Date of Award

12-1991

Document Type

Thesis

Degree Name

Master of Science

Department

Department of Electrical and Computer Engineering

First Advisor

Rob Williams, PhD

Abstract

The purpose of this study was to investigate the use of Artificial Neural Networks to localize sound sources from simulated, human binaural signals. Only sound sources originating from a circle on the horizontal plane were considered. Experiments were performed to examine the ability of the networks to localize using three-different feature sets. The feature sets used were: time-samples of the signals, mena Fast Fourier Transform magnitude and cross correlation data, and auto-correlation and cross correlation data. The two different types of sound source signals considered were tones and gaussian noise. The feature set which yielded the best results in terms of classification accuracy (over 91%) for both tones and noise was the auto-correlation and cross- correlation data. These results were achieved using 18 classes (20 per class). The other two feature sets did not produce accuracy results as high or as consistent between the two signal types. When using time-samples of the signals as features it was observed that in order to accurately classify tones of random-frequency, it was necessary to train with random-frequency tones rather than with tones of one, or a few discrete frequencies.

AFIT Designator

AFIT-GE-ENG-91D-13

DTIC Accession Number

ADA243878

Comments

The author's Vita page is omitted.

Download

Share

COinS