Fall monitoring systems add to seniors’ independence
A new monitoring system being developed by Professor Marjorie Skubic of the Electrical and Computer Engineering Department and her interdisciplinary team at the University of Missouri could help older adults avoid falls before they happen and alert health care professionals when they do.
Major injuries resulting from falls often lead to rapid decline in other aspects of the quality of life for older adults.
Preventing and reacting to falls can be a challenge to health care providers, even in retirement centers. When an older person chooses to stay at home, monitoring their movements has been difficult without a serious breach of privacy, like live video cameras, or extremely costly, like a live-in nurse.
A National Science Foundation grant for $1.4 million is funding Skubic’s research to develop various sensors — silhouette, acoustic and kinect — for fall detection and fall risk management, and colleague Marilyn Rantz, an MU curators professor in the school of nursing, has obtained funding to examine radar sensing. Two previous NSF grants totaling $2.1 million allowed for Skubic’s initial development of image-based silhouette and voxel sensing, successes that resulted in the additional funding.
STAYING HOME AND STAYING SAFE
“The goal is to allow independence and what’s called, ‘aging in place,’” said Skubic.
“People want to stay in their own homes, and this sensor system helps them to do just that. They get the benefits of constant monitoring without the loss of independence and privacy.
“We don’t want them to have to do anything different,” said Skubic. “That’s why the system uses sensors mounted in the environment. Elderly people don’t want to have to wear sensors.”
And what happens if they get up to use the bathroom in the middle of the night? That’s when many falls occur, but people are unlikely to wear sensors to bed or put them on in the middle of the night when they get up, Skubic said.
The non-invasive system Skubic and her team of researchers are investigating uses four separate types of sensing to develop a computer model of how an individual moves about and then watches for changes. Changes in a person’s movement, such as shuffling or limping, could be warning signs that a person is at greater risk for a fall.
This is a big improvement from current systems that are more like “On-Star” for the home, said Skubic. They only allow health professionals to react to possible falls, not prevent them from happening.
“I think the real power is in noticing subtle changes,” Skubic said. “Fixing small health problems can avoid a catastrophic health event.”
By monitoring a person’s motion patterns and gait, then watching for changes in the patterns, the computerized system can look for warning signs that a person’s risk for falling has increased.
The system can also be used to look for changes in gait and movements that might indicate the onset of Parkinson’s disease, Alzheimer’s disease or other dementia health conditions that also affect walking patterns.
It allows such detailed monitoring because the different sensor systems complement each other. A feature of one sensor can be used to compensate for a disadvantage in another. Fusing them together will provide an improved confidence in the accuracy of observations and warnings and keep down the number of false alarms.
HOW IT WORKS:
1. Silhouette imagery
This sensor system can tell whether a person is sitting, standing, or moving and can be used to extract gait parameters such as walking speed, step length, step time, gait symmetry, body sway, and sit-to-stand times. The capture of gait parameters has been validated through the use of a marker-based motion capture system often used for creating animations in movies and video games. In this work, it provides the ground truth to ensure that an accurate estimate is being captured.
It’s important to note that the system is not actually videotaping the patient, said Skubic. The system uses two calibrated cameras to create a 3-D voxel model from silhouettes in each image. The approach uses one-inch voxels, the 3-D version of a pixel. Features are extracted from the voxel model to compute gait parameters.
One of the challenges in developing this system was creating realistic models of what falling looks like. The system uses the models to watch for signs of real falls.
You don’t want older people falling down just so we can capture data, said Skubic. Instead she recruited MU theater graduate students, who had to learn how to fall like older people.
“It was fantastic to see them get into character!” said Skubic. “You could say the theater students really fell for us.”
Fans of video games may have used the Kinect system to play games without using controllers. The system monitors movement using infrared light to compute a depth image, providing a substitute for a stereo camera pair that has the advantage of requiring less computational resources.
The Kinect system works without visible light, unlike the silhouette system, and is not sensitive to lighting changes. However, it has a smaller field of view and is less accurate the further a person moves from the sensor.
3. Acoustic System
Detecting sound origins allows the system to determine the difference between a door slamming and a person hitting the floor. An array of microphones is used to identify the location of the sound source and then recognize an acoustic signature of a person falling down. The sound localization also helps to filter out possible false alarms due to sounds that occur above the ground level.
The challenge with the microphone system is developing algorithms that reliably capture the signature of a fall, regardless of the type of floor covering. The team is investigating acoustic features that can be used to detect a fall on any surface, such as carpeting, a wood floor or a tile floor.
Doppler radar can be used to measure the velocity of moving bodies, including people as they move about the home. In a current project collaborating with the GE Global Research Lab, the group is investigating the use of radar systems for capturing typical gait patterns as well as recognizing falls. One advantage of this system is that the radar signal is not obstructed by furniture placed in the room.
FALL MONITORING IN ACTION:
The integrated system will be tested at Tiger Place, a retirement community in Columbia, Mo., owned by AmeriCare, in collaboration with MU’s Sinclair School of Nursing.
“It’s gratifying to me as an engineer,” said Skubic about knowing that her work is helping people live better lives.
Developing this system had personal relevance for her as well. In her own family she has seen the difference that constant health monitoring can play in helping people stay healthy and active in later life.
Skubic’s elderly parents still live in their own home in Brookings, S. D. During their 65-year marriage they have watched out for each other and watched for changes in each other’s health.
“They do what we want the sensors to do,” said Skubic.
Skubic saw a sad example of what can happen without a monitoring system in the fate of her husband’s mother.
Her mother-in-law suffered from Parkinson’s disease. Late one night, on her way to the kitchen, she tripped and fell. She lay hurt on the floor for hours with a broken shoulder until her husband awoke. He suffers from hearing loss, and hadn’t heard her.
“If she had been at Tiger Place with our monitoring system, we could have sent help immediately,” said Skubic. “That makes it real.”
Others in the College of Engineering that are involved in this research are Professor Jim Keller, associate professor Henry He of electrical and computer engineering, Professor Dominic Ho and assistant professor Tony Han of electrical and computer engineering as well as others outside of engineering.