Speaker adaptation of neural network acoustic models using i-vectors

@article{Saon2013SpeakerAO,
  title={Speaker adaptation of neural network acoustic models using i-vectors},
  author={George Saon and Hagen Soltau and David Nahamoo and Michael Picheny},
  journal={2013 IEEE Workshop on Automatic Speech Recognition and Understanding},
  year={2013},
  pages={55-59}
}
We propose to adapt deep neural network (DNN) acoustic models to a target speaker by supplying speaker identity vectors (i-vectors) as input features to the network in parallel with the regular acoustic features for ASR. For both training and test, the i-vector for a given speaker is concatenated to every frame belonging to that speaker and changes across different speakers. Experimental results on a Switchboard 300 hours corpus show that DNNs trained on speaker independent features and i… CONTINUE READING

Figures, Tables, Results, and Topics from this paper.

Key Quantitative Results

  • Experimental results on a Switchboard 300 hours corpus show that DNNs trained on speaker independent features and i-vectors achieve a 10% relative improvement in word error rate (WER) over networks trained on speaker independent features only.

Citations

Publications citing this paper.
SHOWING 1-10 OF 365 CITATIONS, ESTIMATED 45% COVERAGE

FILTER CITATIONS BY YEAR

2014
2019

CITATION STATISTICS

  • 25 Highly Influenced Citations

  • Averaged 84 Citations per year over the last 3 years

  • 4% Increase in citations per year in 2018 over 2017

References

Publications referenced by this paper.
SHOWING 1-10 OF 15 REFERENCES

Similar Papers

Loading similar papers…