Classification of non-invasive recording of electroencephalography brain signals using hoeffding tree

Abstract EN

There is a considerable advancement in research that concern brain-computer interfaces (BCI).

BCI can be defined as a communication system that is developed for allowing individuals experiencing complete paralysis sending commands or messages with no need to send them via normal output pathways of brain.

EEG recording are affected by cardiac noise, blinks, eye movement, in addition to non-biological sources (such as power-line noise).there will be an obstacle if the subject generates an artifact since will violate the specification of BCI as a non-muscular communication channel and the ability of subjects suffering degenerative diseases could be lost and this artifacts(noise) leads to incorrect classification accuracy .the presented study has the aim of being a sufficient reference in BCI system and also emphasize algorithms which are capable of separating and removing the noise that interferes with the task-related electroencephalography (EEG) signal for the best features .

the task is the motions of the index finger of right or left .the separation process based BSS technique ,this separating would be having an effective speeding impact on classifying patterns of EEG.

and classified using classifier ( Hoeffding Tree).

the proposed algorithm is tested and trained with the use of real recorded signals of EEG .

experiments reveal that the proposed classifier with the stone algorithm leads to high classification results up to the classification accuracy 79%.