Designs of Bipartite Graph Filter Bank Using Graph Fourier Transform and Digital Filter Bank

In this paper, the design of two-channel bipartite graph filter bank is studied. First, the perfect reconstruction two-channel filter bank for bipartite graph is reviewed. For a small-size graph, the filter bank can be implemented by using graph Fourier transform (GFT). For a large-size graph, the filter bank is implemented by using polynomial graph filters which can be realized by distributed computation method. Next, a conversion method is presented to design graph filter bank by using FIR digital filter bank. The optimal parameter of filter conversion design is determined by minimizing the filter approximation errors in spectral domain. Finally, some design examples are demonstrated to show the effectiveness of the conversion design method.