论文信息 - Hierarchical speaker identification using speaker clustering

Hierarchical speaker identification using speaker clustering

We explore an approach to speaker identification called speaker clustering in the GMM-based speaker recognition system in order to reduce the computational complexity. The ISODATA algorithm adapted for our purpose works well when we cluster speakers whose acoustic characteristics are similar to a distance measure. The time spent on HSI (hierarchical speaker identification) is approximately 30.3 percent more than that spent on CSI (conventional speaker identification) when the number of registered speakers is 40 in our experiments. Increasing of the number of speakers decreases the time spent on HSI compared with CSI. It is shown that this approach can improve the speed of the speaker identification system.

[1] Timothy J. Hazen,et al. Rapid speaker adaptation using speaker clustering , 2002, INTERSPEECH.

[2] Michael Picheny,et al. Speaker adaptation based on pre-clustering training speakers , 1997, EUROSPEECH.

[3] D. Rubin,et al. Maximum likelihood from incomplete data via the EM - algorithm plus discussions on the paper , 1977 .

[4] Timothy J. Hazen,et al. Rapid Speaker Adaptation With Speaker Clustering , 2002 .

[5] B. Ripley,et al. Pattern Recognition , 1968, Nature.

[6] Stéphane H. Maes,et al. A distance measure between collections of distributions and its application to speaker recognition , 1998, Proceedings of the 1998 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '98 (Cat. No.98CH36181).

[7] Douglas A. Reynolds,et al. Robust text-independent speaker identification using Gaussian mixture speaker models , 1995, IEEE Trans. Speech Audio Process..