基于音色模型的多基频估计方法  

Multiple fundamental frequency estimation based on timbre model

作　　者：石立新[1] 张俊星[1] 

机构地区：[1]大连民族学院机电信息工程学院,辽宁大连116600

出　　处：《计算机工程与设计》2010年第13期3066-3069,共4页Computer Engineering and Design

摘　　要：为了有效解决复调音乐泛音重叠问题,提出了一种基于音色模型的多基频估计方法。分析了音色的时频特征,指出衰退和维持阶段的谐波结构稳定可靠、适合基频估计。根据谐波匹配率和强度筛选候选基频,采用频谱迭代删除的方法确定确切基频。依据最小谐波匹配率删减谱线,应用比例删除机制解决泛音重叠问题。仿真实验和实际乐曲测试结果表明,该方法精度高,可广泛应用于具有稳定谐波结构的音乐信号分析之中。To solve the partial overlapping problem in polyphonic music effectively,a multiple-f0 estimation method based on the timbre model is proposed.After the time-frequency feature of the timbre is analyzed,it is pointed out that the harmonic structure in delay and sustain state is stable and suitable for fundamental frequency estimation.The candidate f0 is selected according to the harmonic matching rate and the strength,and the exact f0 is determined using the spectrum iterative subtraction method.The spectral line is subtracted based on the minimum harmonic matching rate,and the partial overlapping problem is solved applying the proportional subtraction mechanism.The simulation experiments and actual music testing show that this algorithm is accurate,and can be applied widely to music signal analysis with stable harmonic structure.

关 键 词：多基频估计 音色模型 泛音重叠 谐波匹配 比例删除 

分 类 号：TP391.42[自动化与计算机技术—计算机应用技术]