Open AccessChinese Speech Recognition System based on Deep Learning
Author(s) -
Pengyuan Shao
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1549/2/022012
Subject(s) - speech recognition , computer science , mandarin chinese , acoustic model , transformer , test set , deep learning , artificial intelligence , set (abstract data type) , quiet , natural language processing , speech processing , linguistics , engineering , philosophy , physics , quantum mechanics , voltage , electrical engineering , programming language
This paper builds a complete Chinese speech recognition system, including acoustic model and linguistic model, which can recognize the input audio signal into Chinese characters. The system realizes the modeling of acoustic model and linguistic model in speech recognition based on deep framework, of which the acoustic model is CNN-CTC and linguistic model is transformer. The data set uses THCHS-30, which refers to 30-hour Chinese speech database of Tsinghua University. The experimental results show that the Chinese speech recognition system based on deep learning achieves 90% accuracy on the test set and has an excellent effect on Mandarin speech recognition in quiet environment.