Please use this identifier to cite or link to this item:
Title: Variance reduction in power spectrum using cepstral thresholding appraoch
Researcher: Venkatanarayana M
Guide(s): Jayachandra Prasad T
Keywords: Cepstral thresholding
Electronics and Communication Engineering
Power spectrum
Upload Date: 11-Feb-2014
University: Jawaharlal Nehru Technological University, Anantapuram
Completed Date: 30/08/2012
Abstract: For stationary signals, there are number of power spectral density estimation techniques. The main problem of all these power spectral density estimation methods is high variance. Various forms of windowing or smoothing have been proposed to reduce variance of the periodogram. The use of these techniques requires careful selection of the window, its span and the operation for which there are no clear cut guide lines. The selection of the window based on independence of data may not work properly. Selection of window based on data is also difficult. The most popular method of nonparametric spectral density for variance reduction is the Weighted Overlap Segment Averaging (WOSA), also called the Welch method. As a part of this thesis an attempt has been made to utilize circular overlap in WOSA, known as Weighted Circular Overlap Segment Averaging (WCOSA) to reduce the variance further. The performance evaluation of WOSA and WCOSA are carried out for variance reduction and it is verified that WCOSA is outperforming WOSA. The filter bank concept has been used to estimate spectrum based on the periodogram realization. The filter bank based spectrum estimation with polyphase realization has been studied and is applied to various processes such as dual tone sinusoidal, broadband and narrowband. The results obtained are compared with the conventional spectrum estimation techniques such as periodogram, Welch and Blackman-Tukey. It is observed that the filter bank based technique with polyphase realization using lowpass filter having ix a smooth transition produces spectral estimates with low variability and acceptable resolution. In this research work, an attempt has been made to utilize the lifting scheme to estimate the smooth spectra based on Lifting Wavelet Transform (LWT) via thresholding. The selection of thresholds is based on i) Gaussian approximation ii) Stain s Unbiased Risk Estimator (SURE). The concept of SURE has been applied to estimate the smoothed spectra based on LWT and its performance has been compared
Pagination: xxv, 138p.
Appears in Departments:Department of Electronics and Communication

Files in This Item:
File Description SizeFormat 
01_title.pdfAttached File150.49 kBAdobe PDFView/Open
02_certificates.pdf174.68 kBAdobe PDFView/Open
03_acknowledgements.pdf83.54 kBAdobe PDFView/Open
04_contents.pdf91.29 kBAdobe PDFView/Open
05_list of tables figures.pdf110.94 kBAdobe PDFView/Open
06_chapter 1.pdf307.53 kBAdobe PDFView/Open
07_chapter 2.pdf167 kBAdobe PDFView/Open
08_chapter 3.pdf463.06 kBAdobe PDFView/Open
09_chapter 4.pdf453.84 kBAdobe PDFView/Open
10_chapter 5.pdf472.51 kBAdobe PDFView/Open
11_chapter 6.pdf819.84 kBAdobe PDFView/Open
12_chapter 7.pdf134.98 kBAdobe PDFView/Open
13_references.pdf224.67 kBAdobe PDFView/Open
14_publications.pdf194.2 kBAdobe PDFView/Open

Items in Shodhganga are protected by copyright, with all rights reserved, unless otherwise indicated.