混合F-MFCC参数与多项集成ML算法的音乐情感分类方法研究

Music Sentiment Classification Method Based on Hybrid F-MFCC Parameters and Multi integrated ML Algorithm

刘丹霞 ¹李西萍路惠捷²

扫码查看

作者信息

1. 空军军医大学基础医学院,陕西西安,710032
2. 空军军医大学军事医学心理学系,陕西西安,710032
折叠

摘要

针对目前音乐情感分类方法存在的特征提取不充分、准确率不高的问题,研究提出了一种改进梅尔频率倒谱系数,以更好提取地音乐情感特征,并结合多项集成机器算法来对音乐情感进行分类.结果表明,改进后的梅尔频率倒谱系数参数对愤怒、高兴、放松、悲伤 4种情感特征的提取准确率分别为 72.5%、66.9%、58.2%、56.3%.研究方法对四种情感的整体分类准确率均高于对比算法,分别达到了 90.3%、89.6%、91.4%、92.5%.实验结果显示,通过结合改进的梅尔频率倒谱系数参数和多项集成机器学习算法,显著提高了音乐情感分类的准确率,为智能音乐推荐和情感分析提供了高效的技术支持.

Abstract

Aiming at the problems of insufficient feature extraction and low accuracy in current music emotion classification methods,a study proposes an improved Mel frequency cepstral coefficient to better extract music emotion features,and combines multiple integrated machine algorithms to classify music emotions.The results showed that the improved Mel frequency cepstral coefficient parameters had extraction accuracies of 72.5%,66.9%,58.2%,and 56.3%for the four emotional features of anger,happiness,relaxation,and sadness,respectively.The overall classification accuracy of the research method for the four emotions is higher than that of the comparison algorithm,reaching 90.3%,89.6%,91.4%,and 92.5%,respectively.The experimental results show that by combining the improved Mel frequency cepstral coefficient parameters and multiple integrated machine learning algorithms,the accuracy of music sentiment classification has been significantly improved,providing efficient technical support for intelligent music recommendation and sentiment analysis.

关键词

F-MFCC/ML/音乐情感分类/特征提取/多头注意力机制

Key words

F-MFCC/ML/Music emotion classification/Feature extraction/Multi head attention mechanism

引用本文复制引用

出版年

2024

现代科学仪器

中国分析测试协会

现代科学仪器

CSTPCD

影响因子：0.329

ISSN：1003-8892

段落导航