Optimized mother wavelet in a combination of wavelet packet with detrended fluctuation analysis for controlling a remote vehicle with imagery movement: A brain computer interface study

Abstract

Brain computer interface (BCI) is a critical field in health care to help paralyzed or maim patients back to normal life. This study is focusing on feature extraction based on self-similarity concept in electroencephalography (EEG) signal processing. To this purpose, a combination of discrete Wavelet Packet Transform (WPT) with Detrended Fluctuation Analysis (DFA) is utilized. Also, Event Related Desynchronization (ERD) patterns are used for customizing mother wavelets in the wavelet processing. Therefore, right hand movement imagination ERDs are extracted and used as a mother wavelet in the WPT algorithm and updated automatically for individual subjects. The combination of Optimized WPT with DFA (OWPT-DFA) is utilized for feature extraction for the two classes of right hand imagination and no-imagination. The features are classified and a model is trained for online processing by Soft Margin Support Vector Machine classifier and Generalized Radial basis Function (SSVM-GRBF) kernel. The model is employed to control a remote vehicle for two state of move forward and stop. In the experiment, nine subjects are participated to record data and control the remote vehicle. Results depicted that the OWPT-DFA method’s accuracy reach to 85.33% with p<0.001 and 75.23% with p<0.05 for offline and online processing, respectively. It is concluded that the self-similarity concept in the combination of OWPT and DFA methods with SSVM-GRBF classifier improve the results of movement imagination detection significantly.