Joint source-channel decoder for data-centric network applications

Data-centric networks are gaining importance because of their application in several areas such as sensor networks. One of the key challenges in such networks is the resource constraint under which the nodes must operate. Under these circumstances, the trend is to shift a significant percentage of the computational burden from the individual nodes to a central decoder. The problem of error-resilient communications in data-centric networks is addressed by taking a joint source-channel decoding (JSCD) approach. This improvement in error resilience is achieved by using the source statistics at the decoder, thereby ensuring that the encoder complexity is not increased. Specifically, an optimal JSCD is designed for multiple-access source code (MASC)-encoded Markov sources. This is the first known attempt at designing a JSCD for the general case of correlated sources with memory. Simulation results demonstrate that the proposed decoder outperforms the traditional MASC decoder by up to 10dB in some cases. The results indicate that significant error resilience is possible even without using forward error-correcting codes at the encoder, which is important in encoder-resource constrained applications.