Speaker verification with mostly voiced speech for GMM/UBM and GMM/IBM systems

摘要

The paper proposes the use of just mostly voiced speech (MVS) for speaker verification (SV). The speech is partitioned into an MVS part and a non-MVS part by a simple machine classification. SV experiments were held with a standard Gaussian mixture model (GMM) with universal background model (UBM) system and a GMM with computationally improved individual background model (IBM) system. They demonstrate that that a systems with MVS always outperform a comparable system that uses the whole data. The relative advantage of SV with MVS increased when noise was added to the data.