Temporally Weighted Linear Prediction Features for Speaker Verification in Additive Noise

@inproceedings{Saeidi2010TemporallyWL,
  title={Temporally Weighted Linear Prediction Features for Speaker Verification in Additive Noise},
  author={Rahim Saeidi and Jouni Pohjalainen and Tomi Kinnunen and Paavo Alku},
  booktitle={Odyssey},
  year={2010}
}
We consider text-independent speaker verification under ad ditive noise corruption. In the popular mel-frequency cepstr al coefficient (MFCC) front-end, we substitute the convention al Fourier-based spectrum estimation with weighted linear pr edictive methods, which have earlier shown success in noise-rob ust speech recognition. We introduce two temporally… CONTINUE READING