Open AccessSpeech error recognition based on broad learning
Author(s) -
Yonghong Yan,
Liwen Wang,
Guo Wei
Publication year - 2021
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 - 10.1088/1742-6596/2031/1/012030
Subject(s) - morlet wavelet , computer science , speech recognition , feature (linguistics) , sentence , mel frequency cepstrum , artificial intelligence , wavelet , pattern recognition (psychology) , popularity , signal (programming language) , wavelet transform , feature extraction , discrete wavelet transform , linguistics , psychology , philosophy , programming language , social psychology
In recent years, with the increasing popularity of Chinese in the world, more foreigners have begun to learn Chinese. In this regard, this paper studies the phonetic errors in a single sentence, and designs a phonetic error classification model based on the broad learning network. First, the Laplace wavelet convolutional layer is designed as the first layer of the traditional CNN, and the Morlet-CNN model is used to extract the speech signal. Then, the extracted speech signal is input into broad learning network for training. Finally, the speech with different error types is tested. By comparing with the experimental results of a single speech amplitude feature and MFCC feature, the classification accuracy of the proposed model has been significantly improved, which verifies the effectiveness of the model.