Real-time classification and recognition of EEG/EMG signals for BCI
Development in Brain-Computer Interfaces (BCI) has evolved tremendously in recent years due to the improvement in EEG techniques and the improvement of EEG-capturing technology. The availability of low-cost boards such as the OpenBCI Ganglion vastly increases the amount of people who can develop new...
Saved in:
主要作者: | |
---|---|
其他作者: | |
格式: | Final Year Project |
語言: | English |
出版: |
2017
|
主題: | |
在線閱讀: | http://hdl.handle.net/10356/72792 |
標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
總結: | Development in Brain-Computer Interfaces (BCI) has evolved tremendously in recent years due to the improvement in EEG techniques and the improvement of EEG-capturing technology. The availability of low-cost boards such as the OpenBCI Ganglion vastly increases the amount of people who can develop new systems based on BCI technologies, as well as the people who benefit from such technologies – such as the physically impaired.
The aim of this project is to develop a portable, flexible and practical system to distinguish between various EEG and EMG signals in real-time, with reasonably high accuracies. The input signals chosen for the system include motor imagery, physical motor movements, eye blinks, and jaw clenches.
A Python GUI application was created to perform various functionalities required by the system, such as: data capture, loading and saving of data, the processing and feature extraction of the signal, prediction of input, and a maze game to demonstrate the prediction outputs. The data is processed via baseline removal, Fast Fourier Transform (FFT), and signal power of frequency bands. The machine learning algorithm used is XGBoost, a tree ensemble algorithm.
The performance of the system using the various signal processing operations were discussed, and the results have been presented to show the system’s accuracy. |
---|