Abstract: A method for estimating the target depth in deep water by matching the amplitude and delay structure of multipath echo highlight clusters is proposed. The target echo structure is the convolution of two-way channel transfer function and target acoustic scattering function. It will show the characteristics of highlight clusters under near surface or near bottom condition in deep water, and the delay-amplitude structure of highlight cluster is related to the target depth which can be used to estimate the target depth. A matching cost function based on nonlinear programming model is established to match and estimate the target depth on the distance-depth grid. The effects of inter-cluster time delay deviation and Sound Velocity Profile (SVP) deviation on matching estimation are simulated and analyzed. The results of a winter sea test in the South China Sea show that under the condition of water depth of 1km and distance of 5~8 km, the echo of the target of 40 m depth present three highlight clusters with the time interval of about 0.3 s, and the error range of target depth matching estimation is 0~65 m.