Joint Entropy Criterion for Training Time Domain Equalizer of OFDM System Without CP

This paper describes a new criterion using the joint entropy in time domain to equalize the OFDM system without cyclic prefix (CP). We assume the transmitted and received OFDM symbols are jointly normal random variables, then a finite impulse response (FIR) time domain equalizer (TEQ) is trained given the concept of minimum joint entropy according to the least mean square (LMS) algorithm. We compare the performance of the new criterion against the minimum mean square equalizer (MMSE) in cancelling intersymbol interference (ISI) in terms of the reduction in mean square error (MSE) in the residual error sequence in different types of channels. The results demonstrate the superiority of the new strategy in low SNR and highly frequency selective channels, the new method also possesses faster convergence rate and lower sensitivity