Automated Symbol Rate Estimation Over Frequency-Selective Fading Channel by Using Deep Neural Network

Abstract

The adaptive communication system is going to play a major role for fifth-generation (5G) and beyond wireless communication where the physical layer signal parameters need to be changed at the transmitters as per system requirement and the receiver needs to estimate them to recover the signal. In this paper, we have proposed an efficient and robust automated symbol rate estimation model for single carrier system over frequency-selective fading environment by using deep neural network (DNN) approach. The proposed scheme estimates symbol rate without having any prior knowledge of the signal bandwidth which was the main assumption for existing statistical methods. In the proposed scheme, no additional knowledge such as channel state information (CSI) and synchronization parameters are required to estimate the symbol rate. The proposed model outperforms the existing statistical models in terms of the performance. The performance of the symbol rate estimator is depicted by the normalized mean square error (NMSE).