Matthai Philipose

I work on the Technology for Long-Term Care project, aimed at validating human activity recognition technologies in the long-term care setting. Previously, I worked on the Human Activity Recognition project, which developed many of these technologies.

I am interested in modeling human state based on sensor inputs. I am especially interested in techniques that scale to modeling most of day-to-day life inexpensively and in useful detail. Techniques I have developed to this end include ultra-dense sensing, massive automatically extracted statistical models of daily life, rapid low-overhead data labeling and joint training of models using common sense. I am also interested in applying automated activity monitoring techniques to the problem of care giving, elder care in particular. More generally, I am interested in logic, statistical reasoning and programming languages.