Ph.D. Thesis

My PhD work investigates machine learning methods for wireless communications. Advanced probabilistic inference techniques and the associated graphical model are traditionally used in machine learning, and are recently applied to iterative error control code decoding and scene analysis for digital image processing. It is expected that, in the wireless communications field, this powerful tool can also make a great impact in deriving new and elegant estimation and detection algorithms for many problems which cannot be efficiently resolved by conventional statistical signal processing.

My work focuses on two fronts, as summarized in the following. Please refer to the publication page to review the detailed findings.

1) Frequency Offset and Phase Noise Mitigation in OFDM

Given the recently published IEEE 802.16 standard, OFDM has become a preferred modulation technique in broadband wireless access. However, OFDM suffers from extreme sensitivity to time-varying, multiplicative effects, as a result of the intercarrier interference (ICI) they introduce. These unwanted distortions include the carrier frequency offset (CFO) and phase noise (PHN). To alleviate the detrimental effects of these combined problems, a holistic receiver design must be implemented, with the CFO and PHN taken into account at both the channel estimation stage and the data detection stage. As the first step of the solution, I proposed an optimal (maximum a posteriori) joint estimator for the channel impulse response, CFO and PHN. As such, accurate estimates of the channel and CFO can be obtained in the presence of PHN, and compensated for, before the subsequent data detection stage. (See [J4])

PHN, being time-varying, cannot be removed based on the estimates in the channel estimation stage. Therefore, as the second step of the proposed solution, I derived a near-optimal joint estimator of the random PHN and the desired data signal through an approximate inference technique called variational inference. This novel approach created a paradigm shift for problems of this kind, where rather than generating "hard" estimates of the data and PHN profile, probability distributions for both data and PHN are found at the estimator output, which may be utilized for error control decoding. (See [J3])

2) Unified Framework for Soft-In Soft-Out Detection

I also explored the general problem of joint detection and decoding from the probabilistic inference viewpoint. In particular, I concentrated on the design of soft-in soft-out detectors suitable for turbo receivers. In the absence of a general guiding principle, existing methods often do not allow for insights into design flaws and hence avenues for improvement. The outcome of this research is the establishment of a comprehensive theory, centred on approximate inference, that underlies soft-in soft-out detector designs. A unified treatment of this kind presents rigorous justifications for numerous detectors that were proposed on radically different grounds, and illuminates new and better alternatives. The new framework can be applied to combat multiple access interference (MAI), inter-symbol interference (ISI), and interference in the multiple-antenna (MIMO) environment. (See [J2])

This work can also be extended to the soft-in soft-out detection of Gray coded multilevel modulations. The conventional approach requires the detection of channel symbols first before converting them back to bits for decoding. To avoid the performance loss inherent to such a two-step process, I introduced a bit-level strategy, where the soft information about the bits that make up the symbols are directly estimated. This method combines the detection and demapping operations into one, thereby outperforming the symbol-based method substantially. (See [J1])

Master Thesis

Collision resolution was traditionally viewed as soly a MAC layer issue. From this perspective, various sophisticated access control schemes have been devised. But access control becomes increasingly difficult as the network grows large and as the nodes function in a distributed manner. My master's research was an endeavour to resolve the packet collision problem in a random-access ad hoc network using physical layer techniques. More specifically, the philosophy is to utilize the interference rejection capability of multiuser detectors to allow unscheduled random transmission between any source-destination pair in the network.

My master's research tackles two important problems that enable a multiuser detector to be put into use. First, a spreading code assignment scheme called TRBC is proposed that is particularly suitable for a distributed ad hoc network. Second, a subspace-based active user identification scheme is derived such that each node is able to identify the spreading codes of interest among interfering transmissions. A detailed description of this work can be found in [J5].