29 January 2021
Pan Zhang
Hitachi (China) Research & Development Corporation
China is facing a major challenge as it works to address an aging population. It is predicted that over 480 million Chinese will be over the age of 60 by 2050 [1]. Under the challenge, the Government of China is promoting initiatives for healthy longevity and quality of life improvement for the elderly by issuing nursing and healthcare guidelines, such as "Healthy China 2030" [2]. To support this regional need and government initiative, Hitachi China (HCH) has created a new solution, Digital Care Management (DCM), that utilizes advanced IoT sensors to digitalize the physical or cognitive condition of elderly citizens, to provide devise services for prevention, early diagnosis and intervention [3]. The platform which has been available since 2018, is shown in Figure 1.
Figure 1: DCM solution
To contribute to the DCM solution, my colleagues and I at Hitachi (China) R&D looked at "gait" as it is directly related to the health status of the elderly. As changes or abnormalities in gait may reflect health risks, we proposed an automatic and privacy-considered gait analysis approach using a stereo camera to digitalize gait for DCM solutions.
Figure 2 shows the process flow of our approach. This approach applies a cutting-edge deep learning technology to detect a human subject in 2D images, and then combines 3D sensing data to measure gait features, such as speed and stride length.
Figure 2: Process flow of gait analysis based on stereo camera
The approach was evaluated for the accuracy of gait features compared with different cameras for various walking patterns and camera settings. Figure 3 shows the evaluation results of step length with specified camera settings and walking patterns. It is observed that our approach gave stable and more accurate results than the other approaches.
Figure 3: Evaluation results of step length with a unified step length of 0.5m
(*Kinect is a motion sensor produced by Microsoft, and is a trademark or registered trademark of Microsoft Corporation in the US and other countries.)
We conducted experiments at a daycare facility for the elderly to validate the correlation between gait features and scores indicating the risk of falling over (fall risk). The experimental system and the test environment are shown in Figure 4. Forty elderly persons volunteered for the experiments, and we captured weeks of video datasets containing Timed Up and Go (TUG) [4] and Performance Oriented Mobility Assessment (POMA) [5] tests, which are methods to assess mobility and balance and walking activity. We then used a polynomial regression method to predict TUG and POMA scores based on extracted gait features. The result showed that our gait analysis approach could provide suitable gait features to estimate TUG or POMA scores to predict fall risk.
Figure 4: Experimental system and the test environment at a daycare facility in China
Our gait analysis approach for measuring gait features using a stereo camera is flexible to use and accurate compared to other image sensors. The experiment at the daycare facility showed that our approach can be used in daily life without interfering with the activities of elderly subjects or their comfort. In addition, the proposed approach provides a fall risk prediction method by fall risk scores estimation based on gait to help introduce countermeasure to prevent falls from happening.
For more details, we encourage you to read our paper, "Gait Analysis Using Stereo Camera in Daily Environment", which was presented at the 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), and can be accessed at https://ieeexplore.ieee.org/document/8857494.
Thanks to my co-authors Yuan Li, Yang Zhang and Kunihiko Miyazaki, with whom this research work was jointly executed.
搜索
China