Hossein Ghanei Yakhdan - Department of Electrical and Computer Engineering, Yazd University, Yazd, Iran

Error concealment (EC) is an important technique to recover the lost/damaged video data in transmitting video over error prone networks. A number of EC techniques have beendeveloped to combat the transmission errors. However, the previous techniques are always inefficient when the motions ofthe video object are irregular. This paper proposes a novel dynamic temporal EC approach to conceal the errors due to the motion vectors (MVs) of the damaged macroblocks (MBs) forvideo communications. The proposed method employs a competitive neural network (CNN) as a predictor to estimate theMVs of the damaged MBs, and that can exploit the nonlinearity property of the neural networks to estimate lost MVs moreaccurately. Simulation results show that the proposed technique enhances both subjective and objective quality of reconstructed frames, such as the average PSNR increases about 1.03 dBcompared t o t he BMA m ethod f or t he t est v ideo s equences in some frames.

temporal error concealment, competitive neural network, motion vector estimation, motion compensation