~Mood meter reads human minds (published in ‘Daily Sun’, page 22 on 15/11/2011)
Link to article in Daily Sun:
http://www.daily-sun.com/details_ds-mood-meter-reads-human-minds_393_3_11_1_0.html
Link to print version:
http://daily-sun.com/epaper/pop_up.php?img_name=2011/11/15/newspaper/images/22_100.jpg
Full article:
MIT’s Mood Meter
-A whole new approach of measuring the general mood of a place
by Mahdin Mahboob
Scenario 1: After the A’ Levels or HSC examinations, many of us feel lost as to which universities or colleges to apply to, which subjects to apply to and which country to head to? The university websites do not help much either, most of them portraying happy and content looking students sitting in the campus lawns at summer times apparently without a care about the world. Many of us resort to university / college rankings but those are at times questionable too! Wouldn’t it be nice if there was an objective measure that captures how happy a college campus is throughout the year?
Scenario 2: When looking for various possible destinations for a family holiday trip, we feel lost by the seemingly endless number of places that one can visit. We feel confused by all the good things that we keep hearing about all the different places and with limited budget and time, it gets difficult to finally decide on one. Is it really possible to color code the entire world in terms of happiness so that at one glance we know about the happiest places instantly?
Enter the MIT Mood Meter, recently developed by scientists at MIT (Massachusetts Institute of Technology, USA), which can perhaps provide quantitative measures towards answering those questions and more. Currently working on an experimental basis, the Mood Meter has four cameras set in different parts of MIT’s campus, which continuously send mood data back to a central server to assess the overall mood at different places of MIT. The cameras look at passers-by while they walk down the hallway, detect faces and estimate the mood of the place on the ‘happiness barometer’ in a scale of 0-100 by analysing the expressions of the detected faces. A green smiley face is shown if you are smiling and if you are not, it shows a neutral yellow face (as shown in the pictures).
M. Ehsan Hoque (from Bangladesh) and Javier Hernandez (from Spain), PhD students of the Affective Computing group of the Media Lab at MIT first proposed this project in 2010 when innovative proposals were invited to celebrate the 150th anniversary of MIT.
‘‘When we first proposed this project to the director of SEMO (MIT’s Security and Emergency Management Office), I was very nervous and actually sweating, in spite of sub-zero temperatures outside! There was a very slim chance that the director of SEMO, a grumpy no-nonsense person would allow us to set up cameras around the campus. He hadn’t allocated much time to hear us so I made myself brief and to-the point,’’ says Hoque, when this writer caught up with him.
‘‘I explained our key objectives from this project:
- How friendly MIT is as a campus?
- How stressed out students get during the time of exams?
- Does warmer weather lead to happiness on campus?
- Are people from one academic department happier than others?” explains Hoque.
The idea of setting up 4 cameras in the 4 busiest locations of MIT was, understandably enough, initially faced with stiff resistance by SEMO who saw it as a threat to privacy. But the scientists did not give up, having worked on the technical aspects of the project for almost 2 months by then. Negotiations thus went on, and it was explained to SEMO that 30 images per second from the cameras would be taken and analysed in real time and using computer vision algorithms, the Mood Meter software will be able to detect the number of people in every image and determine how many of them are smiling (or not smiling). For example, if there are 10 people present at the MIT student center at a given instant, and 7 of them are smiling, the software will then record and transmit two relevant statistics (number of people – 10 and how many of them are smiling-7) to the web server. No other information (including images or personal identity) would be recorded beyond that, thus taking care of the privacy issue. Each of the cameras is to be connected with a big projector, which will display the live feed from the camera. The software will detect human faces in every image and draw a yellow blob around every face. If any face appears to be smiling, then colour of the blob will instantly turn green to indicate that their mood has changed.
After being approved by SEMO, the Mood Meter was put into action, and the rest, as they say, is history. This became an instant hit amongst students, teachers, researchers and visitors of MIT. Many people, while accessing the campus, stopped by the Mood Meter to try out making different faces to see its instant effect on the big projector. Many articles were written about it (a quick google search yields almost 1.4 million pages about the ‘MIT Mood Meter’) and the project went on to win awards as well, including the 6th bodies in space award and being nominated for the Laya and Jerome B. Wiesner Award for outstanding achievement in and contributions to the arts at MIT. Established by the Council for the Arts at MIT in 1979, the award honours the late President Emeritus Jerome Wiesner and Mrs. Wiesner for their commitment to the arts at MIT.
While most research in the field of Computer Vision is done in the lab, based on literature review and relevant simulations, the Mood Meter is one of the first to go out to the public and do the collection and analysis of data in real time. Recognizing facial expressions through highly sophisticated algorithms, the Mood Meter does not store any of the images, which makes it impossible to uniquely identify anyone from the recorded heuristic data. Having earned people’s trust in regard to privacy, it is likely to serve as a ‘Proof of Concept’ for future technologies of the sort which may find access to being incorporated in real world applications relatively easier.
Studies of the general mood of a large number of people in a small or big area using Facebook ‘status updates’ or Twitter ‘tweets’ has been done, which is heavily based on the presence/absence of highly subjective positive and negative keywords which define/suggest a person’s mood. That system is however flawed by two facts; the words that people use in social networking sites to describe how they feel may not truly reflect their appropriate state of mind and many people refrain from sharing opinions on sensitive topics since it is not anonymized. The Mood Meter, and technologies like it, which use advanced facial recognition features are a more efficient way of doing that work. The software is made to ‘learn from examples’, identifying a ‘smile’ from the geometric features of the face – eye brow raise, lip widening, wrinkle formation, etc. By using a lot of examples initially, the software is made more accurate and trust-worthy.
The idea of Mood Meter can be extrapolated outside the MIT campus and put into different other fields of research and real life applications in the future, hopes its developers Hoque and Hernandez, and the other people involved in the project.
Video Link How MIT Mood Meter works: http://goo.gl/Vaw7M
MIT Mood Meter Website: http://moodmeter.media.mit.edu/
(The writer is a Lecturer of Electronics and Telecommunication Engineering at the University of Liberal Arts Bangladesh (ULAB) and can be reached at mahdin.mahboob@gmail.com)
50.899837
-1.395606