Abstract
Semi-supervised training and language adversarial transfer learning are two different techniques to improve the Automatic Speech Recognition (ASR) performance in limited resource conditions. In this paper, we combined these two techniques and presented a common framework for the Hindi ASR system. For acoustic modeling, we proposed a hybrid architecture of SincNet-Convolutional Neural Network (CNN)-Light Gated Recurrent Unit (LiGRU), which shows increased interpretability, high accuracy, and fewer parameter size. We investigate the impact of the proposed hybrid model on monolingual Hindi ASR with semi-supervised training, and multilingual Hindi ASR with language adversarial transfer learning. In this work, we have chosen three Indian languages (Hindi, Marathi, Bengali) of the same Indo-Aryan family for multilingual training. All experiments were conducted using Kaldi and Py-Torch Kaldi toolkits. The proposed model with combined learning strategies helps to get the lowest 5.5% Word Error Rate (WER) for Hindi ASR.
Similar content being viewed by others
Notes
Winter school on ASR, 16–28 May 2017, IIT Guwahati.
Abbreviations
- ASR:
-
Automatic Speech Recognition
- BLSTM:
-
Bidirectional long short-term memory
- BPTT:
-
Back-propagation through time
- CD:
-
Context-dependent
- CI:
-
Context-independent
- CNN:
-
Convolutional Neural Network
- DNN:
-
Deep Neural Network
- Fbank:
-
log-Mel Filterbank
- FC:
-
Fully-connected
- GMM:
-
Gaussian mixture model
- GRL:
-
Gradient reversal layer
- GRU:
-
Gated recurrent unit
- HMM:
-
Hidden Markove model
- LF-MMI:
-
Lattice-free maximum mutual information
- LiGRU:
-
Light gated recurrent unit
- LSTM:
-
Long short-term memory
- LVCSR:
-
Large vocabulary continuous speech recognition
- MFCC:
-
Mel-frequency cepstral coefficient
- ML:
-
Maximum-likelihood
- MLP:
-
Multi-layer perceptron
- RNN:
-
Recurrent neural network
- RNNLM:
-
Recurrent neural network language model
- SGD:
-
Stochastic-gradient decent
- SHL:
-
Shared-hidden layer
- SOTA:
-
State-of-the-art
- SRILM:
-
SRI language modeling
- TDNN:
-
Time-delay neural network
- WER:
-
Word error rate
References
Aggarwal RK, Dave M (2013) Performance evaluation of sequentially combined heterogeneous feature streams for Hindi speech recognition system. Telecommun Syst 52(3):1457–1466
Alumäe T, Tsakalidis S, Schwartz RM (2016) Improved multilingual training of stacked neural network acoustic models for low resource languages. In: Interspeech, pp 3883–3887
Barker J, Watanabe S, Vincent E, Trmal J (2018) The fifth’chime’speech separation and recognition challenge: dataset, task and baselines. arXiv preprint arXiv:1803.10609
Biswas A, Menon R, van der Westhuizen E, Niesler T (2019) Improved low-resource somali speech recognition by semi-supervised acoustic and language model training. arXiv preprint arXiv:1907.03064
Biswas A, de Wet F, van der Westhuizen E, Yilmaz E, Niesler T (2018) Multilingual neural network acoustic modelling for ASR of under-resourced English-isizulu code-switched speech. In: Interspeech, pp 2603–2607
Chellapriyadharshini M, Toffy A, Ramasubramanian V et al (2018) Semi-supervised and active-learning scenarios: Efficient acoustic model refinement for a low resource indian language. arXiv preprint arXiv:1810.06635
Chen NF, Lim BP, Hasegawa-Johnson MA et al (2017) Multitask learning for phone recognition of underresourced languages using mismatched transcription. IEEE/ACM Trans Audio Speech Lang Process 26(3):501–514
Chen NF, Lim BP, Ni C, Xu H, HasegawaJohnson M, Chen W, Xiao X, Sivadas S, Chng ES, Ma B et al (2017) Low-resource spoken keyword search strategies in georgian inspired by distinctive feature theory. In: 2017 Asia-Pacific signal and information processing association annual summit and conference (APSIPA ASC), pp 1322–1327. IEEE
Chen X, Shi Z, Qiu X, Huang X (2017) Adversarial multi-criteria learning for chinese word segmentation. arXiv preprint arXiv:1704.07556
Cho K, Van Merriënboer B, Bahdanau D, Bengio Y (2014) On the properties of neural machine translation: encoder–decoder approaches. arXiv preprint arXiv:1409.1259
Chu SM, Povey D, Kuo HK, Mangu L, Zhang S, Shi Q, Qin Y (2010) The 2009 ibm gale mandarin broadcast transcription system. In: 2010 IEEE international conference on acoustics, speech and signal processing, pp 4374–4377. IEEE
