您的位置:首页  > 论文页面

基于HMM的音频多标签分类

发表时间:2010-04-30  浏览量:2118  下载量:938
全部作者: 郑继明,杨会云,吴渝
作者单位: 重庆邮电大学应用数学研究所;重庆邮电大学计算机科学与技术学院
摘 要: 提出了一种基于隐马尔可夫模型(hidden Markov model,HMM)的音频多标签分类算法。运用分类算法先把音频粗分为环境音、音乐和语音(纯语音和带背景音乐的语音)3类,然后运用人的4种情感(愤怒、高兴、平静和伤心)对语音做进一步的分类,从而使一个音频具有了多个标签。实验表明该法取得了不错的分类效果。
关 键 词: 计算机应用;音频分类;隐马尔可夫模型;多标签;情感识别
Title: Audio multi-label classification based on HMM
Author: ZHENG Jiming, YANG Huiyun, WU Yu
Organization: Institute of Applied Mathematics, Chongqing University of Posts and Telecommunications;College of Computer Science and Technology, Chongqing University of Posts and Telecommunications
Abstract: The method of audio multi-label classification on hidden Markov model (HMM) is proposed to classify roughly an audio into environment, music and speech (pure voice and voice with background music), then it has a further classification for speech by emotion recognition (angry, happy, calm and sad) and so each audio has a multi-label. The experiment results show the performance of the approach is good.
Key words: computer application; audio classification; hidden Markov model; multi-label; emotion recognition
发表期数: 2010年4月第8期
引用格式: 郑继明,杨会云,吴渝. 基于HMM的音频多标签分类[J]. 中国科技论文在线精品论文,2010,3(8):791-797.
 
0 评论数 0
暂无评论
友情链接