Discriminative FilterBank selection and EEG information fusion for Brain Computer Interface

Brain computer interface (BCI) provides a direct communication pathway between a human and an external device. In this paper, we propose a new dasiadiscriminative filterbank common spatial pattern (DFBCSP)psila algorithm to select the subject-specific filters automatically during training for a motor imagery based BCI. The subject-specific filters are selected using the fisher ratio values of filtered electroencephalogram (EEG) signal. The channel dasiaC3psila alone could give sufficient information to select the discriminative filterbank for the proposed system. We have also explored the possibility of boosting the system performance by including dynamic temporal features. Fusion of static and dynamic features in the proposed DFBCSP frame work gave an average test accuracy of 92.44%, which is significantly better than conventional filterbank based common spatial pattern algorithms.

I. Introduction

EEG is one of the most common non invasive methods for obtaining brain signal. As EEG signals can be used to interpret human intentions, it provides a potential non muscular communication channel for severely disabled people. The new communication channel called Brain Computer Interface (BCI) extracts the neural information hidden in EEG signal to control external devices such as a computer, wheelchair, neuroprothesis etc. Various neurological phenomena, such as visually evoked potentials, slow cortical potentials, P300 potentials, mu and/or beta rhythms and event related (de-) synchronization (ERD/ERS) are used by various researchers to carry out BCI tasks [1]–[4]. In EEG-based BCIs, motor imagery (MI) patterns are commonly used for extracting the neural information. Its operation is based on a rhythmic power decrease or increase in counter-lateral primary sensorimotor areas during preparation for actual movement or imagination of a movement, which are called event related desynchronization (ERD) and event related synchronization (ERS) respectively [4]. The predominant frequency bands of these patterns are very much subject-dependent and it introduces a high inter-subj ect variability in MI-based BCIs.