基于PCA和CHMM的音频自动分类*

针对DHMM分类器对音频特征进行向量量化引起的误差及特征维数过多导致计算复杂度过大的问题，提出了一种新的基于PCA和CHMM的音频自动分类方法。它先将音频特征组成一个高维向量，然后使用PCA对这些高维向量进行降维，再使用CHMM分类器对降维后的特征进行分类。实验证明了PCA和CHMM音频分类的有效性。

Automatic audio classification based on PCA and CHMM

In order to solve the problems of the vector quantization error and the excessive computation complexity arising from audio features, this paper proposed a new automatic audio classification method based on PCA and CHMM. It concatenated audio features to form a high-dimensional vector, performed the dimensionality reduction on the vector, Used CHMM classifier to classify the audio. The experimental results show the effectiveness of the PCA and CHMM.