We propose two new methods to classify guitar strings for automated tablature transcription using only monophonic audio. The first method estimates the linear regression of log-inharmonicities of guitar strings with respect to their pitches and assigns unseen notes to the strings whose means and variances maximize the probability of their measured inharmonicities. The second method, developed as a baseline, characterizes the inharmonicity distribution of each fretboard position as a normal probability density, and then similarly assigns unseen notes to the fretboard positions that maximize the likelihood of their observed inharmonicities. Results from the standard Real World Corpus of guitar recordings show that exploiting regressions generally improves accuracy compared to our baseline, while both achieve adequate performance in guitar-independent test scenarios.

Authors: Michelson, Jonathan; Stern, Richard; Sullivan, Thomas
Affiliations: Electro-Harmonix / New Sensor Corporation, Brooklyn, NY, USA; Carnegie Mellon University, Pittsburgh, PA, USA(See document for exact affiliation information.)
AES Convention: 145 (October 2018) Paper Number: 10091
Publication Date: October 7, 2018 Import into BibTeX
Subject: Applications in Audio
Permalink: https://www.aes.org/e-lib/browse.cfm?elib=19817

Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!

This paper costs $33 for non-members and is free for AES members and E-Library subscribers.

Learn more about the AES E-Library

E-Library Location:   /conv/145/10091.pdf

Start a discussion about this paper!