基于CQT的乐器独奏单旋律的提取  

Single melody extraction of instrument solo based on CQT

作　　者：张岩 刘哲 ZHANG Yan;LIU Zhe(Computer and Basic Mathematics Education Department, Shenyang Normal University, Shenyang 110034, China)

机构地区：[1]沈阳师范大学计算机与数学基础教学部,沈阳110034

出　　处：《沈阳师范大学学报（自然科学版）》2021年第5期465-469,共5页Journal of Shenyang Normal University:Natural Science Edition

基　　金：国家社会科学基金资助项目(BLA210217)。

摘　　要：乐器独奏的单旋律是音频信号的高级特征,单旋律的提取可以辅助作曲创作、音准分析和校正、音乐检索和音乐情感分析等。利用CQT模型进行时频分析,可以更好地获得频率分辨率和时间分辨率,从而实现乐器独奏的单旋律提取。具体算法是以十二平均律为依据,固定12个频率分量和常系数16.82,计算不同频率分量所对应的窗长,动态调整分帧的长度,产生分帧长度的Hamming窗函数,形成与音乐音频信号相适应的分析模式。为了充分利用FFT和STFT的时频高效计算,先对每一帧的信号进行FFT,进而提取分帧的基频,并根据基频计算窗长,再利用STFT进行基频和采样点的提取。考虑到偶发噪声对CQT方法的影响,为避免窗长过长或过短的问题,在判断出噪声帧后将其窗长设置为1024,以保证提取出正确的基频和采样点特征。在Matlab环境中,以小提琴独奏乐曲为样本,提取该样本前100帧的基频和采样点等特征,并根据基频和采样点特征,绘制出频率轴和时间轴上乐曲旋律的阶梯图。The single melody of instrument solo is the advanced feature of audio signal.The extraction of single melody can assist composition creation,pitch analysis and correction,music retrieval and music emotion analysis.Using CQT model for time-frequency analysis can obtain better frequency resolution and time resolution to achieve single melody extraction of instrument solo.The specific algorithm is programmed based on the twelve average law,fixed 12 frequency components and constant coefficient 16.82,calculated the window length corresponding to different frequency components,dynamically adjusted the length of sub frame,generated Hamming window function of sub frame length,formed the analysis mode suitable for music audio signal.In order to make full use of the efficient time-frequency calculation of FFT and STFT,FFT is used to extract the fundamental frequency of each frame,and the window length is calculated according to the fundamental frequency,and then STFT is used to extract the fundamental frequency and sampling points.Considering the influence of occasional noise on CQT method,in order to avoid the problem of too long or too short window length,after judging the noise frame,the window length is set to 1024 to ensure the extraction of correct fundamental frequency and sampling point features.In the MATLAB environment,taking the violin solo music as the sample,the fundamental frequency and sampling points of the first 100 frames of the sample are extracted.And according to these characteristics,the ladder diagram of the music melody on the frequency axis and time axis is drawn.

关 键 词：CQT 基频 STFT 十二平均律 旋律 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]