Date of Award

8-2018

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical and Computer Engineering (Holcomb Dept. of)

Committee Member

Dr. Carl Baum and Dr. Eric Patterson, Committee Chair

Committee Member

Dr. Harlan B. Russell

Committee Member

Dr. Robert Schalkoff

Committee Member

Dr. Apoorva Kapadia

Abstract

Active Noise Canceling (ANC) is the idea of using superposition to achieve cancellation of unwanted noise and is implemented for many applications such as attempting to reduce noise in a commercial airplane cabin. One of the main traditional techniques for noise cancellation is the adaptive least mean squares (LMS) algorithm that produces the anti-noise signal, or the 180 degree out-of-phase signal to cancel the noise via superposition. This work attempts to compare several neural network approaches against the traditional LMS algorithms. The noise signals that are used for the training of the network are from the Signal Processing Information Base (SPIB) database. The neural network architectures utilized in this paper include the Multilayer Feedforward Neural Network, the Recurrent Neural Network, the Long Short Term Neural Network, and the Convolutional Neural Network. These neural networks are trained to predict the anti-noise signal based on an incoming noise signal. The results of the simulation demonstrate successful ANC using neural networks, and they show that neural networks can yield better noise attenuation than LMS algorithms. Results show that the Convolutional Neural Network architecture outperforms the other architectures implemented and tested in this work.

Share

COinS