Author_Institution :
Dept. of Electr. Eng. & Comput. Sci., California Univ., Berkeley, CA, USA
Abstract :
Feedback may greatly simplify the coding techniques needed to achieve the (information-theoretically) optimum performance. This is true both for the capacity problem and for the joint source-channel coding problem. We are interested in the latter. Gaussian examples involving feedback have appeared in the literature (Cruise, T.J., 1967; Kailath, T., 1967; Schalkwijk, J.P.M. and Bluestein, L.I., 1967). We present a general matching condition for source-channel communication with feedback, extending our previous results for the case without feedback (Gastpar, M. et al., Proc. IEEE Int. Symp. Inf. Theory, p.236, 2000; IEEE Trans. Inf. Theory, 2003). This condition permits, for example, the characterizing of instances of source/channel pairs for which very simple yet optimal feedback coding strategies exist, and leads to an understanding of the potential offered by feedback, and how to exploit it.
Keywords :
Gaussian channels; combined source-channel coding; feedback; radio links; Gaussian source; additive Gaussian noise channel; binary erasure channel; feedback coding strategies; joint source-channel coding; point-to-point link; source-channel communication; Additive noise; Additive white noise; Communication systems; Cost function; Decoding; Distortion measurement; Feedback; Gaussian noise; Noise reduction; Rate-distortion;