Underwater acoustic transients can develop from a variety of sources ranging from the cry of a whale to the sound of a torpedo launch. Accordingly, detection and classification of such transients by automated means can be an exceedingly difficult task. This thesis describes the design and implementation of a new approach to this problem based on adaptive pattern recognition employing neural networks and additional techniques including the Hausdorff metric. This system uses self-organization to both generalize and provide rapid throughput while, in addition, utilizing supervised learning for decision making. The design is based on a concept which temporally partitions acoustic transient signals, and as a result, studies their trajectories through power spectral density space. This method has exhibited a high rate of success for a large set of underwater transients contained in both quiet and noisy ocean environments, and is capable of real-time operation. |