Skip to main content
SpringerLink
Log in

Acoustic model with hybrid Deep Bidirectional Single Gated Unit (DBSGU) for low resource speech recognition

  • Published:
Multimedia Tools and Applications Aims and scope Submit manuscript

Abstract

Nowadays Long Short-Term Memory RNNs (LSTM RNNs) are widely used in Automatic Speech Recognition (ASR) and achieved excellent result in the problem of vanishing gradients. Bidirectional LSTM (BLSTM) will run the inputs in two ways ,both past as well as in future that shows good performance. However implementation of BLSTM is quite difficult because of its high computational requirements and also the problem of vanishing gradients still persist, when we have multiple layer of LSTM. The extensive size of LSTM systems makes them powerless in over fitting issues. The Gated Recurrent Unit (GRU) is the latest generation recurrent neural networks with two gates. The update gate acts similar to forget and input gates of LSTM’s and reset gate responsible to decide how much previous data you should remember. GRU avoids over fitting and also training the GRU is faster compared to LSTM, since size of the GRU network is small. The proposed work is in two- fold architecture. First stage, we tend to reduce the gates in GRU by combining the reset and update gate together to form a Single Gated Unit (SGU). SGU takes half of the parameter compared with LSTM and one third of parameter compared with GRU. It increases the training speed of SGU. Second stage, SGU is combined with Deep Bidirectional design (DBSGU) to build a hybrid acoustic model that takes less number of parameters and increases the learning capability. The proposed model is compared with similarities and differences between Deep Bidirectional GRU (DBGRU) and Deep Bidirectional LSTM (DBLSTM) and found that 2 to 4% decrease in Word Error Rate (WER).The Learning rate of the is increased by 30% The entire work has been evaluated on Crowd Sourced high-quality Multi-Speaker speech (CSMS) data set.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10

Similar content being viewed by others

References

  1. Abandah GA, Graves A, Al-Shagoor B, Arabiyat A, Al-Taee M (2015) Automatic diacritization of Arabic text using recurrent neural networks. Int J Doc Anal Recognit (IJDAR) 18(2):183–197

    Article  Google Scholar 

  2. Barman PP, Boruah A (2018) A RNN based approach for next word prediction in Assamese phonetic transcription. Procedia Comput Sci 143:117–123 (ISSN 1877 – 0509)

    Article  Google Scholar 

  3. Chavandan RS, Sable GS (2013) An overview of speech recognition using HMM. Int J Comput Sci Mob Comput 2(6):233–238

  4. Chung J, Gulcehre C, Cho K, Bengio Y (2014) Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv arXiv:1412.3555

  5. Cheng G, Povey D, Huang L, Xu J, Khudanpur S, Yan Y (2018) Output-gate projected gated recurrent unit for speech recognition. Interspeech, pp 1793–1797

  6. Graves A, Schmidhuber J (2005) Framewise phoneme classification with bidirectional LSTM and other neural network architectures. Neural Netw 18(5–6):602–610

    Article  Google Scholar 

  7. Graves A, Mohamed A, Hinton G (2013) Speech recognitionwith deep recurrent neural networks. In: Proc ICASSP 2013, Vancouver, Canada

  8. Greff K, Srivastava RK, Koutnk J, Steunebrink BR, Schmidhuber J (2015) LSTM: A search space odyssey. arXiv: 1503.04069

  9. He F, Chu SH, Kjartansson O, Rivera C, Katanova A, Gutkin A, Demirsahin I, Johny C, Jansche M, Sain S et al (2020) Open-source Multi-speaker Speech Corpora for Building Gujarati, Kannada, Malayalam, Marathi, Tamil and Telugu Speech Synthesis Systems. In: Proceedings of the 12th LREC Conference, Marseille, France, 11–16

  10. Hochreiter S, Jürgen S (1997) Long short-term memory. Neural Comput 9:1735–80. https://doi.org/10.1162/neco.1997.9.8.1735

