一种稳健的语音/音乐分类法

A Robust Speech/Music Discrimination Approach

提出了一种把音频片段分类成语音或音乐的新系统。系统能自动选取在相应的信噪比下具有最高分类精度的特征参数。将从音频片段提取的特征参数值与门限值相比较进行语音/音乐的分类,其中的门限值与一定的信噪比相适应。介绍了一种新特征参数,即低频带能量率方差,在低信噪比环境下,它对分类精度有很大地提高。考察了系统在不同的信噪比环境下的分类性能。实验结果表明,所提系统分类性能良好。

A new system for classifying audio segments as speech or music is presented. The system selects the features with the highest classification accuracy and corresponding SNR value. The value of this features extracted from each window-level segment arc compared to certain thresholds, which are also adapted to the SNR. Multiexpert method of combining the features is employed to improve the classification accuracy. A new feature, the variance of low-band energy ratio, is also introduced, which produces large improvements in classification accuracy at low SNR. Performance of the proposed system is evaluated for different SNR. The experiment results show that the classification accuracy is excellent.