Developing Adaptive Quantified-Self Applications Using DynaSense

Abstract

There are a number of user-centric applications that use data from sensors in a personal area network. The heavy dependence of such applications on sensors means that if a sensor is not available (e.g. a user forgets to carry a sensor device), some applications might not work properly or even fail. However, the data generated from a sensor that is unavailable can be derived from other devices or a combination of sensors. Since it is impractical and ineffective for application developers to track all such scenarios, user applications generally cannot take advantage of the sensor rich environment of a prospective user. This paper introduces the design of DynaSense which is a middleware system that allows user applications to be agnostic to the data sources or sensors in use. DynaSense provides a unified approach for accessing data from various data sources, which can be sensors or compositions of other data sources. The middleware dynamically decides how to acquire data from available data sources, as well as how to deliver it to requesting user applications. We present the APIs that allow user applications to easily express their needs. We also present four case studies-a heart rate monitoring application, a user behavior anomaly detection application, a calorie tracking application, and a sleep monitoring application-to compare the development of these applications with and without DynaSense. These case studies show that DynaSense can effectively reduce the efforts of developers, in terms of the lines of code written.