Nilpotent bridging for unions of two bases (Q2361558)

The paper is a continuation of a recent work by the authors [J. Fourier Anal. Appl. 21, No. 5, 1146--1167 (2015; Zbl 1377.42035)] on a method of full reconstruction of signals from frame coefficients with erasures that is named nilpotent bridging. Suppose that \((f,g)\) is a dual frame pair consisting of finite frames of \(N\) vectors in an \(n\)-dimensional Hilbert space \(H\). Erasures occur when some set of the frame coefficients is erased or lost. The set of indices for which the coefficients are lost is denoted by \(\Lambda\) and is called the erasure set. It is well known that perfect reconstruction of signals is possible precisely when the erasure set satisfies the minimal redundancy condition which means that the elements of the original frame \(f_j\), \(j\in \Lambda^c\), (where \(\Lambda^c\) denotes the complement of \(\Lambda\)) still span \(H\). Various techniques for such compensation for erasures are known. The nilpotent bridging basically consists of finding another set of indices \(\Omega \subset \Lambda^c\) (so called bridge set) of the same cardinality as \(\Lambda\) such that the matrix \(B(\Lambda,\Omega):=(\langle f_j,g_k\rangle)_{j\in \Lambda,k\in \Omega}\) is invertible. Such bridge set \(\Omega\) exists whenever the erasure set satisfies the minimal redundancy condition; however, some search procedure is needed to identify possible bridge sets. Hence, the most convenient case occurs when no search procedure is needed. This leads to the following notion: dual pairs of frames for which for every set of indices \(\Omega \subset \Lambda^c\) satisfying \(\text{card}\,\Omega=\text{card}\,\Lambda \leq \min\{n,N-n\}\), the bridge matrix \(B(\Lambda,\Omega)\) is invertible are said to satisfy the full skew-spark property. It is known that that such pairs of frames do exist in abundance. In particular, the set of Parseval frames satisfying the full skew-spark property is open and dense in the set of all Parseval frames. In the present paper this property is investigated for frames that are unions of two bases. It turns out that such frames usually do not satisfy the full skew-spark property. The main contribution of the paper is the introduction of a new property of frames which are unions of two bases, that is called the block skew-spark property. As with the full skew-spark property, it turns out that the block skew-spark property of frames which are unions of two bases ensures that a search step is not required in the algorithms to implement nilpotent bridging. Moreover, this property possess most unions of two bases. In addition, the authors show that similar results also hold for infinite frames in infinite-dimensional spaces provided that the erasure set is finite.