Joint Estimation of Articulatory Features and Acoustic Models for Low-Resource Languages

Basil Abraham, S. Umesh, Neethu Mariam Joy


Using articulatory features for speech recognition improves the performance of low-resource languages. One way to obtain articulatory features is by using an articulatory classifier (pseudo-articulatory features). The performance of the articulatory features depends on the efficacy of this classifier. But, training such a robust classifier for a low-resource language is constrained due to the limited amount of training data. We can overcome this by training the articulatory classifier using a high resource language. This classifier can then be used to generate articulatory features for the low-resource language. However, this technique fails when high and low-resource languages have mismatches in their environmental conditions. In this paper, we address both the aforementioned problems by jointly estimating the articulatory features and low-resource acoustic model. The experiments were performed on two low-resource Indian languages namely, Hindi and Tamil. English was used as the high-resource language. A relative improvement of 23% and 10% were obtained for Hindi and Tamil, respectively.


 DOI: 10.21437/Interspeech.2017-1028

Cite as: Abraham, B., Umesh, S., Joy, N.M. (2017) Joint Estimation of Articulatory Features and Acoustic Models for Low-Resource Languages. Proc. Interspeech 2017, 2153-2157, DOI: 10.21437/Interspeech.2017-1028.


@inproceedings{Abraham2017,
  author={Basil Abraham and S. Umesh and Neethu Mariam Joy},
  title={Joint Estimation of Articulatory Features and Acoustic Models for Low-Resource Languages},
  year=2017,
  booktitle={Proc. Interspeech 2017},
  pages={2153--2157},
  doi={10.21437/Interspeech.2017-1028},
  url={http://dx.doi.org/10.21437/Interspeech.2017-1028}
}