Speech Models Training Technologies Comparison Using Word Error Rate

: cc. 74 - 80
Lviv Polytechnic National University, Ukraine
Національний університет «Львівська політехніка»

The main purpose of this work is to analyze and compare several technologies used for training speech models, including traditional approaches as Hidden Markov Models (HMMs) and more recent methods as Deep Neural Networks (DNNs). The technologies have been explained and compared using word error rate metric based on the input of 1000 words by a user with 15 decibel background noise. Word error rate metric has been ex- plained and calculated. Potential replacements for com- pared technologies have been provided, including: Atten- tion-based, Generative, Sparse and Quantum-inspired models. Pros and cons of those techniques as a potential replacement have been analyzed and listed. Data analyzing tools and methods have been explained and most common datasets used for HMM and DNN technologies have been described. Real life usage examples of both methods have been provided and systems based on them have been ana- lyzed.

  1. Borovets D., Pavych T., Paramud Y., (2021). Computer System for Converting Gestures to Text and Audio Mes- sages. Advances in Cyber-Physical Systems. vol. 6, num. 2. Pp. 90—97. DOI: https://doi.org/10.23939/acps2021.02.090
  2. Emiru E. D., Li Y., Xiong S., Fesseha A., (2019). Speech recognition system based on deep neural network acous- tic modeling for low resourced language-Amharic. ICTCE '19: Proceedings of the 3rd International Confer- ence on Telecommunications and Communication Engi- neering. [Online]. Pp. 141—145. DOI: https://dl.acm.org/doi/10.1145/3369555.3369564#sec- terms
  3. Tanaka T., Masumura R., Moriya T., Oba T., Aono Y., (2019). A Joint End-to-End and DNN-HMM Hybrid Automatic Speech Recognition System with Transferring Sharable Knowledge. NTT Media Intelligence Laborato- ries, NTT Corporation. [Online]. Pp. 2210—2214. DOI: http://dx.doi.org/10.21437/Interspeech.2019-226
  4. Shanin I., (2019). Emotion Recognition based on Third- Order Circular Suprasegmental Hidden Markov Model. 2019 IEEE Jordan International Joint Conference on Electrical Engineering and Information Technology (JEEIT). [Online]. Pp. 800—805. DOI: https://doi.org/10.1109/ICASSP.2019.8683172
  5. Dutta A., Ashishkumar G., Rama Rao C. V., (2021). Performance analysis of ASR system in hybrid DNN- HMM framework using a PWL euclidean activation function. Frontiers of Computer Science. [Online]. Pp. 2095—2236. DOI: https://doi.org/10.1007/s11704-020-9419-z
  6. Wang L., Hasegawa-Johnson M., (2020). A DNN-HMM- DNN Hybrid Model for Discovering Word-Like Units from Spoken Captions and Image Regions. Proc. Inter- speech 2020. [Online]. Pp. 1456—1460. DOI: https://doi.org/10.21437/Interspeech.2020-1148
  7. Liu X., Sahidullah M., Kinnunen T., (2021). Learnable MFCCs for Speaker Verification. 2021 IEEE Interna- tional Symposium on Circuits and Systems (ISCAS). [Online]. Pp. 1456—1460. DOI:http://dx.doi.org/10.21437/Interspeech.2020-1148
  8. Delić V., Perić Z., Sečujski M., Jakovljević N., Nikolić  J., Mišković D., Simić N., Suzić S., Delić T., (2019). Speech technology progress based on new machine learn- ing paradigm. Computational Intelligence and Neurosci- ence. [Online]. Pp. 1687—1706. DOI: https://doi.org/10.1155/2019/4368036
  9. Joshi B., Kumar Sharma A., Singh Yadav N., Tiwari S., (2021). DNN based approach to classify Covid’19 using convolutional neural network and transfer learning. International Journal of Computers and Applications. [Online]. Available: https://www.tandfonline.com/doi/abs/10.1080/1206212X.2021.1983289 (Accessed 02/18/2022)
  10. Zhao Y., (2021). Research on Management Model Based on Deep Learning. Complexity. [Online]. Available: https://www.hindawi.com/journals/complexity/2021/999 7662/ (Accessed 02/18/2022)