Activity-aware Sensor Data Processing

There is a lot of research on mobility behavior analysis of moving objects. Such research has traditionally focused on objects such as vehicles, ships etc., and on merely understanding their “movement” patterns. With the advent of people-sensing as an emerging and active area of research, the focus has shifted on “sensing” the everyday activities of an individual using the mobile phone. The primary objective is to use the sensors on the phone (e.g. GPS, accelerometer) and sense different aspects of the environment, effectively bringing people in the sensing loop.

In this project, we will explore the development of algorithms (and an adaptive system), on a mobile phone platform, to efficiently process and control phone sensors for the purposes of activity recognition. The high-level objective is to design energy efficient stream processing and sensor control techniques for on-board acquisition and processing of phone sensor data and progress towards building an adaptive sensing framework exploiting the techniques.

The students will benefit from experience in learning about machine learning algorithms (supervised), time series data processing, system development on an appropriate mobile phone platform (e.g. android).

Interested students are kindly asked to contact Zhixian Yan.

Requirements: C/Java, SQL             Optional: Matlab, Machine learning tools (libsvm, weka)