Blind sequence estimation of MPSK signals using dynamically driven recurrent neural networks

This paper presents a novel blind sequence estimation of multiple phase shift keying (MPSK) signals approach using dynamically driven recurrent neural networks (DDRNN) with the continuous multi-threshold phase activation function (CMTPAF). With the consideration of the characteristics of MPSK signals, a CMTPAF is designed, the parameters of the CMTPAF are illustrated, and the two new concepts of accumulation points and repulsion points are proposed. The weight matrix of DDRNN-CMTPAF is constructed by utilizing the unitary signal space matrix obtained from singular value decomposition for the receiving signal matrix. It is important that the energy functions of synchronous and asynchronous modes in the designed DDRNN-CMTPAF are proposed and proved.