|
Title:
|
FEATURE EXTRACTION BASED ON EMPIRICAL MODE DECOMPOSITION AND BAND POWER APPROACHES FOR MOTOR IMAGERY TASKS CLASSIFICATION |
|
Author(s):
|
Dalila Trad, Tarik Al-ani, Eric Monacelli, Stéphane Delaplace, Mohamed Jemni |
|
ISBN:
|
978-972-8939-52-6 |
|
Editors:
|
Katherine Blashki |
|
Year:
|
2011 |
|
Edition:
|
Single |
|
Keywords:
|
Brain-Machine Interface (BMI); Brain-computer interface (BCI); Electroencephalogram (EEG); Signal processing; Classification; Assistive technology. |
|
Type:
|
Full Paper |
|
First Page:
|
185 |
|
Last Page:
|
192 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
Brain-Computer interface (BCI) is a system that allows paralyzed patients to communicate with their environment. In this work we investigate a nonlinear and non stationary framework for feature extraction and classification of Electroencephalogram (EEG) signals in order to classify motor imagery tasks in BCI. The feature extraction approach is based on the empirical mode decomposition (EMD) and band power (BP). The EMD generates several sets of stationary component series, Intrinsic Mode Functions (IMFs). These IMFs are then analyzed with the power spectral density (PSD) to study the active frequency ranges corresponding to the motor imagery tasks of each subject ( and rhythms) in the EEG signals. Finally, the data were reconstructed with only the specific IMFs and then the band power (BP) is employed on the new database. Once the new feature extraction is applied, the classification of motor imagery is done using Hidden Markov Models (HMMs). The obtained results showed that the EMD method allows the extraction of the most reliable features from EEG and that the obtained classification rate is higher and better than using only the direct BP approach. Then, the recognized mental tasks may be translated into a low-level command in order to help persons with myopathic diseases or muscular dystrophy (MD) to move a joystick to a desired direction. |
|
|
|
|
|
|
