Distributed signal processing using nested lattice codes
Author(s)
Gao, Su
Type
Thesis
Abstract
Multi-Terminal Source Coding (MTSC) addresses the problem of compressing correlated sources
without communication links among them. In this thesis, the constructive approach of this problem
is considered in an algebraic framework and a system design is provided that can be applicable
in a variety of settings. Wyner-Ziv problem is first investigated: coding of an independent and
identically distributed (i.i.d.) Gaussian source with side information available only at the decoder
in the form of a noisy version of the source to be encoded. Theoretical models are first established
and derived for calculating distortion-rate functions. Then a few novel practical code implementations are proposed by using the strategy of multi-dimensional nested lattice/trellis coding. By
investigating various lattices in the dimensions considered, analysis is given on how lattice properties affect performance. Also proposed are methods on choosing good sublattices in multiple
dimensions. By introducing scaling factors, the relationship between distortion and scaling factor
is examined for various rates. The best high-dimensional lattice using our scale-rotate method can
achieve a performance less than 1 dB at low rates from the Wyner-Ziv limit; and random nested
ensembles can achieve a 1.87 dB gap with the limit. Moreover, the code design is extended to
incorporate with distributed compressive sensing (DCS). Theoretical framework is proposed and
practical design using nested lattice/trellis is presented for various scenarios. By using nested
trellis, the simulation shows a 3.42 dB gap from our derived bound for the DCS plus Wyner-Ziv
framework.
without communication links among them. In this thesis, the constructive approach of this problem
is considered in an algebraic framework and a system design is provided that can be applicable
in a variety of settings. Wyner-Ziv problem is first investigated: coding of an independent and
identically distributed (i.i.d.) Gaussian source with side information available only at the decoder
in the form of a noisy version of the source to be encoded. Theoretical models are first established
and derived for calculating distortion-rate functions. Then a few novel practical code implementations are proposed by using the strategy of multi-dimensional nested lattice/trellis coding. By
investigating various lattices in the dimensions considered, analysis is given on how lattice properties affect performance. Also proposed are methods on choosing good sublattices in multiple
dimensions. By introducing scaling factors, the relationship between distortion and scaling factor
is examined for various rates. The best high-dimensional lattice using our scale-rotate method can
achieve a performance less than 1 dB at low rates from the Wyner-Ziv limit; and random nested
ensembles can achieve a 1.87 dB gap with the limit. Moreover, the code design is extended to
incorporate with distributed compressive sensing (DCS). Theoretical framework is proposed and
practical design using nested lattice/trellis is presented for various scenarios. By using nested
trellis, the simulation shows a 3.42 dB gap from our derived bound for the DCS plus Wyner-Ziv
framework.
Date Issued
2011-12
Date Awarded
2012-02
Copyright Statement
Attribution NoDerivatives 4.0 International Licence (CC BY-ND)
Advisor
Ling, Cong
Publisher Department
Electrical and Electronic Engineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)