Publication

Abstract : A machine learning based discrimination modality has been proposed for non stationary signal analysis to be used for sub surface characterization. This proposed modality improves the testability, reliability and present best detection in terms of quantified characteristics pertaining to anomalies. In this paper the region based active contour segmentation along with artificial neural network (ANN) has been employed to facilitate the practitioner in assessing the quality of the test body. The present study carried out over a numerically tested biomedical bone sample with different density variations to simulate Osteoporosis and an experimental Carbon fiber reinforced composite specimen. The performance of the proposed methodology compared with conventional signal processing techniques pertaining to the aspects of defect signal to noise ratio, probability of defect detection and concluded that ANN and Decision tree based processing modalities provide better characterization.