An Optimization of Spatio-Spectral Filter Bank Design for EEG Classification
Masanao Obayashi, Takuya Geshi, Takashi Kuremoto, Shingo Mabu · Journal of Robotics Networking and Artificial Life · 2016
How to select the appropriate frequency band to classify EEG signal by motor imagery is discussed in this paper.Our proposal is an improvement of the conventional Bayesian Spatio-Spectral Filter Optimization (BSSFO).Defect of BSSFO is on the way to generate the renewal particle of the filter bank, such a random number generation.To avoid a local optimum, an evolutional update method of particles is introduced.It is shown that performance of the EEG classification ability is improved.