Automatic Speech Recognition Method Based on Deep Learning Approaches for Uzbek Language

• Published in: Sensors (Basel, Switzerland) Vol. 22; no. 10; p. 3683
• Main Authors: Mukhamadiyev, Abdinabi, Khujayarov, Ilyos, Djuraev, Oybek, Cho, Jinsoo
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI AG 12.05.2022; MDPI
• Subjects: Acoustics; Attention; Automatic speech recognition; Chinese languages; Computational linguistics; convolutional neural network; CTC-attention; Datasets; Deep Learning; Dialects; end-to-end speech recognition; Error analysis; Globalization; Humans; Japanese language; Language; Language processing; Linguistics; Markov analysis; Markov chains; Markov processes; Methods; Natural language interfaces; Neural networks; Spanish language; Speaking; Speech; Speech Perception; Speech recognition; Speech recognition software; transformers; Uzbek language; Voice recognition; Uzbekistan; United Kingdom; deep learning; transformers; hidden Markov model; Uzbek language; convolutional neural network; end-to-end speech recognition; CTC-attention
• ISSN: 1424-8220; 1424-8220
• DOI: 10.3390/s22103683

Communication has been an important aspect of human life, civilization, and globalization for thousands of years. Biometric analysis, education, security, healthcare, and smart cities are only a few examples of speech recognition applications. Most studies have mainly concentrated on English, Spanish, Japanese, or Chinese, disregarding other low-resource languages, such as Uzbek, leaving their analysis open. In this paper, we propose an End-To-End Deep Neural Network-Hidden Markov Model speech recognition model and a hybrid Connectionist Temporal Classification (CTC)-attention network for the Uzbek language and its dialects. The proposed approach reduces training time and improves speech recognition accuracy by effectively using CTC objective function in attention model training. We evaluated the linguistic and lay-native speaker performances on the Uzbek language dataset, which was collected as a part of this study. Experimental results show that the proposed model achieved a word error rate of 14.3% using 207 h of recordings as an Uzbek language training dataset.