Abstract: Music transcription is one of the foundations of audio information retrieval，and it can be used in the areas of music education and music composing.However，in some string music，there is a sound effect called “vibrato” which can express the emotion and the style.If that sound effect is not treated seriously in music transcription，it will lead to a wrong result.A method to recognize the vibrato of string music based on vector quantization is studied in detail，and a method to detect the vibrato in a consecutive music is put forward.Experimental results demonstrate that the method is fairly efficient for solving the automatic string music vibrato detection problem.

Key words: music transcription, vibrato, string music, vector quantization

摘要： 自动音乐标注是音频信息检索的基础，并可广泛应用于辅助音乐教学、辅助音乐创作等许多音乐相关领域。然而，在弦乐器演奏的音乐中存在着大量用于修饰或表现乐曲情感、风格的颤音。在对这类乐器的自动音乐标注中，如果不进行颤音检测而直接进行标注容易出现错误。对基于矢量量化的弦乐颤音识别方法进行了研究，提出了一种用于对整段音乐进行颤音检测的方法。实验证明这种方法是精确有效的。

关键词: 音乐标注, 颤音, 弦乐, 矢量量化