Certain characteristics of acoustic emission (AE) signals from graphite-epoxy composite specimens were correlated last year with the mechanical behavior of the materials. Moisture degradation, which reduced the ultimate strength, resulted in a change in the AE amplitude distribution early in the loading history. Also, AE having distinct frequency spectral types tended to occur at or near singularities in the load curve. In order to transform those observations into a viable NDE tool, two things needed to be done: (1) develop a quicker and more objective method for extracting the pertinent correlations from the data, and (2) develop the statistical relationship between the AE characteristics and some mechanical strength parameter such as ultimate strength or remaining lifetime. The approach taken was to use computer pattern recognition techniques to analyze the data. A problem which was faced in this was the huge amount of data that are available in raw form from a single acoustic emission test. To reduce the amount of data, an intermediate feature extraction step was required and several ways of doing this, based on the prior work, were tried. As a result, the previous correlations between the acoustic emission signal characteristics and the mechanical condition of the composite were confirmed on a more objective basis. This now provides a tool for methodically studying and identifying the specific failure modes which occur in composite materials under various conditions.