乘法模型下西洋乐器音色特征

基于乐音的乘法模型分析研究了可用于西洋乐器音色识别的声学特征：谐波能量比、倒谱与Mel倒谱系数（MFCC）和该文提出的机制声MFCC。对西洋管弦乐器的单音进行了特征提取,通过分析乐器特征的模式,评价它们与音色的关联程度。音色由激励源和共振体共同决定,与共振体特性相比,激励源的特征更能反映音色的不同。研究结果表明：强调激励源本质的机制声MFCC和乐音MFCC能更好地体现音色之间的不同;机制声MFCC尤其适合对乐器大类的识别;谐波能量比仅反映共振体特性,可以在一定程度上反映音色的差别;复倒谱与实倒谱区别不同乐器的能力有限。

Timbre features of western instrument under multiplication model

The acoustic features related to western instrument timbre, including the distribution of harmonic energy, cepstrum and Mel-frequency cepstral coefficients （MFCC）, and the mechanism sound MFCC （MSMFCC） which is proposed in this paper, are investigated under the multiplication model of musical sound. Features are extracted from pure tone of western orchestral instruments, and their influences on timbre are assessed through analyzing each feature pattern. Results show that： MFCC and MSMFCC emphasizing the nature of excitement perception can reflect timbre difference very well; MSMFCC is more capable for instrument classification; The distribution of harmonic energy, which reflects only the properties of resonator, can differ timbre to some degree; Complex and real cepstra have limited ability to distinguish the different instruments.