Mutual Regression in Time-Frequency Domain for System Identification: An Online Equalization Scheme for Uplink Massive MIMO FBMCOQAM Systems

In this paper we proposed Mutual Regression in Time-Frequency Domain (MRTFD) for online regression which is aimed at system identification. If the system input information has been located at a limited number of discrete frequencies and is spread over time (multi-carrier signal), then we can only investigate channel behavior at those frequencies. MRTFD is nothing but Inverse Discrete-Time Fourier Transform (IDTFT) in combinationwith the local property of the system input in frequency domain. The MRTFD algorithm acts similar to adaptive filters and is suitable for the situation that channel information is not available. We applied it to an online Equalization scheme for uplink massive MIMO Offset-QAM-based Filter Bank Multi-Carrier (FBMC-OQAM) and compared its performance with full channel information linear receivers. Moreover, we obtained a MSE floor for the MRTFD algorithm based on convex analysis concepts. Simulations show that the MRTFD performance is independent of the number of base station antennas and the correlation between them and the number of users. Moreover, its BER is decreased by increasing the constellation size of PAM modulation in contrast to full channel information linear receivers and finally, all receivers approach to the same performance.