A feedback framework for improved chord recognition based on NMF-based approximate note transcription

摘要: This paper presents a feedback framework that can improve chord recognition for music audio signals by performing approximate note transcription with Bayesian non-negative matrix factorization (NMF) using prior knowledge on chords. Although the names and compositions of chords are intrinsically linked each other (e.g., C major highly likely to include C, E, G notes, those notes be in chords), (multipitch analysis) have been studied independently. To solve this chicken-and-egg problem, our iterates other's results. More specifically, we first perform based NMF forces basis spectra respectively correspond different semitone-level pitches covering whole range. We then execute hidden Markov models (HMMs) use chroma features obtained from activation patterns pitches. transcription, again encourages certain kinds region activated. Experimental results showed gradually improved accuracy recognition.