Source Turbo Encoder Channel Sink Turbo Decoder BPSK

Abstra t | In this work, we investigate the joint sour ehannel oding issue of transmitting non-uniform independent and identi ally distributed (i.i.d.) sour es over additive white Gaussian noise (AWGN) hannels via Turbo odes. The sour e redundan y in the form of non-uniformity is exploited in the Turbo de oder via a modi ed extrinsi information. In ontrast to previous work, non-systemati re ursive onvolutional en oders are proposed as the onstituent en oders, whi h produ e almost uniform outputs for heavily biased sour es. As a result, unlike the outputs of systemati en oders, they are suitably mat hed to the hannel input sin e a uniformly distributed input maximizes the hannel mutual information and a hieves apa ity. Simulation results show substantial gains a hieved over previously designed systemati Turbo odes, and the gaps to the optimal Shannon limit are therefore signi antly redu ed.