TitleImprovements to percussive component extraction using non-negative matrix factorization and support vector machines
Publication TypeThesis
Year of Publication2008
AuthorsBattenberg, E
Academic DepartmentEECS
Number of Pages48
Date Published12/2008
UniversityUniversity of California, Berkeley
CityBerkeley, CA
Keywordsanalysis,, factorization,, information, machines,, matrix, Music, non-negative, retrieval, rhythm, support, vector

A system for the automatic extraction of percussive components from polyphonic digital audio is presented.Like some previous work, the system uses an iterative non-negative matrix factorization (NMF) algorithmto decompose a songs spectrogram into components, and then it classifies these components as percussive ornon-percussive using a support vector machine (SVM). Our approach attempts to reduce computation timeand improve separation results by incorporating a perceptual dimensionality reduction into the NMF step.In addition, we introduce new features some based on note onset locations extracted from the spectra andgain signals of each component in order to reduce classification errors. Our NMF approach greatly reducescomputation time while retaining the same (or improving) the quality of separation. And using our newfeatures, our component classifier achieves an equal error rate of less than 3.7% on a database of 32 songs.