Scholarlink

<p>Existing music score generation methods are Limited by the scene Limitations and the quality of their generated scores is relatively Limited. To address these problems, an intelligent music score generation method combining short-time Fourier transform and improved convolutional neural network is proposed. The study firstly utilizes short-time Fourier transform to transform the time-frequency of music signals, and then inputs the transformed time-frequency information into an improved convolutional neural network model. The model improves the accuracy and diversity of music score generation by introducing label enhancement strategy and internal convolution structure. The method may effectively increase the quality of music score creation on various music datasets with strong generalization ability, according to the experimental results. The matching rate and complete rate of the generated score of the proposed method were 92% and 95%, respectively, and its score generation time was only 1.05s. The proposed method could improve the efficiency and quality of the music score generation. The intelligent music score generation method can help the drum learners understand their own performance in time, and give feedback on their training to improve the learning efficiency.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Intelligent generation method of drum music scores based on improved CNN and STFT

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月9日

查看全文
  • Yuting Ni

摘要

Existing music score generation methods are Limited by the scene Limitations and the quality of their generated scores is relatively Limited. To address these problems, an intelligent music score generation method combining short-time Fourier transform and improved convolutional neural network is proposed. The study firstly utilizes short-time Fourier transform to transform the time-frequency of music signals, and then inputs the transformed time-frequency information into an improved convolutional neural network model. The model improves the accuracy and diversity of music score generation by introducing label enhancement strategy and internal convolution structure. The method may effectively increase the quality of music score creation on various music datasets with strong generalization ability, according to the experimental results. The matching rate and complete rate of the generated score of the proposed method were 92% and 95%, respectively, and its score generation time was only 1.05s. The proposed method could improve the efficiency and quality of the music score generation. The intelligent music score generation method can help the drum learners understand their own performance in time, and give feedback on their training to improve the learning efficiency.