Electrooculugram based Communication System for People with Locked-in-Syndrome

Locked-in-Syndrome death rate is predicted to be 60% which is quite tragic. Eye movements, blinking, and consciousness of the patient is present. In this study, a method was proposed to process the signals of only vertical eye movement obtained from EOG for twenty-six alphabets in the English language. Using EOG makes this method non-invasive and data can be processed in real-time. The data consisted of 20-30 signals for each alphabet and were collected by BIOPAC System with sampling frequency as 1000 Hz. Empirical mode decomposition was used for segmenting the signal and removing the noise from the signal. Features selection in detail was done, in which Derivatives of indices of peaks, Mean Cycle, and the maximum number of peaks (C-max) showed the best results. The data with selected features were fed into classifier Support Vector Machine-Quadratic (SVM-Q). The highest result obtained was 87.6% classifying only the vertical eye movement for different alphabets. This work can be used in the Brain-Computer Interface system to support and assist LIS patients in communication.