Joint source-channel decoding for robust transmission of video

Motivated by the need for robust transmission of video, an approach to joint source-channel decoding is presented. While this work is focused on transmission of video, the results are applicable to most forms of multimedia. A general overview of the proposed video coding system is given and its individual components are introduced. Since correct reception of synchronization patterns is essential for successful decoding of compressed bitstreams, detection of aperiodically embedded synchronization patterns is first considered for the AWGN and Rayleigh fading channels and detection metrics are provided for both the coherent and noncoherent transmission. It is shown that significant improvements as compared to the standard detection methods are possible with the derived metrics. Next, the problem of joint source-channel decoding is considered and results on joint source-channel decoding of overhead information and variable-length codes are presented. These results show that by exploring the structure imposed by the standard coders, such as that of MPEG-4 video, improved decoding of coded bitstreams is possible. Motivated by these decoding results a new class of error-resilient variable-length codes, called channel optimized source codes, is described and experimental results for these codes are presented.