Recognition of Unspoken Words Using Electrode Electroencephalograhic Signals

摘要

In this paper, the use of electroencephalograhic signals in recognizing unspoken speech is investigated. The aim of the work is to recognize two words, namely, Yes and No by using a single electrode electroencephalograhic device. Yes and No were selected based on the fact that they are the basic answers of any disabled person. The single electrode device was chosen because of its ease of setup, adjustment, and light weight; hence, most suitable also for disabled. Several neural networks were trained with 7 subjects. Online and offline testing were carried out on male and female subjects in semi quiet environment. Average recognition results reached were 57% for online testing and 59% offline, whereas maximum value for offline was 68% and online 90%. The novelty of the work is using a single electrode device, as all previous work was done on multi-electrode devices.