Efficient channel coding for analog sources using chaotic systems

We explore the application of chaotic sequences for encoding and transmission of analog sources over channels with unknown or multiple signal-to-noise ratios, as occur in broadcast and fading scenarios. Lower bounds on the mean-square distortion are derived for codes based on so-called tent-map dynamics, and are compared with those of other codes. For additive white Gaussian noise channels, we show there always exists a power-bandwidth regime in which this code yields lower distortion than any digital (i.e., finite-alphabet) code. We also develop and evaluate three practical decoding algorithms for efficiently exploiting these new codes on intersymbol interference channels.