Sparse Channels Structured Estimation in OFDM Systems

The problem of sparse channels structured estimation is here addressed in the frame of Orthogonal Frequency Division Multiplexing (OFDM) communications. In our approach, the Channel Impulse Response (CIR), to be estimated, is constrained to fit a beforehand detected structure specifying the nonzero valued taps positions. In this contribution, a newly proposed thresholding based scheme, used for CIR structure detection, is compared to the Generalized Akaike Information Criterion (GAIC) method. The tap-dependent thresholds are optimized to minimize the detection error probability, identified to the average of active tap missing and erroneous tap detection probabilities. The proposed scheme is found to outperform the GAIC method with noticeably lower computational load. The performance is evaluated in terms of true structure detection rate and of Channel Frequency Response (CFR) estimation bias and Mean Square Error (MSE).