Treballs Finals de Grau d'Enginyeria Informàtica, Facultat de Matemàtiques, Universitat de Barcelona, Any: 2017, Director: Mariella DimiccoliThis project, in the context of “La Marató de TV3”, pretends to develop a tool based on digital memories for the cognitive training of elderly people affected by mild cognitive impairment. Unfortunately, these people are highly exposed to the risk of suffering from the Alzheimer disease. The aim of the cognitive training based on digital living memories is to delay the deterioration of the memory and the cognitive functions in general. The images of social interactions, due to the emotional charge that they carry on, trigger a strong response of autobiographical memory and, consequently, are very suitable to stimulate it. For this reason, this work will develop an algorithm that, given a group of images captured by a patient with a portable camera during a long period of time (no shorter than two weeks), will determine with how many people the subject usually interacts and with who interacts the most. For each person, a group of images will be selected to be used so that the patient will be able to recognise his relatives in different contexts and so that he could go back to revive the moment and consolidate his memory. For each person who the patient use to interact, it takes several images that are being used for the patients could recognize them in different situations, could revived the moment and consolidate their thoughs. This dissertation has helped me to deepen in two aspects which are vital for the visionIi have of the world. In the first place, being able to help the elderly people, who are the reason why we are who we are. Secondly, computer vision and machine learning which are my passion.