Zero-Delay Joint Source-Channel Coding in the Presence of Interference Known at the Encoder

TL;DR

This paper investigates zero-delay transmission of a Gaussian source over an AWGN channel with known interference at the encoder, proposing non-linear schemes like ICO and 1DL, and demonstrating that non-uniform quantization improves performance.

Contribution

It introduces non-linear transmission schemes for joint source-channel coding with interference, and shows non-uniform quantization enhances performance over traditional scalar methods.

Findings

1DL with non-uniform quantization performs close to the numerically optimized encoder.

Non-uniform quantization improves performance compared to uniform quantization.

Proposed schemes outperform simple linear transmission in the presence of known interference.

Abstract

Zero-delay transmission of a Gaussian source over an additive white Gaussian noise (AWGN) channel is considered in the presence of an additive Gaussian interference signal. The mean squared error (MSE) distortion is minimized under an average power constraint assuming that the interference signal is known at the transmitter. Optimality of simple linear transmission does not hold in this setting due to the presence of the known interference signal. While the optimal encoder-decoder pair remains an open problem, various non-linear transmission schemes are proposed in this paper. In particular, interference concentration (ICO) and one-dimensional lattice (1DL) strategies, using both uniform and non-uniform quantization of the interference signal, are studied. It is shown that, in contrast to typical scalar quantization of Gaussian sources, a non-uniform quantizer, whose quantization intervals become smaller as we go further from zero, improves the performance. Given that the optimal decoder is the minimum MSE (MMSE) estimator, a necessary condition for the optimality of the encoder is derived, and the numerically optimized encoder (NOE) satisfying this condition is obtained. Based on the numerical results, it is shown that 1DL with non-uniform quantization performs closer (compared to the other schemes) to the numerically optimized encoder while requiring significantly lower complexity.