Content/context-adaptive feature selection for environmental sound recognition

Environmental sound recognition (ESR) is a challenging problem that has gained a lot of attention in the recent years. A large number of audio features has been adopted for solving the ESR problem. In this work, we focus on the problem of automatic feature selection. Specifically, we propose two methods, called the content-adaptive and the context-adaptive feature selection schemes to achieve this goal. Finally, the superior performance of the proposed feature selection methods is demonstrated when they are applied to a medium-sized environmental database with a simple Bayesian network classifier.