|
|
2017 » Papers » Volume 3 » An analysis of Electro-oculogram signals processing using an Artificial Neural Network 1. AN ANALYSIS OF ELECTRO-OCULOGRAM SIGNALS PROCESSING USING AN ARTIFICIAL NEURAL NETWORK Authors: Nagy Robert-Bela, Popentiu -Vladicescu Florin, Vesselenyi Tiberiu Volume 3 | DOI: 10.12753/2066-026X-17-257 | Pages: 560-567 | Download PDF | Abstract
Electro-oculograms (EOGs) are signals generated by the movement of the muscles around the user's eyes when the eyes are moving. These signals can be registered by Ag/AgCl electrodes adhered to the user's skin and coupled to a data acquisition device (usually an analog-to-digital converter) and used in different control applications by perfectly healthy or even by disabled people. In this application, we recorded the EOG signals with a 24-bit, 4 channel, 51.2 kS/s per channel sampling rate analog-to-digital converter (ADC), from which we used only 3 input channels, analyzed these signals generated by intentional horizontal and/or vertical movements and also blinking by recording EOG signals from the test subject's eye. The signals for left, right, up and down movements were acquired, filtered and normalized, after which an Artificial Neural Network (ANN), using supervised learning, was trained to classify these recorded signals. MATLAB's ANN was used and in this application we analyzed the obtained results from the classifier and also the possibility to apply this type of signal classification method in order to use it for a portable device which would be used to help disabled people (who still are able to move their eyes) to communicate simple instructions to a computer system or a robotic device. This system can be used also by healthy users, providing them a new way of communication and/or control of different electronic devices, other than the existing usual human communication and/or control channels. The reasons to use EOG signals over EEG signals is that the latter are very prone to electrical noises and also the EEG electrodes are more numerous, is harder to mount them correctly and can be uncomfortable to the user. | Keywords
electro-oculogram, signal processing, artificial neural network, supervised learning |
|
|
|