Design of Lossy Compression of the Gaussian Source with Protograph LDPC Codes

Abstract

The low-density parity-check (LDPC) code has been employed for the lossy compression of the binary symmetric source (BSS). However, the BSS application is limited in digital signal processing. In this paper, the BSS is extended to implement the lossy compression of the Gaussian source (LCGS) by using the protograph LDPC (P-LDPC) code. First, to improve the accuracy of the signal reconstruction, a multilevel coding (MLC) structure is designed with multiple P-LDPC codes for realizing the lossy compression. Then, a novel weighted rate allocation method is proposed to deploy an appropriate code rate at each coding level. Furthermore, the P-LDPC codes are optimized by a short cycle diminishing design to approach the theoretical distortion-rate limit. The simulation results demonstrate that the MLC structure based on the optimized P-LDPC codes outperforms the existing works.