Compactly supported wavelets and representations of the Cuntz relations (Q1975533): Difference between revisions

The operators \( S_i\) defined by \( S_i p(z) = m_i(z)p(z^N)\) in which \(m_1,\dots,m_{N-1}\in L^2(\mathbb T) \) satisfy \( \sum_{i =0}^{N-1}m_i(t+\frac{k}{N})\cong m_j(t+\frac{k}{N}) =\delta_{ij}N \) satisfy the \textit{Cuntz relations} \( S_j^\ast S_i=\delta_{ij}I \) and \(\sum_i S_i^\ast S_i = I\). There is a one-to-one correspondence between operator solutions of the Cuntz relations and irreducible unitary representations of the Cuntz algebra \({\mathcal O}_N\) and the purpose of this paper is to start an investigation of the ramifications that representation theory and wavelet theory can have on one another in this light. The operators \( S_i\), or rather their Fourier transforms, are precisely the quadrature mirror filters that arise in an \(N\)-scale multiresolution analysis. Shifting focus from quadrature mirror filters to abstract operators allows for a coordinate free approach which makes questions of irreducibility less cumbersome and also allows the possibility, for example, of investigating properties of wavelets in higher dimensions by reduction at an abstract level to properties of unitarily equivalent wavelets in one dimension. In the case of polynomial representations the structure of the operators \(S_i \) is analyzed in some detail. In particular, whether and how the restriction of such a representation to the \textit{fixed point subalgebra} decomposes appears to have some bearing on whether a wavelet gives rise to a basis or a tight frame, as is analyzed in detail for the case of third order polynomials on \({\mathcal O}_2\).