Michigan State University Ph.D. student Xi Liu is interning this summer in HP’s Print and 3D Lab, where she’s excited to be taking on new challenges in data mining, her main area of research interest. “Some companies just tell their interns what to work on,” she notes. “But at HP Labs, you can find something that both you and your advisors care about, so it's a great opportunity to get interesting work done.” Xi grew up in the heart of the ancient Chinese city of Xi’an and attended the Xi’an Jiaotong University, where she received her BS in electrical engineering. She’s now a rising fifth year Ph.D. candidate in computer science and engineering at Michigan State. Outside of work, she likes to hike and play violin.
HP: What are you working on at HP Labs this summer?
I’m looking at data from sensors like accelerometers and RGB/D cameras (which detect depth as well as record RGB images) that track human activity. We’re trying to see if we can automatically recognize what people are doing from that data, detecting whether they are sitting or standing, for example, or moving around. I’m creating a framework that allows us to extract the features of different activities from the raw data. And then I’m writing algorithms that let us take the features and identify them as specific activities.
HP: What kind of uses could this work have?
It could help us take care of elderly people who live alone, detecting whether they have fallen in their homes, or be used for other kinds of patient care. The data that we are using comes from the British SPHERE (Sensor Platform for HEalthcare in Residential Environment) project. It’s collected from a smart house they have built – they are encouraging people to find best uses of their data for activity recognition.
HP: What’s challenging about the work?
There are two main challenges. One is the number of distinct activities that we are trying to identify, like sitting, lying, standing, running, or walking. We’re using a set of twenty activity classifications, and with data mining, the more classes you have, the harder the problem is to solve. The second challenge is that people are doing some of these activities – like moving from a sitting position to a standing one, for example – much more rarely than others. So we have much less data about them than activities like standing or sitting for extended periods, and we then have to manage that imbalance.
HP: Do you have any results yet?
Yes, we already have some preliminary results, which are promising, but there’s still a lot of room for improvement. I’m planning to keep working on getting better results until the end of my internship.
HP: What in particular has struck you about working at HP Labs?
It’s been great getting to know both the other interns and the full time researchers here. They’ve taught me a lot about engineering and really broadened my outlook. At school, we learn a lot about theory but we are less concerned with coming up with feasible solutions to real world problems. Also, at university I’m not always sure why I’m being asked to solve a problem. Here it’s much clearer why we are tackling a problem – it’s because it can change people’s lives for the better.
HP: Has being at HP Labs changed how you think about your career?
I’ve always wanted to do research but keep it connected to the real world, which would see me working in an industrial lab like this. Coming to HP Labs has really reinforced that this is the right direction for me.