Deep Unfolding Basis Pursuit: Improving Sparse Channel Reconstruction via Data-Driven Measurement Matrices

Pengxia Wu, Julian Cheng

For massive multiple-input multiple-output (MIMO) systems operating in frequency-division duplex mode, downlink channel state information (CSI) acquisition will incur large overhead. This overhead can be substantially reduced when compressive sensing techniques are employed to estimate downlink channels, owing to the channel sparsity in angular domain. When a compressive sensing method is implemented, the measurement matrix, which is related to the pilot matrix in channel estimation, is essential to the success of channel reconstructions. Existing sparse channel estimation schemes widely adopt random measurement matrices, which have been criticized for their suboptimal reconstruction performances. This paper proposes to acquire data-driven measurement matrices to improve sparse channel reconstructions. In particular, the model-based deep learning approach is proposed and it exploits the deep unfolding technique. Several autoencoder models called deep unfolding basis pursuit autoencoders are customized, and they treat the measurement matrix as the weight matrix such that the measurement matrix can be optimized by backpropagation algorithms. Numerical results show that the acquired data-driven measurement matrices can achieve more accurate reconstructions and use fewer measurements than the existing random matrices, thereby leading to a higher achievable rate for CSI acquisition in massive MIMO systems.

Knowledge Graph



Sign up or login to leave a comment