基于张量模型的音频分类方法研究

The research of audio classification method based on tensor model

本文主要以网络获取和自制音频作为研究对象,采用张量分解作为处理手段,把音频信号映射到高阶子空间完成特征建模和分类器建模,实现音频信号的分类.实验结果表明:张量分解的高阶子空间方法在分类效果上优于传统的支持向量机和高斯混合模型分类器,为提升音频分类性能提供了理论和技术支撑.

Audio is an important class of multimedia data. The massive audio resources need classification in order to effectively manage and utilize them. The classification of audio signals originated from the network and recorded by ourselves was realized. The audio signal was mapped into a high order subspace,and then feature modeling and classifier modeling were obtained by using tensor factorization. Experiment results showed that the performance of classification of tensor factorization proposed in high order subspace is better than that of the support vector machine classifier and Gauss mixture model classifier. The proposed method may provide theoretical and technical support for improving the performance of audio classification.