A Comparison of Melodic Segmentation Techniques for Music Information Retrieval

The scientific research on accessing and retrieval of music documents is becoming increasingly active, including the analysis of suitable features for content description or the development of algorithms to match relevant documents with queries. One of the challenges in this area is the possibility to extend textual retrieval techniques to music language. Music lacks of explicit separators between its lexical units, thus they have to be automatically extracted. This paper presents an overview of different approaches to melody segmentation aimed at extracting music lexical units. A comparison of different approaches is presented, showing their impact on indexes size and on retrieval effectiveness.