results in real time. At the same time, the content ambiguity module marks the ambiguities that arise during the summary process, so that the participating members can discuss and resolve them.

[0018] As another improved supplement of the present application, the input end of the content ambiguity module is connected to an interaction module, which is used to modify and add to the ambiguous content. One of the members participating in the discussion takes the lead and uses the interaction module to modify and resolve the ambiguity, thereby facilitating the advancement of voice data processing.

[0019] As another improvement of the present application, the interaction module is a tablet computer, and the interaction module is connected to any voice recording terminal module via Bluetooth. After one of the members participating in the discussion obtains the interaction module, it is connected to the corresponding voice recording terminal module via Bluetooth, so that the participating member can take the lead in the discussion, which facilitates the transfer of the discussion leader among the participating members, thereby facilitating the members who are good at handling their work.

[0020] In summary, the present invention collects multiple voice information of multiple people through multiple voice collection end modules at the same time, uses the voice classification module to distinguish the voice according to the speaker, and then the content extraction module extracts the content of each person's speech, and then according to the voice characteristics of each participating member, the relative position of the sound-receiving sphere and the directional outer cover is adjusted by telescopic tube extension, so as to adjust the state of the elastic sound-adjusting tube to adapt to the participating members with different voice volumes, thereby effectively improving the voice distinction effect between different members and the quality of voice analysis. The content summary module summarizes the speech content of each person through the content obtained by the content extraction module to obtain the final discussion structure copy, which is convenient for summarizing and exchanging the voice content of multiple people, effectively improving the data quality of input AIGC model, and thus effectively improving the quality of generated works. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a demonstration diagram of the first implementation mode of this application;

[0022] Figure 2 This is a system diagram of the first implementation mode of the present application;

[0023] Figure 3 This is a flowchart of the steps of the first implementation method of this application;

[0024] Figure 4 This is a system diagram of a voice division module in the first embodiment of the present application;

[0025] Figure 5 This is a three-dimensional structural diagram of a microphone body according to a first embodiment of the present application;

[0026] Figure 6 This is a demonstration diagram of the expansion change of the elastic sound-adjusting tube in the first embodiment of the present application;

[0027] Figure 7 This is a demonstration diagram of the contraction change of the elastic sound-tuning tube in the first embodiment of the present application;

[0028] Figure 8 This is a system diagram of the content summary module of the second implementation method of the present application.

[0029] Explanation of the numbers in the figure: 1. Voice recording terminal module; 101. Microphone body; 102. Sound receiving sphere; 103. Directional outer cover; 104. Elastic sound-adjusting tube; 105. Telescopic tube; 106. Support ribs; 2. Real-time display module; 3. Interaction module. DETAILED DESCRIPTION

[0030] Two implementation modes of the present application are described in detail below with reference to the accompanying drawings.

[0031] The first implementation method:

[0032] Figure 1-4 It is shown that a voice data processing method of an AIGC model includes a system processor for voice data processing, the input end of the system processor is connected to a voice collection end module 1, the output end of the system processor is connected to a voice analysis module, the voice analysis module includes a voice division module and a content extraction module, the output end of the voice analysis module is connected to a content summary module, and the output end of the content summary module is connected to a data output module;

[0033] The specific processing steps include:

[0034] S1: recording voice data through voice recording module 1, and preprocessing the voice content during the input process;

[0035] S2: The speech analysis module analyzes the recorded speech content. The speech segmentation module is used to distinguish the speech according to the speaker. Then, the content extraction module extracts the content of each speaker's speech to obtain the sentence text.

[0036] S3: The content summarization module uses the content obtained by the content extraction module to summarize the speech content of each person and obtain the final discussion structure copy. The data output module then processes the copy into code data that adapts to the AIGC model.

[0037] The pre-processing steps of the voice data collection process of the voice collection end module 1 also include: eliminating noise and background noise from the voice data, improving the clarity of the voice content, effectively improving the quality of the collected voice, and facilitating the system to analyze and process the voice. The voice classification module includes a volume comparison unit, a timbre comparison unit, a pitch comparison unit, and a rhythm comparison unit. The voice information is distinguished by the volume, timbre, pitch, and speaking rhythm of the collected voice, thereby effectively improving the effect of distinguishing the voice information of different people;

[0038] Before inputting data to the AIGC model, the theme of the AI-generated work is first determined, and the participating members express their opinions based on the theme. The opinions are collected in the form of voice by the voice collection terminal module 1, and the collected voice is analyzed by the voice analysis module. The voice division module is first used to distinguish the voices of people according to the pronunciation, that is, each member has a voice representing himself. When the number of members is more than the voice collection terminal module 1, multiple participating members can be arranged to share one voice collection terminal module 1. The content extraction module extracts the content of each person's speech. The content summary module summarizes the content of each person's speech through the content obtained by the content extraction module to obtain the final discussion structure copy. The data output module then processes the copy into code data that adapts to the AIGC model, which is convenient for summarizing and exchanging the voice content of multiple people, effectively improving the data quality of the input AIGC model, and thus effectively improving the quality of the generated works.

