Title :
Bootstrap based spectral analysis
Author :
Zoubir, Abdelhak M.
Author_Institution :
Signal Process. Group, Darmstadt Univ. of Technol.
Abstract :
The problem of confidence interval estimation for spectra, coherence functions and transfer functions is addressed. We propose bootstrap based methods as an alternative to asymptotic techniques. Bootstrap methods provide accurate confidence coverage, in particular when asymptotic techniques are invalid, i.e., for small sample sizes or non-Gaussian data
Keywords :
spectral analysis; transfer functions; asymptotic techniques; bootstrap based spectral analysis; confidence interval estimation; transfer functions; Bandwidth; Frequency; Kernel; Limiting; Noise figure; Nonlinear filters; Signal processing; Signal resolution; Spectral analysis; Transfer functions;
Conference_Titel :
Statistical Signal Processing, 2005 IEEE/SP 13th Workshop on
Conference_Location :
Novosibirsk
Print_ISBN :
0-7803-9403-8
DOI :
10.1109/SSP.2005.1628798
