Please use this identifier to cite or link to this item:
|Title:||EEG (brain-wave) recognition using transformers||Authors:||Jin, Tiancheng||Keywords:||Engineering::Electrical and electronic engineering||Issue Date:||2022||Publisher:||Nanyang Technological University||Source:||Jin, T. (2022). EEG (brain-wave) recognition using transformers. Master's thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/157186||Project:||D-258-21221-03469||Abstract:||In today’s world, with the development of biotechnology and brain science, both BCI system and EEG signal become more and more popular. Also, the increasing demand in the field of treating mental illness and realizing control with mind, are also related with the EEG signals, which show the significant role of brainwave signals in reality. However, because of the properties of the EEG signals, like uncertainty and are easily disturbed, the real applications which are achieved are very few. And here, exists a problem that low accuracy in recognizing the EEG signals. Thus, leading to do such research to bring out a new way which can improve the classification performance. Through the literature review, it can be found that the most commonly used classification model for EEG signals is the CNN model. And in 2018, a brand-new model called Transformer and a new mechanism called attention are come out, which show really good performance in NLP. As the similarity between NLP and EEG signals recognition, the idea of applying transformer to classify EEG signals is brought up. And in this dissertation, both CNN model and transformer model are tried to compare the performance between these two models. Before the model training, the CWT and CSP technology have been used to extract the features from the EEG signals. The final results show that the transformer can achieve better accuracy in classifying the EEG signals compared with CNN. The transformer model really can be a potential model in the field of EEG signals classification, and can be widely used in real application in the near future.||URI:||https://hdl.handle.net/10356/157186||Fulltext Permission:||restricted||Fulltext Availability:||With Fulltext|
|Appears in Collections:||EEE Theses|
Files in This Item:
|3.87 MB||Adobe PDF||View/Open|
Updated on May 25, 2022
Updated on May 25, 2022
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.