On the application of pattern recognition techniques to oceanographic signal processing

Pattern recognition techniques can be used in the design of computer programs for interpretation of the oceanographic seismic data. A typical pattern recognition problem is defined by feature measurements, pattern classes, and the decision rule. The physical parameters which describe the structure of the deep sea floors correspond to feature sets in pattern recognition. Examples of pattern classes are various rocks, sediments, mantles, and oceanic layers of low, medium and high porosity. The decision criterion is the minimum number of misclassifications. Specifically, the pattern recognition techniques are applied to processing of a digital seismic data tape supplied by the Woods Hole Oceanograpic Institution. A temporal array data processor combine the time-delayed data by using the stochastic approximation algorithm. Classification is then made between the recent sediments and the older eroded sediments.