Neural Network Based Joint Carrier Frequency Offset and Sampling frequency Offset Estimation and Compensation in MIMO OFDM-OQAM Systems

A derivative of orthogonal frequency division multiplexing, (OFDM) based on offset quadrature amplitude modulation (OQAM), popularly known as OFDM-OQAM, is one of the dominant waveform contenders for 5th generation wireless communication systems. In this paper we investigate on channel estimation for OFDM-OQAM modulations in the case of multiple-input multiple-output (MIMO) transmissions over radio channels. We propose neural network (NN) based joint CFO and SFO estimation along with filter bank compensation in MIMO OFDM-OQAM systems. Here CFO and SFO ranges are evenly divided into a set of sub-ranges, which are being estimated using three layer NN based classifier and being supported by a dedicated CFO and SFO compensation unit. The performance evaluation shows that the proposed joint estimation and compensation technique gives the mean square error value $(10^{-4})$ even at a very low $SNR=0dB$ with $N_{T}=2$ and $N_{R}=4$.