Method for detecting onset times of sounds of string instrument

Abstract

The technology of CG (Computer Graphics) is indispensable in the production of animation especially for playing musical instruments. Generally, in CG animation production of musical instrument performance, recording of musical instrument performance data by a motion capture system and recording of sound source data are performed separately. Therefore, it is inevitable that there are gaps between the sound source and the video. Since high-quality sound is required for sound source data, electronic musical instruments are not used. Therefore, sound source data is recorded in WAV format, which do not include the information of onset times and frequency. Consequently, it is necessary to detect onset times of musical instruments and to stretch the intervals of the onset times in order to synchronize sound source data and video data. There is still no effective method for detecting onset times for sound source of a stringed instrument such as a violin. In this paper, we focus on a unique property that occurs during performance of a stringed instrument. We propose a method for detecting onset times in sound source of a stringed instrument based on the property. Furthermore, we evaluate the effectiveness by experiments using real sound sources.