Research on automatic composing system based on Bi-GRU network and self-attention mechanism
In order to improve the performance performance of an automatic composition system and to enhance the quality and creativity of automatically generated music,the research refines how musical characteristics are represented and creates a model for autonomous music composition that leverages both a bidirectional gated recurrent unit and a self-attention mechanism to more accu-rately capture the temporal features and structure of music.In the Topi accuracy comparison,the research method achieves higher ac-curacy with a shorter number of training rounds.Compared to the other two schemes,the research method improved the accuracy by 10.26％and 14.56％at 100 training rounds,respectively.In subjective evaluations,the research method received scores that mostly indicated better emotional richness and satisfaction,which were also superior to the other two state-of-the-art methods.Finally,in the comparison of composing time,when the length of the generated phrase is 70s,the research method reduces 7.34 s,18.31 s,23.18 s,and 39.92 s,respectively,compared to the other methods.The experimental results verify the validity of this research,and the study provides a strong support for the development of automated composing systems.
万方数据
维普
