Hierarchical structure neural network and it´s application in unusual sound recognition

This paper describes a novel method by using hierarchical structure neural network for unusual sounds recognition. Above all, it discusses the principle of the inspecting system against theft that is used in pipeline transportation. Then the paper presents the structure of neural network, the sampling and the neural network train. The system achieves inspection preventing theft and caution by sound detecting, processing and recognizing. It is attested through actual application that the system capability is stable, the correct recognition ratio attains to 100%, and the error warning ratio is below 1%. In addition, the system can be expanded according to requirement.