Blind channel estimation in OFDM systems by relying on the Gaussian assumption of the input

In an OFDM system, the receiver requires an estimate of the channel to recover the transmitted data. Most channel estimation methods rely on some form of training which reduces the useful data rate. In this paper, we introduce an algorithm that blindly estimates the channel by maximizing the log likelihood of the channel given the output data. Finding the likelihood function of a linear system can be very difficult. However, in the OFDM case, central limit arguments can be used to argue that the time-domain input is Gaussian. This together with the Gaussian assumption on the noise makes the output data Gaussian. The output likelihood function can then be maximized to obtain the maximum likelihood (ML) estimate of the channel. Unfortunately, this optimization problem is not convex and thus finding the global maximum is challenging. In this paper, we propose two methods to find the global maximum of the ML objective function. One is the blind Genetic algorithm and the other is the semi-blind Steepest descent method. The performance of the proposed algorithms is demonstrated by computer simulations.