A method for the channel estimation and tracking of OFDM systems based on a neural network outer loop controller for the LMS training algorithm

The OFDM system is found to be very robust for high data rate communication over multipath channels with a significant intersymbol interference (ISI). The time-domain channel estimation for an OFDM system results in a 3 dB better BER performance. However, the training speed of the LMS needs to be improved since the training sequence has a shorter length and variations on amplitude. In particular, the known waveform of the training sequence forces the LMS to be unstable and perform a poor training when the amplitude is smaller in three regions of the training sequence. This study proposes a neural network (NN) type experience based learning algorithm to design a training trajectory for the LMS algorithm by an outer loop controller. The outer loop controller uses the magnitude of simultaneous error function and time in order to learn a training route for the LMS. The obtained results show that the NN-based LMS performs much better when it is compared to those using the conventional LMS algorithm in the time domain channel estimation of the OFDM system. The introduced complexity does not prohibit the technique since the current DSP processing capabilities are significant for such considered applications.