Hierarchical Sound Classification using Mpeg-7

Abstract

Due to the increasing amount of multimedia contents such as images, audio signals and videos in digital form the need for automatic or semi-automatic classification applications become more and more important. This paper describes a new sound classification technique based on the sound classification algorithm included in the MPEG-7 standard without extending or modifying it. There sequential classification is turned into a hierarchical. Thereby it is possible to use more linear transformations for (lossy) feature vector compression. Thus it is possible to work out differences between sound classes more precisely. This paper also gives a detailed view on how the algorithm is implemented using a XML database to store and request content information of the audio signals and model descriptions of sound classes using the MPEG-7 standard