基于内容的音频自动分类方法

Methods for content-based audio automatic classifition

为了解决基于内容的音频检索、语音文档检索等很多领域中提取音频结构和内容语义的问题时,介绍几种比较典型的音频分类算法包括最小距离方法、支持向量机、神经网络、决策树方法等,指出在基于内容的音频检索技术方面的难点以及发展方向。

Automatic audio classification is one of the key methods to extract content semantics and structure from audio, which is valuable for content-based audio retrieval and spoken document retrieval, etc. This paper describes some typical methods for audio classification, for example minimum distance classifier, support vector machine, and decision tree etc, and points out the difficult point about content-based audio retrieval, and the direction in the future.