A compressed sensing technique for OFDM channel estimation using full-band training symbols

Based on compressed-sensing (CS) technique, a channel estimation scheme OFDM system for double-selected channel is proposed in this work. The variation of each channel path is fitted using a polynomial instead of the Doppler spread model, where a more flexible assumption is allowed, and the sparsity required for the CS technique is still available since each can be characterized using few parameters. In order to mitigate the inter-carrier interference (ICI) and observe the channel state information (CSI) more precisely, full-band training symbols are introduced instead of pilot tones. Combining the full-band training symbols and the polynomial fitting, a determinate measurement matrix is derived for the proposed CS channel estimation constructed using the orthogonal matching pursuit (OMP), and a more accurate channel estimation results can be provided comparing to that based on the conventional least-square (LS) algorithm. The case of multiple transmission antennas is also evaluated, and the improvement can be more significant.