The recent HotMobile conference in San Diego attracted dozens of top mobile technology researchers who presented papers on their latest findings. While we didn't attend the event, a couple of papers jumped out at us while browsing the agenda.
Researchers from Dartmouth College's Mobile Sensing Group and the University of Bologna presented a paper titled "WalkSafe: A Pedestrian Safety App for Mobile Phone Users Who Walk and Talk While Crossing Roads." Their research resulted in a free app called WalkSafe that's been on the Android Market since last year and that "operates at the intersection of vision, learning and mobile computing."
ANDROID APPS: 10 best new ones for fun and, uh, work
The app exploits a mobile phone's back camera to detect approaching vehicles and alert the user, via vibration or sound, of the impending danger. It also uses the phone's accelerometer sensors. The researchers used machine learning algorithms to allow the phone to detect vehicles during active calls. Studies of the app have shown it delivers few false positives, though doesn't identify all oncoming vehicles.
The researchers acknowledge that other apps exist that allow users to text or email and walk by showing the road ahead as a background image, but cite findings that even with such an app phone users are probably too distracted to attend to their own safety. Other apps have been developed to initiate communications between cars and pedestrians, but the researchers say broad implementation would be very involved and "perhaps, unlikely."
Future research includes improving vehicle detection to include bicycles, motor bikes and more, as well as improving detection of vehicles at night, since that's when so many pedestrian accidents occur. Barbara Wang, one of the Dartmouth researchers who is now at Google, says there are no plans to extend the app beyond Android anytime soon.
A fresh spin on wireless localization
Another research paper that caught our eyes was "SpinLoc: Spin Once to Know Your Location," from researchers at Duke University and the University of South Carolina. The researchers say that while location-based apps have exploded of late, most don't involve indoor localization, which could be used, for example, to orient yourself in a shopping mall or big engineering building on a college campus.
Souvik Sen from Duke describes the research as being about indoor localization using existing Wi-Fi deployments. "The technique requires no painful war driving, but requires the user to spin around once to find her location. As the user spins, we find that her body blocks the incoming WiFi signal. The dip in signal strength is the highest when her body is exactly between the access point and her phone. Thus by finding the minimum signal strength and using the corresponding phone compass value, it is possible to find the direction of the AP from her phone. If we can hear multiple APs (a very general setting), we can triangulate with these directions, and find the user's location. Preliminary work shows a localization accuracy of 5 meters, and we are hoping to reduce that in our future work."
Who knows, maybe WalkSafe and SpinLoc will need to be combined some day to keep people from twirling into oncoming shopping carts.