Scholarlink

<p>The classification and identification of music genres hold significant importance in music research, education, and performance. The Guzheng, a typical traditional Chinese ethnic instrument, often faces challenges in its traditional genre classification methods, which are generally subjective, costly, and inefficient. To address these issues, this paper proposes a method for accurate classification and identification of Guzheng genres using deep learning. This method involves converting Guzheng music MP3 audio files into MFCC spectrograms to construct a dataset, followed by training and testing using a CNN network combined with an attention mechanism. After identifying and extracting features, precise classification of Guzheng genres is achieved. The results show that the average recognition rate of the proposed method reaches 95.4291%, with all evaluation metrics being robust, effectively demonstrating the accuracy and effectiveness of this classification and identification method.</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

Classification of Chinese Guzheng genres based on CNN with attention mechanism

  • 期刊论文

  • 发表日期：

    2025年8月21日

查看全文
  • Miao Xiaorui

摘要

The classification and identification of music genres hold significant importance in music research, education, and performance. The Guzheng, a typical traditional Chinese ethnic instrument, often faces challenges in its traditional genre classification methods, which are generally subjective, costly, and inefficient. To address these issues, this paper proposes a method for accurate classification and identification of Guzheng genres using deep learning. This method involves converting Guzheng music MP3 audio files into MFCC spectrograms to construct a dataset, followed by training and testing using a CNN network combined with an attention mechanism. After identifying and extracting features, precise classification of Guzheng genres is achieved. The results show that the average recognition rate of the proposed method reaches 95.4291%, with all evaluation metrics being robust, effectively demonstrating the accuracy and effectiveness of this classification and identification method.