End-To-End deep neural models for Automatic Speech Recognition for Polish Language


  • Karolina Pondel-Sycz Faculty of Electronics and Information Technology, Warsaw University of Technology
  • Agnieszka Paula Pietrzak Faculty of Electronics and Information Technology, Warsaw University of Technology
  • Julia Szymla Faculty of Electronics and Information Technology, Warsaw University of Technology


This article concerns research on deep learning models (DNN) used for automatic speech recognition (ASR). In such systems, recognition is based on Mel Frequency Cepstral Coefficients (MFCC) acoustic features and spectrograms. The latest ASR technologies are based on convolutional
neural networks (CNNs), recurrent neural networks (RNNs) and Transformers. The article presents an analysis of modern artificial intelligence algorithms adapted for automatic recognition of the Polish language. The differences between conventional architectures and ASR DNN End-To-End (E2E) models are discussed. Preliminary tests of five selected models (QuartzNet, FastConformer, Wav2Vec 2.0 XLSR, Whisper and ESPnet Model Zoo) on Mozilla Common Voice, Multilingual LibriSpeech and VoxPopuli databases are demonstrated. Tests were conducted for clean audio signal, signal with bandwidth limitation and degraded. The tested models were evaluated on the basis of Word Error Rate (WER).


R. Ardila et al., “Common Voice: a Massively-Multilingual Speech Corpus,” Proc. of the 12th Conf. on Language Resources and Evaluation (LREC 2020), pp. 4218–4222, 11–16 May 2020. [Online]. Available: https://www.aclweb.org/anthology/2020.lrec-1.520.pdf

J. Mahaveer et al., “Mls: A large-scale multilingual dataset for speech research.”

C. Wang et al., “VoxPopuli: A Large-Scale Multilingual Speech Corpus for Representation Learning, Semi-Supervised Learning and Interpretation,” arXiv (Cornell University), Jan. 2021. [Online]. Available: https://doi.org/10.48550/arXiv.2101.00390

“NVIDIA NeMo conversational AI toolki github repository.” [Online]. Available: https://github.com/NVIDIA/NeMo

“STT Pl Quartznet15x5 model.” [Online]. Available: https://catalog. ngc.nvidia.com/orgs/nvidia/teams/nemo/models/stt pl quartznet15x5

“NVIDIA FastConformer-Hybrid Large (pl) model.” [Online]. Available: https://catalog.ngc.nvidia.com/orgs/nvidia/teams/nemo/models/stt

pl fastconformer hybrid large pc

“Whisper Github repository.” [Online]. Available: https://github.com/ openai/whisper

“Zoo espnet multi-purpose pre-trained model.” [Online]. Available: https://github.com/espnet/espnet model zoo

“Fine-tuned wav2vec2-xlsr-53 large model for speech recognition in polish.” [Online]. Available: https://huggingface.co/jonatasgrosman/ wav2vec2-large-xlsr-53-polish

L. R. Rabiner and B.-H. Juang, “An introduction to hidden Markov models,” IEEE ASSP magazine, vol. 3, no. 1, pp. 4–16, Jan. 1986.

[Online]. Available: https://doi.org/10.1109/massp.1986.1165342

W. Cavnar and J. Trenkle, “N-gram-based text categorization,” Proc. of the 3rd Annu. Symp. on Document Analysis and Information Retrieval, pp. 161–175, 11-13 Apr 1994. [Online]. Available:


M. J. F. Gales and S. Young, “The application of hidden Markov models in speech recognition,” Foundations and Trends in Signal Processing, vol. 1, no. 3, pp. 195–304, Jan. 2007. [Online]. Available: https://doi.org/10.1561/2000000004

J. Holmes, W. Holmes, and P. Garner, “Using formant frequencies in speech recognition,” in Proc. 5th European Conf. on Speech Communication and Technology (Eurospeech 1997). Rhodes, Greece: ISCA Speech, 22-25 Sep. 1997, pp. 2083–2086. [Online]. Available: https://doi.org/10.21437/Eurospeech.1997-551

F. Honig, G. Stemmer, C. Hacker, and F. Brugnara, “Revising Perceptual Linear Prediction (PLP),” in INTERSPEECH 2016. San Francisco, CA, USA: ISCA Speech, 8–12 Sep. 2016, pp. 410–414.