[0039] Figure 5-7 As shown, the voice recording end module 1 is composed of a microphone body 101, the top of the microphone body 101 is rotatably connected to a sound collecting sphere 102, the outside of the sound collecting sphere 102 is fixedly covered with a directional outer cover 103, the open end of the directional outer cover 103 is fixedly connected to an elastic sound-adjusting tube 104, the elastic sound-adjusting tube 104 is sleeved with the sound collecting sphere 102, and a telescopic tube 105 is fixedly connected between the end of the directional outer cover 103 away from the elastic sound-adjusting tube 104 and the sound collecting sphere 102, the directional outer cover 103 is made of sound insulation material, and the elastic sound-adjusting tube 104 is made of elastic rubber sound insulation material, which effectively improves the directional outer cover 103 to the sound collecting sphere 102. The sound insulation effect in other directions effectively reduces the effect of other people's pronunciation on the voice recording effect. The outer surface of the elastic sound-adjusting tube 104 is fixedly inlaid with a supporting rib 106, which is made of elastic plastic. One end of the supporting rib 106 is fixedly connected to the directional outer cover 103, and the other end of the supporting rib 106 is arranged in an arc shape corresponding to the surface of the sound-receiving sphere 102. The supporting rib 106 is used to effectively improve the elastic deformation recovery ability of the elastic sound-adjusting tube 104. At the same time, the arc-shaped supporting rib 106 is easy to be compressed and stretched by the sound-receiving sphere 102, so as to facilitate the adjustment of the sound-receiving ability of the sound-receiving sphere 102 to adapt to members with different speaking volumes, and effectively improve the voice recording quality;

[0040] When the voice recording end module 1 records the voices of the participating members, after the voice division module makes a detailed analysis of the pronunciation of each member, the voice recording end module 1 makes adjustments according to the pronunciation characteristics of the members. When recording the voices of the members with lower volume, the telescopic tube 105 is extended to adjust the relative position of the sound collecting sphere 102 and the directional outer cover 103, so that the sound collecting sphere 102 is closer to the open end of the directional outer cover 103. At the same time, the sound collecting sphere 102 is inserted into the elastic sound adjusting tube 104 to expand the elastic sound adjusting tube 104 and amplify the voice. The sound recording space is used to record the member with a lower volume in detail. Although the influence of other members' voices will be increased in the process, the voice of the member with a lower volume is also recorded in the most detail. On the contrary, when recording the voice of the member with a louder volume, the sound collecting sphere 102 is away from the elastic tuning tube 104, which reduces the voice recording space. While not affecting the voice recording of the member, the voice influence of other members is reduced, thereby adapting to participating members with different voice volumes, effectively improving the voice distinction effect between different members, and effectively improving the quality of voice analysis.

[0041] The second implementation method:

[0042] Compared with the first embodiment, this embodiment mainly adds a real-time display module 2 and a content ambiguity module. The specific newly added structure is as follows, and the remaining structure is consistent with the first embodiment.

[0043] Figure 1and Figure 8 As shown, the content summary module also includes a real-time display module 2 and a content ambiguity module. The real-time display module 2 is used for real-time display of the text of the voice content, and the real-time display module 2 is selected as a projector device and a display screen device. The content ambiguity module is used to mark the ambiguous sentence description information in the real-time copy content. The real-time display module 2 is used to display the results of the discussion in real time, so that the participating members can understand the progress of the results in real time. At the same time, the content ambiguity module marks the ambiguities that appear in the summary process, so that the participating members can discuss and resolve them. The input end of the content ambiguity module is connected to an interaction module 3, and the interaction module 3 is used to modify and delete the ambiguous content. One of the members participating in the discussion is the leader, and the ambiguities that appear are modified and resolved through the interaction module 3, so as to facilitate the advancement of voice data processing. The interaction module 3 is a tablet computer, and the interaction module 3 is connected to any one of the voice collection end modules 1 by Bluetooth. After one of the members participating in the discussion obtains the interaction module 3, it is connected to the corresponding voice collection end module 1 by Bluetooth, so that the participating member can take the lead in the discussion, so that the discussion leader can be transferred among the participating members, so as to facilitate the members who are good at handling their work.

