Research Opportunities for Current Undergraduate and Master's Students
We are often looking for outstanding undergraduate and master's degree students to participate in ongoing research projects, with the goal of getting students experience with academic research and hopefully a co-authorship on a journal or conference article. I typically work with students for a semester or two as sponsor of an independent study project. We would meet weekly to discuss the project.
Master's students: I do not typically offer research assistantships for MS students. If we work together for a year and the collaboration is productive, then we can discuss the possibility. If you are contemplating getting involved in research, the first thing you should do is to begin attending my group's weekly lab meeting. Email me and I'll add you to our google group mailing list. I expect all research students to attend the lab meeting.
Opportunities for Fall 2012 include:
Training visual expertise
Many tasks performed by humans require visual expertise in knowing where to look and what information to focus on, e.g., radiologists need to be able to identify potential tumors, baggage screeners need to look at x-ray images to detect potentially dangerous objects, pilots need to look at a complex instrument panels to assess the status of their planes, FBI fingerprint matchers need to be able to compare pairs of prints. We are interested in developing techniques to train individuals more efficiently. We have a nifty bit of hardware -- a computer monitor that allows us to track the eye positions as an individual looks at the monitor. We have projects that involve collecting data from individuals as they look at a display, projects that involve modeling fixation sequences using machine learning and statistical techniques, and projects that involve modulating the contents of a display in real time to guide individuals to the relevant information in a display.
Improving human learning via artificial intelligence
We are beginning a project with a middle school in Highlands Ranch to provide a tutor that helps Spanish students better learn and retain vocabulary. The tutor is part of a research project that uses artificial intelligence techniques to determine the scheduling of study that optimizes memory retention. We will be collecting lots of data from the students as the semester progresses, and will get involved in analyzing and graphing the data. In addition, we will need back-end support to maintain and monitor the system, and to improve the web site that serves as the interface to the tutor.
Brain-computer interfaces (BCIs) are systems that provide direct means of communication between the human brain and an electronic device. This communication is most commonly achieved via noninvasive electroencephalography (EEG), the recording of electrical activity on the scalp produced by neurons firing within the brain. Measured electrical activity is passed to a computer program which "decodes" it into a binary on-or-off signal. Once the program can appropriately associate certain brain states to the 'on' signal and other states to the 'off' signal, a person wearing a BCI can, to an extent, control the signal by controlling their brain states. This ability to transmit information to a computer via intentional thought allows BCI-wearers to control, albeit at a slow pace, electronic devices.
We are looking for students interested in working on a BCI project using our recently purchased wireless 14 channel EEG. This device lends itself well to real-world BCI projects, unconstrained by a laboratory. Interested students can propose their own project or work with us on our project (details yet to be determined).
We purchased a nifty keyboard whose keys can be individually programmed to light up. We haven't yet determined a legitimate research use for this keyboard, but we're convinced there should be fun things to do with it. For example, we used a corpus of Shakespeare plays to train a machine learning model that predicts what letter or word will come next given the sequence that has been typed so far. The model then lights up keys in accordance with its predicted probability that a key will be struck. You can think of the system as a way of overcoming writer's block -- by suggesting what key should be typed next. If you follow the model's advice, it produces coherent but strange text.