Rebricking frames and bases

Abstract

In 1949, Denis Gabor introduced the “complex signal” (nowadays called “analytic signal”) by combining a real function f with its Hilbert transform Hf to a complex function $f+iHf$. His aim was to extract phase information, an idea that has inspired techniques as the monogenic signal and the complex dual tree wavelet transform. In this manuscript, we consider two questions: When do two real-valued bases or frames $\{f_n:n\in N\}$ and $\{g_n:n\in N\}$ form a complex basis or frame of the form $\{f_n+i{g}_n:n\in N\}$? And for which bounded linear operators A does $\{f_n+iA{f}_n:n\in N\}$ form a complex-valued orthonormal basis, Riesz basis or frame, when $\{f_n:n\in N\}$ is a real-valued orthonormal basis, Riesz basis or frame? We call this approach rebricking. It is well-known that the analytic signals don't span the complex vector space $L^2(R;C)$, hence H is not a rebricking operator. We give a full characterization of rebricking operators for bases, in particular orthonormal and Riesz bases, Parseval frames, and frames in general. We also examine the special case of finite dimensional vector spaces and show that we can use any real, invertible matrix for rebricking if we allow for permutations in the imaginary part.