ISFER home page
     

    Prototype for the Recognition of Emotions an Automatic Way Based on Pictures

    AUTHORS:
    Svatopluk Kraus

    KEYWORDS:

    ABSTRACT:
        We developed a prototype for a visualfeedback control. The central idea of this part of the system is the following:
        A picture of a human face will be captured from a video camera, processed and then the data will be given to a neural network, which will recognize a state of a human mind. We have to solve the following question:

    1. - how often do we want to capture a picture from a video camera,
    2. - how do we process a picture,
    3. - what type of a neural network do we want to use.
        For reasons of simplification we concentrate on photographs of human faces instead of video images. The photographs are classified by human expert in two categories: happy or sad.
        We concentrate ourselves on the second question. We try to deal with all the problems, which can arise in image processing, in particular we try to denote how important the conditions in a capturing of images are and how we can avoid many problems. A special problem was how to pre-process an image, so that we can use it as input data for a neural network.
        The third question is important, because mainly it limits our working capacity. There are two problems in using a neural network:
    1. - how many input neurons can we use,
    2. - how many memory do we need.
        The first one is depending on the software product, which we want to use, and the second one is a big problem for every neural network. It limits the number of image date, which we can use as input data for a neural network.
        At the end of the work we give a summary of our results. We state again that this work has an explorative character, to inventurise the problem
    -*-
    This page was created on January 19, 1998 by Anna Wojdel