Spectrum Intensity Ratio and Thresholding Based SSVEP Detection

The feature extraction of Steady-State Visual Evoked Potential (SSVEP) is developed in a frequency domain regardless of the limitation in hardware architecture. We introduced a spectrum intensity ratio as a simple characterization and separation of SSVEP. However, it is difficult to classify an unseeing state of subjects. In this paper, we propose an adaptive threshold to reject the unseeing state from SSVEP in a narrow frequency band. I. RECORDING CONDITIONS The EEG for flickering visual stimuli whose frequency F is 13, 14, 15, 16, 17 and 18 Hz is recorded in the shielded dark room. The visual stimulus (LED) is located 90cm away from the subject. The subject watches a flickering stimulus over 60 seconds. We employed two bipolar channels O1-P3 for signal processing[1]. All potentials are digitally sampled at 1000Hz. The subject is a male who has a normal vision. II. DATA ANALYSIS AND RESULT A. Unseeing detection with Spectrum intensity ratio (SIR) The SSVEP detection with SIR[2] is performed as: Si(ωk) = Yi(ωk) ∑m j=−m Yi(ωk + j) , (1) Pi(ωk) = Si(ωk) + Si(2ωk), (2) Ωi = arg max ωk (Pi(ωk)). (3) where ωk corresponds to the interest frequencies F , Ωi shows the estimated F of SSVEP. The parameterm indicates the bandwidth for feature extraction. The Yi(ωk) represents the averaged amplitude spectrum for M frames. We propose adaptive threshold using average μ and standard deviation σ of SIR to detect unseeing state. The threshold Ti on ith frame is defined as: