Hybrid Independent Component Analysis and Support Vector Machine Approach for Audio Classification

The recent explosive growth of multimedia information in many applications requires advanced techniques for multimedia information retrieval (MIR). Audio is one of important media in multimedia and contains abundant information. Recent development in MIR has revealed that audio information can provide effective contents to retrieve multimedia information. Consequently, it has been paid more and more attention to how to classify audio information. Many existing and new research issues can lead to improve the retrieval efficiency, effectiveness and accessibility. This paper provides a hybrid classification method for audio indexing and retrieval. It firstly describes main audio characteristics and features and the basic theory of Support Vector Machine (SVM) and Independent Component Analysis (ICA). Finally, the hybrid method that integrates ICA and SVM is presented and applied into audio classification.