Joint source/channel coding for wireless channels

Abstract

Shannon's (1948) fundamental theorem showing that source coding and channel coding can be separated without any loss of optimality does not apply to general time-varying channels. Since the distortion by the source encoder decreases with the data rate, while the channel errors increase with the data rate, the joint source/channel coding problem reduces to allocating bits in an optimal way between the source and channel encoders as the source and channel vary. The author introduces two additional degrees of freedom by allowing both the transmit power and the data rate to vary, subject to an average power constraint. Under these varying power and rate conditions, he first obtains an expression to minimize the end-to-end distortion of general joint source/channel codes for fading channels. He then proposes an adaptive joint source/channel coded modulation technique. The channel code adapts both the transmission rate and power using variable-rate coded MQAM (on a Rayleigh fading channel). He analytically derives the minimum end-to-end distortion of our joint coding scheme. The solution cannot be obtained in closed form, and therefore requires computer search methods. He also obtains a simple upper bound on the distortion by holding the channel error rate constant. Numerical results for this distortion upper bound as a function of the channel coding gain and error rate are obtained. The optimal power control which achieve this bound is also determined.