In this dissertation, we investigate methods of modifying a tight frame sequence on a finite subset of the frame so that the result is a tight frame with better properties. We call this a surgery on the frame. There are basically three types of surgeries: transplants, expansions, and contractions. In this dissertation, it will be necessary to consider surgeries on not-necessarily-tight frames because the subsets of frames that are excised and replaced are usually not themselves tight frames on their spans, even if the initial frame and the final frame are tight. This makes the theory necessarily complicated, and richer than one might expect. Chapter I is devoted to an introduction to frame theory. In Chapter II, we investigate conditions under which expansion, contraction, and transplant problems have a solution. In particular, we consider the equiangular replacement problem. We show that we can always replace a set of three unit vectors with a set of three complex unit equiangular vectors which has the same Bessel operator as the Bessel operator of the original set. We show that this can not always be done if we require the replacement vectors to be real, even if the original vectors are real. We also prove that the minimum angle between pairs of vectors in the replacement set becomes largest when the replacement set is equiangular. Iterating this procedure can yield a frame with smaller maximal frame correlation than the original. Frames with optimal maximal frame correlation are called Grassmannian frames and no general method is known at the present time for constructing them. Addressing this, in Chapter III we introduce a spreading algorithm for finite unit tight frames by replacing vectors three-at-a-time to produce a unit tight frame with better maximal frame correlation than the original frame. This algorithm also provides a "good" orientation for the replacement sets. The orientation part ensures stability in the sense that if a selected set of three unit vectors happens to already be equiangular, then the algorithm gives back the same three vectors in the original order. In chapter IV and chapter V, we investigate two special classes of frames called push-out frames and group frames. Chapter VI is devoted to some mathematical problems related to the "cocktail party problem ".
In this dissertation, we investigate methods of modifying a tight frame sequence on a finite subset of the frame so that the result is a tight frame with better properties. We call this a surgery on the frame. There are basically three types of surgeries: transplants, expansions, and contractions. In this dissertation, it will be necessary to consider surgeries on not-necessarily-tight frames because the subsets of frames that are excised and replaced are usually not themselves tight frames on their spans, even if the initial frame and the final frame are tight. This makes the theory necessarily complicated, and richer than one might expect. Chapter I is devoted to an introduction to frame theory. In Chapter II, we investigate conditions under which expansion, contraction, and transplant problems have a solution. In particular, we consider the equiangular replacement problem. We show that we can always replace a set of three unit vectors with a set of three complex unit equiangular vectors which has the same Bessel operator as the Bessel operator of the original set. We show that this can not always be done if we require the replacement vectors to be real, even if the original vectors are real. We also prove that the minimum angle between pairs of vectors in the replacement set becomes largest when the replacement set is equiangular. Iterating this procedure can yield a frame with smaller maximal frame correlation than the original. Frames with optimal maximal frame correlation are called Grassmannian frames and no general method is known at the present time for constructing them. Addressing this, in Chapter III we introduce a spreading algorithm for finite unit tight frames by replacing vectors three-at-a-time to produce a unit tight frame with better maximal frame correlation than the original frame. This algorithm also provides a "good" orientation for the replacement sets. The orientation part ensures stability in the sense that if a selected set of three unit vectors happens to already be equiangular, then the algorithm gives back the same three vectors in the original order. In chapter IV and chapter V, we investigate two special classes of frames called push-out frames and group frames. Chapter VI is devoted to some mathematical problems related to the "cocktail party problem ".
