This paper analyzes a new dead-reckoning method for pedestrian navigation. In traditional approaches, accelerometers are used to detect steps and to estimate the step length. However, these algorithms must be calibrated for each user and fail in practice when the user moves with an irregular posture like crawling or carrying a victim. In the proposed method the speed is determined from coherence time measurements of the wideband frequency response in Orthogonal Frequency Division Multiplexing (OFDM) systems. The presented algorithm does not require user calibration and it is not based on a particular type of movement. Moreover, it is shown by means of an experimental system that the positioning accuracy matches that of state-of-the-art step length estimation algorithms.