Deep Learning-Based Channel State Information Estimation Technology

With the gradual proliferation of 5G technology and the in-depth research of 6G technology, an increasing number of high-speed mobile scenarios are placing higher demands on mobile networks. In this process, adaptive coding technology has become particularly crucial, as it dynamically adjusts coding schemes based on channel signal-to-noise ratio (SNR) and other information to reduce bit error rates and improve spectral efficiency. Traditional channel state information (CSI) estimation algorithms require the insertion of an increasing number of pilot signals, which not only reduces spectral efficiency but also causes the real-time estimated CSI at the receiver to become outdated in fast-varying channels, significantly impacting the performance of adaptive modulation and coding systems. To address this issue, this paper proposes a deep learning-based CSI estimation model, CTANet, which combines Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks to process channel information while leveraging the Transformer architecture to enhance the SNR estimation obtained through the M2M4 method. By integrating both channel information and SNR data, CTANet model provides more accurate CSI estimation. Experimental results show that, compared to previous deep learning methods, the absolute error of the estimated values is reduced by 1.65.