Citekey: @Lv2011

Lv, Q. (2011). Towards Large-Scale Mobile Sensing and Analytics. Mobile Sensing: Challenges, Opportunities and Future Directions. UbiComp. Retrieved from



The rapid adoption of smartphones and their increased sensing capabilities make it possible to collect a large amount of multi-modality mobile sensing data, which tend to be diverse, dynamic, and may vary significantly across users and applications. (p. 1)

In this paper, we identify the key characteristics and design challenges for large-scale mobile sensing and analytics, which are then demonstrated through three case studies in environmental monitoring, transportation electrification, and mobile video chat. (p. 1)

Large-scale mobile sensing and analytics is becoming a reality. On the one hand, large amounts of multi-modality mobile sensing data (e.g., GPS location, acceleration, lighting, audio, video) can be captured by individual mobile phones. On the other hand, such rich sets of mobile sensing data, together with data available from the Web or other sources (e.g., tweets, news articles, weather information), need to be carefully analyzed and processed (on the fly) in order to support knowledge discovery, user feedback, and decision making. (p. 1)


• Personalized mobile sensing and analytics (p. 1)

Personalization requires the identification (p. 1)

of a user’s specific characteristics (e.g., mobility pattern, driving behavior) and current contexts (e.g., walking, indoor, quiet). (p. 2)

Next, we consider a mobile sensing system for monitoring users’ driving behaviors and analyzing their impacts on the energy efficiency of plug-in hybrid electric vehicles (PHEVs) and the environment [2]. (p. 2)

• Collaborative sensing and analytics (p. 2)

By leveraging social groups and online social communities, socially collaborative data sensing and analytics can be particularly useful. Collaboration can occur in situ (e.g., with neighboring devices) or over the Internet via data sharing and community-based forums (e.g., average fuel efficiency of other people driving a Prius). (p. 2)

Our SafeVchat system aims to determine whether a user is misbehaving through the fusion of multiple image-based classifiers (e.g., face, nose, eye, upper body, skin color) [5]. Sampling video frames and analyzing images can be time consuming, and different classifiers may perform differently for different users’ video chat data. (p. 2)

• Privacy and security (p. 2)

Data gathered by and stored in people’s mobile phones are often highly personal and sensitive, such as geo-tagged and timestamped sensor readings or audio/video recordings. Therefore, remote processing or data mining of mobile sensing data, or collaboration among a large number of mobile devices always raise serious privacy and security concerns. (p. 2)


MAQS, a mobile sensing system for personalized indoor air quality (IAQ) monitoring [1]. IAQ is affected by different air pollutants, and can vary substantially by room, time, location, activity, etc. As people’s mobility patterns also vary, personalized IAQ monitoring and online user feedback are very useful but challenging. (p. 2)

Blog Logo

Bodong Chen



Crisscross Landscapes

Bodong Chen, University of Minnesota

Back to Home