Pengembangan model akustik dengan deep neural network untuk sistem pengenalan wicara bahasa Indonesia

Gunarso Gunarso; Agus Buono; Mushthofa Mushthofa; Mohammad Teduh Uliniansyah

doi:10.24246/aiti.v22i1.84-100

Authors

Gunarso Departemen Ilmu Komputer https://orcid.org/0000-0002-9307-9808
Agus Buono Departemen Ilmu Komputer, FMIPA IPB, Bogor 16680, Indonesia https://orcid.org/0000-0002-4273-0269
Mushthofa Departemen Ilmu Komputer, FMIPA IPB, Bogor 16680, Indonesia
Mohammad Teduh Uliniansyah Pusat Riset Sains Data dan Informasi, Badan Riset dan Inovasi Nasional

DOI:

https://doi.org/10.24246/aiti.v22i1.84-100

Keywords:

hybrid TDNN-LSTM, LSTM, acoustic model, speech recognition, TDNN

Abstract

The Deep Neural Network (DNN)-based approach offers significantly higher accuracy compared to traditional methods such as Hidden Markov Model (HMM)-Gaussian Mixture Model (GMM) in acoustic model development. In this research, three popular DNN variants were evaluated: Time-Delay Neural Network (TDNN), Long Short-Term Memory (LSTM), and a hybrid combination of TDNN-LSTM for acoustic model development in Indonesian speech recognition. Using the KDW-BPPT-50K-ASR1 speech data for over 92 hours, acoustic models were trained, and experiments were conducted to analyze their performance. Research results show that the hybrid TDNN-LSTM model achieved the best performance with a Word Error Rate (WER) of 9.67%, outperforming TDNN with a WER of 12.16% and LSTM with a WER of 10.6%. This finding confirms that the hybrid model is able to improve the accuracy of Indonesian speech recognition compared to using TDNN or LSTM separately. These results provide a significant contribution to the development of more accurate and efficient speech recognition systems.

Downloads

Download data is not yet available.

Metrics

Metrics Loading ...

References

S. Karpagavalli and E. Chandra, “A Review on Automatic Speech Recognition Architecture and Approaches,” International Journal of Signal Processing, Image Processing and Pattern Recognition, vol. 9, no. 4, pp. 393–404, 2016. DOI: https://doi.org/10.14257/ijsip.2016.9.4.34

G. Hinton et al., “Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups,” IEEE Signal Process Mag, vol. 29, no. 6, pp. 82–97, 2012, doi: 10.1109/MSP.2012.2205597. DOI: https://doi.org/10.1109/MSP.2012.2205597

A. B. Nassif, I. Shahin, I. Attili, M. Azzeh, and K. Shaalan, “Speech Recognition Using Deep Neural Networks: A Systematic Review,” IEEE Access, vol. 7, pp. 19143–19165, 2019, doi: 10.1109/ACCESS.2019.2896880. DOI: https://doi.org/10.1109/ACCESS.2019.2896880

V. Peddinti, D. Povey, and S. Khudanpur, “A time delay neural network architecture for efficient modeling of long temporal contexts,” Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH, vol. 2015-Janua, pp. 3214–3218, 2015, doi: 10.21437/interspeech.2015-647. DOI: https://doi.org/10.21437/Interspeech.2015-647

H. Sak, A. Senior, and F. Beaufays, “Long Short-Term Memory Based Recurrent Neural Network Architectures for Large Vocabulary Speech Recognition,” Feb. 2014, [Online]. Available: http://arxiv.org/abs/1402.1128 DOI: https://doi.org/10.21437/Interspeech.2014-80

X. Li and X. Wu, “Constructing long short-term memory based deep recurrent neural networks for large vocabulary speech recognition,” ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings, vol. 2015-Augus, pp. 4520–4524, 2015, doi: 10.1109/ICASSP.2015.7178826. DOI: https://doi.org/10.1109/ICASSP.2015.7178826

N. O. Balula, M. Rashwan, and S. M. Abdo, “TDNN-LSTM-based Acoustic Modeling for verification of Qur’an Recitation for Non-Arabic Speakers using Kaldi Toolkit,” Int J Comput Appl, vol. 183, no. 28, 2021, doi: 10.5120/ijca2021921152. DOI: https://doi.org/10.5120/ijca2021921152

C. W. Chen, Y. F. Yeh, C. L. Lin, S. P. Tseng, and J. F. Wang, “Hybrid deep neural network acoustic model for taiwanese speech recognition,” 2020 8th International Conference on Orange Technology, ICOT 2020, Dec. 2020, doi: 10.1109/ICOT51877.2020.9468762. DOI: https://doi.org/10.1109/ICOT51877.2020.9468762

A. Waibel, T. Hanazawa, G. Hinton, K. Shikano, and K. J. Lang, “Phoneme Recognition Using Time-Delay Neural Networks,” IEEE Trans Acoust, vol. 37, no. 3, pp. 328–339, 1989, doi: 10.1109/29.21701. DOI: https://doi.org/10.1109/29.21701

S. Hochreiter and J. Schmidhuber, “Long Short-Term Memory,” Neural Comput, vol. 9, no. 8, pp. 1735–1780, Nov. 1997, doi: 10.1162/neco.1997.9.8.1735. DOI: https://doi.org/10.1162/neco.1997.9.8.1735

A. L. Georgescu, H. Cucu, and C. Burileanu, “Kaldi-based DNN architectures for speech recognition in Romanian,” 2019 10th International Conference on Speech Technology and Human-Computer Dialogue, SpeD 2019, no. i, pp. 1–6, 2019, doi: 10.1109/SPED.2019.8906555. DOI: https://doi.org/10.1109/SPED.2019.8906555

M. Teduh Uliniansyah, H. Riza, A. Santosa, Gunarso, M. Gunawan, and E. Nurfadhilah, “Development of text and speech corpus for an Indonesian speech-to-speech translation system,” in 2017 20th Conference of the Oriental Chapter of International Committee for Coordination and Standardization of Speech Databases and Assessment Techniques, O-COCOSDA 2017, Institute of Electrical and Electronics Engineers Inc., Jun. 2018, pp. 53–57. doi: 10.1109/ICSDA.2017.8384448. DOI: https://doi.org/10.1109/ICSDA.2017.8384448

D. S. Mistry, “Speech Recognition using MFCC and Neural Networks,” 2014, [Online]. Available: www.ijedr.org

D. Povey et al., “The Kaldi Speech Recognition Toolkit,” IEEE 2011 Workshop on Automatic Speech Recognition and Understanding, Hilton Waikoloa Village, Big Island, Hawaii, US, 2011.

A. Ali and S. Renals, “Word error rate estimation for speech recognition: E-wer,” ACL 2018 - 56th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference (Long Papers), vol. 2, no. 2014, pp. 20–24, 2018, doi: 10.18653/v1/p18-2004. DOI: https://doi.org/10.18653/v1/P18-2004