    Article  Google Scholar 

  11. Jozefowicz R, Zaremba W, Sutskever I (2015) An empirical exploration of recurrent network architectures. In Proceedings of the 32nd International Conference on Machine Learning, Lille, France, vol 37, pp 2342–2350

  12. Kim J, Kim J, Thu HLT, Kim H (2016) Long short term memory recurrent neural network classifier for intrusion detection. International Conference on Platform Technology and Service (PlatCon), Jeju, pp 1–5

  13. Kingma DP, Ba J (2014) Adam: A method for stochastic optimization, CoRR, vol abs/1412.6980

  14. Kumar J, Goomer R, Singh AK (2018) Long Short Term Memory Recurrent Neural Network (LSTM-RNN) based workload forecasting model for cloud datacenters. Procedia Comput Sci 125:676–682 (ISSN 1877 – 0509)

    Article  Google Scholar 

  15. Kumar S, Hussain L, Banarjee S, Reza M (2018) Energy load forecasting using deep learning approach-LSTM and GRU in spark cluster. Fifth International Conference on Emerging Applications of Information Technology (EAIT), Kolkata, pp 1–4

  16. Li X, Xianyu H, Tian J, Chen W, Meng F, Xu M et al (2016) A deep bidirectional long short-term memory based multi-scale approach for music dynamic emotion prediction. In: IEEE International Conference in Acoustics, Speech and Signal Processing (ICASSP); Shanghai, China, p 544–548

  17. Panayotov V, Chen G, Povey D, Khudanpur S. Librispeech: an ASR corpus based on public domain audio books. In: International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Brisbane, Australia, pp 5206–5210

  18. Panzner M, Cimiano P (2016) Comparing hidden Markov models and long short term memory neural networks for learning action representations. In: Pardalos P, Conca P, Giuffrida G, Nicosia G (eds) Machine Learning, Optimization, and Big Data. MOD 2016, vol 10122. Springer, Cham

    Google Scholar 

  19. Povey D, Ghoshal A, Boulianne G, Goel N, Hannemann M, Qian Y, Schwarz P, Stemmer G (2011) The kaldi speech recognitiontoolkit. In: Workshop on Automatic Speech Recognition and Understanding (ASRU), Hawaii, US, pp 1–4

  20. Ravanelli M, Brakel P, Omologo M, Bengio Y (2018) Light gated recurrent units for speech recognition. IEEE Trans Emerg Top Comput Intell 2(2):92–102

    Article  Google Scholar 

  21. Schuster M, Paliwal KK (1997) Bidirectional recurrent neural networks. Signal Process IEEE Trans 45(11):2673–2681

    Article  Google Scholar 

  22. Stolcke A (2002) SRILM-An extensible language modeling toolkit. In: International Conference on Spoken Language Processing (ICSLP), Denver, Colorado, pp 901–904

  23. Thireou T, Reczko M (2007) Bidirectional Long Short-Term Memory Networks for Predicting the Subcellular Localization of Eukaryotic Proteins. IEEE/ACM Trans Comput Biol Bioinform 4(3):441–446

  24. Zhang Y, Chen G, Yu D, Yao K, Khudanpur S, Glass JR (2016) Highway long short-term memory RNNS for distant speech recognition. In: Proc. of ICASSP 2016, pp 5755–5759

  25. Zhou G-B, Wu J, Zhang C-L, Zhou Z-H (2016) Minimal gated unit for recurrent neural networks. Int J Automat Comput 13(3):226–234

Download references

Author information

Authors and Affiliations

  1. Department of Computer Science and Engineering, School of Computing, SRM Institute of Science and Technology, Chennai, India

    S. Girirajan & A. Pandian

Authors
  1. S. Girirajan
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. A. Pandian
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to S. Girirajan.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Girirajan, S., Pandian, A. Acoustic model with hybrid Deep Bidirectional Single Gated Unit (DBSGU) for low resource speech recognition. Multimed Tools Appl 81, 17169–17184 (2022). https://doi.org/10.1007/s11042-022-12723-4

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11042-022-12723-4

Keywords

Search

Navigation