Using a Pitch Detector for Onset Detection

A segmentation strategy is explored for monophonic instrumental pitched non-percussive material (PNP) which proceeds from the assertion that human-like event analysis can be founded on a notion of stable pitch percept. A constant-Q pitch detector following the work of Brown and Puckette provides pitch tracks which are post processed in such a way as to identify likely transitions between notes. A core part of this preparation of the pitch detector signal is an algorithm for vibrato suppression. An evaluation task is undertaken on slow attack and high vibrato PNP source files with human annotated onsets, exemplars of a difficult case in monophonic source segmentation. The pitch track onset detection algorithm shows an improvement over the previous best performing algorithm from a recent comparison study of onset detectors. Whilst further timbral cues must play a part in a general solution, the method shows promise as a component of a note event analysis system.

• Type: Conference
• Year: 2005
• Where: ISMIR
• Authors: Nick Collins
• Keywords: Constant-Q Pitch Detector, Information Retrieval, ISMIR 2005, Pitch Detector, Pitch Track