Erman Köken

Gaussian HDA coding with bandwidth expansion and side information at the decoder

Lossy transmission of a Gaussian source over an additive white Gaussian (AWGN) channel with side information at the decoder is tackled under the regime of bandwidth expansion. A previously known scheme, hybrid digital/analog Wyner-Ziv (HDA-WZ) coding is shown to remain optimal when extended from the bandwidth matched case to the bandwidth expansion case. The extended scheme also exhibits similar robustness properties under mismatched side information and/or channel quality. Finally, under the criterion of min-max distortion loss, the extended HDA-WZ scheme is shown to outperform a purely-digital scheme known as the common description scheme (CDS).

Energy-Distortion Exponents in Lossy Transmission of Gaussian Sources Over Gaussian Channels

Lossy transmission of Gaussian sources over energy-limited Gaussian point-to-point and broadcast channels is studied under the infinite bandwidth regime, i.e., when the number of channel uses is unlimited. Using previously known asymptotic achievability and converse results, the energy-distortion exponent, defined as the rate of decay of the square-error distortion as the available energy-to-noise ratio increases without bound, is completely characterized for both the point-to-point and broadcast channel cases. Turning then to the scenario of zero-delay transmission, where outage events with arbitrarily small probability are allowed, it is shown that the same energy-distortion exponent as in the infinite-delay case can be achieved in all the studied scenarios.

Joint Source–Channel Coding for Broadcasting Correlated Sources

This paper studies lossy transmission of a memoryless bivariate Gaussian source over a bandwidth-mismatched memoryless Gaussian broadcast channel with two receivers, where each receiver is interested in reconstructing only one source component. For both bandwidth expansion and compression regimes, novel hybrid digital/analog (HDA) coding schemes are proposed. With appropriate choice of parameters, our schemes are shown to specialize to separate source–channel coding studied by Gao and Tuncel, and is, therefore, superior to it in both bandwidth regimes. Our scheme for bandwidth expansion also outperforms the HDA coding scheme of Behroozi et al. On the other hand, if a proposed conjecture (supported by numerical observations) is indeed true, the same superiority follows for the bandwidth compression regime as well. Finally, when the bandwidth expansion/compression ratio approaches 1, both of our schemes become optimal as their performance approaches that of the bandwidth-matched scheme of Tian et al.

On the asymptotic distortion-energy tradeoff for zero-delay transmission of a Gaussian source over the AWGN channel

An achievable scheme for zero-delay transmission of an i.i.d. Gaussian source over an additive white Gaussian noise channel with no bandwidth limitation is introduced, and its energy-distortion performance is analyzed. By the nature of the problem, one must transmit each source sample separately but can use the channel infinitely many times. The proposed scheme builds on separation of source and channel coding, whereby the source is quantized into “equiprobable” cells so that the output can be seen as a message suitable for channel coding. Moreover, as the number of quantization cells go to infinity, the channel capacity can be approached with arbitrarily small error. In the high energy-to-noise ratio regime, the minimum energy required to obtain a given distortion level in the proposed scheme can come as close as 3dB to the Shannon bound, which can only be achieved using infinite delay.

Joint source-channel coding for broadcasting correlated sources

We consider the lossy transmission of a memoryless bivariate Gaussian source over an average-power-constrained bandwidth-mismatched Gaussian broadcast channel with two receivers where each receiver is interested in only one component. We propose new hybrid digital/analog coding schemes which are demonstrated to outperform the previously known schemes.

On the energy-distortion tradeoff for the Gaussian broadcast problem

The energy-distortion tradeoff for the transmission of a white Gaussian source over the additive white Gaussian broadcast channel is investigated by translating the known upper and lower bounds into the infinite bandwidth regime. While a gap continues to exist between the bounds in this regime, it is shown that in a certain region on the distortion plane, the energy difference between the best known upper and lower bounds is quantifiably small.

Spectral Mask Compliance and Amplifier Nonlinearity in Single Carrier and OFDM Systems

In this paper we investige the comparative merits of two transmission techniques, OFDM and single carrier, with respect to spectral mask compliance when a nonlinear amplifier is utilized in transmission. Two different nonlinear power amplifier models are used for analyzing the behavior of the two techniques in terms of average transmitted power and distortion.

On minimum energy for robust Gaussian joint source-channel coding with a distortion-noise profile

Minimum energy required to achieve a distortion-noise profile, i.e., a function indicating the maximum allowed distortion value for each noise level, is studied for robust transmission of Gaussian sources over Gaussian channels. It is shown that for the inversely linear profile, uncoded transmission is optimal. On the other hand, it turns out that exponential profiles are not achievable with finite energy. Finally, using a family of lower bounds and a proposed coding scheme, the minimum energy behavior for the square-law profile is understood up to a multiplicative constant.

On the distortion-energy tradeoff for zero-delay transmission of a Gaussian source over the AWGN channel

An achievable scheme for zero-delay transmission of an i.i.d. Gaussian source over an additive white Gaussian channel with no bandwidth limitation is introduced, and its energy-distortion performance is analyzed. By the nature of the problem, one must transmit each source sample separately but can use the channel infinitely many times. We introduce an outage concept, and analyze the expected distortion conditioned on no outage. We show that the proposed scheme can approach to the asymptotical decay for large enough energy for arbitrary outage probability. The proposed scheme builds on separation of source and channel coding, whereby the source is quantized with a high-resolution optimal quantizer. In the high energy-to-noise ratio (ENR) regime, the minimum energy required to obtain a given distortion level in the proposed scheme can approach arbitrarily close the Shannon bound, which can only be achieved using infinite delay.

On robustness of hybrid digital analog source-channel coding with bandwidth mismatch

For transmission of memoryless Gaussian sources over channels with additive white Gaussian noise, the tradeoff between the distortion when the channel quality is good versus bad is investigated under the constraint that the distortion is optimal when the channel has the targeted median quality. The problem was proposed by Tian and Shamai, who subsequently proved remarkable achievability results for bandwidth expansion ratios k, which are integers or unit fractions, i.e., k = 2,3,4, ... or k =1/2,1/3,1/4,... Their result is extended here to all k≥2 and k≤1/2 by generalizing the hybrid digital/analog (HDA) scheme of Wilson et al. to the case of bandwidth mismatch. Finally, novel schemes are proposed for 1/2≤k<;1and 1<;k≤ 2 achieving nontrivial tradeoffs outperforming all known schemes. These latter schemes rely on another extension of the HDA scheme by Wilson et al., namely, the relaxation of the independence of source and channel input sequences.