Utterance Verification Detection and Fusio

In this paper, we consider an application of distributed detection and fusion framework to utterance verification (UV) and confidence measure (CM) objectives. We formulate the UV as a distributed detection and Bayesian fusion problem by combining various individual UV methods. We essentially design an optimal fusion rule that achieves minimum error rate. In the relevant isolated word OOV rejection experiments, the proposed method consistently outperforms over the individual UV methods.