TitleMapping Spectral Frames to Pitch with the Support Vector Machine
Publication TypeConference Paper
Year of Publication2004
AuthorsSchmeder, AW
Conference NameInternational Computer Music Conference
PublisherInternational Computer Music Association
Conference LocationMiami, FL

The Support Vector Machine algorithm is studied in the context of pitch estimation. Its learning capacity is analyzed using an artificial dataset of harmonic spectra. We propose an architecture for learning pitch in difficult real-world scenarios, and demonstrate its application with a database of guitar sounds. Domain-specific aspects of kernel methods are discussed, and a method for extracting structural knowledge via visualization is examined.