A System on Chip Design of Online Recursive ICA Based Real-time Multi-channel EEG Acquisition System with Automatic Eye Blink Artifacts Rejection

• Main Authors: Liao, Jui-Chieh, 廖瑞傑
• Other Authors: Fang, Wai-Chi
• Format: Others
• Language: en_US
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/49758965093284384363

碩士 === 國立交通大學 === 電子工程學系 電子研究所 === 101 === This thesis presents a system on chip design of online recursive ICA based real-time multi-channel EEG acquisition system with automatic eye blink artifacts rejection. EEG signal is one of the feeblest physiological electrical signals. It is easily contaminated by artifacts caused by noncerebral electrical activity. Previously, ICA was used to extract artifacts from a time period of EEG data. After processing of ICA, automatic artifact detection and elimination were performed. Then, artifact free EEG signals can be reconstructed. Recently, brain computer interfaces (BCIs) are developed to control machines through EEG directly. In order to enhance the feasibility, reliability, and accuracy of BCIs, EEG signals used for BCI applications should be acquired from human without artifacts in real-time. For the real-time requirement, online recursive ICA (ORICA) is adopted for real-time artifacts extraction because it can immediately find the ICA result right after each EEG sample. There are two kinds of artifacts. The one which is caused from the inside of the human body is called as biological artifacts. The other one which is caused from outside of the human body is named as environment artifacts. Since the eyes of human are very close to brain, eye blink artifact is one of the most harmful artifact to EEG signals. Therefore, in this work we focus on automatic eye blink artifact elimination and the algorithm used for eye blink artifact detection is sample entropy. In order to fully take advantage of ORICA, the real-time processing flow is proposed to automatically remove the eye blink artifact without detection misses in real-time. The system with these algorithms and the proposed real-time processing flow are implemented on a chip using TSMC 90nm CMOS technology. Since the good hardware sharing arrangement, the core size, which is 1200 × 1200 μm2, is lower than previous work even though containing additional eye blink artifact rejection. With the proposed real-time processing flow, artifact free EEG signals are acquired in 0.2638 s after each EEG sample. The performance of eye blink artifact elimination is evaluated through correlation coefficient between original artifact free EEG signals and processed artifact free EEG signal which is 0.9135 on average. The processed results with real EEG signals are also provided and shown to remove eye blink artifacts exactly.