Abstract—This paper presents a promising Deep-Learning (DL) approach for accurate symbol detection in a slow Frequency Hopping (SFH) wireless communication System under a Narrow Band (NB) multipath channel fading. A feedforward neural network with three layers of input, hidden, and output was employed for deep learning. The neural network is designed to take 80 features as input, representing the received signal samples at the receiver. The neural network is trained to anticipate the transmitted symbol based on the provided training dataset, utilizing different modulation techniques, including Binary Phase Shift Keying (BPSK), Quadrature Phase Shift Keying (QPSK), 8-PSK, and 16-PSK. Additionally, computer simulations are conducted to verify the effectiveness of the proposed method across various modulation schemes. The generated training loss and validation loss curves confirmed the ability of the receiver to learn.