基于谐波结构的民族乐器音色特征提取  

Extraction of timbre features of national Musical Instruments based on harmonic structure

作　　者：牛育谦 杨艺媛[1] NIU Yuqian;YANG Yiyuan(Xi’an University,Xi’an 710065,China)

机构地区：[1]西安文理学院,西安710065

出　　处：《自动化与仪器仪表》2023年第4期34-38,共5页Automation & Instrumentation

摘　　要：针对民族乐器音色特征提取准确率问题,提出谐波特征提取结合支持向量机的音色分类识别方法。其中,考虑到谐波结构对音色的重要性,提出离散谐波变换的音色谐波提取方法,以此构建音色表达谱;然后算融合LPCC、MPCC等特征作为SVM的输入,最终实现不同民族乐器的分类识别。结果表明,在识别准确率上,组合特征的识别率明显高于单一特征输入;在单音色识别上,本方法在96%左右,高于KNN等其他方法;在音乐片段识别方面,对全部乐器的识别准确率和针对音乐片段数据库的测度平均分别提高了7.3%和2.23%,识别效果更好。Aiming at the problem of accuracy in extracting the timbre features of ethnic musical instruments,a timbre classification and recognition method based on harmonic feature extraction and support vector machines is proposed.Considering the importance of harmonic structure to timbre,a discrete harmonic transform method for extracting timbre harmonics is proposed to construct a timbre expression spectrum;Then,LPCC,MPCC and other features are fused as input to SVM to ultimately achieve the classification and recognition of different ethnic musical instruments.The results show that in terms of recognition accuracy,the recognition rate of combined features is significantly higher than that of single feature input;In single tone recognition,this method is around 96%higher than other methods such as KNN;In terms of music segment recognition,the recognition accuracy for all musical instruments and the measurement for the music segment database have improved by an average of 7.3%and 2.23%,respectively,with better recognition results.

关 键 词：谐波结构 特征提取 支持向量机 民族乐器 分类识别 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]