Electrooculogram Signal Identification For Elderly Disabled Using Elman Network

This study proposes an algorithm for classifying the different eye movements gathered from 20 subjects through Electrooculogram (EOG) by means of neural networks. Reference Power and Plancherel theorem were used to excavate outstanding attributes from raw signals to recognize the eye movement tasks performed by individuals. Elman Recurrent Neural Network (ERNN) was projected for categorization of filtered EOG signals. Classification performance of Reference Power using the ERNN was observed to be better with a range of 90.28 to 91.95% in comparison with Plancherel features. Recognition performance was identified using single trial analysis. The single trial analysis shows that subject S20 and S14 has the maximum recognition and identification rate compared to the other subjects involved in this study for reference power and plancherel theorem, and also this study proves that even though the classification accuracy was the maximum the average recognition rate was minimum for some of the subjects at the same time minimum classification accuracy subject shows the maximum recognition and identification rate for some subjects. Result depicted that average recognition and identification rate was maximum for plancherel features compared to reference features using ERNN and also single trial analysis evident that depends upon the subject performance and involvement the identification and recognition rate was vary for all the subjects.