Blind Channel Estimation and Tracking of Massive MIMO THz Wireless System Based on PA-OSTBC Coding and Deep Learning

Terahertz (THz) wireless communication offers immense promise for ultra-high data rate applications due to its large available bandwidth. However, it faces significant challenges, primarily high path loss and atmospheric absorption, as well as the difficulty of acquiring accurate channel state information (CSI) due to unknown and rapidly varying channel parameters. This work addresses these limitations through the integration of newly proposed techniques, including Massive MIMO, blind channel estimation, and a novel coding framework. Massive MIMO enhances spatial multiplexing and link robustness using large antenna arrays, while blind estimation methods enable CSI recovery without reliance on pilot symbols, thereby reducing overhead. To further support reliable communication, we propose a Phase Algebraic Orthogonal Space-Time Block Code (PA-OSTBC), which combines a rotation matrix with the conventional Alamouti scheme, embedding a phase distinction that improves block separation and aids blind initial channel estimation. The code's orthogonality also enables simplified decoding. Additionally, Long Short-Term Memory (LSTM) neural networks are employed to model temporal channel dynamics, enhancing tracking accuracy over time. The proposed framework significantly improves estimation performance and decoding efficiency, marking a critical step toward practical THz Massive MIMO deployment.