Channel Estimation by Tensor-Train Deep Neural Networks for OCDM Communications over Time-varying Channels
Dongzhuo Lu, Yiyin Wang · 2023
Orthogonal chirp division multiplexing (OCDM) has been introduced as an innovative modulation scheme, which is characterized with dual-spreading in both time and frequency domains. In scenarios involving swift mobile communication, channels are time-varying and encompass diverse multipath components, which often degrade communication performance. Therefore, channel estimation of OCDM systems on time-varying multipath channels is a challenging task. In this paper, a tensortrain deep neural network (TT-DNN) based channel estimation method is developed to address this challenge. The proposed TT-DNN based method reduces hardware demands by decomposing hidden layers into tensors with fewer parameters, relaxes the prerequisites for pilot symbols and enhances bandwidth efficiency. Simulation results validate the effectiveness and superior performance of our TT-DNN based approach in comparison to the state-of-the-art.