Method for estimating the number of sound sources using SVM in sound source separation

摘要

In sound source separation, the estimation of the number of sources is an important factor which greatly affects the performance of the separation system. It is known that the eigenvalue distribution of the spatial correlation matrix reflects the number of sources. In this report, a method of estimating the number of sources by classifying the eigenvalue distribution using Support Vector Machines (SVM) is proposed. An evaluation of this method using real world data is also reported.