Detection and Classification Using Fractal Models: Analysis,

Methods are developed and investigated which will exploit the fractal and nonlinear characteristics of measured sensor data. Novel autoregressive and fractal signal processing methods are applied to the detection and classification of pertinent signals. Autoregressive and other techniques are generalized to take advantage of potential nonlnear and fractal properties of data. Furthermore, it is essential to have objective measures of processor performance on which to base assessments. Toward this end, we calculatee statistically correct and objective measures of detection and classification. Receiver operating characteristics (ROC) curves are calculated to quantify detection results; classification results are quantified by the Bayes error calculated using the K-th Nearest Neighbor method.