Cui J, Kingsbury B, Ramabhadran B, Sethy A, Audhkhasi K, Cui X, Kislal E, Mangu L, Nussbaum-Thom M, Picheny M et al (2015) Multilingual representations for low resource speech recognition and keyword search. In: 2015 IEEE workshop on automatic speech recognition and understanding (ASRU), pp 259–266. IEEE
Dahl GE, Yu D, Deng L, Acero A (2011) Context-dependent pre-trained deep neural networks for large-vocabulary speech recognition. IEEE Trans Audio Speech Language Process 20(1):30–42
Dash D, Kim MJ, Teplansky K, Wang J (2018) Automatic speech recognition with articulatory information and a unified dictionary for Hindi, Marathi, Bengali and Oriya. In: Interspeech, pp 1046–1050
Dua M, Aggarwal RK, Biswas M (2018) Discriminative training using noise robust integrated features and refined hmm modeling. J Intell Syst 29(1):327–344
Dua M, Aggarwal RK, Biswas M (2019) Discriminatively trained continuous Hindi speech recognition system using interpolated recurrent neural network language modeling. Neural Comput Appl 31(10):6747–6755
Fathima N, Patel T, Mahima C, Iyengar A (2018) Tdnn-based multilingual speech recognition system for low resource Indian languages. In: Interspeech, pp 3197–3201
Gales MJ, Knill KM, Ragni A, Rath SP (2014) Speech recognition and keyword spotting for low-resource languages: Babel project research at cued. In: Fourth International workshop on spoken language technologies for under-resourced languages (SLTU-2014), pp 16–23. International Speech Communication Association (ISCA)
Ganin Y, Lempitsky V (2015) Unsupervised domain adaptation by backpropagation. In: International conference on machine learning, pp 1180–1189
Ganin Y, Ustinova E, Ajakan H, Germain P, Larochelle H, Laviolette F, Marchand M, Lempitsky V (2016) Domain-adversarial training of neural networks. J Mach Learn Res 17(1):2026–2030
Ghoshal A, Swietojanski P, Renals S (2013) Multilingual training of deep neural networks. In: 2013 IEEE international conference on acoustics, speech and signal processing, pp 7319–7323. IEEE
Grézl F, Karafiat M, Janda M (2011) Study of probabilistic and bottle-neck features in multilingual environment. In: 2011 IEEE workshop on automatic speech recognition & understanding, pp 359–364. IEEE
Hain T, Woodland P, Evermann G, Povey D (2000) The CU-HTK march 2000 hub5e transcription system. In: Proc. speech transcription workshop, vol 1. Citeseer
Hartmann W, Hsiao R, Tsakalidis S (2017) Alternative networks for monolingual bottleneck features. In: 2017 IEEE international conference on acoustics, speech and signal processing (ICASSP), pp 5290–5294. IEEE
Heigold G, Vanhoucke V, Senior A, Nguyen P, Ranzato M, Devin M, Dean J (2013) Multilingual acoustic models using distributed deep neural networks. In: 2013 IEEE international conference on acoustics, speech and signal processing, pp 8619–8623. IEEE
Hernandez F, Nguyen V, Ghannay S, Tomashenko N, Estève Y (2018) Ted-lium 3: twice as much data and corpus repartition for experiments on speaker adaptation. In: International conference on speech and computer, pp 198–208. Springer
Hinton G, Deng L, Yu D, Dahl GE, Mohamed A, Jaitly N, Senior A, Vanhoucke V, Nguyen P, Sainath TN et al (2012) Deep neural networks for acoustic modeling in speech recognition: the shared views of four research groups. IEEE Signal Process Mag 29(6):82–97
Hochreiter S, Schmidhuber J (1997) Long short-term memory. Neural Comput 9(8):1735–1780
Hoshen Y, Weiss RJ, Wilson KW (2015) Speech acoustic modeling from raw multichannel waveforms. In: 2015 IEEE international conference on acoustics, speech and signal processing (ICASSP), pp 4624–4628. IEEE
Huang JT, Li J, Yu D, Deng L, Gong Y (2013) Cross-language knowledge transfer using multilingual deep neural network with shared hidden layers. In: 2013 IEEE international conference on acoustics, speech and signal processing, pp 7304–7308. IEEE
Jung JW, Heo HS, Yang IH, Shim HJ, Yu HJ (2018) Avoiding speaker overfitting in end-to-end DNNS using raw waveform for text-independent speaker verification. Extraction 8(12):23–24
Kadyan V, Mantri A, Aggarwal R (2018) Refinement of HMM model parameters for Punjabi automatic speech recognition (PASR) system. IETE J Res 64(5):673–688
Ko T, Peddinti V, Povey D, Khudanpur S (2015) Audio augmentation for speech recognition. In: Sixteenth annual conference of the international speech communication association
