An Efficient Bit Allocation Scheme for Weighted Random Graph Signal Sampling and Quantization
Bin Wang, Lin Wang, P. Takis Mathiopoulos · 2020
We study the quantization effect and bit allocation problem for the reconstruction of bandlimited graph signals from a subset of nodes, selected by a weighted random sampling strategy. Under a limited total rate condition, we design an optimal rate allocation scheme for the sampling set which minimizes the reconstruction error due to quantization. However, this scheme requires the estimation of the eigenvectors of a large graph Laplacian matrix. In order to reduce the computational complexity associated with such matrix calculation, a computational efficient method was then proposed. By considering various graph examples and by means of computer simulations, it is shown that the proposed scheme achieves better performance than the uniform rate assignment especially under a 1-bit restrict condition.