Modelling on Human Intelligence a Machine Learning System

Recently, a huge set of systems, devoted to emotions recognition has been built, especially due to its application in many work domains, with the aims to understand human behaviour and to embody this knowledge into human-computer interaction or human-robot interaction. The recognition of human expressions is a very complex problem for artificial systems, caused by the extreme elusiveness of the phenomenon that, starting from six basic emotions, creates a series of intermediate variations, difficult to recognize by an artificial system. To overcome these difficulties, and expand artificial knowledge, a Machine Learning (ML) system has been designed with the specific aim to develop a recognition system modelled on human cognitive functions. Cohn-Kanade database images was used as data set. After training the ML, it was tested on a representative sample of unstructured data. The aim is to make computational algorithms more and more efficient in recognizing emotional expressions in the faces of human subjects.