A parameterized model for efficient over-sampled filter bank architectures

There is much literature on maximally decimated filter banks, polyphase structures and specific areas of interest such as filter optimization techniques for perfect or near perfect reconstruction systems. There is less on the broader topic of architectures for hardware implementations. This paper lays a more basic foundation for over-sampled filter banks along with their practical considerations. The objective is a universal architecture for efficient hardware implementation of DFT filter banks, whether over-sampled or critically sampled. To that end, a parameterized model that is, for the most part, independent of filter and optimization techniques will be proposed. The discussion deals primarily with analysis banks of the DFT variety, but can easily be extended to the synthesis side as well as other varieties of filter banks, where applicable