Abstract
An iterative turbo decoder based cross layer error recovery scheme for compressed video is presented in this paper. The soft information exchanged between two convolutional decoders are reinforced both by channel coded parity and video compression syntactical information. An algorithm to identify the video frame boundaries in corrupted compressed sequences is formulated. The paper continues to propose algorithms to deduce the correct values for selected fields in the compressed stream. Modifying the turbo extrinsic information using these corrections act as reinforcements in the turbo decoding iterative process. The optimal number of turbo iterations suitable for the proposed system model is derived using EXIT charts. Simulation results reveal that a transmission power saving of 2.28% can be achieved using the proposed methodology. Contrary to typical joint cross layer decoding schemes, the additional resource requirement is minimal since the proposed decoding cycle does not involve the decompression function.