Channel estimation 


Channel estimation is an essential task in wireless communication, with compressed sensing (CS) the channel can be estimated using much fewer received signal samples. For single multi-input multi-output (MIMO) system, the channel can be estimated in virtual beam space domain by solving a regularized least-squares problem. Iterative Shrinkage Thresholding Algorithm (ISTA) is commonly used for solving such problem, however, unrolling ISTA into a neural network allows more benefit from data-driven approach. ISTA-CS-Net is a model-driven neural network channel estimator that is derived from the unrolling ISTA solving CS problem. This project is meant for undergraduate students who have working knowledge of Python for machine and deep learning, with a background in linear algebra.


Build up ISTA-CS-Net in the architecture of multi-layer perceptron (MLP). Train and Test the model using received signal as the input and having estimated coefficient vector as the output of the model. Compare the simulation result with other given conventional channel estimators.