[0044] When the factors are discussed according to the topic, the results of the discussion are displayed by the real-time display module 2. Figure 1 As shown, the real-time display module 2 in the figure uses a projector device, which uses the projector device to display the discussion results in real time, so that the members participating in the discussion can proceed in an orderly manner according to the progress of the results. In the process of summarizing in the content summary module, the ambiguities that arise during the discussion are marked, so that the participating members can discuss based on the ambiguities, and a member who is good at this problem takes the interaction module 3. The interaction module 3 in the attached figure is a tablet computer. The interaction module 3 is connected to the corresponding voice recording terminal module 1 via Bluetooth connection, so that the ambiguous problems can be modified and resolved, which is convenient for the continued advancement of the discussion results.

[0045] In view of current practical needs, the above-mentioned implementation mode adopted in this application is not limited to the scope of protection. Various changes made within the knowledge scope of technical personnel in this field without departing from the concept of this application still fall within the scope of protection of the present invention.

Claims

1. A speech data processing method of an AIGC model, characterized in that: It comprises a system processor for voice data processing, wherein the input end of the system processor is connected to a voice collection end module (1), the output end of the system processor is connected to a voice analysis module, the voice analysis module contains a voice division module and a content extraction module, the output end of the voice analysis module is connected to a content summary module, and the output end of the content summary module is connected to a data output module; The specific processing steps include: S1: The voice data is collected through the voice collection terminal module (1) and the voice content is pre-processed during the input process; S2: The speech analysis module analyzes the recorded speech content. First, the speech classification module is used to distinguish the speakers according to their voices. Then, the content extraction module extracts the content of each speaker's speech to obtain the sentence text. S3: The content summarization module uses the content obtained by the content extraction module to summarize the speech content of each person and obtain the final discussion structure copy. The data output module then processes the copy into code data that adapts to the AIGC model. The voice recording end module (1) is composed of a microphone body (101), the top end of the microphone body (101) is rotatably connected to a sound receiving sphere (102), the outside of the sound receiving sphere (102) is fixedly covered with a directional outer cover (103), the open end of the directional outer cover (103) is fixedly connected to an elastic sound-adjusting tube (104), the elastic sound-adjusting tube (104) is sleeved with the sound receiving sphere (102), and a telescopic tube (105) is fixedly connected between the end of the directional outer cover (103) away from the elastic sound-adjusting tube (104) and the sound receiving sphere (102).

2. The method for processing speech data of an AIGC model according to claim 1, characterized in that: The pre-processing step of the voice data collection process by the voice collection end module (1) also includes: eliminating noise and background noise from the voice data to improve the clarity of the voice content.

3. The voice data processing method of the AIGC model according to claim 1, characterized in that: The voice division module contains a volume comparison unit, a timbre comparison unit, a pitch comparison unit and a rhythm comparison unit.

4. The speech data processing method of the AIGC model according to claim 1, characterized in that: The directional outer cover (103) is made of sound insulation material, and the elastic sound-adjusting tube (104) is made of elastic rubber sound insulation material.

5. The voice data processing method of the AIGC model according to claim 1, characterized in that: The outer surface of the elastic sound-tuning tube (104) is fixedly inlaid with a supporting rib (106), the supporting rib (106) is made of elastic plastic, one end of the supporting rib (106) is fixedly connected to the directional outer cover (103), and the other end of the supporting rib (106) is arranged in an arc shape corresponding to the surface of the sound-receiving sphere (102).

6. The speech data processing method of the AIGC model according to claim 1, characterized in that: The content summarizing module also includes a real-time display module (2) and a content ambiguity module. The real-time display module (2) is used for real-time display of the text of the speech content, and the real-time display module (2) is selected as a projector device and a display screen device. The content ambiguity module is used to mark ambiguous sentence description information in the real-time text content.

7. The method for processing speech data of an AIGC model according to claim 6, characterized in that: The input end of the content ambiguity module is connected to an interaction module (3), and the interaction module (3) is used to modify the ambiguous content by adding or deleting.

8. The method for processing speech data of an AIGC model according to claim 7, characterized in that: The interaction module (3) is a tablet computer, and the interaction module (3) is connected to any one of the voice recording end modules (1) via Bluetooth.

Images