Skip to Navigation Skip to Page Content

Graduate students work on fall detection system

Master’s student Daniel Nabelek (left) and doctoral candidate Yun Li show a device Li conceived to use sound to detect falls. Li has worked on sound analysis since 2008 when he was working as a research assistant while earning his master’s degree.

Sound has always been important to Yun Li. His parents are both musicians in China, and he plays the piano in his rare free time to relax.

Sound also plays a key role in his research interests.

Li, a doctoral student in electrical engineering, has been working on analyzing sounds to detect one very specific type of noise — the sound of someone falling to the ground. He has been working on this topic since 2008, when he worked as a research assistant while getting his master’s from the University of Missouri.

Most recently, he worked with Daniel Nabelek, who is pursuing a master’s in electrical engineering, to create a circular microphone array to detect fall sounds.

“If older people fall on the ground my sensor can detect that and send an alarm to get help,” Li said. “The goal is to increase the probability of detecting falls while disregarding other noises.”

This research, published in IEEE Transactions on Biomedical Engineering, used an array of eight microphones arranged in a circle. This positioning makes it easier to use algorithms to determine the source of the noise more easily. Testing the array with stunt actors simulating falls and other non-fall sounds produced the best results of any sound-detection system Li has used yet.

“For this particular case of fall detection — I had some success in recovering the fall sound,” Li said. “We need to minimize the number of false alarms but at the same time we don’t want to miss any.”

All of the falls were identified but the system also misidentified three percent of non-fall sounds. To increase the accuracy, useful information from other sensors could be fused with the sounds.

The leading cause of emergency room visits for older adults, about 30 percent of people over 65 have accidental falls each year. If an elderly person living alone falls and does not receive assistance quickly, the success of treatment decreases.

The aging population in the United States makes this a serious concern. Li works in the Center for Eldercare and Rehabilitation Technology, which is run by Marjorie Skubic, a professor of electrical and computer engineering.

The microphone array is now being tested in a more realistic environment. Nabelek said they built a system over a weekend when they learned there was an opportunity to install it at TigerPlace. The active retirement community gives researchers the opportunity to test new technologies to help seniors live independently.

“We did a few shopping trips,” Nabelek said.

The sensor was placed in an apartment. Li and Nabelek covered the front of the rectangular sensor, where the microphones are visible on top of a sheet of plywood, with green-and-blue striped fabric.

“We put this in the living room,” Li said. “So we wanted to hide the cables so the people living there would be more comfortable.”

Li said the next challenge is to separate sounds that occur at the same time based on the source. The current system has not been tested with sounds occurring simultaneously.

“Now we are switching to doing more realistic situations,” Li said. “We have to deal with cases where fall sounds overlap with other sounds.”

Nabelek said if a television is on, it’s difficult to separate that noise from other sounds and if someone falls on the television audio, it may register as a fall. The solution is a technique called sound source separation.

“The source separation techniques developed so far are not that good,” Li said. He has been reading articles on the topic to find ways to improve the accuracy.

Following a growing trend in research, Li is also working to adapt what he’s achieved with the circular array to the Microsoft Kinect. The Kinect has only four microphones arranged in a line, which makes it more difficult to determine the location, especially the height of a sound. The depth sensor and the camera on the Kinect could be used to determine source location information, Li said.

Li expects to graduate in May 2013, so he doesn’t have much time to accomplish everything he wants to do. “I have to work hard,” he said. After he receives his doctorate, Li said he’d like to work in industry.

“I’d like to work at a company with lots of challenging work, that will push me every day,” Li said. “It would be best if I could develop my own company and do research and development.”

Nabelek is considering going into industry when he gets his master’s. Currently he works with detection and classification problems for the Computational Intelligence Signal Processing Laboratory.  Nabelek is the president of Mizzou Engineers Without Borders and also plays soccer.

Both Nabelek and Li are IEEE members. In addition to playing the piano, Li also swims and plays tennis and karaoke with friends.

Li said the research he’s done has been challenging but enjoyable.

“A good project like this is a good opportunity for students,” Li said. “In industry, they like to see that you can actually develop something.”