Efficient KLT based on overlapped subframes for speaker identification

摘要

This letter integrates the Karhunen-Loeve (KL) transform with the overlapped subframe techniques for efficient speaker identification under an additive noise environment. Based on the overlapped subframe partition method, we present an efficient technique to construct an eigenvector matrix and achieve the advantages of the KL transform technique. Computer simulations demonstrate the effectiveness of the proposed technique.