Akademik Məlumat İdarəetmə Sistemi
7th International Symposium on Computational Intelligence and Industrial Applications, ISCIIA 2016, Beijing, China, 3 - 06 November 2016, (Full Text)
Authors measured electroencephalograms (EEGs) as subjects recognized and recalled ten types of images of robot (PLEN.D, dmm.com) movement presented on a CRT monitor. During the experiment, electrodes were fixed on the scalps of the subjects. Four EEG channels allocated on the right frontal and temporal cortices (Fp2, F4, C4 and F8 according to the international 10-20 system) were used in the discrimination. The authors analyzed a single trial EEGs of the subjects precisely after the latency at 400ms, and determined effective sampling latencies for the discriminant analysis to ten types of images. Sampling data 1 was collected at latencies from 400ms to 900ms at 25ms intervals for each trial. Sampling data 2 was collected from 399ms to 899ms at the same interval and sampling data 3 was collected from 398ms to 898ms at the same interval. Thus, data was an 84 dimensional vector (21 time point x 4 channels). The number of external criteria was 10 (the number of different movement), and the number of explanatory variables was thus 84. The canonical discriminant analysis was applied to those tripled single trial EEGs. Results from the canonical discriminant analysis were obtained using the jackknife method. And discrimination ratio was 100% for each of three subjects. The discriminant results were transmitted to the robot PLEN.D by the blue tooth. We could control the robot with ten commands by single trial EEGs of the subjects who only recalled corresponding robot images.