Kriman S, Beliaev S, Ginsburg B, Huang J, Kuchaiev O, Lavrukhin V, Leary R, Li J, Zhang Y (2020) Quartznet: Deep automatic speech recognition with 1d time-channel separable convolutions. In: ICASSP 2020-2020 IEEE international conference on acoustics, speech and signal processing (ICASSP), pp 6124–6128. IEEE
Lazaridis A, Himawan I, Motlicek P, Mporas I, Garner PN (2016) Investigating cross-lingual multi-level adaptive networks: The importance of the correlation of source and target languages. In: Proceedings of the international workshop on spoken language translation, CONF
Liu D, Xu J, Zhang P, Yan Y (2019) Investigation of knowledge transfer approaches to improve the acoustic modeling of vietnamese asr system. IEEE/CAA J Autom Sin 6(5):1187–1195
Miao Y, Metze F (2013) Improving low-resource CD-DNN-HMM using dropout and multilingual DNN training. Interspeech 13:2237–2241
Ni C, Leung CC, Wang L, Chen NF, Ma B (2017) Efficient methods to train multilingual bottleneck feature extractors for low resource keyword search. In: 2017 IEEE international conference on acoustics, speech and signal processing (ICASSP), pp 5650–5654. IEEE
Palaz D, Collobert R et al (2015) Analysis of CNN-based speech recognition system using raw speech as input. Tech. rep, Idiap
Panayotov V, Chen G, Povey D, Khudanpur S (2015) Librispeech: an ASR corpus based on public domain audio books. In: 2015 IEEE international conference on acoustics, speech and signal processing (ICASSP), pp 5206–5210. IEEE
Parcollet T, Morchid M, Linarès G, De Mori R (2019) Bidirectional quaternion long short-term memory recurrent neural networks for speech recognition. In: ICASSP 2019-2019 IEEE international conference on acoustics, speech and signal processing (ICASSP), pp 8519–8523. IEEE
Parcollet T, Ravanelli M, Morchid M, Linarès G, Trabelsi C, De Mori R, Bengio Y (2018) Quaternion recurrent neural networks. arXiv preprint arXiv:1806.04418
Passricha V, Aggarwal RK (2019) Convolutional support vector machines for speech recognition. Int J Speech Technol 22(3):601–609
Ravanelli M, Bengio Y (2018) Interpretable convolutional filters with sincnet. arXiv preprint arXiv:1811.09725
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
Ravanelli M, Parcollet T, Bengio Y (2019) The Pytorch-Kaldi speech recognition toolkit. In: ICASSP 2019-2019 IEEE international conference on acoustics, speech and signal processing (ICASSP), pp 6465–6469. IEEE
Ravanelli M (2018) Interpretable convolutional filters with sincnet. arXiv preprint arXiv:1811.09725
Rebai I, BenAyed Y, Mahdi W, Lorré JP (2017) Improving speech recognition using data augmentation and acoustic model fusion. Proc Comput Sci 112:316–322
Roger V, Farinas J, Pinquier J (2020) Deep neural networks for automatic speech processing: a survey from large corpora to limited data. arXiv preprint arXiv:2003.04241
Sahraeian R, Van Compernolle D (2016) Using weighted model averaging in distributed multilingual DNNS to improve low resource ASR. Proc Comput Sci 81:152–158
Sahraeian R, Van Compernolle D (2018) Cross-entropy training of DNN ensemble acoustic models for low-resource ASR. IEEE/ACM Trans Audio Speech Lang Process 26(11):1991–2001
Sailor HB, Krishna MVS, Chhabra D, Patil AT, Kamble MR, Patil HA (2018) DA-IICT/IIITV system for low resource speech recognition challenge 2018. In: Interspeech, pp 3187–3191
Samudravijaya K, Rao P, Agrawal S (2000) Hindi speech database. In: Sixth International conference on spoken language processing
Saon G, Kurata G, Sercu T, Audhkhasi K, Thomas S, Dimitriadis D, Cui X, Ramabhadran B, Picheny M, Lim LL et al (2017) English conversational telephone speech recognition by humans and machines. arXiv preprint arXiv:1703.02136
Scanzio S, Laface P, Fissore L, Gemello R, Mana F (2008) On the use of a multilingual neural network front-end. In: Ninth annual conference of the international speech communication association
Sercu T, Puhrsch C, Kingsbury B, LeCun Y (2016) Very deep multilingual convolutional neural networks for LVCSR. In: 2016 IEEE international conference on acoustics, speech and signal processing (ICASSP), pp 4955–4959. IEEE
Shangguan Y, Li J, Qiao L, Alvarez R, McGraw I (2019) Optimizing speech recognition for the edge. arXiv preprint arXiv:1909.12408
Shetty VM, Sharon RA, Abraham B, Seeram T, Prakash A, Ravi N, Umesh S (2018) Articulatory and stacked bottleneck features for low resource speech recognition. In: Interspeech, pp 3202–3206
