The State-of-the-art speaker recognition system is now using the i-vector framework to make the supervector of the UBM to a low-dimensional vector. In this paper, we propose a new method to do the ...