Abstract
At the beginning, we had started to develop a Czech telephone acoustic model by evaluating various Kaldi recipes. We had a 500-h Czech telephone Switchboard-like corpus. We had selected the Time-Delay Neural Network (TDNN) model variant “d” with the i-vector adaptation as the best performing model on the held-out set from the corpus. The TDNN architecture with an asymmetric time-delay window also fulfilled our real-time application constrain. However, we were wondering why the model totally failed on a real call center task. The main problem was in the i-vector estimation procedure. The training data are split into short utterances. In the recipe, 2-utterance pseudospeakers are made and i-vectors are evaluated for them. However, the real call center utterances are much longer, in order of several minutes or even more. The TDNN model was trained from i-vectors that did not match the test ones. We propose two ways how to normalize statistics used for the i-vector estimation. The test data i-vectors with the normalization are better compatible with the training data i-vectors. In the paper, we also discuss various additional ways of improving the model accuracy on the out-of-domain real task including using LSTM based models.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Chen, X., et al.: Recurrent neural network language model adaptation for multi-genre broadcast speech recognition. In: Sixteenth Annual Conference of the International Speech Communication Association (2015)
Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural Comput. 9(8), 1735–1780 (1997)
Hsu, W.N., Zhang, Y., Glass, J.: Unsupervised domain adaptation for robust speech recognition via variational autoencoder-based data augmentation. In: 2017 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), pp. 16–23. IEEE (2017)
Karafiát, M., Burget, L., Matějka, P., Glembek, O., Černockỳ, J.: iVector-based discriminative adaptation for automatic speech recognition. In: 2011 IEEE Workshop on Automatic Speech Recognition & Understanding, pp. 152–157. IEEE (2011)
Karanasou, P., Wang, Y., Gales, M.J., Woodland, P.C.: Adaptation of deep neural network acoustic models using factorised I-Vectors. In: Fifteenth Annual Conference of the International Speech Communication Association (2014)
Ko, T., Peddinti, V., Povey, D., Khudanpur, S.: Audio augmentation for speech recognition. In: Sixteenth Annual Conference of the International Speech Communication Association (2015)
Park, D.S., et al.: Specaugment: a simple data augmentation method for automatic speech recognition. arXiv preprint arXiv:1904.08779 (2019)
Pollak, P., et al.: SpeechDat(E) - Eastern European telephone speech databases. In: Proceedings LREC 2000 Satelite workshop XLDB - Very Large Telephone Speech Databases, pp. 20–25. European Language Resources Association, Athens (2000)
Povey, D., et al.: The kaldi speech recognition toolkit. In: IEEE 2011 Workshop on Automatic Speech Recognition and Understanding. IEEE Signal Processing Society (Dec 2011), iEEE Catalog No.: CFP11SRW-USB
Pražák, A., Müller, L., Šmídl, L.: Real-time decoder for LVCSR system. In: The 8th World Multi-Conference on Systemics, Cybernetics and Informatics: vol. VI : Image, Acoustic, Signal Processing and Optical Systems, technologies and applications, pp. 450–454. International Institute of Informatics and Systemics, Orlando, Florida (2004). http://www.kky.zcu.cz/en/publications/PrazakA_2004_Real-timedecoderfor
Saon, G., Soltau, H.: Speaker adaptation of neural network acoustic models using I-Vectors. In: IEEE Workshop on Automatic Speech Recognition and Understanding pp. 55–59 (2013). https://doi.org/10.1109/ASRU.2013.6707705,http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6707705
Sim, K.C., et al.: Domain adaptation using factorized hidden layer for robust automatic speech recognition. In: Proceedings of the INTERSPEECH (2018)
Sim, K.C., Qian, Y., Mantena, G., Samarakoon, L., Kundu, S., Tan, T.: Adaptation of deep neural network acoustic models for robust automatic speech recognition. In: Watanabe, S., Delcroix, M., Metze, F., Hershey, J.R. (eds.) New Era for Robust Speech Recognition, pp. 219–243. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-64680-0_9
Waibel, A., Hanazawa, T., Hinton, G., Shikano, K., Lang, K.J.: Phoneme recognition using time-delay neural networks. Backpropagation: Theory, Architectures and Applications, pp. 35–61 (1995)
Yu, D., Li, J.: Recent progresses in deep learning based acoustic models. IEEE/CAA J. Automatica Sin. 4(3), 396–409 (2017)
Acknowledgement
This research was supported by the LINDAT/CLARIN, project of the Ministry of Education of the Czech Republic No. CZ.02.1.01/0.0/0.0/16_013/0001781.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2019 Springer Nature Switzerland AG
About this paper
Cite this paper
Vaněk, J., Michálek, J., Psutka, J. (2019). Tuning of Acoustic Modeling and Adaptation Technique for a Real Speech Recognition Task. In: Martín-Vide, C., Purver, M., Pollak, S. (eds) Statistical Language and Speech Processing. SLSP 2019. Lecture Notes in Computer Science(), vol 11816. Springer, Cham. https://doi.org/10.1007/978-3-030-31372-2_20
Download citation
DOI: https://doi.org/10.1007/978-3-030-31372-2_20
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-31371-5
Online ISBN: 978-3-030-31372-2
eBook Packages: Computer ScienceComputer Science (R0)