Windowed DFT-based Channel Estimation for 5G

DFT-based channel estimation has been widely studied for its excellent denoising ability in OFDM systems. But in 5G standard, the channel is usually non-sample-spaced which cause the energy leakage of channel impulse response. In addition, Bandwidth Part (BWP) also destroys the orthogonality of DFT and leads to the Gibbs phenomenon. These makes the original DFT-based estimators experience a MSE floor at high SNRs. To solve this, we proposed an improved estimators based on spectrum extrapolation and frequency domain windowing which are conducive to denoise in time domain. The simulation results show that the proposed method eliminates the MSE floor and improves the performance with limited complexity.