Electronic Theses and Dissertations

Identifier

661

Date

2012

Document Type

Thesis

Degree Name

Master of Science

Major

Electrical and Computer Engr

Concentration

Computer Engineering

Committee Chair

Mohammed Yeasin

Committee Member

Bonny Banerjee

Committee Member

Aaron L Robinson

Abstract

The aim of this study is to develop robust algorithm to automatically detect the Tone and Break Indices(ToBI) from the speech signal and explore their applications. iLAST was introduced to analyze the acoustic and prosodic features to detect the ToBI indices. Both "expert" and "data" driven rules were used to improve the robustness. The integration of multi-scale signal analysis with rule-based classification has helped in robustly identifying tones that can be used in applications, such as identifying Vowel triangle, emotions from speech etc. Empirical analyses using labeled dataset were performed to illustrate the utility of the proposed approach. Further analyses were conducted to identify the inefficiencies with the proposed approach and address those issues through co-analyses of prosodic features in identifying the major contributors to robust detection of ToBI. It was demonstrated that the proposed approach performs robustly and can be used for developing a wide variety of applications.

Comments

Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.

Download

Share

COinS