期刊名称：International Journal of Computer Science Issues
印刷版ISSN：1694-0784
电子版ISSN：1694-0814
出版年度：2013
卷号：10
期号：1
出版社：IJCSI Press
摘要：In this paper, we propose a fast and accurate transcription method of a musical signal. It consists of extracting the musical information from the temporal evolution of the generated signal by the instrument. Each note (primary) will mainly be represented by a finite set of basic attributes (pitch, partial, energy, duration,). To do so, we begin by extracting each note by selecting the beginning of its appearance (onset detection), then proceeding by segmenting the signal, in order to delimit each note, which is to be identified later by determining its remaining features. The proposed method is an extension of the well known spectral based method. It is specially designed for oriental music which is characterized by its richness in tone that can be extended to tone. It aims to detect and isolate notes from a real audio signal recorded from an Oriental lute in an ordinary environment, then exploits the constraints of the lutes sound to improve the performance of the proposed transcriber. This method also includes, preprocessing and post processing based mainly on the surrounding noise, and echo. Subsequently, we present an interpretation of the results and rigorous assessment of the method through modeling the lute string motion.
