Joint Data Detection and Phase Noise Mitigation for Light Field Video Transmission in MIMO-OFDM Systems

Omar Hazim Salim, Wei Xiang, Ali A. Nasir, Gengkun Wang, Hani MehrpouyanarXiv 2016 · arXiv

Introduction

Next-generation wireless networks (5G and beyond) are designed to support immersive 3D Light Field (LF) video streaming. LF video captures both spatial and angular information, creating massive datasets that necessitate gigabit-per-second transmission speeds. MIMO-OFDM (Multiple-Input Multiple-Output Orthogonal Frequency Division Multiplexing) is the standard technology to achieve these rates. However, hardware imperfections, specifically oscillator phase noise (PHN), present a significant bottleneck. In MIMO systems where antennas may have independent local oscillators, time-varying PHN introduces Common Phase Error (CPE) and Inter-Carrier Interference (ICI), which can severely degrade video quality.

System Architecture and Methodology

The proposed system addresses PHN through a joint data detection and mitigation framework. As illustrated in the system diagram below, the transmitter encodes the LF video sequence using H.264/AVC and LDPC codes before modulating it via M-QAM and transmitting it through a MIMO-OFDM chain.

Figure 1: Block diagram of the proposed MIMO-OFDM system for LF video transmission, featuring joint data detection and PHN mitigation at the receiver.

At the receiver, an iterative algorithm employing an Extended Kalman Filter (EKF) tracks the non-linear, time-varying phase noise directly from the data. This approach minimizes the need for bandwidth-intensive pilot symbols. The process involves a two-step loop:

  1. Interference Cancellation: The signal from a specific transmit antenna is isolated by subtracting estimated interference from other antennas.
  2. EKF Update: The EKF updates the PHN estimate for the specific antenna pair and refines data detection. This loop repeats until convergence, typically within 2 to 3 iterations at moderate-to-high Signal-to-Noise Ratios (SNR).

Light Field Video Processing

The complex 3D LF image matrix, which contains multiple horizontal and vertical views over time, is processed for transmission. The 3D structure is transposed into a 2D sequence of frames to facilitate standard video encoding.

Figure 2: Illustration of the Light Field video sequence structure and its conversion into a temporal sequence of frames for encoding.

For evaluation, the authors utilized a "Candle" LF video sequence (378 × 378 pixels, 30 fps) captured with a Lytro camera. The sequence features a rotating candle, providing a dynamic test case for the transmission system.

Key Insights and Results

The study reveals several critical findings regarding hardware-aware multimedia transmission:

  • Performance Benchmarks: The proposed detector achieves a Peak Signal-to-Noise Ratio (PSNR) exceeding 30 dB at SNR ≥ 23 dB with 64-QAM. The coded Bit Error Rate (BER) remains within 0.8 dB of a theoretically perfect system at a BER of $10^{-6}$.
  • The Antenna Paradox: While increasing antennas usually improves spatial diversity, high PHN variance can reverse this benefit. The compounding interference from multiple un-trackable oscillators can actually degrade performance, presenting a trade-off between spectral efficiency and hardware quality.
  • Robustness: Unlike traditional pilot-based methods that fail when ICI corrupts the pilots themselves, the joint EKF approach remains robust by estimating data and noise simultaneously.

Limitations and Future Work

Despite its effectiveness, the system encounters "error floors" at extremely high PHN variances (e.g., $10^{-4}$ rad$^2$), where residual estimation errors cap the maximum achievable video quality. Future research will need to address these scaling limits, either through improved oscillator hardware or more advanced mitigation techniques for massive MIMO configurations.

Summary updated

Abstract

Previous studies in the literature for video transmission over wireless communication systems focused on combating the effects of additive channel noise and fading channels without taking the impairments in the physical layer such as phase noise (PHN) into account. Oscillator phase noise impairs the performance of multi-input multi-output- orthogonal frequency division multiplexing (MIMO-OFDM) systems in providing high data rates for video applications and may lead to decoding failure. In this paper, we propose a light field (LF) video transmission system in wireless channels, and analyze joint data detection and phase mitigation in MIMO-OFDM systems for LF video transmission. The signal model and rate-distortion (RD) model for LF video transmission in the presence of multiple PHNs are discussed. Moreover, we propose an iterative algorithm based on the extended Kalman filter for joint data detection and PHN tracking. Numerical results show that the proposed detector can significantly improve the average bit-error rate (BER) and peak-to-noise ratio (PSNR) performance for LF video transmission compared to existing algorithms. Moreover, the BER and PSNR performance of the proposed system is closer to that of the ideal case of perfect PHN estimation. Finally, it is demonstrated that the proposed system model and algorithm are well suited for LF video transmission in wireless channels.
BibTeX
@misc{salim2016joint,
  author = {Omar Hazim Salim and Wei Xiang and Ali A. Nasir and Gengkun Wang and Hani Mehrpouyan},
  title = {{Joint Data Detection and Phase Noise Mitigation for Light Field Video Transmission in MIMO-OFDM Systems}},
  year = {2016},
  eprint = {1602.02834},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1602.02834},
}
Figure