Large Transformer Model With Haar Wavelet Transform for Channel Estimation in Massive MIMO Systems

In recent scenario, the accurate Channel State Information (CSI) is essential for reliable performance of massive Multiple Input Multiple Output (MIMO) systems, particularly in spatially correlated channels of wireless communications. Traditional channel estimation techniques heavily relied on the pilot symbol transmission from all antennas which resulted in pilot overhead. Therefore, this research proposes Large Transformer Model with Haar Wavelet Transform (LT‐HWT) for channel estimation in massive MIMO systems. The proposed LT‐HWT model captures both spatial and temporal dependencies by processing the sequences of partial CSI with a dual‐stage encoder. Initially, the temporal patterns are extracted and then integrated with spatial embeddings to reduce the missing inference of CSI. Consequently, the pretrained transformer which is a large model decoder is utilized for robust sequence modeling and improved generalization. The proposed LT‐HWT model significantly reduced the pilot overhead while maintain high estimation accuracy with high results when compared to existing Deep Neural Network (DNN) in terms of Bit Error Rate (BER) of 0.955 and throughput of 0.891, respectively.