A new approach to acoustic signal monitoring based on the generalized probabilistic descent method

摘要

We propose a new acoustic signal monitoring method, which enables one to selectively detect the confusers (irregular signal events) and ignore the other true (regular) events, based on a previous pattern recognizer design methodology called the generalized probabilistic descent method. The utility of the proposed method is demonstrated in the task of detecting confusers such as sounds that are due to tool misoperation in a noisy environment.