Three-Description Scalar And Lattice Vector Quantization Techniques For Efficient Data Transmission

• Main Author: Teo, Hui Ting
• Format: Thesis
• Language: English
• Published: 2015
• Subjects: T Technology; TK1-9971 Electrical engineering. Electronics. Nuclear engineering
• Online Access: http://eprints.usm.my/46909/1/Three-Description%20Scalar%20And%20Lattice%20Vector%20Quantization%20Techniques%20For%20Efficient%20Data%20Transmission.pdf; http://eprints.usm.my/46909/

In twenty-first century, it has been witness the tremendous growth of communication technology and has had a profound impact on our daily life. Throughout history, advancements in technology and communication have gone hand-in-hand, and the most recent technical developments such as the Internet and mobile devices have achieved in the development of communication to a new phase. Majority of researches who work in Multiple Description Coding (MDC) are interested with only two description coding. However, most of the practical applications necessitate more than two packets of transmission to acquire preferable quality. The goals of this work are to develop three description coding system of scalar quantizers using modified nested index assignment technique at the number of diagonals used in the index assignment of two. Furthermore, this work aims to develop three description lattice vector quantizers using designed labeling function in the four dimensional lattice 4 since it offers more lattice points as neighbours that lead the central decoder to achieve better reconstruction quality. This thesis put emphasis on exploiting three description MDC system using scalar quantizers and lattice vector quantizers. The proposed three description system consists of three encoders and seven decoders (including of one central decoder). A three dimensional modified nested index assignment is implemented in the proposed three description scalar quantization scheme. The index assignment algorithm utilizes a matrix, to indicate the mapping process in the proposed three description scalar quantization scheme. As this thesis suggests a new labeling algorithm that uses lattice 4 for three description MDC system. Projection of a tesseract in four-dimensional space of lattice 4 yields four outputs and the data are transmitted via three channels where one of the outputs is defined as time. The three description quantization system is efficient that provides low distortion and good peak signal-to-noise ratio (PSNR) reconstruction quality. The greater the number of diagonals used in the index assignment, k in MDSQ scheme, the higher quality of the central reconstruction can be accomplished. Simulation results show that the central PSNR is promoted to 34.53 dB at rate of 0.1051 bpp and 38.07 dB at 0.9346 bpp for the proposed three description with 2k= Multiple Description Scalar Quantization (MDSQ) scheme. The percentage gain for the central reconstruction quality is improved from 6.36 % to 18.97 % by the proposed three description scalar quantizer which is at 2k= compared to the renownedMDSQ schemes.Moreover, the proposed three description lattice vector quantization (3DLVQ- 4) scheme outperforms the renowned MDC schemes from 4.4 % to 11.43 %. The central reconstruction quality is promoted to 42.63 dB and the average side reconstruction quality inaugurates 32.13 dB, both at bit rate of 1.0 bpp for the proposed 3DLVQ- 4 scheme.