James Brock

Abstract

This thesis research investigates and demonstrates the feasibility of performing computationally efficient, high-dimensional acoustic classification using Mel-frequency cepstral coefficients and independent component analysis for temporal feature extraction. A process was developed to calculate Mel-frequency cepstral coefficients from samples of acoustic data grouped by either musical genre or spoken world language. Then independent component analysis was employed to extract the higher level temporal features of the coefficients in each class. These sets of unique independent features represent themes, or patterns, over time that are the underlying signals in that class of acoustic data. The results obtained from this process clearly show that the uniqueness of the independent components for each class of acoustic information are legitimate grounds for separation and classification of the data.

Library of Congress Subject Headings

Sound--Classification; Music--Acoustics and physics--Data processing; Music--Data processing; Speech--Data processing; Principal components analysis

Publication Date

2006

Document Type

Thesis

Department, Program, or Center

Computer Science (GCCIS)

Advisor

Reynolds, Carl

Advisor/Committee Member

Anderson, Peter

Comments

Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: QC226 .B76 2006

Campus

RIT – Main Campus

Download

Share

COinS