When: Thursday 30th March, 2017 @ 5:10 PM
Where: Room 4.31/4.33, Informatics Forum, University of Edinburgh
Automatic transcription of a cappella recordings from multiple singers
Rodrigo Schramm (Visiting Fellow, C4DM, Queen Mary University of London // Federal University of Rio Grande do Sul (UFRGS), Brazil)
This research focuses on a new method for multi-pitch detection and voice assignment applied to automatic music transcription of a cappella performances with multiple singers.
Our technique uses Probabilistic Latent Component Analysis (PLCA) for spectrogram factorisation, with the help of a 6-dimensional sparse dictionary which contains spectral templates of vowel vocalisations. A post-processing step is proposed to remove false positive pitch detections through a binary classifier, where overtone-based features are used as input into this step.
Preliminary experiments have shown promising multi-pitch detection results when applied to audio recordings of Bach Chorales and Barbershop music. Comparisons made with alternative methods have shown that our approach increases the number of true positive pitch detections while the post-processing step keeps the number of false positives lower than those measured in comparative approaches. Voice assignment is driven by the integration of an HMM-based method into the PLCA model, improving concomitantly the accuracy of multi-pitch detection and voice separation.
Rodrigo Schramm received his PhD in Computer Science from the Federal University of Rio Grande do Sul (UFRGS)/Brazil in 2015, where he is currently a faculty member. Between 2013 and 2014, he was visiting fellow at ICCMR – Interdisciplinary Centre for Computer Music Research – Plymouth University/UK. In 2016, he was awarded by the Royal Academy of Engineering with the Newton Research Collaboration Programme Award. Currently, he is conducting research at the C4DM (QMUL) in London, focusing his activities on the development of techniques for automatic transcription of audio recordings containing multiple singers.