ACTIVITY RECOGNITION ON SMART DEVICES: Dealing with diversity in the wild

Abstract

Smart consumer devices are omnipresent in our everyday lives. We find novel interesting applications of smart devices, such as mobile phones, tablets, smartwatches and smartglasses, in monitoring personal health, tracking sporting performances, identifying physical activities and obtaining navigation information among many others. These novel applications make use of a large variety of the smart devices' sensors, such as accelerometer, gyroscope and GPS. However, the usefulness of these applications relies often primarily on the abilities of interpreting noisy, and often biased, measurements from said sensors in order to extract high-level context information, such as the activity currently performed by the user.