This dissertation focuses on channel estimation and symbol detection in two-way relay systems. First, we assume that channels exhibits block flat fading and derive generalized likelihood ratio test (GLRT) detector along with a convergent iterative (blind) joint estimation and detection (JED) method to resolve its computational complexity. We also propose simple least squares (LS) and weighted LS (WLS) estimators, for low SNRs. Then, we proceed to temporally correlated channels and orthogonal signaling to propose a channel estimation technique without pilot assistance. In this method, users apply known periodic phase shifts to their transmit symbols and adopt maximum likelihood (ML) or linear minimum mean square error (LMMSE) criteria to estimate the channels. Optimal and suboptimal phase shifts are designed to minimize MSE. Third, we assume the channels are fast Rayleigh fading and, derive the optimal detector and two closed-form suboptimal structures to circumvent its hindering numerical integration. We investigate non-coherent detection in a denoise-and-forward (DNF) relay scheme and, design the optimal and suboptimal detectors for fast fading channels. With additional practical orientation we derive optimal detector for (amplitude) block fading channels and, propose efficient estimators for the amplitudes. Key Words Amplify-and-Forward, Channel Estimation, Denoise-and-Forward, Detection, Two-Way
