With the advent of extreme massive multiple-input multiple-output (X-MIMO) systems and the emerging Frequency Range 3 (FR3, 7.125–24.25 GHz) for 6G networks, the dimensionality of channel state information (CSI) has increased significantly. However, to ensure scalable operation and avoid an excessive increase in overhead, the feedback size is expected to increase only marginally. This constraint necessitates an extremely high compression ratio during CSI feedback, which in turn severely degrades the reconstruction performance of standardized codebooks. Moreover, even when employing AI methods, training models from scratch under such severe compression conditions leads to vanishing or exploding gradients, resulting in unstable training and diminished performance. In this paper, we propose a novel hierarchical transfer learning with progressive compression (HTL-PC) method to address these challenges. Our approach leverages a pretrained autoencoder extraction block from a large-feedback model and progressively transfers it to models with smaller latent spaces, thereby enabling stable training under severe compression conditions. Extensive system-level simulations demonstrate that the proposed HTL-PC method achieves remarkable improvements in squared generalized cosine similarity and downlink cell throughput over benchmark techniques.

To mitigate the growing overhead resulting from the increasing number of antennas in massive multiple-input multiple-output (MIMO) systems, various autoencoder-based schemes employing two-sided artificial intelligence (AI) models for channel state information (CSI) feedback-specifically, feedback of the precoding matrix-have been proposed. In these twosided AI models, the encoder and decoder are separately deployed at the user equipment (UE) and the base station (BS), posing significant challenges in maintaining synchronization between them in real-world scenarios with multiple UEs and BSs. In this paper, we propose a novel transformer-based temporaldomain multi-length optimized (TMO) decoder that effectively leverages temporal correlations in MIMO channels to enhance CSI reconstruction. Unlike existing schemes that require different models or additional signaling overhead for re-synchronization between the encoder and decoder under varying conditions, the transformer-based TMO decoder can robustly recover CSI by utilizing feedback of the precoding matrix with temporal information using a single fixed model for inputs of varying sequential time steps. Performance evaluations demonstrate that the proposed transformer-based TMO decoder achieves significantly higher squared generalized cosine similarity and average downlink throughput compared to state-of-the-art techniques.

The application of hyper-wideband is required by 6G sub-THz wireless communication, to enable it provide a peak data rate of up to hundreds of Gbps. However, simultaneously achieving a large bandwidth and high signal quality at the receiver remains a challenge owing to a various hardware non-idealities; hence, additional signal processing techniques are necessary to compensate for these issues. Although conventional analog circuits for the signal compensation cause extra power consumption and hardware implementation complexity, the digital-domain signal compensation technique at the receiver modem exploits both the low power consumption per computation and side-information provided by a sub-THz communication system. In this paper, we propose a digital-domain signal compensation method comprising a compensation filter and parameter estimation, achieving both hyper-wideband and high signal quality. Numerical evaluations were conducted with hyper-wideband and high signal quality scenarios, and the performance of the proposed method was further demonstrated with the spectrum variability of sub-THz systems.

We propose a time synchronization method which is applicable to both offset quadrature amplitude modulation based filter-bank multicarrier (FBMC-OQAM) system and quadrature amplitude modulation based filter-bank multicarrier (FBMC-QAM) system. Based on the relation between the transmitted data symbols and FBMC signal, we design the scattered symbols to make a specific time domain region empty. Then, we insert the preamble in that region so as to transmit it perfectly. The performance of the proposed synchronization method is compared with the conventional time synchronization methods for the FBMC systems in terms of root mean square error (RMSE) of symbol timing offset (STO), bit error ratio (BER) and correctly detected data bit ratio per frame. Since the proposed method can reduce the resource usage for the synchronization, the resources for data transmission can be increased, thereby improving the spectral efficiency of the FBMC systems.