Performance of Time-Invariant ML Channel Estimator and its Extension for OFDM Systems in Time-Variant Channels

In time-invariant (TI) channels, the maximum likelihood channel estimator (MLCE) needs no matrix inversion if equally spaced pilots or a training symbol is used in an OFDM system. This optimality vanishes in time-variant channels since CFR dispersion destroys the orthogonality between subcarriers. However, our analysis results show that TI-MLCE is still useful to provide complete CIRs under our low-complexity modification. We provide both analysis and simulation results to demonstrate our claim. Satisfactory performance is observed that more than 60% complexity is saved in the extension.