Sonar was invented by copying bats' and dolphins' echolocation systems, but the dominant approaches taken to sonar systems today tend to focus on imagery and resolution. In this paper we explore a bio-inspired approach to sonar and sonar signal processing. Conventional sonars use narrowband pulses. The bio-mimetic pulses developed for this work are based on studies of dolphins echolocation clicks and cover a large frequency band. Classical sonar systems only provide the amplitude of the backscattered echo and consequently classical sonar signal processing is focused on amplitude. In fact much more information concerning the target is hidden in the phase and, working with the wideband bio-mimetic pulses, we describe new methods to extract highly discriminant features. Our approach is demonstrated by applying these techniques in an ATR (automatic target recognition) trial in a real environment.