Iterative H.264 Source and Channel Decoding Using Sphere Packing Modulation Aided Layered Steered Space-Time Codes

Abstract

Abstract-I. MOTIVATION AND BACKGROUND Shannon's source-channel separation principle [1] states that under idealised, but only asymptotically valid conditions source coding and channel coding schemes may be designed independently. However, in reality the source codes designed without taking into account the presence of channel decoding errors tend to have a poor performance, owing to the finite length, finite complexity source and channel codes employed. In a high compression source codec a low number of erroneous bits in an entropy coded sequence typically is mapped to a high video distortion in the reconstructed sequence. This results in low target error rate requirements for channel coding, in order to prevent the source coded video from becoming unduly impaired. In such a situation, joint source-channel decoding (JSCD) results in substantial performance improvements at a given channel coding rate, by exploiting any residual redundancy inherent in the resultant bit-stream after source coding. In recent years, the turbo-principle was extended to JSCD by invoking iterative source-channel decoding (ISCD) by iteratively exchanging extrinsic information between two or more concatenated codes. In [2,3], it was revealed that ISCD schemes indeed exhibit superior bit-error correction capability incomparison to those of the non-iterative schemes. However, the number of useful iterations is typically limited to two or three iterations [4,5] due to the limited residual redundancy left in the source coded stream. In [4], it was shown that the achievable ISCD performance can be improved by introducing artificial redundancy in the source coded stream. Therefore, in our proposed scheme we advocate a novel class of short block codes [6] invoked, in order to improve the attainable iterative decoding performance. A turbo-detected multidimensional Sphere Packing (SP) modulation scheme using two transmit antennas combined with Differentially encoded Space-Time Spreading (DSTS) is presented in [7]. DSTS is a non-coherent MIMO scheme capable of providing a diversity gain, while attaining no multiplexing gain for the case of two transmit antennas and a throughput loss in the four transmit antennas case compared to a single transmit antenna system. However, throughput improvements can be achieved using a multi-layer MIMO structure designed for achieving a multiplexing gain, which is known as the vertical bell labs layered space-time (V-BLAST) scheme [8]. More explicitly, the advantage of V-BLAST is that it is capable of providing an increased effective bit-rate without any increase in the transmitted power or in the system's bandwidth. By contrast, a high transmit diversity gain can be achieved using Alamouti's low-complexity spacetime code (STC) [9]. On the other hand, beamforming [10] provides an effective technique of reducing the multiple-access interference (MAI), where the antenna gain is increased in the direction of the target user, while reducing the gain towards the interfering users. In order to combine the benefits of the ...

Averting BER Floor with Iterative Source and Channel Decoding for Layered Steered Space-Time Codes

“…Most often an interleaver is deployed before modulating the encoded stream over a wireless channel [ 14 ], and we have depicted that second interleaver as in Figure 1 . We have used SP modulator for the purpose of exploiting the joint space-time symbols of the LSSTC scheme by efficient selection of single point [ 7 ]. Furthermore, the SP modulator is very much beneficial for the best possible minimum Euclidean distance in the real-valued space for constructing efficient error correcting codes [ 23 ].…”

“…SBSD exploits the natural redundancy to improve the convergence behavior of iterative decoding. SBSD gleans the residual redundancy for generating extrinsic LLRs, and hence beneficially transforms the coded video into short time frames, labelled as index [ 7 ]. This SBSD operation results in significant performance improvement of the iteratively decoded system and is a modified version of the SBSD presented in [ 47 , 48 ].…”

“…The chief pitfall associated with independent designs is the poor performance due to the finite length and complexity of source and channel codes deployed, deriving low target error rates [ 4 ]. The technique of joint source-channel decoding (JSCD) was formulated [ 5 ] for improvements in performance by utilizing the residual redundancy in source coded stream [ 6 , 7 ].…”

“…Until now, turbo codes result in the highest performance and are deemed as the best codes with little required encoding and decoding complexity [ 8 ]. The principle of JSCD was adopted in turbo coding by iteratively exchanging information between the constituent decoders and referred to as iterative source-channel decoding (ISCD) [ 7 ]. Systems with ISCD exhibited lower bit-error rate (BER) and better performance gain than the systems with non-iterative schemes [ 9 , 10 , 11 ].…”

“…These techniques are simultaneously employed in order to enhance the reliability of compressed information at the destination with lower requirements. The system is referred to as layered steered space-time code (LSSTC), as demonstrated in [ 7 , 33 ]. We have utilized iterative decoding for enhancing the performance of our proposed three-stage LSSTC system.…”