Blind Channel Estimation for Power-line Communications by a Kohonen Neural Network

In this paper a new technique is proposed for blind channel identification and equalization in orthogonal frequency-division multiplexing (OFDM) communication over power-lines. The proposed scheme is based on a competitive neural network which utilizes the frequency domain data in the receiver performing a nonlinear decision-directed equalization. This method allows for blind tracking of the time variant channel response and is robust to impulsive noise and non-linearities. Numerical simulations for a realistic indoor power-line show the potentiality of the proposed estimation procedure.