In hydrographic survey, sound speed refraction errors are best corrected by measuring the velocity profile using a dedicated profiler sensor. However, it is expected that hydrography survey will be more and more often realized using autonomous surface or underwater vehicles. Therefore, the development of an automatic, robust and real-time refraction correction method is a key interest. We will show that a practical solution is to use a dual axis multibeam echosounder which can record along and across track profiles-a system that was first proposed (Hydrosweep) in the 1980's. The main contribution of this paper is the development of a new algorithm. A new soundings refraction correction algorithm is first presented, based on the mean velocity profile model. This mean velocity profile can be estimated in real-time using the dual axis multibeam. The method enables estimation and correction of refraction induced errors at each sounding depth. This has been evaluated experimentally where the dual axis multibeam echosounder is simulated by using a standard echosounder rotated 90° to the side. Furthermore, from the estimated mean velocity profile, the true sound velocity profile over the soundings may be obtained. A first inversion result is given.