Hyperparameter Optimized DNN for OFDM Signal Detection in the Presence of CFO and PO

In any multi-carrier modulation (MCM) system, it is crucial to have accurate channel estimation. Typically, the choice of channel state information (CSI) based on the least-square (LS) and minimum-mean square error (MMSE) estimators has been the traditional approach. With the advancement of deep learning algorithms, it has become a state-of-the-art algorithm for signal detection. However, deep learning comes with various hyperparameters that should be tuned for maximum performance. In this paper, an optimized Deep Neural Network (DNN) based signal detection approach has been proposed for the orthogonal frequency division-multiplexing (OFDM) system in the carrier frequency offset (CFO) and phase offset (PO)impaired environment at the receiver end. Thorough experiments were carried out to determine the optimal architecture for accurate signal detection. The optimum architecture is of shape 64-12-78-22-64, where 12,78,22 are the number of neurons in the hidden layer.