Do you like my TV show?

Abstract

Nowadays people do several side activities while watching television. They use several devices, like smartphones, laptops and tablets, or they eat or read a magazine. It is interesting for broadcasters or television manufacturers to investigate whether people's postures and behaviour show how they feel about what is on TV. In this master thesis we will present a method for detecting people's body postures while watching TV. This system is called AffecTV. We investigated if these postures are related to specific affections and activities. We use a depth camera to detect sitting postures. To train AffecTV and to evaluate its accuracy, we have performed a pre-study and a more elaborate experiment with emotion elicitation. Results from the experiment show that AffecTV can detect certain trained activities with an average accuracy of 64.6%. The experiment conditions were not right for training affect detection.