ABSTRACT
This chapter provides anoverview of pattern recognition as it applies to the classifications of acoustic sounds in the ocean. For classification of underwater sounds, the measurement space consists of the analog voltage time wave form generated by various acoustic events as an output from the hydrophones. In dealing with the classification information in the underwater pressure wave, it is useful to note the existence of the sampling theorem that specifies the least number of discrete samples of an unknown wave form necessary for its complete and unambiguous definition. The concatenation approach is important because it often encompasses redundant information, thus minimizing the presence of noise in the system. Also, the concatenation may remove ambiguities in final classification by making the total string ineligible for assignment to certain classes. Feature selection provides a formalism to reduce the high-dimensional space to a more tractable low-dimensional space by the elimination of redundant information and certain types of noise.
