Content Based Music Information Retrieval

Estimates put the canon of traditional Irish dance tunes at at least seven thousand compositions. Given this diversity, a common problem faced by musicians is identifying tunes from playing. This work solves this problem by proposing Content Based Music Information Retrieval (CBMIR) systems adapted to the characteristics of traditional Irish music. The systems described in this book include an MIR prototype - MATT2 (Machine Annotation of Traditional Tunes), a segmentation system - TANSEY (Turn ANnotation from SEts using SimilaritY profiles), tunepal.org (a browser hosted MIR system) and Tunepal for iPhone/iPod touch devices. Tunepal for iPhone is a QBP system that can be used in situ in traditional music sessions. In experiments, these systems provide 93% accuracy in identifying a tune from playing. These latter systems use a backend corpus of 13,290 traditional Irish, Scots, Welsh, Breton, Canadian and American Old Time tunes drawn from community sources and standard references. This book provides extensive background information on both traditional Irish Music and Music Information Retrieval Systems in addition to detailed explanations of how these popular tools work.