Eastman School of Music, University of Rochester
Title: Acoustic Models and Composed Structures
Abstract: In the mid-60ies a novel approach emerged in contemporary composition, which integrated spectrographic sound analysis with composition. This technique, which has since become known as “spectral composition”, opened up new ways of combing computer-based sound analysis and resynthesis with instrumental composition, and made the science of sound, acoustics, and psychoacoustics part of the compositional process. Extracting data from sound analysis to use in a piece of composed music raises several interesting challenges and possibilities. In my talk, I will describe this approach in some detail, using several examples from other composers as well as my work.
University at Buffalo
Title: Cognitive and sensorimotor factors in singing: Implications for music information retrieval
Abstract: Singing is distinctive among forms of music performance in that it requires no formal training. As such, singing is a broadly accessible source of data for music information retrieval, such as in ‘query by humming’ algorithms. However, singing accuracy of pitch is also notoriously subject to large individual differences, which may compromise the effectiveness of such retrieval algorithms. In this talk I discuss how singing accuracy is manifested in the general population, drawing on data from both controlled experiments and from a large online database (The Seattle Singing Accuracy Protocol, www.ssap.northwestern.edu). I will discuss results based on different measures of accuracy, as well as different timescales within the structure of melodies ranging from melodic contour (pitch direction) to stability of pitch within sung notes. These data will be considered in light of the underlying mechanisms that may guide the planning and control of pitch in the human voice, and also with respect to how the empirical results and theoretical underpinnings alike may contribute to music information retrieval.