Abstract
Texture features of a signal reflect its shape characteristics and have been used for signal classification. However, little has been done in using morphological texture features for signal classification in general and biological sound signal classification in particular. In this thesis, the performance of these features in the classification of biological sound signals is examined. This is done by carrying out a comparative study with two other sets of features used in signal classification-the widely used Linear Predictive Coefficients (LPCS) and the statistical shape features of the frequency spectrum. The Hidden Markov Model (HMM) is used as the classifier. Both the LPCs and the morphological texture features are implemented in real-time on a DSP processor and their respective classification rates are presented.
D'Souza, Carol Shilpa (1998). Evaluation of morphological texture features for real-time biological signal classification. Master's thesis, Texas A&M University. Available electronically from
https : / /hdl .handle .net /1969 .1 /ETD -TAMU -1998 -THESIS -D76.