Professor Zhu Tingshao and colleagues at the Institute of Psychology have developed an accurate method of determining sleep quality through its impact on gait, with the help of a machine learning algorithm and a video game component, according to a paper published by online scientific journal PLOS One.
Zhu and his team found upper body parts such as head and shoulders betrayed more about a persons lack of sleep than lower parts like hips and legs. The researchers claim their method could flag the number of people in a crowd who have had insufficient sleep, providing a low cost way for public health authorities to measure the scale of a growing health problem, leading to more timely and effective policies to address the issue.
Poor sleep quality has been linked to everything from bad temper to obesity, work accidents and heart attacks. In China, about a quarter of the population has sleeping difficulties, according to a survey conducted by government hospitals last year. Other countries such as the US have reported similar or worse problems.
The soundness of sleep can be measured by wearing a smart watch or bracelet, donning a brainwave monitor or taking an MRI scan in hospital. These methods, however, require people's collaboration and cannot be used to monitor a population, especially when people are wide-awake.
Previous studies by other researchers suggested that inadequate sleep affected walking speed but the correlation had been too weak for practical use.
In contrast, the latest research paper suggests the system developed by Zhu and his team could monitor 360 types of movement taking place in all major joints of the human body.
To collect gait data quickly and on a large scale, Zhu's team turned to Kinect, the motion sensor add-on for Microsofts Xbox gaming consoles and, after extensive experiments, found two of the sensors were sufficient to make the eyes of their system.
However, the data captured straight out of the video game peripheral contained too much noise which hampered results so the researchers used a mathematical tool to get rid it by blurring the images without losing the useful signal.
Another challenge, especially in a crowd setting, was that each person approached the camera from a different distance and angle. The researchers solved this problem by tracking the hip joint of each participant and using a mathematical formula to convert the movements of all other body parts around the hip in to a standardised gait which could be recognised by a computer.
The system was designed to track each person for two minutes and select just two seconds of footage, captured when the subject was facing the camera, for in-depth analysis.
More than 50 first-year postgraduate students from the University of the Chinese Academy of Sciences in Beijing took part in the training and testing of the system, completing a sleep quality questionnaire before walking in front of the camera.
The correlation coefficient between the machine estimate and conclusion of the questionnaire was a highly positive 0.78, the researchers reported. The results indicate that gait patterns can reveal sleep quality quite well, they wrote.
The new method can measure sleep [quality] when people are awake in real time and remotely, and these advantages allow us to measure peoples sleep condition and obtain scores every day in a very short period of time without disturbances, which cannot be achieved by a questionnaire, they added.
Huo Yongquan, professor of psychology with the Shaanxi Normal University in Xian, central China, said existing methods of measuring sleep quality including questionnaires, surveys and brainwave monitoring had many limitations, including intrusiveness and inconvenience.
This new technology could be a useful item in our toolkit, said Huo, who was not involved in the study.
Zhu and his team are not the first to study human gait, with researchers around the world looking into its role in personal identification and emotion detection.
Assessing people by their gait also has the advantage that it can be used remotely, while other biometrics might become obscured, the new research paper said.
The authors said the only limit of their study had been the relatively small number of participants. In future studies, we plan to include large sample populations with individuals of different occupations, ages and cultural groups.