| 摘要 |
A system, method, and apparatus for identifying a speaker of an utterance, particularly when the utterance has portions of it missing due to packet losses. Different packet loss models are applied to each speaker's training data in order to improve accuracy, especially for small packet sizes.
|
