Robust speaker recognition based on multi-stream features

In this paper, we investigate the effect of the G.723.1 (6.3kbps) on speaker recognition system. In order to improve the robustness of codec mismatch, we used the Power Normalized Cepstral Coefficients (PNCC) which is a new robustness acoustic feature, to improve the performance of speaker verification system. And a modified SCF speech feature is propose to improve the robustness under codec mismatch. We proposed a new method to improving the performance of I-vector based speaker recognition system by combining PNCC and the modified SCF feature. Three type of fusion method is introduced and compared in this paper. The experiment results of speaker recognition towards G.723.1 resynthesized coded speech demonstrate the effectiveness of our proposed method. Compared with traditional speaker recognition system, the EER improved 72% by the multi-stream features based speaker recognition system.