Title :
Structured Prediction Models for Chord Transcription of Music Audio
Author :
Weller, Adrian ; Ellis, Daniel ; Jebara, Tony
Author_Institution :
Columbia Univ., New York, NY, USA
Abstract :
Chord sequences are a compact and useful description of music, representing each beat or measure in terms of a likely distribution over individual notes without specifying the notes exactly. Transcribing music audio into chord sequences is essential for harmonic analysis, and would be an important component in content-based retrieval and indexing, but accuracy rates remain fairly low. In this paper, the existing 2008 LabROSA Supervised Chord Recognition System is modified by using different machine learning methods for decoding structural information, thereby achieving significantly superior results. Specifically, the hidden Markov model is replaced by a large margin structured prediction approach (SVMstruct) using an enlarged feature space. Performance is significantly improved by incorporating features from future (but not past) frames. The benefit of SVMstruct increases with the size of the training set, as might be expected when comparing discriminative and generative models. Without yet exploring non-linear kernels, these improvements lead to state-of-the-art performance in chord transcription. The techniques could prove useful in other sequential learning tasks which currently employ HMMs.
Keywords :
audio coding; content-based retrieval; harmonic analysis; learning (artificial intelligence); music; support vector machines; HMM; LabROSA supervised chord recognition system; chord transcription; content-based retrieval; harmonic analysis; hidden Markov model; indexing; machine learning methods; music audio; nonlinear kernels; structural information; structured prediction models; Content based retrieval; Decoding; Harmonic analysis; Hidden Markov models; Indexing; Kernel; Labeling; Machine learning; Music information retrieval; Predictive models; acoustic features; chord transcription; classification and prediction; structured and relational data;
Conference_Titel :
Machine Learning and Applications, 2009. ICMLA '09. International Conference on
Conference_Location :
Miami Beach, FL
Print_ISBN :
978-0-7695-3926-3
DOI :
10.1109/ICMLA.2009.132