基于LDA模型的音频分类方法  

Audio Classification Based on LDA Model

作　　者：张翔[1] 孙伟[1] 余璇[1] 

机构地区：[1]上海海事大学信息工程学院,上海201306

出　　处：《现代计算机（中旬刊）》2017年第6期16-20,共5页Modern Computer

摘　　要：随着网络的发展海量音频文件涌现,音频分类系统也越来越普及。音频分类,尤其是语音和音乐的分类是提取音频结构和内容语义的重要手段,是基于内容的音频检索和分析的基础。介绍一种基于音频内容根据音频内容间的相似度对音频进行分类的方法。用每个音频的音高集代表该音频文件,以LDA主题模型对音频进行分类。With the development of the network mass audio files come to the fore, audio classification system is becoming more and more popular. Audio classification, especially classification between pronunciation and music is an important means of extract audio structure and content semantics. It is the foundation of audio retrieval and analysis based on content. Introduces a method which classifies audio based on audio contents. And the method classifies audio according to degree of similarity between audio content. In this method, pitch set of every song behalves the song and audio is classification by LDA model.

关 键 词：相似度 音频内容 音高 LDA主题模型 

分 类 号：TP317[自动化与计算机技术—计算机软件与理论]