Per-symbol ICI mitigation for frequency non-selective time-varying channels

A time domain window optimized at the symbol level of an OFDM signal is proposed for equalizing the ICI over frequency non-selective max-speed bounded time-varying channels. ICI becomes significant when the time variation over the OFDM symbol exhibits multiple cycle oscillations. At practical speeds of motion, the induced channel time variation has a concentrated Doppler spectrum. Knowing the max-speed bound on moving objects in the channel, a parsimonious discrete prolate spheroidal (DPS) based BEM approximation can be attained. A parsimonious representation exposes the true dimensionality of the channel which in many cases is low compared to the number of samples in the OFDM symbol. Leveraging the low dimensionality, an optimization problem is formulated to attain the time domain ICI-mitigation window. For an OFDM signal, pilot-based LS estimation is compromised due to the contamination of pilots by ICI. Our proposed method exceeds the LS method in performance with channel estimation MSE gains up to 10 dB and 4-fold reduction in SER at 40 dB SNR.