June 4, 2010
Joint Source-Channel Encoding/Decoding Techniques For Reliable Communications
Joint source-channel coding/decoding (JSCC/JSCD) techniques have become state-of-the-art and one of the challenging research subjects in the spatial communication area. This paper addresses the basic principles of joint source-channel optimal design and classification of various existing JSCC/JSCD methods. It presents a JSCD scheme based on variable-length coding, which is capable of providing reliable resolutions and has shown superior performance compared with the existing ones.
The research team in the School of Information Science and Engineering, Graduate University of the Chinese Academy of Sciences (GUCAS) in Beijing summarizes the important achievements in joint source-channel encoding/decoding (JSCC/JSCD) techniques from the aspects of fundamental theory, and flow-media communication applications. This paper also presents a JSCD scheme based on variable-length coding and symbol-level trellis decoding.
Source and channel coders are usually implemented sequentially and independently based on Shannon's well-known separation theory. However, practical communication systems are constrained by complexity and latency. As a result, the separation coding principle does not hold even theoretically in some practical communication systems. Aiming at the limitation of the Shannon separation theory in practical applications, many researchers have focused their attention on the study of joint source-channel coding/decoding (JSCC/JSCD) techniques. Through jointly optimized source and channel parameters, they have achieved a significant number of important results for optimal transmission performance.
Currently, JSCC/JSCD techniques are challenging research subjects, with great theoretical significance and application prospects. On the encoder side, the channel coder is controlled by the source significance information (SSI) from the source coder, which improves the overall encoding efficiency through joint source-channel coding. On the decoder side, the channel decoder utilizes the source a priori information (SAI) from the encoder side and the channel state information (CSI) obtained from the channel estimator, to carry out the joint source-channel decoding. Through the repetitive and iterative measurement and adjustment of the decoder reliability information (DRI) and the a posteriori information (API) between the channel decoder and the source decoder, a minimum joint decoding error ratio is achieved. Existing joint source-channel optimized design can be divided into three classes.
The first is "Joint source-channel coding (JSCC)", which usually focuses on the optimized design of source coding and channel coding on the encoder side. The existing JSCC is included mainly for rate allocation, unequal error protection, optimized quantization, error-resiliency source coding, rate distortion optimization coding and optimal rate control. The channel coder can be controlled by the source significant information (SSI) from the source coder, which helps to improve the overall encoding efficiency. However, joint source-channel coding is not as efficient for JSCD.
The second is "Joint source-channel decoding (JSCD)". It utilizes a priori information (SAI) for bit-level encoding/decoding, which is mainly for fixed-length encoding (FLC), such as codebook-excited linear prediction (CELP). The advantage of the FLC lies in its simple implementation and low complexity. However, the compression is not as efficient. The JSCD based on the FLC is mainly applied to speech coding systems. JSCD can be implemented by exploring source a priori information (SAI) for bit-level encoding/decoding, which helps to improve system performance. One advantage of the algorithm is that it only requires modification on the decoder side, and does not interfere with the source encoder However, without using variable-length coding, the JSCD based on SAI for bit-level encoding/decoding is mainly applied to data coding systems. To improve the robustness of the VLC, the JSCD for variable-length coding with bit-level decoding technology can be used to enhance the transmission performance.
The third is "Variable-length JSCC with variable-length JSCD". For variable-length JSCC with variable-length JSCD, the authors have constructed a new symbol-level joint trellis with compound states by merging a VLC trellis with a convolutional trellis. Based on this joint trellis, a symbol-level a posteriori probability (APP) decoding algorithm is also derived, which leads to a joint iterative decoding approach with symbol-level soft outputs. The experimental results show that the joint source-channel encoding/decoding scheme has obtained better performance than existing joint iterative decoding based on the bit-level super trellis.
Existing joint decoding methods based on turbo codes focus on the traditional source-channel separate encoding model. The modification carries out the JSCD using residual source redundancy on the receiver side. The APP decoding algorithms in current JSCD methods are based on the bit-level trellis. In fact, according to the optimal coding theorem, binary sequences from the source variable-length encoder are composed of several VLC symbols with different lengths. Hence, using the VLC symbols as channel decoding units is a better fit with the source variable-length coding method, with which JSCC can be implemented and system encoding/decoding performance can be enhanced. Based on this idea, the authors present a variable-length turbo encoding/decoding system. On the encoder side, a new hybrid concatenated encoding structure is proposed. The variable-length encoder is sequentially concatenated with the first RSC code of the turbo coder, which is seen as the horizontal constituent code. The second RSC code of the turbo coder is concatenated in parallel as the vertical constituent code through a quantizer and an interleaver. The authors further developed the VLC and the convolutional coding by the proposed joint KT-trellis. This joint trellis representation provides a trellis structure, which can be utilized to decode the module for the VLC combined with the first RSC code by a symbol-level APP algorithm.
A journal reviewer stated that "This paper first introduces the basic principles of joint source-channel coding/decoding technologies and then gives a very comprehensive survey of the current research on the related techniques applied in flow media communications", while another said "This is a review paper which describes the issues of joint source-channel coding in detail".
On the Net: