A life-transforming home monitoring system for the elderly can predict falls three weeks in advance.
Motion sensors detect tiny changes in walking speed, stride length, and speed of sitting up or down that warn of trouble ahead.
Help can then be summoned to prevent a tumble that could result in disability or even death.
Scientists who developed the artificially intelligent (AI) system found that an elderly person's risk of falling can increase more than four-fold if their walking speed slows.
The system, which uses an infrared projector, camera, and computing technology to track movement, has undergone preliminary tests at a residential apartment block in the US.
But its developers hope to produce a more simplified version that can be fitted into a person or couple's own home.
It would be capable of sending out email or text message alerts to carers or relatives.
Professor Marjorie Skubic, from the University of Missouri in the US, who has pioneered the project, said: "Our goal is to help people age in the home of their choice, which in many cases will be their existing home.
"There has been an assumption that there has to be a decline with age and a loss of independence.
"What we're showing is there doesn't have to be a decline. You can square the age curve."
With their full permission, Prof Skubic has tested the system on her elderly parents Lou, 95, and Mary-Ann, 93.
"Three weeks ago for my mother's 93rd birthday I installed the sensor system in their home," she said.
"They don't want to move. I hope and pray they spend the rest of their days there."
The system, known as Eldercare, produces moving 3D silhouettes that are analysed by AI software designed to flag up health warnings.
Walking speed, stride pattern, trunk sway and how long it takes a person to stand up or sit down are all taken into account.
"If your walking speed decreases by about five centimetres per second over a week, then there is an 86% probability that you are going to fall in the next three weeks," said Prof Skubic.
This compared with a 19.5% risk for those whose walking speed had stayed the same.
Similarly, a drop in stride length was associated with a 50.6% chance of falling over a three-week period.
Prof Skubic spoke about the system at the annual meeting of the American Association for the Advancement of Science (AAAS) in Boston, Massachusetts.
In the home in Columbia, Missouri, where Eldercare was trialled, using the sensor system allowed residents to remain independent of constant care for an extra 1.7 years.
Another part of the system monitored bathroom visits to look for signs of likely bladder infections, which can be serious for older people.
Acting on the health alerts, carers could step in with interventions such as balance and strength exercises, as well as medical prescriptions.
"There are treatments that can help people overcome a fall risk," said Prof Skubic.
One in three pensioners in the UK has had at least one fall in the past year, and falls cost the NHS an estimated £2.3 billion.
Caroline Abrahams, from the charity Age UK, said: " Falls are the most frequent cause of emergency hospital admissions for older people and a serious threat to their health and independence, causing pain, distress and loss of confidence.
"However, despite having serious consequences, falls in later life are often dismissed as an inevitable part of growing older, when the reality is they are preventable.
"Older people who have poor balance or difficulty walking are more likely than others to fall and this could be for a whole host of reasons, many of which can be addressed to reduce the risk.
"That's why we need a good falls service in every area but at the moment that's not where we are as they are really patchy.
"Investing in falls services could save some older people from the misery and distress of a fall and take some of the pressure off our hospitals too."