Using the Barometer to Provide Better Context Awareness

Many of the most popular applications today on Android are using context awareness of what the user is currently doing to improve their user experience. For example, Google Now uses GPS to determine when a user is away from home, and provides the time that it would take to get back home. One of the most recent apps that takes advantage of knowing what the user is doing is Cover ( https://play.google.com/store/apps/details?id=com.coverscreen.cover&hl=en ), which displays different apps on the lock screen depending on if the user is at home, in the car, or somewhere else. Being able to understand how a user is transporting themselves in their commute is vital to being able to improve context awareness. For example, a user may want to use Waze to report traffic when they are driving, but this app is much less useful when walking.

 

Most often, to provide context awareness of user’s transportation, the accelerometer is used. The accelerometer is a relatively low-power sensor that is present on nearly every Android device, from phones, to tablets, to even Android Wear devices. However, in order to understand what you are currently doing, the algorithm must also know the phone’s orientation and position, and it must sample the accelerometer’s values nearly ten times per second. This is a problem that is both difficult to program and also requires more power consumption. However, Sankaran et al. have found a more efficient way to determine context awareness using a sensor that has recently been introduced to some devices: the barometer. The barometer, which is a sensor that measures air pressure, is being included in more and more Android devices as an altimeter to assist the GPS and to help determine elevation change.

 

Barometers solve many of the programmatic problems and power consumption problems brought by the accelerometer. Barometers are both position independent and orientation independent, meaning that it can avoid many of the instances when a false positive is issued when a user moves around the phone to use it. Additionally, the authors found that barometers, which only need to be queried every second, only result in a 2% increase over base power consumption, while relying on an accelerometer sampling at 10 times per second resulted in a 67% increase over base power consumption. This means that if the barometers can be used in place of accelerometers, overall power consumption can drop. However, the barometer is not without its problems. As barometers measure air pressure, they can also detect changing weather patterns. Although the authors did not find that this impacted their results significantly as the change in air pressure is slow, this is a problem that the accelerometer does not face.

 

In order to use the barometer to detect transportation context, the authors used the profiles of three different states (idle, walking, and vehicle) to determine how air pressure changes when a user is in each one. The first thing that the authors use is the fact that roads are not perfectly flat, even when they appear to be. This allows the program to easily see whether or not the user is walking or in a vehicle, by seeing whether or not there are peaks and troughs in the barometer output over time. When driving, these peaks and troughs are closer together, and when walking, these peaks still exist, but are farther apart. When idle, these peaks are very small and sometimes do not even exist.

 

To determine accuracy of their method, the authors compared their method to both the Google Activity Recognition, which is tied into Google Now and relies on only the accelerometer, and the Future Urban Mobility Survey (FMS) app, which is a GPS and accelerometer based solution that was developed by the Singapore-MIT Alliance for Research. The authors found that in a case study in Singapore, the barometer was 69% accurate overall, with its largest deficiency in detection of walking. This was better than both FMS’s 68% accuracy and Google’s 56% accuracy. Most importantly, using the barometer required only 3% more power consumption than the base, while using the Accelerometer with the Google solution required 34% over base power.

 

Overall, it seems that barometers, when available, may be able to be a better replacement for accelerometers in context awareness. These sensors have use far beyond elevation and weather detection, and may be used in more applications in the future.