The traveling-wave amplifier (TWA) model of the cochlea [A. Hubbard, Science 259, 68–71 (1993)] has been shown to produce outputs that compare quite well with experimental data. A TWA model with parameters adjusted to fit the physiological properties of the dolphin cochlea was used as part of a sonar signal discrimination system. The system was tested on a cylinder wall thickness discrimination problem. Broadband echoes from cylinders with different wall thicknesses were aligned using a matched filter and envelope detection. The aligned signals were used as inputs to the TWA model and energy from different locations along the cochlea model were computed in time increments. Data were reduced by principal component analysis and finally classified by a neural network. Results from the experiment demonstrate the potential of the framework.