Machine learning empowered context-aware receiver for high-band transmission

With the 5G Evolution/6G, the carrier frequency is expected to increase compared to the current 4G/5G operations. Transmission at higher carrier frequency provides new opportunities for larger spectrum allocations. However, transmission at higher frequencies is more challenging due to the impact of hardware impairments such as oscillator's phase noise on the performance of communication systems. At these frequencies, in one hand, phase noise power increases as the carrier frequency increases, which can lead to detrimental impact on performance. On the other hand, design of oscillators with low phase noise at high frequencies is complex which can lead to an increase in production cost. Using machine learning techniques to model and compensate for the phase noise at the receiver can significantly improve performance and/or allow for low-cost hardware components in future transmitters/receivers. Considering that base stations have more computational resources compared to user equipment, receivers in uplink transmission scenarios would be suitable for the operation of machine learning techniques. For high-band transmission, DFT-s-OFDM is a candidate modulation scheme for uplink transmission. This paper proposes a context-aware receiver for DFT-s-OFDM based transmission to perform soft bit computation based on trained models. The paper presents performance evaluation of the proposed techniques using link level simulations. In addition, the paper provides the performance versus complexity trade-off based on a detailed complexity analysis of the methods. The results show that the proposed techniques provide large performance gains in scenarios where the radio link is severely phase noise limited while keeping complexity at reasonable levels.