This paper concerns the use of complex multipath propagation for discriminating the depth of a moving target with a low-frequency active sonar. Although matched-field processing techniques, which exploit full-wave modeling of multipath propagation, have been previously demonstrated in passive sonar applications, their use in active sonar has been precluded by uncertainty in the complex backscattering characteristics of the target. In this paper, a matched-field technique is proposed which exploits changes in the complex moving target return seen between consecutive sonar pings at a horizontal receive array. Because the method exploits only changes in the coherent combination of multipath arrivals, it does not require a priori knowledge of the target backscattering characteristics assuming its aspect is only slowly changing from ping to ping. In this paper, simulation results are presented which are modeled after a realistic Mediterranean scenario taken from the Shallow-Water Active Classification Experiment performed in November 1995. Matched-field depth estimation simulation results indicate that target depth accuracies of better than 20% of the channel depth can be achieved at ranges of 10 to 15 km. [Work supported by ONR.]