The classification of acoustic emission signals via artificial neural network

Automatic identification of wood species is a long desired goal in the pulp and paper industry. Motivated by this, researchers have investigated Pattern Recognition (PR) for the classification of wood chip species based on acoustic emission signals. In this thesis, a new Artificial Neural Network (ANN) approach is proposed to perform this task. One of the purposes of the thesis is to explore the connection between traditional methods of statistics and modern approaches of neural networks regarding pattern recognition. We attempt to understand how the neural networks perform signal processing functions such as the Karhunen-Loeve Transform (KLT) and pattern recognition functions such as the Principal Component Analysis (PCA). The configuration of the classifier is a multilayer feed forward network in which the supervised and unsupervised learning algorithms are implemented. Based on the established orthogonal data space, feature extraction and data compression are accomplished through the training process. The reconstructed data, which is in orthogonal and compressed form with maximal variance, are used to improve the classification efficiency. Since the primary goal of the thesis is to identify wood species from their acoustic emission, simulations are used to characterize classification accuracy, classification efficiency, noise immunity, and generalization capability.

Name thesis:degree_name

Master of Applied Science - MASc

Level thesis:degree_level

master's

Discipline thesis:degree_discipline

Electrical and Computer Engineering

Grantor dc:publisher

University of British Columbia

Year dc:date

1992

Author dc:creator

• Yang, Jian

Statement dc:rights

• For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

Language dc:language

eng

Handle dc:identifier

http://hdl.handle.net/2429/1875

OAI identifier oai:identifier

oai:circle.library.ubc.ca:2429/1875