Channel estimation and data detection of OFDM systems under unknown channel order doppler frequency: from point-to-point to relaying systems

Abstract

Recently, there has been an increasing demand for OFDM system operating in

high mobility environment. In such situation, wireless channel is both

frequency-selective and time-varying, a.k.a. doubly-selective, making it hard for the

receiver to keep track of the channel state information (CSI). Moreover, the statistical

information of channel, e.g., tap positions, channel length, Doppler shifts and noise

power, is generally unknown to the receiver. In this thesis, two kinds of mobile

OFDM systems are investigated for data detection and channel estimation. Different

from previous works, which highly depend on the statistical information of the doubly

selective channel to deliver accurate channel estimation and data detection results, we

focus on more practical scenarios with unknown channel orders and Doppler

frequencies.

Firstly, point-to-point OFDM system with high mobility is considered. Due to

the unknown channel characteristics, we formulate the channel using GCE-BEM with

a large oversampling factor. The resulted GCE-BEM coefficients are sparse on

delay-Doppler domain and contain only a few nonzero elements. To enable the

identification of nonzero entries, sparsity enhancing Gaussian priors with Gamma

hyperpriors are adopted. An iterative algorithm is developed under variational

inference (VI) framework. The proposed algorithm iteratively estimate the channel,

recover the unknown data using Viterbi algorithm and learn the channel and noise

statistical information, using only limited number of pilot subcarrier in one OFDM

symbol.

Secondly, we investigate multihop amplify-and-forward (AF) OFDM system,

where system structure is generally unknown to the receiver due to the variable

number of hops and relaying paths in high mobility environment. We notice that in AF

relaying systems, the composite source-relay-destination channel is sufficient for data

detection. Then we integrate the multilink, multihop channel matrices into one

composite channel matrix, which turns out to have the same structure as the

point-to-point OFDM channel. The reformulated system model is more concise and a

similar iterative algorithm to that of the point-to-point case can be derived to estimate

the composite channel and detect data. This means that the proposed framework

applies to OFDM system under high mobility regardless of the system structure.

Simulation results show that the performance of the proposed algorithm is very

close to that of the optimal channel estimation and data detection algorithm, which

requires specific information of system structure, channel tap positions, channel

lengths, Doppler shifts as well as noise powers. It is worth noting that, the

close-to-ideal performance of the proposed algorithms is achieved with none of the

above information.