摘要
按照MEPG-7的音频特征规范,设计了帧层次和段层次上的音频特征提取算法.在深入分析足球比赛中各类音频信息的不同特点的基础上,通过有针对性地选取特征,构造特征向量,设计并实现了一种基于决策树的层次化分类算法.该算法可以自动将足球比赛中的音频信息分为噪音、解说员语音、哨音、欢呼声和带背景音的解说员语音等5个类别.实验表明:该算法提取的特征有效,分类效果良好.
The extraction algorithm of frame-level and clip-level audio features is designed according to the MPEG-7 audio standard.By deeply analyzing the characteristics of audio information in soccer game,the feature vector is formed and a hierarchical audio classifier is realized based on decision tree.The decision tree audio classifier divides audio information in soccer game into five classes： noise,commentator speech,whistle,cheer and commentator speech with background noise.The experiment results show that the selected features are effective for audio classification,and the classification accuracy is reasonable.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2007年第10期35-38,45,共5页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
湖北省自然科学基金资助项目(2005ABA256)
华为基金资助项目(YJCB20050241N)
关键词
特征提取
音频分类
决策树
足球比赛
feature extraction
audio classification
decision tree
soccer game
