Project Hawaii Applications (supported by Microsoft Research)

The application collects ECG and EKG data from our inhouse nanosensors sensors over Bluetooth. The application works in two modalities. It can display realtime ECG data on the mobile phone screen and it can relay data to our backend server. The data sent over to the backend is geo-tagged. The data at the backend server is stored in a mysql database. The data and the location of the patient is displayed on a web based portal. Although the present prototype does not support it, we are in the process of designing machine learning services that can detect abnormalities in the EKG data. The goal is to automatically inform doctors through a text message when the patient experiences heart problems (Student: Prashanth Shyamkumar, William Wilkins).

We are developing a zombie based game where users can infect each other when they are physically proximate. Every user of the application uploads his latitude and longitude to a central database and a script resident on the server finds users that are close to each other. The webserver resident script notifies users devices of the phones that are close to it. While the front-end of the application is a simple game, we will use the location data and how users infect each other, to understand the semantics underlying spread of infectitious diseases among humans (Student: Christian Williams).

MobiSafe is an application that uses the Microsoft Hawaii GPS Wrapper, Skype4COM API, and a neural network service running on a server. MobiSafe is used to detect whether a driver is in a “danger zone”. Every driver offloads his location and speed to the server and the neural network engine and uses other input parameters (such as the number of past accidents around that location and the number of red lights around that location) to decide whether the driver is in a danger zone. Then the application intercepts an incoming call and reroutes it back to the caller, notifying that the driver is in a “danger zone”. By using MobiSafe, drivers are safe from being distracted by incoming calls. Also, it lets the caller know that the driver is driving and in a “danger zone” (Student: Tri Nyugen).

The goal of this application is to minimize the energy consumption of localization using a GPS unit. The project uses an accelerometer to augment a GPS unit on a Windows Mobile phone. We are developing a HMM based algorithm to convert raw and noisy accelerometer readings into meaningful distance measures. Using an adaptive algorithm, our system will determine when to switch on the GPS unit such that accuracy is within acceptable limits (as determined by the application)--- when the GPS unit is off, location is inferred from the accelerometer (Student: Haibo Zhang).

In this application, we use geo-tagged images taken using the windows mobile phone to generate 3D models outdoors and indoors. The mobile phones takes GPS annotated images outdoors, and accelerometer annotated images indoors and transfers it to back-end server. The backend server runs a clustering algorithm, an image stitching, and consequently a 3D model generator to generate 3D models of indoors and outdoors. Although not supported by the application yet, these models can be used to augment street views of Google Maps. The major advantage of the system is that it provides temporal consistency--the models are updated frequently based on how often pics are uploaded to the server. (Student: Anirudh Ladha).

The goal of this application is to collect data on important contexual information that can be used to build a personalized search engine for mobile phones. Personalized search refers to building user specific models and augmenting search queries with additional information specific to the user. Such a search engine on the phone can help minimize the number of unnecessary results retrieved from the backend search service like Bing or Google. Our application profiles data on web browsing history, search history, location, to-do and calendar items. We plan to use the collected data to study the efficacy of personalized search. (Student: Chris Gaetely)