No abstract

No abstract

No abstract

This paper demonstrate the combined approach of Blind Source Separation (BSS) and canonical correlation analysis (CCA) to detect the frequency component of Steady state Visual evoked Potential (SSVEP) based Brain computer Interface (BCI) system from non-invasive recorded electroencephalography (EEG) signal. Detection of SSVEP frequency component with great accuracy is most challenging and difficult task for the development of SSVEP based BCI system. Canonical correlation analysis (CCA) is the most widely and rigorously employed method to detect the SSVEP frequency component from multichannel recorded EEG signal. But the presence of spontaneous EEG signal and artifacts that often occurs during the recording of scalp-based EEG signal may deteriorate the detection accuracy of SSVEP frequency component from recorded EEG signal. This work investigates the blind source separation (BSS) as preprocessing technique to decorrelate the source signal (SSVEP) from the recorded, mixed-signal (EEG) to improve the detection accuracy of SSVEP based BCI Inference system. In this paper, the author proposes second-order AMUSE-based BSS algorithms as preprocessing methods for multichannel EEG signals. The CCA technique employs the preprocessed signal to detect the SSVEP frequency components from the recorded EEG signal. The obtained finding indicates that the proposed BSS-CCA method significantly improved the SSVEP detection accuracy as compared to the standard CCA method. In addition to this, the author has also reported that selection of stimulus frequency also plays a vital role in order to improve the detection accuracy of SSVEP BCI system. The obtained result indicate that average detection is much more improved using when stimulus frequency is the range of alpha band (8Hz – 16Hz) as compared to stimulus frequency beyond alpha band (above 16Hz) using both CCA and BSS-CCA approach.

No abstract