The first topic that I have dedicated my time is the sensing of residential water usage. Namely, using a single pressure sensor attached to any point on the water supply line of a residential home, it is possible to infer all water activities throughout the home. Water will be one of the most pressing issues on the global and national agenda over the next 20 years (World Water Forum, 2008). The United Nations predicts that by 2025, more than 2.8 billion people living in 48 countries will face water shortages. The United States is not immune to these concerns: according to US government estimates, 36 states will face serious water shortages in the next five years (EPA, 2008).
In addition to impacting sustainability, the sensing under naturalistic conditions enables the use of stochastic signal processing techniques, such as Markov Random Fields and Bayesian Networks. The goal being to model not only water flow and fixture classification, but also modeling sophisticated water trends and activities of the consumer. This promises to further the state of the art and awareness of discriminative learning in graphical models and efficient optimization over sparse datasets (as water usage graphs tend to be dense during certain times of the day, followed by long periods of sparseness). This type of stochastic implementation, because it can be used for activity recognition, lends itself well to elder care applications.
This is an ongoing collaborative project. Team members include Jon Froehlich and myself (project leaders), Elliot Saba, Tim Campbell, Gabe Cohn, and Conor Haggerty - and our advising professor Shwetak Patel. Professors James Fogarty and Les Atlas are also active in the collaboration.
As thermal imaging becomes less expensive, it comes within reach of a number of new applications such as human computer interaction. We are exploring ways to leverage thermal imaging for sensing touches and gestures in interactive projection systems (i.e., using an overhead projector to display content onto an arbitrary surface). At its core, this technology has the potential to turn an arbitrary surface into a multi-touch, multi-user, pressure aware system. Example interactions include (1) distinguishing hovering above a surface from touch events, (2) shape-based gestures similar to ink strokes, (3) pressure based gestures, and (4) multi-finger gestures.
I am currently developing a method of analyzing radiant sound from the lips via a microphone as a way of measuring airflow and flow-volume from the lungs. The goal here is to develop signal processing techniques for analyzing the lip reverberation capable of monitoring pulmonary ailments such as asthma, chronic obstructive pulmonary disease, and cystic fibrosis.
One endemic problem in current lung monitoring systems is patient usage compliance. For example, one 1998 study by the Cleveland VA hospital reported only 52% patient compliance. In 2000, the British Thoracic Society reported compliance even lower, <50%. This is largely due to system portability and cost (as identified by the same article). As such, our target monitoring platform is the mobile phone—an increasingly ubiquitous device that has substantial computing power. In addition, because of wireless internet capabilities, our system can upload results and trends directly to the healthcare professionals, replacing expensive visits to the pulmonologist. Finally, because children are often most affected by pulmonary ailments such as asthma, we are investigating methods to make the lung testing procedure part of a simple, yet engaging, game on the phone. This should further increase compliance and reduce perceptions of burden.
This project is part of a larger mobile healthcare tool known as MDPhone. In the future, this mobile spirometer will be part of the open source suite currently being developed by Waylon Brunnette at the University of Washington.
The most apparent distortion (MAD): This was the subject of my masters thesis in regard to subjective image quality prediction. This algorithm, to date, is the single best performing image quality assesment available. It appeared in the Journal of Electronic Imaging January-March 2010 issue as the cover article. The performance of MAD has been assesed on a variety of different databases including LIVE, CSIQ, TID, and Toyama. See additional information at vision.okstate.edu and http://vision.okstate.edu/mad/. Also a part of my masters thesis was the creation of a subjective image quality database, Categorical Subjective Image Quality or CSIQ, which contains over 5000 subjective ratings over 30 different original images and their distorted versions. This database is freely available to the research community: CSIQ Database.
The beta release of MAD for Matlab is available here: [Mad_beta_release] Updated: October 2011
The zip file contains the mad_index.m and .mex files for a windows computer pre-compiled. The raw .c files are also included should you need to compile your own .mex files for linux or mac.