Blind Known Interference Cancellation

Shengli Zhang, Soung Chang Liew, Hui WangIEEE JSAC 2011 · DOI

TL;DR

This paper introduces Blind Known Interference Cancellation (BKIC), a novel technique for wireless receivers to eliminate known interference without needing to estimate the interference channel. BKIC achieves near-perfect performance, comparable to a system without interference, by cleverly using adjacent symbols and then employing either smoothing or belief propagation to recover the desired signal.

What Problem Does This Paper Address?

In wireless relay networks, a receiver often encounters interference where the interfering signal's data is known beforehand. This is common, for instance, when a relay node retransmits data that was previously sent by another node, and this retransmitted signal overlaps with a new transmission. Traditional methods for dealing with this "known interference" rely on accurately estimating the interference channel. However, this estimation is a significant challenge in real-world wireless environments due to factors like fast fading, low signal-to-noise ratios, and overlapping training sequences, which can lead to substantial performance degradation. This paper addresses the critical need for an interference cancellation method that bypasses the problematic channel estimation step.

Methods

The proposed Blind Known Interference Cancellation (BKIC) scheme operates in two main steps:

  1. Interference Cancellation: This step leverages the observation that wireless channel coefficients remain relatively stable between adjacent symbols. By combining the received signal with a slightly shifted and weighted version of itself, the interference from adjacent symbols can effectively cancel each other out. This clever manipulation exploits the known interference data.
  2. Signal Recovery: The first step, while reducing interference, distorts the desired signal. This second step focuses on recovering the original "Desired Signal plus Noise" (DSN) from this distorted signal. The authors propose two algorithms for this crucial stage:
  3. BKIC-S (Smoothing): This algorithm uses a smoothing principle to recover the DSN. It's simpler and performs well in slow fading scenarios.
  4. BKIC-RBP (Real-valued Belief Propagation): This algorithm employs a novel real-valued belief propagation framework. It achieves MAP-optimal performance, converges quickly, and is robust even in fast fading conditions.

A key feature of BKIC is its compatibility with legacy wireless systems, as it doesn't require modifications to the transmitter or frame structure. It can be implemented as an add-on module in the receiver's signal processing path.

Datasets

Not discussed in the paper. The paper focuses on theoretical analysis and numerical simulations.

Key Insights and Ideas

  • Exploiting Symbol Correlation: The core innovation lies in using the near-constancy of channel coefficients between adjacent symbols to cancel interference without explicit channel estimation. This is a powerful insight that bypasses a major practical hurdle.
  • Blind Cancellation: The "blind" aspect is crucial – the scheme works without needing to know the interference channel's characteristics, making it more robust and practical.
  • Two-Stage Approach: The separation into interference cancellation and signal recovery is elegant. The challenge of signal distortion introduced by the cancellation step is addressed by sophisticated recovery algorithms.
  • Belief Propagation for Real-Valued Signals: The development of a real-valued belief propagation algorithm (BKIC-RBP) for this specific problem is a significant contribution, enabling near-optimal performance in complex fading scenarios.
  • Compatibility with Legacy Systems: The ability to integrate BKIC as an add-on module without altering existing infrastructure is a major practical advantage.

Key Results

  • Near-Perfect Performance: Both BKIC-S and BKIC-RBP achieve performance very close to that of a pure point-to-point communication link without any interference.
  • Superiority over Traditional Methods: BKIC schemes significantly outperform traditional known interference cancellation schemes, even those with perfect initial channel information, by a large margin.
  • Low Complexity: The algorithmic complexities of both BKIC-S and BKIC-RBP are linear with respect to packet length, making them computationally efficient. BKIC-S requires only one multiplication and two additions per symbol.
  • Robustness: BKIC-RBP demonstrates robustness against fast fading and is less sensitive to packet length compared to BKIC-S.
  • Block Fading Performance: In block fading, BKIC-S shows a residual interference that is independent of the interfering signal's power, a desirable trait.
  • Continuous Fading Performance: BKIC-RBP consistently outperforms other schemes, including traditional KIC, in continuous fading channels, with only minor degradation compared to block fading.

Limitations and Future Work

  • Quantization in BKIC-RBP: While BKIC-RBP achieves optimal MAP performance, its real-valued processing requires quantization, which can introduce minor performance losses, especially with larger packet lengths and coarser quantization. The authors note that smaller quantization steps improve performance at the cost of complexity.
  • Channel Variation in BKIC-RBP: Although BKIC-RBP is robust to channel variation, the paper suggests that further improvements could be made by designing new MAC and network layer protocols to fully exploit BKIC's potential.
  • Full-Duplex Communication: The authors highlight the potential of applying BKIC to wireless full-duplex communication, where a node transmits and receives simultaneously, creating self-interference. This is identified as a promising area for future research.
  • Turbo-like Detection-Decoding: The paper suggests that the output of BKIC can be combined with channel decoding for Turbo-like detection-decoding, potentially leading to even better performance.

Summary updated

BibTeX
@article{zhang2011blind,
  author = {Shengli Zhang and Soung Chang Liew and Hui Wang},
  title = {{Blind Known Interference Cancellation}},
  journal = {IEEE J. Sel. Areas Commun.},
  year = {2011},
  doi = {10.1109/jsac.2013.130819},
}
Figure