Abstract—Soundtracks of multimedia files are information rich sources, from which much content-related information and metadata can be extracted. There exist many individual algorithms for the recognition and analysis of speech, music or event sounds, allowing for information embedded in audio format files to be retrieved or represented in a semantic fashion. However, soundtracks are typically a mixture these three different types of signals, and sometimes overlapped. Segmentation and classification therefore become essential pre-processors for audio based information retrieval and metadata generation. This paper stresses the importance of a universal audio indexing and segmentation pre-processor, proposes a high-level architecture for such a system, and presents signal processing algorithms based on soft-computing and two important but neglected feature spaces to improve the accuracy of classification.

Index Terms—Audio indexing, classification, metadata, information extraction, soft-computing.

Francis F. Li is with the University of Salford, M5 4WT, UK (e-mail: f.f.li@salford.ac.uk).

Cite:Francis F. Li, "Soft-Computing Audio Classification as a Pre-Processor for Automated Content Descriptor Generation," International Journal of Computer and Communication Engineering vol. 3, no. 2, pp. 101-104, 2014.