摘要
针对音频信号的特点,提出了一种将灰关联分析应用于语音/音乐信号的分类方法,并给出了对音频信号进行灰关联分析的方法和步骤。利用语音和音乐信号的短时能量均方根的概率统计特征值建立了目标的参考数据和比较数据,进行了不同语音和音乐信号的灰关联分析,确定了目标分类的判据,并对两类信号以及含有独立两类信号的音频信号进行了分类。仿真结果表明:基于单特征值的音频信号灰关联分析方法实现过程简单,分类平均准确率达到90%,优于文献[4]中基于1种和2种特征值的分类性能。
Application of gray correlation analysis method to speech/music discrimination based on statistic features of short energy root mean square (RMS) is presented. Steps of the method are described. The characteristic values of comparative and referenced data are then realized. Discriruination criterions are defined and speech/music is recognized the described algorithm is more effective and accurate than that on one and two features in real-time multimedia applications.
出处
《声学技术》
CSCD
北大核心
2007年第2期262-267,共6页
Technical Acoustics
关键词
灰关联
特征
语音和音乐分类
gray correlation
features
speech/music discrimination Simulation results indicate that introduced in the literature based