Score-Informed Voice Separation For Piano Recordings.
Year: 2011
Type: conference-paper
Abstract: The decomposition of a monaural audio recording into mu-sically meaningful sound sources or voices constitutes a fun-damental problem in music information retrieval. In this paper, we consider the task of separating a monaural piano recording into two sound sources (or voices) that correspond to the left hand and the right hand. Since in this scenario the two sources share many physical properties, sound sep-aration approaches identifying sources based on their spec-tral envelope are hardly applicable. Instead, we propose a score-informed approach, where explicit note events speci-fied by the score are used to parameterize the spectrogram of a given piano recording. This parameterization then al-lows for constructing two spectrograms considering only the notes of the left hand and the right hand, respectively. Fi-nally, inversion of the two spectrograms yields the separa-tion result. First experiments show that our approach, which involves high-resolution music synchronization and para-metric modeling techniques, yields good results for real-world non-synthetic piano recordings. 1.
Authors: Ewert, Sebastian, Müller, Meinard
Institutions: Max Planck Society, Max Planck Institute for Informatics
Language: English
Topic: Music and Audio Processing
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W1807124684
