Pitch Estimation of Musical Signals

Can a computer algorithm assess musical pitch with results analogous to those of a trained human musician? This thesis attempts to answer that question by demonstration. It starts by examining the nature of pitch, and human spectrotemporal perception of music. In pitch vibrato, the frequency varies up and down over time, usually sinusoidally. At vibrato frequencies over about 4Hz, most listeners perceive pitch vibrato tones as having a single pitch. However, few controlled studies have been conducted to explore the question of just what that pitch is. A software tool was developed to generate arbitrary pitch vibrato tones, and administer controlled pitch matching tests. With respect to the longstanding debate as to whether pitch perception is based on spectral or temporal mechanisms, some support was found for temporal neural coding being the primary, but not singular pitch detection mechanism. A novel method for detecting musical pitch is demonstrated, based on spectrotemporal analysis of the signal. A Fourier analysis algorithm infers perceived pitch from the distance between the harmonics that are present in the signal. A weighted voting technique was developed to assess inter-peak distances, and merge candidate distances where possible. Individual spectral peaks are also considered in the weighting process. A technique developed at CERN for interpolating Fourier analysis results is applied to musical signals for the first time. The temporal analysis algorithm improves on the author’s previous research by adding a new confidence metric, which improves octave resolution. It also uses a weighted voting technique similar to the one used by the Fourier analysis algorithm. A confidence metric is used to select from multiple algorithms. The resulting algorithm improves on previously published methods for estimating musical pitch in several ways: improved accuracy, improved immunity to octave errors, improved ability to follow transient pitch changes, and improved frequency range. The algorithm was validated against the extensive instrumental recording library at the University of Iowa, and against a small set of vocal recordings. A case study methodology was used to assess the experimental software tool’s potential for use as an educational tool.

Name thesis:degree_name

Doctor of Philosophy (PhD)

Discipline thesis:degree_discipline

Interdisciplinary Graduate Program

Grantor dc:publisher.institution

Graduate Studies

Year dc:date.issued

2017

Author dc:creator

• Heerema, John

Advisor dc:contributor.advisor

• Parker, James Robert

Committee members dc:contributor.committeemember

• Eggermont, Jos Jan
• Kopp, Gail Diane
• Keenan, Thomas Patrick
• Bell, Donald Munro
• Osborn, Wendy Kathleen

• Music
• Neuroscience
• Audiology
• Computer Science
• Engineering--Biomedical
• Psychology--Cognitive

Statement dc:rights

• University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission.

Language dc:language.iso

eng

DOI

http://dx.doi.org/10.11575/PRISM/27363

OAI identifier oai:identifier

oai:ucalgary.scholaris.ca:11023/4008