A biologically inspired adaptive underwater target classification using a multi-aspect decision feedback unit

Buried Underwater Object Classification Using a Collaborative Multiaspect Classifier

“…This implies that the system is able to use collaboration among the agents to reduce the chance of generating an incorrect final decision. In contrast, in other multiple-ping fusion systems [4], [5], [10], no collaborative decision making takes place, either implicitly or explicitly, among the decision-making agents. The lack of collaboration prevents certain discriminatory evidence about the relationship among the feature vectors to be used, and hence stymies these systems' classification performance.…”

“…The development of the CMAC system is motivated by its collaborative ability to minimize a cost function based on overall misclassifications. This property is not shared by any of the other multiple-ping classifiers [4]- [10]. CMAC is inspired from the distributed detection method developed in [11] for sensor networks.…”

“…The framework developed for CMAC can also be used to implement a DF system for multiaspect classification [10]. The idea behind DF-based multiaspect classification is to make a final classification decision for an aspect (or sonar ping) using not only the current feature vector , but also the final decisions made at previous pings, , .…”

“…The second approach is classifier based and can be performed using either decision-level fusion (DLF) [4], feature-level fusion [5]- [9], or a combination of feature-level fusion and DLF [10]. In DLF [4], intermediate decisions obtained using a single-ping classifier are fused to yield a final decision.…”

“…Multiple-ping classification can also be performed using a combination of feature-level fusion and DLF [10]. In this system, the class label is determined using not only the current feature vector, but also the decisions made at the previous pings.…”