Principal component filter banks for optimum wavelet analysis
M.K. Tsatsanis, Georgios B. Giannakis · 2003
Optimality issues are addressed when multiresolution methods are applied to random signals. An optimal P-band perfect reconstruction filter bank is derived, which decomposes the input signal into its lower-resolution principal components. Results are developed for stationary processes and a class of nonstationary processes, which includes cyclostationary signals. FIR designs are discussed and some examples presented.>