Clasificación de señales electro-oculográficas para el control de una pinza mecánica usando máquinas de soporte vectorial y lenguaje de programación PYTHON.

Abstract

Inside the group of bio-electrical signals there is a type that is generated on the muscles related to the eyes movement, which can be used for the activation of mechanical and electronic parts that let the ability to realize actions to patients with loss of extremities, by use signal processing and classification methods on the Python programming language. The features extraction techniques provide a number of options that allow to enhance the areas of interest for each of the eye movements, the following movements will be classified: look to the right, look to the left, wink right eye, wink left eye , double wink right eye and will be parametrized to a mechanical clamp that gives the user prehensile ability, making it possible for the user to perform tasks in a personal, academic and professional environment. The techniques that showed the best results in this work are level lines, peak counting, wavelet transform, energy, normalization, feature addiction. All these techniques were applied to improve the classification percentage of the vector support machine (from now on SVM), for the training of the SVM 30 signals were recorded by movements of which 15 were used to train and 15 to classify obtaining thus a classification percentage of 97.78%.