Smartwatch data help identify daily activities in real life

Scientists have long been able to use information from smartwatches to identify physical movement, such as sitting or walking, which are wearing in a controlled laboratory.

Now scientists from Washington State University have developed a way using a computer algorithm and a large set of data collected from smartwatches to comprehensively determine what people do in everyday conditions such as work, food, hobby or dealing with matters.

Work, published in, could once improve the assessment and understanding of cognitive health, rehabilitation, disease management or surgical revival. In their research, scientists were able to accurately identify 78% of the time.

“If we want to determine if someone needs care help in their home or elsewhere and at what level of help, we need to know how well this person can perform critical actions,” said Diane Cook, professor of regent WSU at the School of Electrical Engineering and WSU computer science, who conducted work. “How well they can bathe, feed, deal with finances or deal with your own things? They are things you really need to be independent.”

One of the biggest challenges in healthcare is the assessment of how sick or older people manage their daily lives. Medical specialists often need more comprehensive information about how a person performed functional activities or behavior focused on a goal to really assess their health. As he knows that everyone who tries to help a distant parent with aging or health, information about how well a person results from paying bills, arranging matters or cooking meals, is complex, variable and difficult to collect – in a doctor’s office or with a smartwatch.

Lack of awareness of the cognitive and physical status of man is one of the obstacles that we stand with as we age, and thus having an automated way to indicate where a person is, will allow us to better intervene for them and keep not only healthy, but perfectly independent. This work is the basis for more advanced, conscious behavior of application in the field of digital health and artificial human intelligence. “

Diane Cook, WSU Regents Professor, WSU’s School of Electrical Engineering and Computer Science

For their research, WSU researchers collected information about activity for several years from several studies.

“Whenever we had a study that collected the smartwatch, we added a question to our application for collecting data that asked the participants for their own marking of their current activity, and so we finished with so many participants on so many research,” she said. “And then we just dug up whether we can do the recognition of activity.”

503 study participants within eight years were asked at random times during the day to select the 12 categories from the scrolling list to describe what they were doing. Categories included such matters, sleeping, traveling, work, food, social contacts or relaxation. Scientists have analyzed various methods of artificial intelligence in terms of their ability to generalize in the population of research participants.

Scientists have developed a large set of data, which covers over 32 million marked data points, with each point represented one minute of activity. Then they trained the AI model to predict what functional activity took place. They were able to predict actions up to 77.7% of the time.

“The basic step is to recognize activity, because if we can describe the behavior of a person in terms of activities in terms that are well recognized, we can start talking about their behavioral patterns and changes in their patterns,” Cook said. “We can use what we sense to get a closer to traditional measures of health, such as cognitive health and functional independence.”

Scientists hope that they use their model in future research in such areas as the ability to automate clinical diagnoses and search for connections between behavior, health, genetics and environment. Methods and set of data without any identifying information are also publicly available to other researchers. The work was financed by the National Institutes of Health.