Week 2: the thesis adventure continues and I am even more motivated than last week! Searching for papers myself has given me the freedom I need to fully grasp the topic of (semi-)automatic emotion detection.
First, let’s start by looking back at my first progress meeting with Karsten. Naturally, I came to the meeting with a lot of questions: what kind of prototype they expect me to develop by December, how to objectively compare different physiological sensor systems, whether the user is tied to a desktop or free to roam around in a more mobile environment, … Karsten was happy to answer all of these questions and even sent me a feedback mail with all his answers right after the meeting.
Secondly, there is of course a lot of progress to report about. I’ve managed to obtain an enormous amount of relevant papers regarding emotion detection, a few of which I have actually read or skimmed. The most important thing that I have learned up until now is that most physiological sensors are inherently noisy and that this noise is very difficult to be filtered out accurately when conducting experiments outside of a controlled environment (e.g. a significant movement of the head may cause any attached elektrodes to also shift and hence invalidate the underlying measurements). An alternative would be to detect the user’s emotions based on his or her touch screen behavior. However, since the final product will be a mobile music recommender system, not much interactivity should be required of the user (i.e. little input is available to work with). I guess this will be the next topic of discussion between me and Karsten during our meeting later today.
Finally, since Karsten recommended me to start looking at the Android development basics in Android Studio last week, I also looked at a few tutorials and managed to build my first mobile application! I must say that I am really looking forward to building my own mobile recommender app!