Abstract. Most of the examples of wavelet sets are for dilation sets which are groups. We find a necessary and sufficient condition under which subspace wavelet sets exist for dilation sets of the form AB, which are not necessarily groups. We explain the construction by a few examples.
IntroductionWavelets and frames have become a widely studied tool in mathematics and applied science. One of the obvious questions is the construction of wavelets with given properties. The classical wavelet system on the line is given by a function ψ ∈ L 2 (R), such that the dyadic dilates and translates of ψ form an orthonormal basis for L 2 (R). Thus, {ψ j,n } j,n∈Z with ψ j,n (t) = 2 j/2 ψ(2 j t + n), is an orthonormal basis for L 2 (R). There are several obvious generalizations: One can replace 2 by any integer N ; one can allow several wavelet functions ψ 1 , . . . , ψ L ; and one can consider an orthonormal basis for a closed subspace of L 2 (R). There have also been several publications of wavelets in higher dimensions, cf [1,2,3,5,10,12,13,14,21,22,23] to name few. One of the differences in higher dimensions is that we now have many more choices in the sets of dilations and translations. So, to fix the notation, let D ⊆ GL(n, R) and T ⊆ R n be countable sets. A (D, T )-wavelet is a square integrable function ψ with the property that the set of functionsforms an orthonormal basis for L 2 (R n ). The set D is called the dilation set and the set T is called the translation set. If we replace L 2 (R n ) in the above definition bywe get a (D, T )-subspace wavelet. Here F stands for the Fourier transformWe will often writef for the Fourier transform of f . The most natural starting point is to consider groups of dilations and full rank lattices as translation sets. The simplest examples would then be groups generated by one element D = {a k | k ∈ Z}, see [24] and the reference therein. In [22,23] 1991 Mathematics Subject Classification. 42C40,43A85.