Abstract
ECC (Error-Correcting Codes) is used to recover loss data due to error in network data transmission. The traditional ECC has fixed code rate, the code rate will affect the ability of error correction, if channel condition is so bad that the quantity of data loss is too large to recover, the overall system decoding performance will drop quickly due to error propagation in P2P streaming system. In the thesis, we use fountain code combine with P2P system, fountain code doesn’t have the fixed code rate, so there is no problems of error checking and correction. We proposed a decoding distribution which is an approximation to statistic results for fountain code, and we put the distribution’s information into our distributed decoding probability predicted model which considers error propagation. Furthermore, we can do peer selection by using the predicted decoding probability when the upload bandwidth is not enough. Our experimental results show that the proposed predicted model has high accuracy, and the overall system decoding probability also increases after our peer selection method.