Superimposed training aided channel estimation in OFDM system

Superimposed training has raised lots of attentions due to its great spectral efficiency and its relatively fast channel estimation algorithms. In this paper, we present channel estimation methods for OFDM system using periodic superimposed training added in frequency domain. At the receiver, channel estimation is done both in time domain (pre-FFT) and in frequency domain (post-FFT). We prove that the estimation asymptotically converges to the MMSE solution. Compared to subspace based blind channel estimation methods, this one manifests higher precision especially when SNR is lower and data length is short. In addition, peak to average (PAR) problem is analyzed with superimposed training, which will act as a guide to training arrangement. Simulations testify the effectiveness of these proposed algorithms.