Sample Identification in Hip Hop Music

Sampling is a creative tool in composition that is widespread in popular music production and composition since the 1980's. However, the concept of sampling has for a long time been unaddressed in Music Information Retrieval. We argue that information on the origin of samples has a great musicological value and can be used to organise and disclose large music collections. In this paper we introduce the problem of automatic sample identification and present a first approach for the case of hip hop music. In particular, we modify and optimize an existing fingerprinting approach to meet the necessary requirements of a real-world sample identification task. The obtained results show the viability of such an approach, and open new avenues for research, especially with regard to inferring artist influences and detecting musical reuse.