Music Retrieval based on Melodic Similarity

This book describes a method for retrieving pieces of music that contain musical material which is melodically similar to a given query. The method uses a continuous dissimilarity measure and works for symbolic representations of music such as MIDI files. The algorithmic complexity is logarithmic, and therefore the method scales well. The more data added to the searchable database, the smaller the increase in response time. This method produced the best results for the MIREX 2006 “Symbolic Melodic Similarity” retrieval competition. Other topics include a survey of music information retrieval systems and a description of how well vantage indexing works for music. Further, the book contains a chapter about evaluating music information retrieval systems and a measure, “average dynamic recall”, that is well suited to this task.