Date of Award

Spring 1992

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical & Computer Engineering

Program/Concentration

Electrical Engineering

Committee Director

Stephen A. Zahorian

Committee Member

David L. Livingston

Committee Member

John W. Stoughton

Call Number for Print

Special Collections LD4331.E55B43

Abstract

Teaching the hearing-impaired to speak is both a challenging and rewarding task. Computer-based training aids are valuable teaching tools and complement the speech therapist in the speech training process. We have developed a computer based vowel articulation training aid designed to provide visual feedback for vowel production to the hearing-impaired user. This system transforms speech spectra from a multi-dimensional acoustic space to a two-dimensional phonetic space which is displayed on a computer monitor. The system signal processing includes feature extraction designed to mimic the properties of the human ear, and a transformation to a two-dimensional display space. The dimensionality transformation is performed by a combination nonlinear/linear multi-layer feedforward neural network.

This thesis details the development of the real-time training aid with particular attention to the neural network transformation. Methods of reducing the training time of the neural network are explored. One such method utilizes multi-stage processing for training the neural network. The result is an increase in performance on test data of over 5% and a decrease in the time needed to train the network. Various feedback displays including games have been developed as well as data collection techniques for easily adding speakers to the training database.

Rights

In Copyright. URI: http://rightsstatements.org/vocab/InC/1.0/ This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s).

DOI

10.25777/xmfc-4g92

Download

Share

COinS