Quadrature mirror filters with perfect reconstruction and reduced computational complexity

This paper introduces a new quadrature mirror filter technique allowing to split a signal in sub-bands, with subsequent perfect reconstruction of this signal. The analysis is performed by a tree structure based on a set of non-linear phase FIR filters that split the signal into N equally spaced sub-band signals sub-sampled by 1/N. The synthesis makes use of a set of non-linear phase filters that are linear phase complementary to the analysis filters. This new scheme replaces the shifted sampling technique used in an earlier approach, by a synchronous sampling technique, in such a way that the perfect reconstruction property is retained, with a reduced computational complexity.