Flame acceleration and explosion of hydrogen/air mixtures remain a key problem for severe accident management in nuclear power plants. Empirical criteria have been developed in the early 2000s by Dorofeev et al. (2001) providing effective tool to discern possible flame acceleration (FA) or deflagration-to-detonation transition (DDT) scenarios. A large experimental database, composed mainly by middle-scale experiments in smooth tubes and obstacle-laden ducts, has been used to validate these criteria. In these devices the position of the reaction front is usually detected by photo-diodes or photomultiplier tubes uniformly distributed along the tube axis. As a result, only a coarse representation of the velocity profile can be achieved. In this paper we develop a new technique to track the flame position along the tube at any time. This method consists in performing time-resolved IR absorption measurements by doping the fresh mixture with an alkane. The velocity profile is then derived by measuring the variation of the extension in depth of the unburnt gas along the tube axis. Correction factors are eventually drawn from the comparison between longitudinal (IR absorption measurements) and cross-sectional (photomultiplier tubes) flame velocity diagnosis techniques. Finally, experimental results are compared to numerical simulations and analytical models proposed by V. Bychkov group.