Date of Award

Fall 1989

Degree Type

Thesis - Restricted

Degree Name

Master of Science (MS)

Department

Mathematics, Statistics and Computer Science

First Advisor

Barnard, Mark

Second Advisor

Corliss, George F.

Third Advisor

Feng, Xin

Abstract

This thesis describes, experiments in developing artificial neural networks, using a feedforward architecture and the backpropagation learning algorithm for classifying acoustic signals from three appliances: a washer, a dryer, and a furnace. The artificial neural network was trained to give the on/off status of these three appliances based on amplitudes of acoustic frequencies. Tests were performed on the network which varied the number of hidden units, learning rates, tolerances, and the number of training spectra. Two of the best artificial neural networks were compared to three classical pattern classification techniques: Euclidean distance, direction cosines, and the Tanimoto similarity measure. The performance of the network was also compared to that of human subjects visually classifying the spectra.

Share

COinS