Integrated Fuzzy Decision Logic and Artificial Neural Network for Submarine Classification

American GNC Corporation (AGNC) proposes to develop early submarine classification methodology. The proposed target discrimination in the decision hierarchy, i.e., detection, classification, recognition, identification, and characterization, will eventually lead to reliable, timely, and accurate submarine classification. Anticipated innovations include: l) Spectral signature extraction which takes into account the target signature variability, environment, and sonar operation conditions to generate most submarine acoustic signals for early classification in various vibration modes; 2) Optimized recognition approach which results from the integration of different methods based on distinct signature domain and knowledge of the targets of interest; 3) Classification with artificial neural network which take advantages of the learning feature of neural network to both identify submarine signals which were learned before, and to classify suspicious submarine signals which may or may not be known before; 4) Classification with integrated fuzzy decision logic and artificial neural network which blends the approximate reasoning capability of the fuzzy decision logic and learning feature of neural network. The deliverable for Phase I will be the discrimination methodology for tactical applications. Capability extension and performance enhancement of the proposed system will also be made along Phase I period as the research progresses. In Phase II, the validated target recognition system and associated algorithms will be fully developed, tested, and documented. Hardware and software necessary to implement the early submarine classification system will be developed with various submarine models. The end product will be tailored ready for embedded applications.