Exploring structural sections in EDM DJ-created radio mixes with the help of automatic music descriptors

TitleExploring structural sections in EDM DJ-created radio mixes with the help of automatic music descriptors
Publication TypeMaster Thesis
Year of Publication2016
AuthorsZurita, V.
AbstractDisc Jockeys are the most likely the ultimate experts at mixing music to “move the crowd”. In this work, structural sections based on “levels of emotional experience” in electronic dance music (EDM) are investigated by analyzing DJ-created radio mixes using automatic music descriptors in order to get audio content information. Mixes are first manually annotated to categorise structural sections and then, their characteristics and patterns between them are investigated. Finally, various evaluations are performed to explore the accuracy with which a system could automatically detect, with the given information, the different sections within randomly chosen independent ten seconds excerpts. Results prove the reliability of the chosen descriptors to achieve a successful classification with moderate number of errors. Moreover, a solution for improving the prediction’s accuracy by taking in account patterns between structural sections is proposed.
intranet