Song conversion method, device and equipment and medium
A conversion method and song technology, applied in the computer field, can solve the problems of difficult song style conversion and replacement, and achieve the effect of efficient song style conversion
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0041] Embodiment 1 of the present invention provides a song conversion method, figure 1 Show the flowchart of Embodiment 1 of the song conversion method of the present invention, including:
[0042] Step S101: Obtain the word part and the melody part of the first song, encode the word part and the melody part respectively, and obtain the word feature vector and the melody feature vector;
[0043] First get the first song Y that you want to convert the style of r , get the first song Y r Part of X 1 and Kobe X r , respectively encode the part of the word and the part of the song to obtain the word feature vector h 1 and the eigenvector h r , here both the word part and the melody can be regarded as a sequence, and the Transformer model can be used to encode the word part and the melody to obtain the word feature vector h 1 and the eigenvector h r .
[0044] Step S102: Acquiring the style feature vector corresponding to the target conversion genre;
[0045] Select the ...
Embodiment approach
[0058] As a specific implementation manner, the word feature vector and the converted song feature vector are spliced, and after obtaining the second spliced vector, it also includes:
[0059] A harmony degree judgment vector is obtained, and a style conversion harmony degree value is calculated according to the harmony degree judgment vector and the second splicing vector.
[0060] After the vectors are concatenated during model training, obtain the harmony judgment vector v, and use the vector v and the second concatenation vector z r’ Find the cosine value of the included angle to get the harmony value of style conversion. The larger the cos value, the more harmonious the lyrics and music after style conversion. When the harmony of the lyrics and music after the style conversion of the model is guaranteed, the actual style conversion process is already After the trained model is obtained, step S302 is directly performed, and the degree of harmony of lyrics and music can n...
Embodiment 2
[0066] The song conversion device described below and the song conversion method described above can be referred to in correspondence with each other.
[0067] The embodiment of the present invention also provides a song conversion device, Figure 4 A schematic structural view of an embodiment of the song conversion device of the present invention is shown, including:
[0068] Lyrics and music coding module 101, for obtaining the word part and the music part of the first song, described word part and described music part are coded respectively, obtain word feature vector and song feature vector;
[0069] The style acquisition module 102 is used to acquire the style feature vector corresponding to the target conversion genre;
[0070] Conversion vector acquisition module 103, for obtaining the converted song feature vector according to the style feature vector, the word feature vector and the song feature vector;
[0071] The genre conversion module 104 is configured to obtai...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More - R&D
- Intellectual Property
- Life Sciences
- Materials
- Tech Scout
- Unparalleled Data Quality
- Higher Quality Content
- 60% Fewer Hallucinations
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.
© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com



