[Paper Review] Zero-Delay Gaussian Joint Source-Channel Coding for the Interference Channel

Xuechen Chen|arXiv (Cornell University)|Jan 24, 2018

Wireless Communication Security Techniques7 references3 citations

TL;DR

This paper proposes a zero-delay hybrid digital-analog (HDA) joint source-channel coding scheme for correlated Gaussian sources over a Gaussian interference channel, where each transmitter superimposes a scaled version of the source and its scalar-quantized output. The scheme outperforms uncoded transmission in terms of distortion when the channel signal-to-noise ratio exceeds a threshold, depending on correlation and interference levels.

ABSTRACT

This paper studies zero-delay joint source channel coding (JSCC) for transmission of correlated Gaussian sources over a Gaussian interference channel (GIC). We propose to adopt delay-free hybrid digital and analog (HDA) scheme, which is, transmitting the superposition of scaled source and its quantized version after applying scalar quantization to the source at each transmitter. At the corresponding receiver, two kinds of estimators are presented. It is shown that both the schemes, when optimized, beat the uncoded transmission if the channel signal-to-noise ratio (CSNR) is higher than a threshold value for different correlation coefficients and interference values.

Motivation & Objective

• To address the need for low-complexity, zero-delay coding in real-time wireless sensor networks with correlated sources.
• To overcome the limitations of separate source and channel coding (SSCC) in multi-terminal scenarios with interference.
• To design a practical JSCC scheme that leverages source correlation and interference for improved distortion performance.
• To evaluate performance gains of the proposed HDA scheme over uncoded transmission under varying correlation and interference conditions.

Proposed method

• Uses a hybrid digital-analog (HDA) encoding strategy where each transmitter sends a superposition of the original source and its scalar-quantized version.
• Applies midtread scalar quantization to the source, with quantization step Δ and reconstruction levels tk = kΔ.
• Scales the source and its quantized version using coefficients βi and δi, respectively, to form the channel input Xi = δiTi + βiSi.
• Employs two receiver estimation schemes: one based on quantized estimates and another using joint estimation of source and quantized values.
• Derives analytical expressions for distortion by modeling the joint distribution of source, quantized values, and noise, integrating over quantization intervals.
• Uses numerical integration to compute probabilities and expectations over continuous intervals, accounting for interference and noise in the Gaussian interference channel.

Experimental results

Research questions

• RQ1 Under what channel SNR conditions does the proposed HDA JSCC scheme outperform uncoded transmission for correlated Gaussian sources over a Gaussian interference channel?
• RQ2 How does the performance of the HDA scheme vary with different levels of source correlation and interference power?
• RQ3 What is the optimal trade-off between analog and digital components in the HDA transmission strategy to minimize end-to-end distortion?
• RQ4 Can the proposed scheme achieve better distortion than uncoded transmission in a zero-delay setting with interference?
• RQ5 How do the two proposed receiver estimation methods compare in terms of distortion performance under varying channel and source parameters?

Key findings

• The proposed HDA JSCC scheme achieves lower distortion than uncoded transmission when the channel signal-to-noise ratio (CSNR) exceeds a threshold value, which depends on source correlation and interference level.
• For all tested correlation coefficients and interference values, the HDA scheme provides performance gains over uncoded transmission above the SNR threshold.
• The distortion performance of the HDA scheme is sensitive to the choice of scaling coefficients βi and δi, which are optimized to minimize mean-squared error.
• The two proposed receiver estimation schemes—based on quantized and joint estimation—both outperform uncoded transmission when CSNR is sufficiently high.
• Numerical results confirm that the HDA approach effectively exploits source correlation and interference to improve reconstruction quality in zero-delay settings.
• The analytical framework enables precise computation of distortion by integrating over quantization intervals and modeling noise and interference effects through Gaussian distributions.