A least-squares (LS) channel estimation (CE) technique for mobile OFDM communications over a rapidly time-varying frequency-selective fading channel is investigated in this paper. The proposed technique keeping the comb-type pilot arrangement can achieve low error probability by accurately estimating channel impulse response (CIR) and effectively tracking rapid CIR variations. In addition, the LS CE technique proposed here is conducted in the time domain (TD); meanwhile, a generic estimator is, thus, performed serially block by block without assistance from a priori channel information and without increasing computational complexity. By taking advantage of linearly frequency-modulated (LFM) or pseudorandom signals that are transceived for the purpose of sounding pilot subchannels, the proposed LS channel estimator (CE) can inherently perform pseudo-noise (PN) matched Altering (MF) to suppress multipath interference (MPI) caused by frequency-selective fading and inter-carrier-interference (ICI) resulting from data subchannels. The dual optimality of the LFM and PN pilot symbols is verified for both time-domain (TD) and frequency-domain (FD) CEs. Furthermore, the proposed technique also exhibits good resistance against residual timing errors occurring with the DFT demodulation. Extensive computer simulations in conjunction with statistical derivations show the superiority of the proposed technique.