Resumen de Similarity and style in electronic dance music drum rhythms
This thesis presents original research carried out in the topic of electronic dance music (EDM) drum sequencing, a fundamental and yet underdeveloped subject in the music production literature. The work undertaken is focused in two main areas: similarity between drum patterns and modeling of drumming style. The study of pattern similarity is rooted in current knowledge on human processing of monophonic rhythms, and is expanded until a model capable of predicting similarity sensations of polyphonic drum rhythms is reached. With this model, RhythmSpace, a graphical system for the continuous real-time exploration of drum pattern collections, is developed. The second area of research, drumming style modeling, is approached from a statistical perspective, developing a generative model capable of learning styles from examples and creating original drum patterns in the learned styles. This model allows high-level musical flexibility, letting a musician combine and transform styles in real-time during the generative process. Taking advantage of this model, a style-based drum machine application, DrDrums, is implemented and evaluated in subject-based experiments.