[Online]. Available: https://doi.org/10.21437/Interspeech.2016-1446

R. M. S. Chanwoo Kim, “Power-Normalized Cepstral Coefficients (PNCC) for Robust Speech Recognition,” IEEE/ACM Transactions on Audio Speech and Language Processing, vol. 24, no. 7, pp. 1315 – 1329, Jul. 2016. [Online]. Available: https://doi.org/10.1109/TASLP. 2016.2545928

F. Zheng, G. Zhang, and Z. Song, “Comparison of different implementations of MFCC,” Journal of Computer Science and Technology, vol. 16, no. 6, pp. 582–589, 11 2001. [Online].

Available: https://doi.org/10.1007/bf02943243

A. Pohl and B. Ziołko,´ Using part of speech N-Grams for improving automatic speech recognition of Polish, Jan. 2013. [Online]. Available: https://doi.org/10.1007/978-3-642-39712-7 38

K. Marasek et al., “Spoken language translation for polish,” arXiv (Cornell University), Nov. 2015. [Online]. Available: https://doi.org/10. 48550/arXiv.1511.07788

A. Graves and N. Jaitly, “Towards End-To-End Speech Recognition with Recurrent Neural Networks,” Proc. of 31st Int. Conf. on Machine Learning, pp. 1764–1772, 21–26 Jun. 2014. [Online]. Available: http://proceedings.mlr.press/v32/graves14.pdf

O. Vinyals, S. V. Ravuri, and D. Povey, “Revisiting recurrent neural networks for robust asr,” in 2012 IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP). Kyoto, Japan: IEEE, 25 - 30 Mar. 2012, pp. 4085–4088. [Online]. Available: https://doi.org/10.1109/ ICASSP.2012.6288816

Dan Jurafsky and James H. Martin, “Speech and Language Processing (3rd ed. draft).” [Online]. Available: https://web.stanford.edu/∼jurafsky/ slp3/

Y. Zhang et al., “Towards End-to-End Speech Recognition with Deep Convolutional Neural Networks,” in INTERSPEECH 2005 - Eurospeech, 9th European Conf. on Speech Communication and Technology. Lisbon, Portugal: ISCA Speech, 4-8 Sep. 2005, pp. 2997–3000. [Online].

Available: https://doi.org/10.21437/Interspeech.2005-138

A. V. Oppenheim, “Speech spectrograms using the fast Fourier transform,” IEEE Spectrum, vol. 7, no. 8, pp. 57–62, 8 1970. [Online]. Available: https://doi.org/10.1109/mspec.1970.5213512

S. William and C. Jim, “End-to-end deep neural network for automatic speech recognition,” Stanford CS224D Reports, 2015.

[Online]. Available: http://cs224d.stanford.edu/reports/SongWilliam.pdf

B. X. Linhao Dong, Shuang Xu, “Speech-transformer: A norecurrence sequence-to-sequence model for speech recognition,” in 2018 IEEE 40th Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 15-20 Apr. 2018, pp. 5884–5888. [Online]. Available: https://doi.org/10.1109/ICASSP.2018.8462506

A. Vaswani et al., “Attention is all you need,” in Advances in Neural Information Processing Systems 30 (NIPS 2017), I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, Eds., vol. 30. Curran Associates, Inc., 2017.

[Online]. Available: https://proceedings.neurips.cc/paper files/paper/ 2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf

“Common Voice open source, multi-language dataset of voices.”

[Online]. Available: https://commonvoice.mozilla.org/en/datasets

“Dataset Card for CommonVoice) Hugging Face.” [Online]. Available: https://huggingface.co/datasets/common voice

V. o. Panayotov, “Librispeech: an asr corpus based on public domain audio books,” in 2015 IEEE 40th Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 19-24 Apr. 2015, pp. 5206–5210. [Online]. Available: https://doi.org/10.1109/ICASSP.2015.7178964

“LibriVox free public domain audiobooks.” [Online]. Available: https://librivox.org/

“Project Gutenberg free eBooks.” [Online]. Available: https://www. gutenberg.org/

“Multilingual LibriSpeech (MLS) Website.” [Online]. Available: https://www.openslr.org/94/

“VoxPopuli Github repository.” [Online]. Available: https://github.com/ facebookresearch/voxpopuli

“Github The AI-powered developer platform to build, scale, and deliver secure software.” [Online]. Available: https://github.com/

“Hugging Face The AI community building the future.” [Online].

