Análise comparativa de redes neurais artificiais e modelos escondidos de Markov para aplicações de reconhecimento de gestos em superfícies.

Abstract

Many human activities are tactile. Recognizing how a person touches an object or a surface that surrounds them daily is an active area of research and has generated a strong interest within the interactive surfaces community. In this thesis, we compare two machine learning techniques, namely Artificial Neural Network (ANN) and Hidden Markov Models (HMM), as they are some of the most common techniques with low computational cost used to classify an acoustic-based input that relies on the unique sound produced when a fingernail is dragged over a surface. We employ a small and low cost microphone that could be easily incorporated into a surface on which it rests to be applied as a passive gesture recognition input. Our contribution is to analyze the advantages and limitations of these techniques in the context of gesture recognition using a simple alphabet of three geometrical figures: circle, square and triangle. To do so, we use Matlab’s toolboxes to implement the models and evaluate the dataset used to train the ANN and the HMM.