Shinohara Y (2016) Adversarial multi-task learning of deep neural networks for robust speech recognition. In: Interspeech, pp 2369–2372. San Francisco, CA, USA
Stolcke A (2002) Srilm-an extensible language modeling toolkit. In: Seventh international conference on spoken language processing
Tong S, Garner PN, Bourlard H (2017) An investigation of deep neural networks for multilingual speech recognition training and adaptation. In: Proc. of Interspeech, conf
Trmal J, Wiesner M, Peddinti V, Zhang X, Ghahremani P, Wang Y, Manohar V, Xu H, Povey D, Khudanpur S (2017) The kaldi openkws system: Improving low resource keyword search. In: Interspeech, pp 3597–3601
Tüske Z, Golik P, Schlüter R, Ney H (2014) Acoustic modeling with deep neural networks using raw time signal for LVCSR. In: Fifteenth annual conference of the international speech communication association
Tüske Z, Pinto J, Willett D, Schlüter R (2013) Investigation on cross-and multilingual MLP features under matched and mismatched acoustical conditions. In: 2013 IEEE international conference on acoustics, speech and signal processing, pp 7349–7353. IEEE
Tzeng E, Hoffman J, Saenko K, Darrell T (2017) Adversarial discriminative domain adaptation. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 7167–7176
Veselỳ K, Karafiát M, Grézl F, Janda M, Egorova E (2012) The language-independent bottleneck features. In: 2012 IEEE spoken language technology workshop (SLT), pp 336–341. IEEE
Vu NT, Schultz T (2013) Multilingual multilayer perceptron for rapid language adaptation between and across language families. In: Interspeech, pp 515–519
Vydana HK, Gurugubelli K, Vegesna VVR, Vuppala AK (2018) An exploration towards joint acoustic modeling for Indian languages: Iiit-h submission for low resource speech recognition challenge for Indian languages, interspeech 2018. In: Interspeech, pp 3192–3196
Wilkinson N, Biswas A, Yılmaz E, De Wet F, van der Westhuizen E, Niesler TR (2020) Semi-supervised acoustic modelling for five-lingual code-switched ASR using automatically-segmented soap opera speech. arXiv preprint arXiv:2004.06480
Xu H, Do VH, Xiao X, Chng ES (2015) A comparative study of BNF and DNN multilingual training on cross-lingual low-resource speech recognition. In: Sixteenth annual conference of the international speech communication association
Xu H, Li K, Wang Y, Wang J, Kang S, Chen X, Povey D, Khudanpur S (2018) Neural network language modeling with letter-based features and importance sampling. In: 2018 IEEE international conference on acoustics, speech and signal processing (ICASSP), pp 6109–6113. IEEE
Xu H, Su H, Ni C, Xiao X, Huang H, Chng ES, Li H (2016) Semi-supervised and cross-lingual knowledge transfer learnings for DNN hybrid acoustic models under low-resource conditions. In: Interspeech, pp 1315–1319
Yi J, Tao J, Wen Z, Bai Y (2018) Language-adversarial transfer learning for low-resource speech recognition. IEEE/ACM Trans Audio Speech Lang Process 27(3):621–630
Yılmaz E, van den Heuvel H, van Leeuwen D (2016) Investigating bilingual deep neural networks for automatic recognition of code-switching Frisian speech. Proc Comput Sci 81:159–166
Yin W, Kann K, Yu M, Schütze H (2017) Comparative study of CNN and RNN for natural language processing. arXiv preprint arXiv:1702.01923
Yu D, Li J (2017) Recent progresses in deep learning based acoustic models. IEEE/CAA J Autom Sin 4(3):396–409
Zeghidour N, Usunier N, Synnaeve G, Collobert R, Dupoux E (2018) End-to-end speech recognition from the raw waveform. arXiv preprint arXiv:1806.07098
Zhang M, Liu Y, Luan H, Sun M (2017) Adversarial training for unsupervised bilingual lexicon induction. In: Proceedings of the 55th annual meeting of the association for computational linguistics (Volume 1: Long Papers), pp 1959–1970
Zhou S, Zhao Y, Xu S, Xu B et al (2017) Multilingual recurrent neural networks with residual learning for low-resource speech recognition
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Kumar, A., Aggarwal, R.K. An exploration of semi-supervised and language-adversarial transfer learning using hybrid acoustic model for hindi speech recognition. J Reliable Intell Environ 8, 117–132 (2022). https://doi.org/10.1007/s40860-021-00140-7
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s40860-021-00140-7