Available: https://huggingface.co/

S. Watanabe et al., “Espnet: End-to-end speech processing toolkit,” in INTERSPEECH 2018. ISCA Speech, 2-6 Sep. 2018, pp. 2207–2211.

[Online]. Available: https://doi.org/10.21437/Interspeech.2018-1456

S. Kriman et al., “Quartznet: Deep automatic speech recognition with 1d time-channel separable convolutions,” in 2020 IEEE International Conf. on Acoustics, Speech and Signal Processing (ICASSP). Barcelona, Spain: IEEE, 04–08 May 2020, p. 6124–6128. [Online]. Available: https://doi.org/10.1109/ICASSP40776.2020.9053889

J. Li et al., “Jasper: An end-to-end convolutional neural acoustic model,” arXiv (Cornell University), Apr. 2019. [Online]. Available: https://doi.org/10.48550/arXiv.1904.03288

A. Waibel et al., “Phoneme recognition using time-delay neural networks,” IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 37, no. 3, pp. 328–339, Mar. 1989. [Online]. Available: https://doi.org/10.1109/29.21701

A. ulati et al., “Conformer: Convolution-augmented Transformer for Speech Recognition,” in INTERSPEECH 2020. Shanghai, China: ISCA Speech, 25-29 Oct. 2020, pp. 5036–5040. [Online]. Available: https://doi.org/10.21437/Interspeech.2020-3015

“NVIDIA Models - NeMo Core.” [Online]. Available: https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/ main/asr/models.html#hybrid-transducer-ctc

A. Baevski et al., “wav2vec 2.0: A Framework for

Self-Supervised Learning of Speech Representations,” Neural Information Processing Systems, vol. 33, pp. 12449–12460, 6

[Online]. Available: https://proceedings.neurips.cc/paper/2020/ file/92d1e1eb1cd6f9fba3227870bb6d7f07-Paper.pdf

“MetaAI Wav2vec 2.0: Learning the structure of speech from raw audio.” [Online]. Available: https://ai.meta.com/blog/ wav2vec-20-learning-the-structure-of-speech-from-raw-audio/

C. Alexis et al., “Unsupervised cross-lingual representation learning for speech recognition,” in INTERSPEECH 2021. Brno, Czechia: ISCA Speech, 30 Aug. - 3 Sep. 2021, pp. 2426–2430. [Online]. Available: https://doi.org/10.21437/Interspeech.2021-329

P. Roach et al., “Babel: An eastern european multi-language database,” in Proc. of 4th International Conf. on Spoken Language Processing. ICSLP ’96. IEEE, 3-6 Oct. 1996. [Online]. Available: https://doi.org/10.1109/ICSLP.1996.608002

A. Conneau et al., “FLEURS: FEW-Shot Learning Evaluation of Universal Representations of Speech,” 2022 IEEE Spoken Language Technology Workshop (SLT), 1 2023. [Online]. Available: https: //doi.org/10.1109/slt54892.2023.10023141

“Espnet2 Espnet major update.” [Online]. Available: ttps://espnet.github. io/espnet/espnet2 tutorial.html

“List of ESPnet2 corpora.” [Online]. Available: https://github.com/ espnet/espnet/blob/master/egs2/README.md

A. Ahmed and R. Steve, “Word Error Rate Estimation for Speech Recognition: e-WER,” in Proc. of the 56th Annu. Meeting of the Association for Computational Linguistics (Volume 2: Short Papers).

Melbourne, Australia: -, Association for Computational Linguistics

, pp. 20—-24. [Online]. Available: https://aclanthology.org/ P18-2004/

“Google Colaboratory hosted Jupyter Notebook service.” [Online].

Available: https://colab.google/

“Audio degradation toolbox in python, with a command-line tool. github repository.” [Online]. Available: https://github.com/ emilio-molina/audio degrader

“Jiwer ASR evaluation library.” [Online]. Available: https://jitsi.github.


Additional Files