An Overview

Mobile wireless devices, such as cell phones, PDA's, music players, notebook computers, are becoming increasingly powerful and sophisticated. Not only are their computing power and communication abilities expanding, but they now routinely incorporate a variety of sensors, including accelerometers, microphones, cameras, light sensors, and proximity sensors. Millions of people now carry these sensor-enabled mobile devices, which can either transmit the sensor data wirelessly to a server for remote processing or can process the data locally. This situation provides dramatic new opportunities for data mining and also makes it possible to deploy intelligent data mining applications on a scale not previously imaginable. This is where our WISDM (WIreless Sensor Data Mining) Lab comes in.

In this lab we collect and mine sensor data from these wireless devices. The scope of this project is enormous and there will be many opportunities for us to make an impact on how these smart mobile devices are used. Our initial focus was mainly on mining the accelerometer data, for tasks such as biometric identification and activity recognition (summarized below), but while we are continuing on in that area, we are now also actively looking into mining other sensor data, such as GPS data. Our initial platform will be based on the phones running Google's Android operating system. Our preliminary work, which was based on a research oriented mobile device from Sun Microsystems, showed that we could accurately and reliably:

• Identify a user or verify the identity of a user based on his/her accelerometer data (behavioral biometric identification)
• Identify the physical task that a user is performing (e.g., walking, jogging) based on the accelerometer data (activity recognition)

This work overlaps with a related project that is jointly being conducted with Dr. Warren Tryon from the Psychology department. While the WISDM project is quite general and spans all potential data mining applications of all forms of wireless sensor data, the project with Dr. Tryon focuses mainly on the accelerometer data and health care applications, leveraging Dr. Tryon's work on actigraphy.