Text-indicated speaker recognition using kernel mutual subspace method

Abstract

We propose a novel speaker recognition method that is used to compare the trajectories of continuous phonemes. The Gaussian Mixture Model has already been developed as a speaker recognition algorithm. However, Gaussian Mixture Model assume continuous speaker recognition of using only one input sample. To apply continuous observation approach, we propose a novel speaker recognition method to compare the trajectories of continuous phoneme. To compare nonlinear and complicated trajectories, we propose a speaker recognition method based on the kernel mutual subspace method. We experimentally demonstrate the proposed method's effectiveness with simulation results and show that the method achived higher accuracy than that of using the Gaussian Mixture Model.