Loading...
Single Trial Event Related Potential Extraction Using Tensor Decompositions
Taghi Beyglou, Behrad | 2019
982
Viewed
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 52561 (05)
- University: Sharif University of Technology
- Department: Electrical Engineering
- Advisor(s): Shamsollahi, Mohammad Bagher
- Abstract:
- Event related potentials (ERPs), are potentials that arise from the occurrence of an event in the electroencephalogram signals and have very small amplitude compared to the Electroencephalogram (EEG) signal. For that reason, to access ERPs, the experiment is repeated several times under similar conditions and then the are extracted by synchronized averaging, but in this way information such as Amplitude and Delay (Lag) which reflect Mental fatigue and Task habituation of subject is disappeared. Many methods for extracting the ERP components from the EEG signals have been presented as matrices. However, due to the twodimensional information (time and space) available, resource extraction is limited. Tensors are multidimensional arrays that can provide a variety of information depending on their arrangement. So, The tensor analysis can be used to extract information from EEG signals, and to extract ERPs particularly. In following thesis, four methods are proposed for tensorization and adding new information to the matrix data. Next, two new methods, namely ETucker and ETucker2, are proposed based on Tucker decomposition and spatiotemporal filtering. Evaluation of the proposed methods has been done with use of Simulation data and Real one. In the Simulation data, the proposed methods in addition to the good performance compared to the basic methods in low noise condition, showed good performance in high noise values. In real data, BCI Competition IIDataset IIb and BCI Competition IIIDataset II were used and the proposed method (ETucker) revealed an acceptable performance compared to winners of tournaments methods
- Keywords:
- Event Related Potential (ERP) ; Tensor Decomposition ; Electroencphalogram Signal ; ETrucker 2 Method ; E Trucker Method ; Single Trial Extraction
-
محتواي کتاب
- view
- فهرست اشکال
- فهرست جداول
- مقدمه
- مفاهیم پایه
- پیشینهی پژوهش
- مقدمه
- متوسطگیری سنکرون
- متوسطگیری Woody
- فیلتر زمانی-مکانی برای مؤلفههای ناهمبسته
- فیلتر زمانی-مکانی برای دو زیرمؤلفهی همبسته
- فیلتر زمانی-مکانی برای چند زیرمؤلفهی همبسته
- تخمین مؤلفه با روش RIDE
- ICAتکرارشونده (iICA)
- ICA با مرجع زمانی (ICA-R)
- تجزیهی تانسوری PARAFAC
- تجزیه تانسوری CP شیفتیافته (SCP)
- جمعبندی
- مفاهیم نظری
- رویکردهای پیشنهادی استخراج تک مؤلفه و زیرمؤلفهها
- نتایج
- جمعبندی، نتیجهگیری و پیشنهادات
- منابع و مراجع
