In this paper, we design and evaluate the performance of non-data-aided timing recovery methods for a filter bank multicarrier modulation technique called filtered multitone (FMT). Starting from a maximum likelihood approach as optimality criterion, we illustrate the derivation of a closed-loop tracker and demonstrate how to derive a relatively low-complexity scheme with reasonable performance loss. The features of the proposed timing estimators (S-curve, acquisition time, estimation accuracy) are derived through a combination of analysis and simulation both on the additive white Gaussian noise channel and a typical static frequency-selective wireless channel. Unlike previously published works, a remarkable acquisition range equal to plus and minus one-half the FMT signaling interval is achieved. We also show by simulation a significant robustness with respect to the multipath channel and nonideal carrier frequency estimation