PDCON:Conference/A Framework for Real-time Instrumental Sound Segmentation and Labeling

A FRAMEWORK FOR REAL-TIME INSTRUMENTAL SOUND SEGMENTATION AND LABELING

Author: Adriano Monteiro

This paper presents a collection of Pure Data abstractions for real-time transcription of the audio signal produced by musical instruments. It is a tool 1) for musical notes labeling and targeting in time; 2)for handling the representation of musical data obtained in 1). It is part of the PDescriptors, a library for audio features extracting developed by the author, mostly based on BSP technique (Barknecht, 2010). In the text we start introducing literature review on automatic music transcription and its central issues. it follows the models parametric transcription we adopted in the present research regarding onset detection, extraction of harmonic content and timbre classification of percussion instruments. Finally, we discuss and detail our Pure Data implementation and present some results related its aplications in a human-computer interaction system.