摘要
音频分类在基于内容的音频、视频检索和分析中起着重要作用。文章针对静音、语音、音乐和环境背景音4类音频提出基于VQ-GMM的分类算法。首先通过阈值判决区分静音和非静音,然后利用VQ-GMM分类器将非静音进而分为语音、音乐和环境背景音。实验结果表明该方法的分类性能良好,平均正确率可达95%。
Audio classification plays an important role in audio/video retrieval and analysis. This paper presents a classification algorithm based on VQ-GMM to classify audio into silence, speech, music and background. Firstly silence and non-silence are classified based on threshold judging and then VQ-GMM is used to further classify non-silence into speech, music and background. Experimental results show that the proposed algorithm produces satisfactory results and the average accuracy reaches 95 %.
出处
《信息工程大学学报》
2008年第4期423-426,454,共5页
Journal of Information Engineering University
基金
国家863计划资助项目(2006AA01Z146)
