Research On Adaptive Demodulation Based On Fountain Codes

Fountain code is a rateless code. The encoding packets that can be generated from the source packets on the fly are limitless. The destination can recover the source packets with high probability as long as it receives a sufficient number of encoding packets. To employ this advantage of fountain codes, we present a novel adaptive demodulation(ADM) scheme.In the ADM scheme, the transmitter would send a fixed high-level modulated and a fixed coded data stream, and the receiver would demodulate the symbols at a nonfixed rate, where the demodulation rate at any condition would be determined by the observed channel state information at the receiver. Thus, the receiver would demodulate only those bites which have higher degree of certainty, and would treat those which have an increased chance of being error as erasures. The ADM preserves the advantages of rate adaptive schemes, while avoiding a feedback path.In this paper, our ADM is designed based on LT codes and Raptor codes. We propose a low-complexity soft demapper and three algorithms to select the bits to be erased. The first method is called Decision Region Set, in which the selection is in each symbol. The second method is called Free Demodulation Rate, in which the selection is in the data packet s, therefore, the probability of discarding incorrect bit will increase, resulting in bit error rate(BER) performance improvement. By setting threshold we also propose the third method called Threshold Set. In the third method, we get the threshold of different channel state information by simulating. In practical applications, we can compare the threshold with the calculated likelihood ratio for each bit to decide whether to delete it or not. The third method decreases the computing load. In our paper, all of these three methods are simulated over an additive white Gaussian noise channel, and our simulation results also demonstrate that the methods outperform the conventional system in terms of BER.