New Channel Estimation Method Using Singular Spectrum Analysis for OFDM Systems

Abstract

This paper presents a new method for OFDM channel estimation (CE) using singular spectrum analysis (SSA). In this method, the conventional LMMSE procedure is used for CE and then SSA based optimum low rank approximation is performed on channel correlation matrix. Detailed simulations for bit error rate and mean square error have shown good performance for this algorithm compared to the well-known LSE, MMSE, and SVD based CE methods. In the proposed method, the SSA is also employed for denoising the signal aimed at further improvement in the performance of CE. SSA decomposes the received OFDM symbol into empirical orthogonal functions (EOFs). A de-noised signal is reconstructed using selective EOFs based on Pearson correlation coefficients. The proposed method has outperformed LSE and MMSE based methods and shown an average 2 dB improvements over SVD based method.