Adaptive Pruning in Compressed Sensing Channel Estimation for OFDM System Enabled by a Novel Loss Function

Channel estimation in an Orthogonal Frequency Division Multiplexing system using compressed sensing theory allows for a significant reduction in the number of pilot subcarriers and improves the utilization of spectrum resources. Currently, a large number of greedy algorithms have been developed, many of which have proven to be highly effective for solving this problem. However, these algorithms require prior knowledge of the channel sparsity order. Additionally, greedy algorithms typically use such metrics as error functions, that do not allow to set reliable threshold for algorithm stopping criteria. Information about sparsity may not be available in real communication system, and the number of paths can change with time. Using the standard error function metrics for channel estimation is insufficient, as it can often lead to overfitting of the algorithm and degradation in the quality of the reconstructed channel impulse response. Adaptive Pruning Orthogonal Matching Pursuit was proposed in this work for channel estimation in OFDM systems. The algorithm is adaptive and does not require prior knowledge of the channel sparsity. APOMP is based on the Orthogonal Matching Pursuit algorithm, which itself is not adaptive. Adaptability to channel sparsity is achieved by introducing a new loss function. This loss function utilizes more available information, thereby avoiding overfitting and enabling the determination of the sparsity order of the channel impulse response. The effectiveness of the algorithm was verified through simulation using the QuaDRiGa channel generator.