Today's cities are a complex system of interconnected elements such as of schools, shopping malls as well as our homes. Our lifestyle defines their temporal use, when and how we move, reside or interact with this urban system. The man-made urban layout has been adapted over thousands of years to our demands and make our lives easier, and by this, creates spatial-temporal constraints to use urban elements.
Within this topic we research methods to model human activity and behavior from social media and mobility.
As smartphones essentially enable any time use of social media platforms, relevant properties emerge for collecting evidence about the user’s activity. First, content is shared in real-time and focuses on experiences that “happen right now”. Second, “daily chatters” share content multiple times a day. Third, and most importantly, it has been shown that the majority of users focus on themselves, rather than on, for example, sharing plain information or opinions. Moreover, social media usage is widespread geographically and has become a natural part of people’s daily lives, much of it taking place on smartphones. As a consequence, an abundance of data revealing a user’s activities is generated implicitly by the user. Through social media platforms that record such data, we can obtain rich signals for activity recognition without any extra instrumentation.
However, since users are not systematically submitting their life activities onto social platforms, the data content we access is unstructured, ambiguous or can contain manifold activities. This opens the question as to how to define a common scheme for activity. We make use of a standardized activity taxonomy from the American Time-Use Survey (ATUS) and classify tweets into predefined activities. Training data is obtained by an crowdsourced labelling task using mechanical turk. A resulting model takes unstructured tweet data, and optional meta data such as the venue type and tweet time, to classify the data into the most likely activity.
Probing an entire population, we can then create likelihood maps for activities to be performed at different locations of an urban area. With such insights, we can then conduct real-time recognition of people's activities given geo-referenced trajectory data.
Large scale events with tens of thousands or millions of visitors contain an immanent risk for stampedes as they occurred, for instance, at the Love Parade in Duisburg in 2010 or at the Hindu-Fest Maha Kumbh Mela in India in 2013. A constrained space and the movement of the crowd can develop critical dynamics leading to such incidents. A careful design of the event area and planning of the event schedule are crucial factors to influence the crowd dynamics in order to minimise this risk.
At the Wearable Computing Lab, we developed an app that collects location information of individual visitors. Aggregating the observation of individuals enables behavior, emergent behavior and crowd dynamics can be obtained.
During the Züri Fäscht the app was downloaded more than 56.000 times. Offering event information, attractive features such as a friend friender and a trophy collection game.
We collected GPS positions from 29.000 users during the event leading to 24M GPS datapoints.
Based on the collected data, crowd densities have been esimtated and shown in real time to police authorities. Furthermore parameters such as crowd speeds and velocities have been calculated.
To increase potential App users, we developed a communication strategy that resulted in significant traction on "multimodal" media outlets. Opening this channel to potential users, it helped us most to boost AppStore downloads. A selection of numerous press releases is given below.
Contact Ulf Blanke
Diese Website wird in älteren Versionen von Netscape ohne graphische Elemente dargestellt. Die Funktionalität der Website ist aber trotzdem gewährleistet. Wenn Sie diese Website regelmässig benutzen, empfehlen wir Ihnen, auf Ihrem Computer einen aktuellen Browser zu installieren. Weitere Informationen finden Sie auf
The content in this site is accessible to any browser or Internet device, however, some graphics will display correctly only in the newer versions of Netscape. To get the most out of our site we suggest you upgrade to a newer browser.