Electrooculogram based detection of visual memory recall process

Detection of visual memory recall finds applications in context aware ubiquitous computing systems capable of assisting people with memory oblivion. The present work is aimed at identification of visual memory recall of human beings from the analysis of their eye movements through Electrooculogram signals. These signals are represented through Adaptive Autoregressive Parameters, Power Spectral Density, Hjorth Parameters and Wavelet Coefficients as signal features. Classification of the obtained feature spaces is carried out using Support Vector Machine with Radial Basis Function Kernel, K-Nearest Neighbour and Naïve Bayes classifiers to distinctly identify previously seen and new images from a series of images presented as visual stimuli. Performance of classification is evaluated in terms of classification accuracy, sensitivity and specificity. A maximum accuracy of 89.50% is obtained on an average over ten participating subjects using SVM-RBF classifier on a combined feature space comprising all four signal features.