Analysis of EEG for Classification Vowel Sounds

摘要

Electroencephalogram is one the common method to acquire EEG data for any purposes like Brain Computer Interfaces. Many people are suffering from disabilities like locked in syndrome or Amyotrophic Lateral Sclerosis and which impair normal communication. Here provide a control scheme method for brain computer interfaces using vowel speech imaginary. Three healthy subjects were selected and EEG were recorded during the performance of three tasks, imaginary speech of the English vowels /a/, /u/ and no action state as control. To enhance the classification performance, wavelet transform were used and Linear Discriminant Analysis (LDA), Support Vector Machine (SVM) and k-Nearest Neighbor (KNN) were used as classifier. Features namely mean, standard deviation, band power, entropy and skewness are extracted from each EEG data. The performance of classifier was evaluated by measuring sensitivity, specificity and accuracy. The overall accuracies ranged from (72-80) %. Results indicate significant potential for the use of vowel speech imagery as a speech prosthesis controller.