|
Authors: | Urs Niesen |
Group: | Computer Engineering |
Type: | Techreport |
Title: | Speaker Verification Using Artificial Neural Networks |
Year: | 2003 |
Month: | September |
Pub-Key: | Nie03 |
Institution: | Institut TIK, ETH Zurich |
Abstract: | In the present work the use of Artificial Neural Networks (ANN) for text dependent speaker verification is analyzed. While the method currently used at the TIK relies on the Euclidean
cepstral distances to discriminate between a pair of signals spoken by the same and by two different speakers, in the current study a distance metric learned by an ANN is used for the same task.
The results of the conducted experiments show that a significant improvement in correct classification rate can be achieved with this new approach. The experimental results also confirm that the proposed method can as well be applied in the case where the speech signals to be compared are uttered by speakers not present in the training set. |
Resources: | [BibTeX] |