"This thesis focuses on the application of statistical techniques for pattern classification where the patterns are power spectra of the recorded acoustic signals of occluded objects. Different approaches used to classify are : The Bayes linear and piece-wise linear classifiers, the correlation classifier, classification using error spectra and the clustering algorithm. Classifiers were designed and experiments were conducted to investigate the performance. The performance is evaluated in terms of the percentage efficiency by counting the number of vectors identified correctly and the number of vectors which were misclassified. The feature vectors from primitive measurement space is then reduced to lesser lengths using the techniques of minimum mean square error (MMSE, feature selection using largest eigenvalues) and the Gram-Schmidt procedure. The classification is again performed on the processed set of feature vectors in the new vector space. The performance of feature extractors was compared by observing the capability of vectors to contribute towards correct classification even after their lengths are reduced. Results of all the experiments performed are explained"--Abstract, page iii.
Stuller, John A.
Moss, Randy Hays, 1953-
Dagli, Cihan H., 1949-
Electrical and Computer Engineering
M.S. in Electrical Engineering
University of Missouri--Rolla
x, 81 pages
© 2000 Vivek Ramaprasad, All rights reserved.
Thesis - Restricted Access
Print OCLC #
Electronic OCLC #
Link to Catalog Record
Electronic access to the full-text of this document is restricted to Missouri S&T users. Otherwise, request this publication directly from Missouri S&T Library or contact your local library.http://laurel.lso.missouri.edu/record=b4414631~S5
Ramaprasad, Vivek, "Investigation of pattern classification and feature selection of Gaussian processes" (2000). Masters Theses. 1896.
Share My Thesis If you are the author of this work and would like to grant permission to make it openly accessible to all, please click the button above.